summaryrefslogtreecommitdiffstats
path: root/xlators/cluster/afr/src
diff options
context:
space:
mode:
authorGluster Ant <bugzilla-bot@gluster.org>2018-09-12 17:52:45 +0530
committerNigel Babu <nigelb@redhat.com>2018-09-12 17:52:45 +0530
commite16868dede6455cab644805af6fe1ac312775e13 (patch)
tree15aebdb4fff2d87cf8a72f836816b3aa634da58d /xlators/cluster/afr/src
parent45a71c0548b6fd2c757aa2e7b7671a1411948894 (diff)
Land part 2 of clang-format changes
Change-Id: Ia84cc24c8924e6d22d02ac15f611c10e26db99b4 Signed-off-by: Nigel Babu <nigelb@redhat.com>
Diffstat (limited to 'xlators/cluster/afr/src')
-rw-r--r--xlators/cluster/afr/src/afr-common.c10685
-rw-r--r--xlators/cluster/afr/src/afr-dir-read.c493
-rw-r--r--xlators/cluster/afr/src/afr-dir-write.c2270
-rw-r--r--xlators/cluster/afr/src/afr-inode-read.c2873
-rw-r--r--xlators/cluster/afr/src/afr-inode-write.c3534
-rw-r--r--xlators/cluster/afr/src/afr-lk-common.c1683
-rw-r--r--xlators/cluster/afr/src/afr-open.c526
-rw-r--r--xlators/cluster/afr/src/afr-read-txn.c752
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-common.c4108
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-data.c1469
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-entry.c1902
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-metadata.c820
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-name.c1091
-rw-r--r--xlators/cluster/afr/src/afr-self-heald.c2151
-rw-r--r--xlators/cluster/afr/src/afr-transaction.c3887
-rw-r--r--xlators/cluster/afr/src/afr.c2045
16 files changed, 19717 insertions, 20572 deletions
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index 3a32ebc31a4..eb0e7330a91 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -42,157 +42,155 @@
#include "afr-messages.h"
int32_t
-afr_quorum_errno (afr_private_t *priv)
+afr_quorum_errno(afr_private_t *priv)
{
- return ENOTCONN;
+ return ENOTCONN;
}
int
-afr_fav_child_reset_sink_xattrs (void *opaque);
+afr_fav_child_reset_sink_xattrs(void *opaque);
int
-afr_fav_child_reset_sink_xattrs_cbk (int ret, call_frame_t *frame,
- void *opaque);
+afr_fav_child_reset_sink_xattrs_cbk(int ret, call_frame_t *frame, void *opaque);
static void
-afr_discover_done (call_frame_t *frame, xlator_t *this);
+afr_discover_done(call_frame_t *frame, xlator_t *this);
gf_boolean_t
-afr_is_consistent_io_possible (afr_local_t *local, afr_private_t *priv,
- int32_t *op_errno)
-{
- if (priv->consistent_io && local->call_count != priv->child_count) {
- gf_msg (THIS->name, GF_LOG_INFO, 0,
- AFR_MSG_SUBVOLS_DOWN, "All subvolumes are not up");
- if (op_errno)
- *op_errno = ENOTCONN;
- return _gf_false;
- }
- return _gf_true;
+afr_is_consistent_io_possible(afr_local_t *local, afr_private_t *priv,
+ int32_t *op_errno)
+{
+ if (priv->consistent_io && local->call_count != priv->child_count) {
+ gf_msg(THIS->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOLS_DOWN,
+ "All subvolumes are not up");
+ if (op_errno)
+ *op_errno = ENOTCONN;
+ return _gf_false;
+ }
+ return _gf_true;
}
call_frame_t *
-afr_copy_frame (call_frame_t *base)
+afr_copy_frame(call_frame_t *base)
{
- afr_local_t *local = NULL;
- call_frame_t *frame = NULL;
- int op_errno = 0;
+ afr_local_t *local = NULL;
+ call_frame_t *frame = NULL;
+ int op_errno = 0;
- frame = copy_frame (base);
- if (!frame)
- return NULL;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local) {
- AFR_STACK_DESTROY (frame);
- return NULL;
- }
+ frame = copy_frame(base);
+ if (!frame)
+ return NULL;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local) {
+ AFR_STACK_DESTROY(frame);
+ return NULL;
+ }
- return frame;
+ return frame;
}
/* Check if an entry or inode could be undergoing a transaction. */
gf_boolean_t
-afr_is_possibly_under_txn (afr_transaction_type type, afr_local_t *local,
- xlator_t *this)
-{
- int i = 0;
- int tmp = 0;
- afr_private_t *priv = NULL;
- GF_UNUSED char *key = NULL;
-
- priv = this->private;
-
- if (type == AFR_ENTRY_TRANSACTION)
- key = GLUSTERFS_PARENT_ENTRYLK;
- else if (type == AFR_DATA_TRANSACTION)
- /*FIXME: Use GLUSTERFS_INODELK_DOM_COUNT etc. once
- * pl_inodelk_xattr_fill supports separate keys for different
- * domains.*/
- key = GLUSTERFS_INODELK_COUNT;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].xdata)
- continue;
- if (dict_get_int32 (local->replies[i].xdata, key, &tmp) == 0)
- if (tmp)
- return _gf_true;
- }
+afr_is_possibly_under_txn(afr_transaction_type type, afr_local_t *local,
+ xlator_t *this)
+{
+ int i = 0;
+ int tmp = 0;
+ afr_private_t *priv = NULL;
+ GF_UNUSED char *key = NULL;
+
+ priv = this->private;
+
+ if (type == AFR_ENTRY_TRANSACTION)
+ key = GLUSTERFS_PARENT_ENTRYLK;
+ else if (type == AFR_DATA_TRANSACTION)
+ /*FIXME: Use GLUSTERFS_INODELK_DOM_COUNT etc. once
+ * pl_inodelk_xattr_fill supports separate keys for different
+ * domains.*/
+ key = GLUSTERFS_INODELK_COUNT;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].xdata)
+ continue;
+ if (dict_get_int32(local->replies[i].xdata, key, &tmp) == 0)
+ if (tmp)
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static void
-afr_inode_ctx_destroy (afr_inode_ctx_t *ctx)
+afr_inode_ctx_destroy(afr_inode_ctx_t *ctx)
{
- int i = 0;
+ int i = 0;
- if (!ctx)
- return;
+ if (!ctx)
+ return;
- for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
- GF_FREE (ctx->pre_op_done[i]);
- }
+ for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
+ GF_FREE(ctx->pre_op_done[i]);
+ }
- GF_FREE (ctx);
+ GF_FREE(ctx);
}
int
-__afr_inode_ctx_get (xlator_t *this, inode_t *inode, afr_inode_ctx_t **ctx)
-{
- uint64_t ctx_int = 0;
- int ret = -1;
- int i = -1;
- int num_locks = -1;
- afr_inode_ctx_t *ictx = NULL;
- afr_lock_t *lock = NULL;
- afr_private_t *priv = this->private;
-
- ret = __inode_ctx_get (inode, this, &ctx_int);
- if (ret == 0) {
- *ctx = (afr_inode_ctx_t *)ctx_int;
- return 0;
- }
-
- ictx = GF_CALLOC (1, sizeof (afr_inode_ctx_t), gf_afr_mt_inode_ctx_t);
- if (!ictx)
- goto out;
-
- for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
- ictx->pre_op_done[i] = GF_CALLOC (sizeof *ictx->pre_op_done[i],
- priv->child_count,
- gf_afr_mt_int32_t);
- if (!ictx->pre_op_done[i]) {
- ret = -ENOMEM;
- goto out;
- }
- }
-
- num_locks = sizeof(ictx->lock)/sizeof(afr_lock_t);
- for (i = 0; i < num_locks; i++) {
- lock = &ictx->lock[i];
- INIT_LIST_HEAD (&lock->post_op);
- INIT_LIST_HEAD (&lock->frozen);
- INIT_LIST_HEAD (&lock->waiting);
- INIT_LIST_HEAD (&lock->owners);
- }
-
- ctx_int = (uint64_t)ictx;
- ret = __inode_ctx_set (inode, this, &ctx_int);
- if (ret) {
- goto out;
- }
-
- ictx->spb_choice = -1;
- ictx->read_subvol = 0;
- ictx->write_subvol = 0;
- ictx->lock_count = 0;
- ret = 0;
- *ctx = ictx;
+__afr_inode_ctx_get(xlator_t *this, inode_t *inode, afr_inode_ctx_t **ctx)
+{
+ uint64_t ctx_int = 0;
+ int ret = -1;
+ int i = -1;
+ int num_locks = -1;
+ afr_inode_ctx_t *ictx = NULL;
+ afr_lock_t *lock = NULL;
+ afr_private_t *priv = this->private;
+
+ ret = __inode_ctx_get(inode, this, &ctx_int);
+ if (ret == 0) {
+ *ctx = (afr_inode_ctx_t *)ctx_int;
+ return 0;
+ }
+
+ ictx = GF_CALLOC(1, sizeof(afr_inode_ctx_t), gf_afr_mt_inode_ctx_t);
+ if (!ictx)
+ goto out;
+
+ for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
+ ictx->pre_op_done[i] = GF_CALLOC(sizeof *ictx->pre_op_done[i],
+ priv->child_count, gf_afr_mt_int32_t);
+ if (!ictx->pre_op_done[i]) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ }
+
+ num_locks = sizeof(ictx->lock) / sizeof(afr_lock_t);
+ for (i = 0; i < num_locks; i++) {
+ lock = &ictx->lock[i];
+ INIT_LIST_HEAD(&lock->post_op);
+ INIT_LIST_HEAD(&lock->frozen);
+ INIT_LIST_HEAD(&lock->waiting);
+ INIT_LIST_HEAD(&lock->owners);
+ }
+
+ ctx_int = (uint64_t)ictx;
+ ret = __inode_ctx_set(inode, this, &ctx_int);
+ if (ret) {
+ goto out;
+ }
+
+ ictx->spb_choice = -1;
+ ictx->read_subvol = 0;
+ ictx->write_subvol = 0;
+ ictx->lock_count = 0;
+ ret = 0;
+ *ctx = ictx;
out:
- if (ret) {
- afr_inode_ctx_destroy (ictx);
- }
- return ret;
+ if (ret) {
+ afr_inode_ctx_destroy(ictx);
+ }
+ return ret;
}
/*
@@ -226,1723 +224,1670 @@ out:
*/
int
-__afr_set_in_flight_sb_status (xlator_t *this, afr_local_t *local,
- inode_t *inode)
-{
- int i = 0;
- int txn_type = 0;
- int count = 0;
- int index = -1;
- uint16_t datamap_old = 0;
- uint16_t metadatamap_old = 0;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint16_t tmp_map = 0;
- uint16_t mask = 0;
- uint32_t event = 0;
- uint64_t val = 0;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- txn_type = local->transaction.type;
-
- if (txn_type == AFR_DATA_TRANSACTION)
- val = local->inode_ctx->write_subvol;
- else
- val = local->inode_ctx->read_subvol;
-
- metadatamap_old = metadatamap = (val & 0x000000000000ffff);
- datamap_old = datamap = (val & 0x00000000ffff0000) >> 16;
- event = (val & 0xffffffff00000000) >> 32;
-
- if (txn_type == AFR_DATA_TRANSACTION)
- tmp_map = datamap;
- else if (txn_type == AFR_METADATA_TRANSACTION)
- tmp_map = metadatamap;
-
- count = gf_bits_count (tmp_map);
-
- if (count == 1)
- index = gf_bits_index (tmp_map);
-
- for (i = 0; i < priv->child_count; i++) {
- mask = 0;
- if (!local->transaction.failed_subvols[i])
- continue;
-
- mask = 1 << i;
- if (txn_type == AFR_METADATA_TRANSACTION)
- metadatamap &= ~mask;
- else if (txn_type == AFR_DATA_TRANSACTION)
- datamap &= ~mask;
- }
-
- switch (txn_type) {
+__afr_set_in_flight_sb_status(xlator_t *this, afr_local_t *local,
+ inode_t *inode)
+{
+ int i = 0;
+ int txn_type = 0;
+ int count = 0;
+ int index = -1;
+ uint16_t datamap_old = 0;
+ uint16_t metadatamap_old = 0;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint16_t tmp_map = 0;
+ uint16_t mask = 0;
+ uint32_t event = 0;
+ uint64_t val = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ txn_type = local->transaction.type;
+
+ if (txn_type == AFR_DATA_TRANSACTION)
+ val = local->inode_ctx->write_subvol;
+ else
+ val = local->inode_ctx->read_subvol;
+
+ metadatamap_old = metadatamap = (val & 0x000000000000ffff);
+ datamap_old = datamap = (val & 0x00000000ffff0000) >> 16;
+ event = (val & 0xffffffff00000000) >> 32;
+
+ if (txn_type == AFR_DATA_TRANSACTION)
+ tmp_map = datamap;
+ else if (txn_type == AFR_METADATA_TRANSACTION)
+ tmp_map = metadatamap;
+
+ count = gf_bits_count(tmp_map);
+
+ if (count == 1)
+ index = gf_bits_index(tmp_map);
+
+ for (i = 0; i < priv->child_count; i++) {
+ mask = 0;
+ if (!local->transaction.failed_subvols[i])
+ continue;
+
+ mask = 1 << i;
+ if (txn_type == AFR_METADATA_TRANSACTION)
+ metadatamap &= ~mask;
+ else if (txn_type == AFR_DATA_TRANSACTION)
+ datamap &= ~mask;
+ }
+
+ switch (txn_type) {
case AFR_METADATA_TRANSACTION:
- if ((metadatamap_old != 0) && (metadatamap == 0) &&
- (count == 1)) {
- local->transaction.in_flight_sb_errno =
- local->replies[index].op_errno;
- local->transaction.in_flight_sb = _gf_true;
- metadatamap |= (1 << index);
- }
- if (metadatamap_old != metadatamap) {
- event = 0;
- }
- break;
+ if ((metadatamap_old != 0) && (metadatamap == 0) && (count == 1)) {
+ local->transaction.in_flight_sb_errno = local->replies[index]
+ .op_errno;
+ local->transaction.in_flight_sb = _gf_true;
+ metadatamap |= (1 << index);
+ }
+ if (metadatamap_old != metadatamap) {
+ event = 0;
+ }
+ break;
case AFR_DATA_TRANSACTION:
- if ((datamap_old != 0) && (datamap == 0) && (count == 1)) {
- local->transaction.in_flight_sb_errno =
- local->replies[index].op_errno;
- local->transaction.in_flight_sb = _gf_true;
- datamap |= (1 << index);
- }
- if (datamap_old != datamap)
- event = 0;
- break;
+ if ((datamap_old != 0) && (datamap == 0) && (count == 1)) {
+ local->transaction.in_flight_sb_errno = local->replies[index]
+ .op_errno;
+ local->transaction.in_flight_sb = _gf_true;
+ datamap |= (1 << index);
+ }
+ if (datamap_old != datamap)
+ event = 0;
+ break;
default:
- break;
- }
+ break;
+ }
- val = ((uint64_t) metadatamap) |
- (((uint64_t) datamap) << 16) |
- (((uint64_t) event) << 32);
+ val = ((uint64_t)metadatamap) | (((uint64_t)datamap) << 16) |
+ (((uint64_t)event) << 32);
- if (txn_type == AFR_DATA_TRANSACTION)
- local->inode_ctx->write_subvol = val;
- local->inode_ctx->read_subvol = val;
+ if (txn_type == AFR_DATA_TRANSACTION)
+ local->inode_ctx->write_subvol = val;
+ local->inode_ctx->read_subvol = val;
- return 0;
+ return 0;
}
gf_boolean_t
-afr_is_symmetric_error (call_frame_t *frame, xlator_t *this)
+afr_is_symmetric_error(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int op_errno = 0;
- int i_errno = 0;
- gf_boolean_t matching_errors = _gf_true;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int op_errno = 0;
+ int i_errno = 0;
+ gf_boolean_t matching_errors = _gf_true;
+ int i = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret != -1) {
- /* Operation succeeded on at least one subvol,
- so it is not a failed-everywhere situation.
- */
- matching_errors = _gf_false;
- break;
- }
- i_errno = local->replies[i].op_errno;
-
- if (i_errno == ENOTCONN) {
- /* ENOTCONN is not a symmetric error. We do not
- know if the operation was performed on the
- backend or not.
- */
- matching_errors = _gf_false;
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret != -1) {
+ /* Operation succeeded on at least one subvol,
+ so it is not a failed-everywhere situation.
+ */
+ matching_errors = _gf_false;
+ break;
+ }
+ i_errno = local->replies[i].op_errno;
- if (!op_errno) {
- op_errno = i_errno;
- } else if (op_errno != i_errno) {
- /* Mismatching op_errno's */
- matching_errors = _gf_false;
- break;
- }
+ if (i_errno == ENOTCONN) {
+ /* ENOTCONN is not a symmetric error. We do not
+ know if the operation was performed on the
+ backend or not.
+ */
+ matching_errors = _gf_false;
+ break;
+ }
+
+ if (!op_errno) {
+ op_errno = i_errno;
+ } else if (op_errno != i_errno) {
+ /* Mismatching op_errno's */
+ matching_errors = _gf_false;
+ break;
}
+ }
- return matching_errors;
+ return matching_errors;
}
int
-afr_set_in_flight_sb_status (xlator_t *this, call_frame_t *frame,
- inode_t *inode)
+afr_set_in_flight_sb_status(xlator_t *this, call_frame_t *frame, inode_t *inode)
{
- int ret = -1;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ int ret = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- /* If this transaction saw no failures, then exit. */
- if (AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count) == 0)
- return 0;
+ /* If this transaction saw no failures, then exit. */
+ if (AFR_COUNT(local->transaction.failed_subvols, priv->child_count) == 0)
+ return 0;
- if (afr_is_symmetric_error (frame, this))
- return 0;
+ if (afr_is_symmetric_error(frame, this))
+ return 0;
- LOCK (&inode->lock);
- {
- ret = __afr_set_in_flight_sb_status (this, local, inode);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_set_in_flight_sb_status(this, local, inode);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-__afr_inode_read_subvol_get_small (inode_t *inode, xlator_t *this,
- unsigned char *data, unsigned char *metadata,
- int *event_p)
-{
- afr_private_t *priv = NULL;
- int ret = -1;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint32_t event = 0;
- uint64_t val = 0;
- int i = 0;
- afr_inode_ctx_t *ctx = NULL;
-
- priv = this->private;
-
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret < 0)
- return ret;
+__afr_inode_read_subvol_get_small(inode_t *inode, xlator_t *this,
+ unsigned char *data, unsigned char *metadata,
+ int *event_p)
+{
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint32_t event = 0;
+ uint64_t val = 0;
+ int i = 0;
+ afr_inode_ctx_t *ctx = NULL;
+
+ priv = this->private;
+
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret < 0)
+ return ret;
- val = ctx->read_subvol;
+ val = ctx->read_subvol;
- metadatamap = (val & 0x000000000000ffff);
- datamap = (val & 0x00000000ffff0000) >> 16;
- event = (val & 0xffffffff00000000) >> 32;
+ metadatamap = (val & 0x000000000000ffff);
+ datamap = (val & 0x00000000ffff0000) >> 16;
+ event = (val & 0xffffffff00000000) >> 32;
- for (i = 0; i < priv->child_count; i++) {
- if (metadata)
- metadata[i] = (metadatamap >> i) & 1;
- if (data)
- data[i] = (datamap >> i) & 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (metadata)
+ metadata[i] = (metadatamap >> i) & 1;
+ if (data)
+ data[i] = (datamap >> i) & 1;
+ }
- if (event_p)
- *event_p = event;
- return ret;
+ if (event_p)
+ *event_p = event;
+ return ret;
}
-
int
-__afr_inode_read_subvol_set_small (inode_t *inode, xlator_t *this,
- unsigned char *data, unsigned char *metadata,
- int event)
+__afr_inode_read_subvol_set_small(inode_t *inode, xlator_t *this,
+ unsigned char *data, unsigned char *metadata,
+ int event)
{
- afr_private_t *priv = NULL;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint64_t val = 0;
- int i = 0;
- int ret = -1;
- afr_inode_ctx_t *ctx = NULL;
+ afr_private_t *priv = NULL;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint64_t val = 0;
+ int i = 0;
+ int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
- priv = this->private;
+ priv = this->private;
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- goto out;
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret)
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- if (data[i])
- datamap |= (1 << i);
- if (metadata[i])
- metadatamap |= (1 << i);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (data[i])
+ datamap |= (1 << i);
+ if (metadata[i])
+ metadatamap |= (1 << i);
+ }
- val = ((uint64_t) metadatamap) |
- (((uint64_t) datamap) << 16) |
- (((uint64_t) event) << 32);
+ val = ((uint64_t)metadatamap) | (((uint64_t)datamap) << 16) |
+ (((uint64_t)event) << 32);
- ctx->read_subvol = val;
+ ctx->read_subvol = val;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-__afr_inode_event_gen_reset_small (inode_t *inode, xlator_t *this)
+__afr_inode_event_gen_reset_small(inode_t *inode, xlator_t *this)
{
- int ret = -1;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint32_t event = 0;
- uint64_t val = 0;
- afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint32_t event = 0;
+ uint64_t val = 0;
+ afr_inode_ctx_t *ctx = NULL;
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- return ret;
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret)
+ return ret;
- val = ctx->read_subvol;
+ val = ctx->read_subvol;
- metadatamap = (val & 0x000000000000ffff) >> 0;
- datamap = (val & 0x00000000ffff0000) >> 16;
- event = 0;
+ metadatamap = (val & 0x000000000000ffff) >> 0;
+ datamap = (val & 0x00000000ffff0000) >> 16;
+ event = 0;
- val = ((uint64_t) metadatamap) |
- (((uint64_t) datamap) << 16) |
- (((uint64_t) event) << 32);
+ val = ((uint64_t)metadatamap) | (((uint64_t)datamap) << 16) |
+ (((uint64_t)event) << 32);
- ctx->read_subvol = val;
+ ctx->read_subvol = val;
- return ret;
+ return ret;
}
-
int
-__afr_inode_read_subvol_get (inode_t *inode, xlator_t *this,
- unsigned char *data, unsigned char *metadata,
- int *event_p)
+__afr_inode_read_subvol_get(inode_t *inode, xlator_t *this, unsigned char *data,
+ unsigned char *metadata, int *event_p)
{
- afr_private_t *priv = NULL;
- int ret = -1;
+ afr_private_t *priv = NULL;
+ int ret = -1;
- priv = this->private;
+ priv = this->private;
- if (priv->child_count <= 16)
- ret = __afr_inode_read_subvol_get_small (inode, this, data,
- metadata, event_p);
- else
- /* TBD: allocate structure with array and read from it */
- ret = -1;
+ if (priv->child_count <= 16)
+ ret = __afr_inode_read_subvol_get_small(inode, this, data, metadata,
+ event_p);
+ else
+ /* TBD: allocate structure with array and read from it */
+ ret = -1;
- return ret;
+ return ret;
}
int
-__afr_inode_split_brain_choice_get (inode_t *inode, xlator_t *this,
- int *spb_choice)
+__afr_inode_split_brain_choice_get(inode_t *inode, xlator_t *this,
+ int *spb_choice)
{
- afr_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret < 0)
- return ret;
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret < 0)
+ return ret;
- *spb_choice = ctx->spb_choice;
- return 0;
+ *spb_choice = ctx->spb_choice;
+ return 0;
}
int
-__afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
- unsigned char *metadata, int event)
+__afr_inode_read_subvol_set(inode_t *inode, xlator_t *this, unsigned char *data,
+ unsigned char *metadata, int event)
{
- afr_private_t *priv = NULL;
- int ret = -1;
+ afr_private_t *priv = NULL;
+ int ret = -1;
- priv = this->private;
+ priv = this->private;
- if (priv->child_count <= 16)
- ret = __afr_inode_read_subvol_set_small (inode, this, data,
- metadata, event);
- else
- ret = -1;
+ if (priv->child_count <= 16)
+ ret = __afr_inode_read_subvol_set_small(inode, this, data, metadata,
+ event);
+ else
+ ret = -1;
- return ret;
+ return ret;
}
int
-__afr_inode_split_brain_choice_set (inode_t *inode, xlator_t *this,
- int spb_choice)
+__afr_inode_split_brain_choice_set(inode_t *inode, xlator_t *this,
+ int spb_choice)
{
- afr_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- goto out;
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret)
+ goto out;
- ctx->spb_choice = spb_choice;
+ ctx->spb_choice = spb_choice;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-__afr_inode_event_gen_reset (inode_t *inode, xlator_t *this)
+__afr_inode_event_gen_reset(inode_t *inode, xlator_t *this)
{
- afr_private_t *priv = NULL;
- int ret = -1;
+ afr_private_t *priv = NULL;
+ int ret = -1;
- priv = this->private;
+ priv = this->private;
- if (priv->child_count <= 16)
- ret = __afr_inode_event_gen_reset_small (inode, this);
- else
- ret = -1;
+ if (priv->child_count <= 16)
+ ret = __afr_inode_event_gen_reset_small(inode, this);
+ else
+ ret = -1;
- return ret;
+ return ret;
}
-
int
-afr_inode_read_subvol_get (inode_t *inode, xlator_t *this, unsigned char *data,
- unsigned char *metadata, int *event_p)
+afr_inode_read_subvol_get(inode_t *inode, xlator_t *this, unsigned char *data,
+ unsigned char *metadata, int *event_p)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_read_subvol_get (inode, this, data,
- metadata, event_p);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_read_subvol_get(inode, this, data, metadata, event_p);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
int
-afr_inode_get_readable (call_frame_t *frame, inode_t *inode, xlator_t *this,
+afr_inode_get_readable(call_frame_t *frame, inode_t *inode, xlator_t *this,
unsigned char *readable, int *event_p, int type)
{
-
- afr_private_t *priv = this->private;
- afr_local_t *local = frame->local;
- unsigned char *data = alloca0 (priv->child_count);
- unsigned char *metadata = alloca0 (priv->child_count);
- int data_count = 0;
- int metadata_count = 0;
- int event_generation = 0;
- int ret = 0;
-
- ret = afr_inode_read_subvol_get (inode, this, data, metadata,
- &event_generation);
- if (ret == -1)
+ afr_private_t *priv = this->private;
+ afr_local_t *local = frame->local;
+ unsigned char *data = alloca0(priv->child_count);
+ unsigned char *metadata = alloca0(priv->child_count);
+ int data_count = 0;
+ int metadata_count = 0;
+ int event_generation = 0;
+ int ret = 0;
+
+ ret = afr_inode_read_subvol_get(inode, this, data, metadata,
+ &event_generation);
+ if (ret == -1)
+ return -EIO;
+
+ data_count = AFR_COUNT(data, priv->child_count);
+ metadata_count = AFR_COUNT(metadata, priv->child_count);
+
+ if (inode->ia_type == IA_IFDIR) {
+ /* For directories, allow even if it is in data split-brain. */
+ if (type == AFR_METADATA_TRANSACTION || local->op == GF_FOP_STAT ||
+ local->op == GF_FOP_FSTAT) {
+ if (!metadata_count)
return -EIO;
-
- data_count = AFR_COUNT (data, priv->child_count);
- metadata_count = AFR_COUNT (metadata, priv->child_count);
-
- if (inode->ia_type == IA_IFDIR) {
- /* For directories, allow even if it is in data split-brain. */
- if (type == AFR_METADATA_TRANSACTION ||
- local->op == GF_FOP_STAT || local->op == GF_FOP_FSTAT) {
- if (!metadata_count)
- return -EIO;
- }
- } else {
- /* For files, abort in case of data/metadata split-brain. */
- if (!data_count || !metadata_count) {
- return -EIO;
- }
}
-
- if (type == AFR_METADATA_TRANSACTION && readable)
- memcpy (readable, metadata, priv->child_count * sizeof *metadata);
- if (type == AFR_DATA_TRANSACTION && readable) {
- if (!data_count)
- memcpy (readable, local->child_up,
- priv->child_count * sizeof *readable);
- else
- memcpy (readable, data, priv->child_count * sizeof *data);
+ } else {
+ /* For files, abort in case of data/metadata split-brain. */
+ if (!data_count || !metadata_count) {
+ return -EIO;
}
- if (event_p)
- *event_p = event_generation;
- return 0;
+ }
+
+ if (type == AFR_METADATA_TRANSACTION && readable)
+ memcpy(readable, metadata, priv->child_count * sizeof *metadata);
+ if (type == AFR_DATA_TRANSACTION && readable) {
+ if (!data_count)
+ memcpy(readable, local->child_up,
+ priv->child_count * sizeof *readable);
+ else
+ memcpy(readable, data, priv->child_count * sizeof *data);
+ }
+ if (event_p)
+ *event_p = event_generation;
+ return 0;
}
int
-afr_inode_split_brain_choice_get (inode_t *inode, xlator_t *this,
- int *spb_choice)
+afr_inode_split_brain_choice_get(inode_t *inode, xlator_t *this,
+ int *spb_choice)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_split_brain_choice_get (inode, this,
- spb_choice);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_split_brain_choice_get(inode, this, spb_choice);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
-
int
-afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
- unsigned char *metadata, int event)
+afr_inode_read_subvol_set(inode_t *inode, xlator_t *this, unsigned char *data,
+ unsigned char *metadata, int event)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_read_subvol_set (inode, this, data, metadata,
- event);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_read_subvol_set(inode, this, data, metadata, event);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
-
int
-afr_inode_split_brain_choice_set (inode_t *inode, xlator_t *this,
- int spb_choice)
+afr_inode_split_brain_choice_set(inode_t *inode, xlator_t *this, int spb_choice)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_split_brain_choice_set (inode, this,
- spb_choice);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_split_brain_choice_set(inode, this, spb_choice);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
-
/* The caller of this should perform afr_inode_refresh, if this function
* returns _gf_true
*/
gf_boolean_t
-afr_is_inode_refresh_reqd (inode_t *inode, xlator_t *this,
- int event_gen1, int event_gen2)
+afr_is_inode_refresh_reqd(inode_t *inode, xlator_t *this, int event_gen1,
+ int event_gen2)
{
- gf_boolean_t need_refresh = _gf_false;
- afr_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ gf_boolean_t need_refresh = _gf_false;
+ afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- goto unlock;
-
- need_refresh = ctx->need_refresh;
- /* Hoping that the caller will do inode_refresh followed by
- * this, hence setting the need_refresh to false */
- ctx->need_refresh = _gf_false;
- }
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret)
+ goto unlock;
+
+ need_refresh = ctx->need_refresh;
+ /* Hoping that the caller will do inode_refresh followed by
+ * this, hence setting the need_refresh to false */
+ ctx->need_refresh = _gf_false;
+ }
unlock:
- UNLOCK(&inode->lock);
+ UNLOCK(&inode->lock);
- if (event_gen1 != event_gen2)
- need_refresh = _gf_true;
+ if (event_gen1 != event_gen2)
+ need_refresh = _gf_true;
out:
- return need_refresh;
+ return need_refresh;
}
-
static int
-afr_inode_need_refresh_set (inode_t *inode, xlator_t *this)
+afr_inode_need_refresh_set(inode_t *inode, xlator_t *this)
{
- int ret = -1;
- afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- goto unlock;
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret)
+ goto unlock;
- ctx->need_refresh = _gf_true;
- }
+ ctx->need_refresh = _gf_true;
+ }
unlock:
- UNLOCK(&inode->lock);
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
int
-afr_inode_event_gen_reset (inode_t *inode, xlator_t *this)
+afr_inode_event_gen_reset(inode_t *inode, xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_event_gen_reset (inode, this);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_event_gen_reset(inode, this);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
int
-afr_spb_choice_timeout_cancel (xlator_t *this, inode_t *inode)
+afr_spb_choice_timeout_cancel(xlator_t *this, inode_t *inode)
{
- afr_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- if (!inode)
- return ret;
+ if (!inode)
+ return ret;
- LOCK(&inode->lock);
- {
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret < 0 || !ctx) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- "Failed to cancel split-brain choice timer.");
- goto out;
- }
- ctx->spb_choice = -1;
- if (ctx->timer) {
- gf_timer_call_cancel (this->ctx, ctx->timer);
- ctx->timer = NULL;
- }
- ret = 0;
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret < 0 || !ctx) {
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Failed to cancel split-brain choice timer.");
+ goto out;
}
+ ctx->spb_choice = -1;
+ if (ctx->timer) {
+ gf_timer_call_cancel(this->ctx, ctx->timer);
+ ctx->timer = NULL;
+ }
+ ret = 0;
+ }
out:
- UNLOCK(&inode->lock);
- return ret;
+ UNLOCK(&inode->lock);
+ return ret;
}
void
-afr_set_split_brain_choice_cbk (void *data)
+afr_set_split_brain_choice_cbk(void *data)
{
- inode_t *inode = data;
- xlator_t *this = THIS;
+ inode_t *inode = data;
+ xlator_t *this = THIS;
- afr_spb_choice_timeout_cancel (this, inode);
- inode_invalidate (inode);
- inode_unref (inode);
- return;
+ afr_spb_choice_timeout_cancel(this, inode);
+ inode_invalidate(inode);
+ inode_unref(inode);
+ return;
}
-
int
-afr_set_split_brain_choice (int ret, call_frame_t *frame, void *opaque)
-{
- int op_errno = ENOMEM;
- afr_private_t *priv = NULL;
- afr_inode_ctx_t *ctx = NULL;
- inode_t *inode = NULL;
- loc_t *loc = NULL;
- xlator_t *this = NULL;
- afr_spbc_timeout_t *data = opaque;
- struct timespec delta = {0, };
- gf_boolean_t timer_set = _gf_false;
- gf_boolean_t timer_cancelled = _gf_false;
- gf_boolean_t timer_reset = _gf_false;
- gf_boolean_t need_invalidate = _gf_true;
- int old_spb_choice = -1;
-
- frame = data->frame;
- loc = data->loc;
- this = frame->this;
- priv = this->private;
-
+afr_set_split_brain_choice(int ret, call_frame_t *frame, void *opaque)
+{
+ int op_errno = ENOMEM;
+ afr_private_t *priv = NULL;
+ afr_inode_ctx_t *ctx = NULL;
+ inode_t *inode = NULL;
+ loc_t *loc = NULL;
+ xlator_t *this = NULL;
+ afr_spbc_timeout_t *data = opaque;
+ struct timespec delta = {
+ 0,
+ };
+ gf_boolean_t timer_set = _gf_false;
+ gf_boolean_t timer_cancelled = _gf_false;
+ gf_boolean_t timer_reset = _gf_false;
+ gf_boolean_t need_invalidate = _gf_true;
+ int old_spb_choice = -1;
+
+ frame = data->frame;
+ loc = data->loc;
+ this = frame->this;
+ priv = this->private;
+
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ delta.tv_sec = priv->spb_choice_timeout;
+ delta.tv_nsec = 0;
+
+ if (!loc->inode) {
+ ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ if (!(data->d_spb || data->m_spb)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Cannot set "
+ "replica.split-brain-choice on %s. File is"
+ " not in data/metadata split-brain.",
+ uuid_utoa(loc->gfid));
+ ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ /*
+ * we're ref'ing the inode before LOCK like it is done elsewhere in the
+ * code. If we ref after LOCK, coverity complains of possible deadlocks.
+ */
+ inode = inode_ref(loc->inode);
+
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Failed to get inode_ctx for %s", loc->name);
+ goto unlock;
}
- delta.tv_sec = priv->spb_choice_timeout;
- delta.tv_nsec = 0;
+ old_spb_choice = ctx->spb_choice;
+ ctx->spb_choice = data->spb_child_index;
- if (!loc->inode) {
- ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (!(data->d_spb || data->m_spb)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR, "Cannot set "
- "replica.split-brain-choice on %s. File is"
- " not in data/metadata split-brain.",
- uuid_utoa (loc->gfid));
- ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- /*
- * we're ref'ing the inode before LOCK like it is done elsewhere in the
- * code. If we ref after LOCK, coverity complains of possible deadlocks.
+ /* Possible changes in spb-choice :
+ * valid to -1 : cancel timer and unref
+ * valid to valid : cancel timer and inject new one
+ * -1 to -1 : unref and do not do anything
+ * -1 to valid : inject timer
*/
- inode = inode_ref (loc->inode);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- "Failed to get inode_ctx for %s", loc->name);
- goto unlock;
+ /* ctx->timer is NULL iff previous value of
+ * ctx->spb_choice is -1
+ */
+ if (ctx->timer) {
+ if (ctx->spb_choice == -1) {
+ if (!gf_timer_call_cancel(this->ctx, ctx->timer)) {
+ ctx->timer = NULL;
+ timer_cancelled = _gf_true;
}
-
- old_spb_choice = ctx->spb_choice;
- ctx->spb_choice = data->spb_child_index;
-
- /* Possible changes in spb-choice :
- * valid to -1 : cancel timer and unref
- * valid to valid : cancel timer and inject new one
- * -1 to -1 : unref and do not do anything
- * -1 to valid : inject timer
+ /* If timer cancel failed here it means that the
+ * previous cbk will be executed which will set
+ * spb_choice to -1. So we can consider the
+ * 'valid to -1' case to be a success
+ * (i.e. ret = 0) and goto unlock.
*/
-
- /* ctx->timer is NULL iff previous value of
- * ctx->spb_choice is -1
- */
- if (ctx->timer) {
- if (ctx->spb_choice == -1) {
- if (!gf_timer_call_cancel (this->ctx,
- ctx->timer)) {
- ctx->timer = NULL;
- timer_cancelled = _gf_true;
- }
- /* If timer cancel failed here it means that the
- * previous cbk will be executed which will set
- * spb_choice to -1. So we can consider the
- * 'valid to -1' case to be a success
- * (i.e. ret = 0) and goto unlock.
- */
- goto unlock;
- }
- goto reset_timer;
- } else {
- if (ctx->spb_choice == -1)
- goto unlock;
- goto set_timer;
- }
-
-reset_timer:
- ret = gf_timer_call_cancel (this->ctx, ctx->timer);
- if (ret != 0) {
- /* We need to bail out now instead of launching a new
- * timer. Otherwise the cbk of the previous timer event
- * will cancel the new ctx->timer.
- */
- ctx->spb_choice = old_spb_choice;
- ret = -1;
- op_errno = EAGAIN;
- goto unlock;
- }
- ctx->timer = NULL;
- timer_reset = _gf_true;
-
-set_timer:
- ctx->timer = gf_timer_call_after (this->ctx, delta,
- afr_set_split_brain_choice_cbk,
- inode);
- if (!ctx->timer) {
- ctx->spb_choice = old_spb_choice;
- ret = -1;
- op_errno = ENOMEM;
- }
- if (!timer_reset && ctx->timer)
- timer_set = _gf_true;
- if (timer_reset && !ctx->timer)
- timer_cancelled = _gf_true;
- need_invalidate = _gf_false;
+ goto unlock;
+ }
+ goto reset_timer;
+ } else {
+ if (ctx->spb_choice == -1)
+ goto unlock;
+ goto set_timer;
}
+
+ reset_timer:
+ ret = gf_timer_call_cancel(this->ctx, ctx->timer);
+ if (ret != 0) {
+ /* We need to bail out now instead of launching a new
+ * timer. Otherwise the cbk of the previous timer event
+ * will cancel the new ctx->timer.
+ */
+ ctx->spb_choice = old_spb_choice;
+ ret = -1;
+ op_errno = EAGAIN;
+ goto unlock;
+ }
+ ctx->timer = NULL;
+ timer_reset = _gf_true;
+
+ set_timer:
+ ctx->timer = gf_timer_call_after(this->ctx, delta,
+ afr_set_split_brain_choice_cbk, inode);
+ if (!ctx->timer) {
+ ctx->spb_choice = old_spb_choice;
+ ret = -1;
+ op_errno = ENOMEM;
+ }
+ if (!timer_reset && ctx->timer)
+ timer_set = _gf_true;
+ if (timer_reset && !ctx->timer)
+ timer_cancelled = _gf_true;
+ need_invalidate = _gf_false;
+ }
unlock:
- UNLOCK(&inode->lock);
- if (!timer_set)
- inode_unref (inode);
- if (timer_cancelled)
- inode_unref (inode);
- /*
- * We need to invalidate the inode to prevent the kernel from serving
- * reads from an older cached value despite a change in spb_choice to
- * a new value.
- */
- if (need_invalidate)
- inode_invalidate (inode);
+ UNLOCK(&inode->lock);
+ if (!timer_set)
+ inode_unref(inode);
+ if (timer_cancelled)
+ inode_unref(inode);
+ /*
+ * We need to invalidate the inode to prevent the kernel from serving
+ * reads from an older cached value despite a change in spb_choice to
+ * a new value.
+ */
+ if (need_invalidate)
+ inode_invalidate(inode);
out:
- GF_FREE (data);
- AFR_STACK_UNWIND (setxattr, frame, ret, op_errno, NULL);
- return 0;
+ GF_FREE(data);
+ AFR_STACK_UNWIND(setxattr, frame, ret, op_errno, NULL);
+ return 0;
}
int
-afr_accused_fill (xlator_t *this, dict_t *xdata, unsigned char *accused,
- afr_transaction_type type)
+afr_accused_fill(xlator_t *this, dict_t *xdata, unsigned char *accused,
+ afr_transaction_type type)
{
- afr_private_t *priv = NULL;
- int i = 0;
- int idx = afr_index_for_transaction_type (type);
- void *pending_raw = NULL;
- int pending[3];
- int ret = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int idx = afr_index_for_transaction_type(type);
+ void *pending_raw = NULL;
+ int pending[3];
+ int ret = 0;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_get_ptr (xdata, priv->pending_key[i],
- &pending_raw);
- if (ret) /* no pending flags */
- continue;
- memcpy (pending, pending_raw, sizeof(pending));
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_get_ptr(xdata, priv->pending_key[i], &pending_raw);
+ if (ret) /* no pending flags */
+ continue;
+ memcpy(pending, pending_raw, sizeof(pending));
- if (ntoh32 (pending[idx]))
- accused[i] = 1;
- }
+ if (ntoh32(pending[idx]))
+ accused[i] = 1;
+ }
- return 0;
+ return 0;
}
int
-afr_accuse_smallfiles (xlator_t *this, struct afr_reply *replies,
- unsigned char *data_accused)
+afr_accuse_smallfiles(xlator_t *this, struct afr_reply *replies,
+ unsigned char *data_accused)
{
- int i = 0;
- afr_private_t *priv = NULL;
- uint64_t maxsize = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ uint64_t maxsize = 0;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid && replies[i].xdata &&
- dict_get (replies[i].xdata, GLUSTERFS_BAD_INODE))
- continue;
- if (data_accused[i])
- continue;
- if (replies[i].poststat.ia_size > maxsize)
- maxsize = replies[i].poststat.ia_size;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid && replies[i].xdata &&
+ dict_get(replies[i].xdata, GLUSTERFS_BAD_INODE))
+ continue;
+ if (data_accused[i])
+ continue;
+ if (replies[i].poststat.ia_size > maxsize)
+ maxsize = replies[i].poststat.ia_size;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (data_accused[i])
- continue;
- if (AFR_IS_ARBITER_BRICK(priv, i))
- continue;
- if (replies[i].poststat.ia_size < maxsize)
- data_accused[i] = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (data_accused[i])
+ continue;
+ if (AFR_IS_ARBITER_BRICK(priv, i))
+ continue;
+ if (replies[i].poststat.ia_size < maxsize)
+ data_accused[i] = 1;
+ }
- return 0;
+ return 0;
}
int
-afr_readables_fill (call_frame_t *frame, xlator_t *this, inode_t *inode,
- unsigned char *data_accused,
- unsigned char *metadata_accused,
- unsigned char *data_readable,
- unsigned char *metadata_readable,
- struct afr_reply *replies)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- int i = 0;
- int ret = 0;
- ia_type_t ia_type = IA_INVAL;
-
- local = frame->local;
- priv = this->private;
+afr_readables_fill(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ unsigned char *data_accused, unsigned char *metadata_accused,
+ unsigned char *data_readable,
+ unsigned char *metadata_readable, struct afr_reply *replies)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ int i = 0;
+ int ret = 0;
+ ia_type_t ia_type = IA_INVAL;
+
+ local = frame->local;
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ data_readable[i] = 1;
+ metadata_readable[i] = 1;
+ }
+ if (AFR_IS_ARBITER_BRICK(priv, ARBITER_BRICK_INDEX)) {
+ data_readable[ARBITER_BRICK_INDEX] = 0;
+ metadata_readable[ARBITER_BRICK_INDEX] = 0;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies) { /* Lookup */
+ if (!replies[i].valid || replies[i].op_ret == -1 ||
+ (replies[i].xdata &&
+ dict_get(replies[i].xdata, GLUSTERFS_BAD_INODE))) {
+ data_readable[i] = 0;
+ metadata_readable[i] = 0;
+ continue;
+ }
- for (i = 0; i < priv->child_count; i++) {
- data_readable[i] = 1;
- metadata_readable[i] = 1;
- }
- if (AFR_IS_ARBITER_BRICK (priv, ARBITER_BRICK_INDEX)) {
- data_readable[ARBITER_BRICK_INDEX] = 0;
- metadata_readable[ARBITER_BRICK_INDEX] = 0;
+ xdata = replies[i].xdata;
+ ia_type = replies[i].poststat.ia_type;
+ } else { /* pre-op xattrop */
+ xdata = local->transaction.changelog_xdata[i];
+ ia_type = inode->ia_type;
}
- for (i = 0; i < priv->child_count; i++) {
- if (replies) {/* Lookup */
- if (!replies[i].valid || replies[i].op_ret == -1 ||
- (replies[i].xdata && dict_get (replies[i].xdata,
- GLUSTERFS_BAD_INODE))) {
- data_readable[i] = 0;
- metadata_readable[i] = 0;
- continue;
- }
-
- xdata = replies[i].xdata;
- ia_type = replies[i].poststat.ia_type;
- } else {/* pre-op xattrop */
- xdata = local->transaction.changelog_xdata[i];
- ia_type = inode->ia_type;
- }
+ afr_accused_fill(this, xdata, data_accused,
+ (ia_type == IA_IFDIR) ? AFR_ENTRY_TRANSACTION
+ : AFR_DATA_TRANSACTION);
- afr_accused_fill (this, xdata, data_accused,
- (ia_type == IA_IFDIR) ?
- AFR_ENTRY_TRANSACTION : AFR_DATA_TRANSACTION);
+ afr_accused_fill(this, xdata, metadata_accused,
+ AFR_METADATA_TRANSACTION);
+ }
- afr_accused_fill (this, xdata,
- metadata_accused, AFR_METADATA_TRANSACTION);
- }
+ if (replies && ia_type != IA_INVAL && ia_type != IA_IFDIR &&
+ /* We want to accuse small files only when we know for
+ * sure that there is no IO happening. Otherwise, the
+ * ia_sizes obtained in post-refresh replies may
+ * mismatch due to a race between inode-refresh and
+ * ongoing writes, causing spurious heal launches*/
+ !afr_is_possibly_under_txn(AFR_DATA_TRANSACTION, local, this)) {
+ afr_accuse_smallfiles(this, replies, data_accused);
+ }
- if (replies && ia_type != IA_INVAL && ia_type != IA_IFDIR &&
- /* We want to accuse small files only when we know for
- * sure that there is no IO happening. Otherwise, the
- * ia_sizes obtained in post-refresh replies may
- * mismatch due to a race between inode-refresh and
- * ongoing writes, causing spurious heal launches*/
- !afr_is_possibly_under_txn (AFR_DATA_TRANSACTION, local, this)) {
- afr_accuse_smallfiles (this, replies, data_accused);
+ for (i = 0; i < priv->child_count; i++) {
+ if (data_accused[i]) {
+ data_readable[i] = 0;
+ ret = 1;
}
-
- for (i = 0; i < priv->child_count; i++) {
- if (data_accused[i]) {
- data_readable[i] = 0;
- ret = 1;
- }
- if (metadata_accused[i]) {
- metadata_readable[i] = 0;
- ret = 1;
- }
+ if (metadata_accused[i]) {
+ metadata_readable[i] = 0;
+ ret = 1;
}
- return ret;
+ }
+ return ret;
}
int
-afr_replies_interpret (call_frame_t *frame, xlator_t *this, inode_t *inode,
- gf_boolean_t *start_heal)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- int event_generation = 0;
- int i = 0;
- unsigned char *data_accused = NULL;
- unsigned char *metadata_accused = NULL;
- unsigned char *data_readable = NULL;
- unsigned char *metadata_readable = NULL;
- int ret = 0;
-
- local = frame->local;
- priv = this->private;
- replies = local->replies;
- event_generation = local->event_generation;
-
- data_accused = alloca0 (priv->child_count);
- data_readable = alloca0 (priv->child_count);
- metadata_accused = alloca0 (priv->child_count);
- metadata_readable = alloca0 (priv->child_count);
-
- ret = afr_readables_fill (frame, this, inode, data_accused,
- metadata_accused, data_readable,
- metadata_readable, replies);
-
- for (i = 0; i < priv->child_count; i++) {
- if (start_heal && priv->child_up[i] &&
- (data_accused[i] || metadata_accused[i])) {
- *start_heal = _gf_true;
- break;
- }
- }
- afr_inode_read_subvol_set (inode, this, data_readable,
- metadata_readable, event_generation);
- return ret;
+afr_replies_interpret(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ gf_boolean_t *start_heal)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ int event_generation = 0;
+ int i = 0;
+ unsigned char *data_accused = NULL;
+ unsigned char *metadata_accused = NULL;
+ unsigned char *data_readable = NULL;
+ unsigned char *metadata_readable = NULL;
+ int ret = 0;
+
+ local = frame->local;
+ priv = this->private;
+ replies = local->replies;
+ event_generation = local->event_generation;
+
+ data_accused = alloca0(priv->child_count);
+ data_readable = alloca0(priv->child_count);
+ metadata_accused = alloca0(priv->child_count);
+ metadata_readable = alloca0(priv->child_count);
+
+ ret = afr_readables_fill(frame, this, inode, data_accused, metadata_accused,
+ data_readable, metadata_readable, replies);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (start_heal && priv->child_up[i] &&
+ (data_accused[i] || metadata_accused[i])) {
+ *start_heal = _gf_true;
+ break;
+ }
+ }
+ afr_inode_read_subvol_set(inode, this, data_readable, metadata_readable,
+ event_generation);
+ return ret;
}
int
-afr_refresh_selfheal_done (int ret, call_frame_t *heal, void *opaque)
+afr_refresh_selfheal_done(int ret, call_frame_t *heal, void *opaque)
{
- if (heal)
- AFR_STACK_DESTROY (heal);
- return 0;
+ if (heal)
+ AFR_STACK_DESTROY(heal);
+ return 0;
}
int
-afr_inode_refresh_err (call_frame_t *frame, xlator_t *this)
+afr_inode_refresh_err(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int err = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int err = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].valid && !local->replies[i].op_ret) {
- err = 0;
- goto ret;
- }
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].valid && !local->replies[i].op_ret) {
+ err = 0;
+ goto ret;
+ }
+ }
- err = afr_final_errno (local, priv);
+ err = afr_final_errno(local, priv);
ret:
- return err;
+ return err;
}
gf_boolean_t
-afr_selfheal_enabled (xlator_t *this)
+afr_selfheal_enabled(xlator_t *this)
{
- afr_private_t *priv = NULL;
- gf_boolean_t data = _gf_false;
- int ret = 0;
+ afr_private_t *priv = NULL;
+ gf_boolean_t data = _gf_false;
+ int ret = 0;
- priv = this->private;
+ priv = this->private;
- ret = gf_string2boolean (priv->data_self_heal, &data);
- GF_ASSERT (!ret);
+ ret = gf_string2boolean(priv->data_self_heal, &data);
+ GF_ASSERT(!ret);
- return data || priv->metadata_self_heal || priv->entry_self_heal;
+ return data || priv->metadata_self_heal || priv->entry_self_heal;
}
-
int
-afr_txn_refresh_done (call_frame_t *frame, xlator_t *this, int err)
-{
-
- call_frame_t *heal_frame = NULL;
- afr_local_t *heal_local = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- inode_t *inode = NULL;
- int event_generation = 0;
- int read_subvol = -1;
- int op_errno = ENOMEM;
- int ret = 0;
-
- local = frame->local;
- inode = local->inode;
- priv = this->private;
-
- if (err)
- goto refresh_done;
-
- if (local->op == GF_FOP_LOOKUP)
- goto refresh_done;
-
- ret = afr_inode_get_readable (frame, inode, this, local->readable,
- &event_generation,
- local->transaction.type);
-
- if (ret == -EIO || (local->is_read_txn && !event_generation)) {
- /* No readable subvolume even after refresh ==> splitbrain.*/
- if (!priv->fav_child_policy) {
- err = EIO;
- goto refresh_done;
- }
- read_subvol = afr_sh_get_fav_by_policy (this, local->replies,
- inode, NULL);
- if (read_subvol == -1) {
- err = EIO;
- goto refresh_done;
- }
-
- heal_frame = copy_frame (frame);
- if (!heal_frame) {
- err = EIO;
- goto refresh_done;
- }
- heal_frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
- heal_local = AFR_FRAME_INIT (heal_frame, op_errno);
- if (!heal_local) {
- err = EIO;
- AFR_STACK_DESTROY (heal_frame);
- goto refresh_done;
- }
- heal_local->xdata_req = dict_new();
- if (!heal_local->xdata_req) {
- err = EIO;
- AFR_STACK_DESTROY (heal_frame);
- goto refresh_done;
- }
- heal_local->heal_frame = frame;
- ret = synctask_new (this->ctx->env,
- afr_fav_child_reset_sink_xattrs,
- afr_fav_child_reset_sink_xattrs_cbk,
- heal_frame,
- heal_frame);
- return 0;
- }
+afr_txn_refresh_done(call_frame_t *frame, xlator_t *this, int err)
+{
+ call_frame_t *heal_frame = NULL;
+ afr_local_t *heal_local = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ inode_t *inode = NULL;
+ int event_generation = 0;
+ int read_subvol = -1;
+ int op_errno = ENOMEM;
+ int ret = 0;
+
+ local = frame->local;
+ inode = local->inode;
+ priv = this->private;
+
+ if (err)
+ goto refresh_done;
+
+ if (local->op == GF_FOP_LOOKUP)
+ goto refresh_done;
+
+ ret = afr_inode_get_readable(frame, inode, this, local->readable,
+ &event_generation, local->transaction.type);
+
+ if (ret == -EIO || (local->is_read_txn && !event_generation)) {
+ /* No readable subvolume even after refresh ==> splitbrain.*/
+ if (!priv->fav_child_policy) {
+ err = EIO;
+ goto refresh_done;
+ }
+ read_subvol = afr_sh_get_fav_by_policy(this, local->replies, inode,
+ NULL);
+ if (read_subvol == -1) {
+ err = EIO;
+ goto refresh_done;
+ }
+
+ heal_frame = copy_frame(frame);
+ if (!heal_frame) {
+ err = EIO;
+ goto refresh_done;
+ }
+ heal_frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
+ heal_local = AFR_FRAME_INIT(heal_frame, op_errno);
+ if (!heal_local) {
+ err = EIO;
+ AFR_STACK_DESTROY(heal_frame);
+ goto refresh_done;
+ }
+ heal_local->xdata_req = dict_new();
+ if (!heal_local->xdata_req) {
+ err = EIO;
+ AFR_STACK_DESTROY(heal_frame);
+ goto refresh_done;
+ }
+ heal_local->heal_frame = frame;
+ ret = synctask_new(this->ctx->env, afr_fav_child_reset_sink_xattrs,
+ afr_fav_child_reset_sink_xattrs_cbk, heal_frame,
+ heal_frame);
+ return 0;
+ }
refresh_done:
- afr_local_replies_wipe (local, this->private);
- local->refreshfn (frame, this, err);
+ afr_local_replies_wipe(local, this->private);
+ local->refreshfn(frame, this, err);
- return 0;
+ return 0;
}
static void
-afr_fill_success_replies (afr_local_t *local, afr_private_t *priv,
- unsigned char *replies)
+afr_fill_success_replies(afr_local_t *local, afr_private_t *priv,
+ unsigned char *replies)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].valid && local->replies[i].op_ret == 0)
- replies[i] = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].valid && local->replies[i].op_ret == 0)
+ replies[i] = 1;
+ }
}
int
-afr_inode_refresh_done (call_frame_t *frame, xlator_t *this, int error)
-{
- call_frame_t *heal_frame = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t start_heal = _gf_false;
- afr_local_t *heal_local = NULL;
- unsigned char *success_replies = NULL;
- int op_errno = ENOMEM;
- int ret = 0;
-
- if (error != 0) {
- goto refresh_done;
- }
-
- local = frame->local;
- priv = this->private;
- success_replies = alloca0 (priv->child_count);
- afr_fill_success_replies (local, priv, success_replies);
-
- if (!afr_has_quorum (success_replies, this)) {
- error = afr_final_errno (frame->local, this->private);
- if (!error)
- error = afr_quorum_errno(priv);
- goto refresh_done;
- }
-
- if (priv->thin_arbiter_count && local->is_read_txn &&
- AFR_COUNT (success_replies, priv->child_count) !=
- priv->child_count) {
- /* We need to query the good bricks and/or thin-arbiter.*/
- error = EINVAL;
- goto refresh_done;
- }
-
- ret = afr_replies_interpret (frame, this, local->refreshinode,
- &start_heal);
-
- if (ret && afr_selfheal_enabled (this) && start_heal) {
- heal_frame = copy_frame (frame);
- if (!heal_frame)
- goto refresh_done;
- heal_frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
- heal_local = AFR_FRAME_INIT (heal_frame, op_errno);
- if (!heal_local) {
- AFR_STACK_DESTROY (heal_frame);
- goto refresh_done;
- }
- heal_local->refreshinode = inode_ref (local->refreshinode);
- heal_local->heal_frame = heal_frame;
- if (!afr_throttled_selfheal (heal_frame, this)) {
- AFR_STACK_DESTROY (heal_frame);
- goto refresh_done;
- }
- }
+afr_inode_refresh_done(call_frame_t *frame, xlator_t *this, int error)
+{
+ call_frame_t *heal_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t start_heal = _gf_false;
+ afr_local_t *heal_local = NULL;
+ unsigned char *success_replies = NULL;
+ int op_errno = ENOMEM;
+ int ret = 0;
+
+ if (error != 0) {
+ goto refresh_done;
+ }
+
+ local = frame->local;
+ priv = this->private;
+ success_replies = alloca0(priv->child_count);
+ afr_fill_success_replies(local, priv, success_replies);
+
+ if (!afr_has_quorum(success_replies, this)) {
+ error = afr_final_errno(frame->local, this->private);
+ if (!error)
+ error = afr_quorum_errno(priv);
+ goto refresh_done;
+ }
+
+ if (priv->thin_arbiter_count && local->is_read_txn &&
+ AFR_COUNT(success_replies, priv->child_count) != priv->child_count) {
+ /* We need to query the good bricks and/or thin-arbiter.*/
+ error = EINVAL;
+ goto refresh_done;
+ }
+
+ ret = afr_replies_interpret(frame, this, local->refreshinode, &start_heal);
+
+ if (ret && afr_selfheal_enabled(this) && start_heal) {
+ heal_frame = copy_frame(frame);
+ if (!heal_frame)
+ goto refresh_done;
+ heal_frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
+ heal_local = AFR_FRAME_INIT(heal_frame, op_errno);
+ if (!heal_local) {
+ AFR_STACK_DESTROY(heal_frame);
+ goto refresh_done;
+ }
+ heal_local->refreshinode = inode_ref(local->refreshinode);
+ heal_local->heal_frame = heal_frame;
+ if (!afr_throttled_selfheal(heal_frame, this)) {
+ AFR_STACK_DESTROY(heal_frame);
+ goto refresh_done;
+ }
+ }
refresh_done:
- afr_txn_refresh_done (frame, this, error);
+ afr_txn_refresh_done(frame, this, error);
- return 0;
+ return 0;
}
void
-afr_inode_refresh_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *buf,
- dict_t *xdata, struct iatt *par)
-{
- afr_local_t *local = NULL;
- int call_child = (long) cookie;
- int8_t need_heal = 1;
- int call_count = 0;
- int ret = 0;
-
- local = frame->local;
- local->replies[call_child].valid = 1;
- local->replies[call_child].op_ret = op_ret;
- local->replies[call_child].op_errno = op_errno;
- if (op_ret != -1) {
- local->replies[call_child].poststat = *buf;
- if (par)
- local->replies[call_child].postparent = *par;
- if (xdata)
- local->replies[call_child].xdata = dict_ref (xdata);
- }
- if (xdata) {
- ret = dict_get_int8 (xdata, "link-count", &need_heal);
- local->replies[call_child].need_heal = need_heal;
- } else {
- local->replies[call_child].need_heal = need_heal;
- }
-
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_need_heal (this, local);
- ret = afr_inode_refresh_err (frame, this);
- afr_inode_refresh_done (frame, this, ret);
- }
-
+afr_inode_refresh_subvol_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *buf,
+ dict_t *xdata, struct iatt *par)
+{
+ afr_local_t *local = NULL;
+ int call_child = (long)cookie;
+ int8_t need_heal = 1;
+ int call_count = 0;
+ int ret = 0;
+
+ local = frame->local;
+ local->replies[call_child].valid = 1;
+ local->replies[call_child].op_ret = op_ret;
+ local->replies[call_child].op_errno = op_errno;
+ if (op_ret != -1) {
+ local->replies[call_child].poststat = *buf;
+ if (par)
+ local->replies[call_child].postparent = *par;
+ if (xdata)
+ local->replies[call_child].xdata = dict_ref(xdata);
+ }
+ if (xdata) {
+ ret = dict_get_int8(xdata, "link-count", &need_heal);
+ local->replies[call_child].need_heal = need_heal;
+ } else {
+ local->replies[call_child].need_heal = need_heal;
+ }
+
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ afr_set_need_heal(this, local);
+ ret = afr_inode_refresh_err(frame, this);
+ afr_inode_refresh_done(frame, this, ret);
+ }
}
int
-afr_inode_refresh_subvol_with_lookup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret,
- int op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *par)
+afr_inode_refresh_subvol_with_lookup_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *par)
{
- afr_inode_refresh_subvol_cbk (frame, cookie, this, op_ret, op_errno,
- buf, xdata, par);
- return 0;
+ afr_inode_refresh_subvol_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ xdata, par);
+ return 0;
}
-
int
-afr_inode_refresh_subvol_with_lookup (call_frame_t *frame, xlator_t *this,
- int i, inode_t *inode, uuid_t gfid,
- dict_t *xdata)
+afr_inode_refresh_subvol_with_lookup(call_frame_t *frame, xlator_t *this, int i,
+ inode_t *inode, uuid_t gfid, dict_t *xdata)
{
- loc_t loc = {0, };
- afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- loc.inode = inode;
- if (gf_uuid_is_null (inode->gfid) && gfid) {
- /* To handle setattr/setxattr on yet to be linked inode from
- * dht */
- gf_uuid_copy (loc.gfid, gfid);
- } else {
- gf_uuid_copy (loc.gfid, inode->gfid);
- }
+ loc.inode = inode;
+ if (gf_uuid_is_null(inode->gfid) && gfid) {
+ /* To handle setattr/setxattr on yet to be linked inode from
+ * dht */
+ gf_uuid_copy(loc.gfid, gfid);
+ } else {
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ }
- STACK_WIND_COOKIE (frame, afr_inode_refresh_subvol_with_lookup_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->lookup, &loc, xdata);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_inode_refresh_subvol_with_lookup_cbk,
+ (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->lookup, &loc, xdata);
+ return 0;
}
int
-afr_inode_refresh_subvol_with_fstat_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *buf, dict_t *xdata)
+afr_inode_refresh_subvol_with_fstat_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- afr_inode_refresh_subvol_cbk (frame, cookie, this, op_ret, op_errno,
- buf, xdata, NULL);
- return 0;
+ afr_inode_refresh_subvol_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ xdata, NULL);
+ return 0;
}
int
-afr_inode_refresh_subvol_with_fstat (call_frame_t *frame, xlator_t *this, int i,
- dict_t *xdata)
+afr_inode_refresh_subvol_with_fstat(call_frame_t *frame, xlator_t *this, int i,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- STACK_WIND_COOKIE (frame, afr_inode_refresh_subvol_with_fstat_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->fstat, local->fd, xdata);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_inode_refresh_subvol_with_fstat_cbk,
+ (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->fstat, local->fd, xdata);
+ return 0;
}
int
-afr_inode_refresh_do (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
- int ret = 0;
- dict_t *xdata = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- unsigned char *wind_subvols = NULL;
-
- priv = this->private;
- local = frame->local;
- wind_subvols = alloca0 (priv->child_count);
-
- afr_local_replies_wipe (local, priv);
-
- if (local->fd) {
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- if (!fd_ctx) {
- afr_inode_refresh_done (frame, this, EINVAL);
- return 0;
- }
- }
+afr_inode_refresh_do(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int i = 0;
+ int ret = 0;
+ dict_t *xdata = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ unsigned char *wind_subvols = NULL;
- xdata = dict_new ();
- if (!xdata) {
- afr_inode_refresh_done (frame, this, ENOMEM);
- return 0;
- }
+ priv = this->private;
+ local = frame->local;
+ wind_subvols = alloca0(priv->child_count);
- ret = afr_xattr_req_prepare (this, xdata);
- if (ret != 0) {
- dict_unref (xdata);
- afr_inode_refresh_done (frame, this, -ret);
- return 0;
- }
+ afr_local_replies_wipe(local, priv);
- ret = dict_set_str (xdata, "link-count", GF_XATTROP_INDEX_COUNT);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set link-count in dict ");
+ if (local->fd) {
+ fd_ctx = afr_fd_ctx_get(local->fd, this);
+ if (!fd_ctx) {
+ afr_inode_refresh_done(frame, this, EINVAL);
+ return 0;
}
+ }
- ret = dict_set_str (xdata, GLUSTERFS_INODELK_DOM_COUNT, this->name);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set inodelk-dom-count in dict ");
+ xdata = dict_new();
+ if (!xdata) {
+ afr_inode_refresh_done(frame, this, ENOMEM);
+ return 0;
+ }
- }
+ ret = afr_xattr_req_prepare(this, xdata);
+ if (ret != 0) {
+ dict_unref(xdata);
+ afr_inode_refresh_done(frame, this, -ret);
+ return 0;
+ }
- if (local->fd) {
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i] &&
- fd_ctx->opened_on[i] == AFR_FD_OPENED)
- wind_subvols[i] = 1;
- }
- } else {
- memcpy (wind_subvols, local->child_up,
- sizeof (*local->child_up) * priv->child_count);
- }
+ ret = dict_set_str(xdata, "link-count", GF_XATTROP_INDEX_COUNT);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to set link-count in dict ");
+ }
- local->call_count = AFR_COUNT (wind_subvols, priv->child_count);
+ ret = dict_set_str(xdata, GLUSTERFS_INODELK_DOM_COUNT, this->name);
+ if (ret) {
+ gf_msg_debug(this->name, -ret,
+ "Unable to set inodelk-dom-count in dict ");
+ }
- call_count = local->call_count;
- if (!call_count) {
- dict_unref (xdata);
- if (local->fd && AFR_COUNT(local->child_up, priv->child_count))
- afr_inode_refresh_done (frame, this, EBADFD);
- else
- afr_inode_refresh_done (frame, this, ENOTCONN);
- return 0;
+ if (local->fd) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] && fd_ctx->opened_on[i] == AFR_FD_OPENED)
+ wind_subvols[i] = 1;
}
- for (i = 0; i < priv->child_count; i++) {
- if (!wind_subvols[i])
- continue;
+ } else {
+ memcpy(wind_subvols, local->child_up,
+ sizeof(*local->child_up) * priv->child_count);
+ }
- if (local->fd)
- afr_inode_refresh_subvol_with_fstat (frame, this, i,
- xdata);
- else
- afr_inode_refresh_subvol_with_lookup (frame, this, i,
- local->refreshinode,
- local->refreshgfid, xdata);
+ local->call_count = AFR_COUNT(wind_subvols, priv->child_count);
- if (!--call_count)
- break;
- }
+ call_count = local->call_count;
+ if (!call_count) {
+ dict_unref(xdata);
+ if (local->fd && AFR_COUNT(local->child_up, priv->child_count))
+ afr_inode_refresh_done(frame, this, EBADFD);
+ else
+ afr_inode_refresh_done(frame, this, ENOTCONN);
+ return 0;
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!wind_subvols[i])
+ continue;
- dict_unref (xdata);
+ if (local->fd)
+ afr_inode_refresh_subvol_with_fstat(frame, this, i, xdata);
+ else
+ afr_inode_refresh_subvol_with_lookup(
+ frame, this, i, local->refreshinode, local->refreshgfid, xdata);
- return 0;
-}
+ if (!--call_count)
+ break;
+ }
+
+ dict_unref(xdata);
+ return 0;
+}
int
-afr_inode_refresh (call_frame_t *frame, xlator_t *this, inode_t *inode,
- uuid_t gfid, afr_inode_refresh_cbk_t refreshfn)
+afr_inode_refresh(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ uuid_t gfid, afr_inode_refresh_cbk_t refreshfn)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->refreshfn = refreshfn;
+ local->refreshfn = refreshfn;
- if (local->refreshinode) {
- inode_unref (local->refreshinode);
- local->refreshinode = NULL;
- }
+ if (local->refreshinode) {
+ inode_unref(local->refreshinode);
+ local->refreshinode = NULL;
+ }
- local->refreshinode = inode_ref (inode);
+ local->refreshinode = inode_ref(inode);
- if (gfid)
- gf_uuid_copy (local->refreshgfid, gfid);
- else
- gf_uuid_clear (local->refreshgfid);
+ if (gfid)
+ gf_uuid_copy(local->refreshgfid, gfid);
+ else
+ gf_uuid_clear(local->refreshgfid);
- afr_inode_refresh_do (frame, this);
+ afr_inode_refresh_do(frame, this);
- return 0;
+ return 0;
}
-
int
-afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req)
+afr_xattr_req_prepare(xlator_t *this, dict_t *xattr_req)
{
- int i = 0;
- afr_private_t *priv = NULL;
- int ret = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_uint64 (xattr_req, priv->pending_key[i],
- AFR_NUM_CHANGE_LOGS * sizeof(int));
- if (ret < 0)
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Unable to set dict value for %s",
- priv->pending_key[i]);
- /* 3 = data+metadata+entry */
- }
- ret = dict_set_uint64 (xattr_req, AFR_DIRTY,
- AFR_NUM_CHANGE_LOGS * sizeof(int));
- if (ret) {
- gf_msg_debug (this->name, -ret, "failed to set dirty "
- "query flag");
- }
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int ret = 0;
- ret = dict_set_int32 (xattr_req, "list-xattr", 1);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set list-xattr in dict ");
- }
+ priv = this->private;
- return ret;
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_set_uint64(xattr_req, priv->pending_key[i],
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Unable to set dict value for %s", priv->pending_key[i]);
+ /* 3 = data+metadata+entry */
+ }
+ ret = dict_set_uint64(xattr_req, AFR_DIRTY,
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ if (ret) {
+ gf_msg_debug(this->name, -ret,
+ "failed to set dirty "
+ "query flag");
+ }
+
+ ret = dict_set_int32(xattr_req, "list-xattr", 1);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to set list-xattr in dict ");
+ }
+
+ return ret;
}
int
-afr_lookup_xattr_req_prepare (afr_local_t *local, xlator_t *this,
- dict_t *xattr_req, loc_t *loc)
-{
- int ret = -ENOMEM;
-
- if (!local->xattr_req)
- local->xattr_req = dict_new ();
-
- if (!local->xattr_req)
- goto out;
-
- if (xattr_req && (xattr_req != local->xattr_req))
- dict_copy (xattr_req, local->xattr_req);
-
- ret = afr_xattr_req_prepare (this, local->xattr_req);
-
- ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "%s: Unable to set dict value for %s",
- loc->path, GLUSTERFS_INODELK_COUNT);
- }
- ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "%s: Unable to set dict value for %s",
- loc->path, GLUSTERFS_ENTRYLK_COUNT);
- }
-
- ret = dict_set_uint32 (local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "%s: Unable to set dict value for %s",
- loc->path, GLUSTERFS_PARENT_ENTRYLK);
- }
-
- ret = dict_set_str (local->xattr_req, "link-count",
- GF_XATTROP_INDEX_COUNT);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set link-count in dict ");
- }
-
- ret = 0;
+afr_lookup_xattr_req_prepare(afr_local_t *local, xlator_t *this,
+ dict_t *xattr_req, loc_t *loc)
+{
+ int ret = -ENOMEM;
+
+ if (!local->xattr_req)
+ local->xattr_req = dict_new();
+
+ if (!local->xattr_req)
+ goto out;
+
+ if (xattr_req && (xattr_req != local->xattr_req))
+ dict_copy(xattr_req, local->xattr_req);
+
+ ret = afr_xattr_req_prepare(this, local->xattr_req);
+
+ ret = dict_set_uint64(local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Unable to set dict value for %s", loc->path,
+ GLUSTERFS_INODELK_COUNT);
+ }
+ ret = dict_set_uint64(local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Unable to set dict value for %s", loc->path,
+ GLUSTERFS_ENTRYLK_COUNT);
+ }
+
+ ret = dict_set_uint32(local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Unable to set dict value for %s", loc->path,
+ GLUSTERFS_PARENT_ENTRYLK);
+ }
+
+ ret = dict_set_str(local->xattr_req, "link-count", GF_XATTROP_INDEX_COUNT);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to set link-count in dict ");
+ }
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-afr_least_pending_reads_child (afr_private_t *priv)
-{
- int i = 0;
- int child = 0;
- int64_t read_iter = -1;
- int64_t pending_read = -1;
-
- pending_read = GF_ATOMIC_GET (priv->pending_reads[0]);
- for (i = 1; i < priv->child_count; i++) {
- if (AFR_IS_ARBITER_BRICK(priv, i))
- continue;
- read_iter = GF_ATOMIC_GET(priv->pending_reads[i]);
- if (read_iter < pending_read) {
- pending_read = read_iter;
- child = i;
- }
+afr_least_pending_reads_child(afr_private_t *priv)
+{
+ int i = 0;
+ int child = 0;
+ int64_t read_iter = -1;
+ int64_t pending_read = -1;
+
+ pending_read = GF_ATOMIC_GET(priv->pending_reads[0]);
+ for (i = 1; i < priv->child_count; i++) {
+ if (AFR_IS_ARBITER_BRICK(priv, i))
+ continue;
+ read_iter = GF_ATOMIC_GET(priv->pending_reads[i]);
+ if (read_iter < pending_read) {
+ pending_read = read_iter;
+ child = i;
}
+ }
- return child;
+ return child;
}
int
-afr_hash_child (afr_read_subvol_args_t *args, afr_private_t *priv)
+afr_hash_child(afr_read_subvol_args_t *args, afr_private_t *priv)
{
- uuid_t gfid_copy = {0,};
- pid_t pid;
- int child = -1;
+ uuid_t gfid_copy = {
+ 0,
+ };
+ pid_t pid;
+ int child = -1;
- switch (priv->hash_mode) {
+ switch (priv->hash_mode) {
case 0:
- break;
+ break;
case 1:
- gf_uuid_copy (gfid_copy, args->gfid);
- child = SuperFastHash((char *)gfid_copy,
- sizeof(gfid_copy)) % priv->child_count;
- break;
+ gf_uuid_copy(gfid_copy, args->gfid);
+ child = SuperFastHash((char *)gfid_copy, sizeof(gfid_copy)) %
+ priv->child_count;
+ break;
case 2:
- if (args->ia_type != IA_IFDIR) {
- /*
- * Why getpid? Because it's one of the cheapest calls
- * available - faster than gethostname etc. - and
- * returns a constant-length value that's sure to be
- * shorter than a UUID. It's still very unlikely to be
- * the same across clients, so it still provides good
- * mixing. We're not trying for perfection here. All we
- * need is a low probability that multiple clients
- * won't converge on the same subvolume.
- */
- pid = getpid();
- memcpy (gfid_copy, &pid, sizeof(pid));
- }
- child = SuperFastHash((char *)gfid_copy,
- sizeof(gfid_copy)) % priv->child_count;
- break;
+ if (args->ia_type != IA_IFDIR) {
+ /*
+ * Why getpid? Because it's one of the cheapest calls
+ * available - faster than gethostname etc. - and
+ * returns a constant-length value that's sure to be
+ * shorter than a UUID. It's still very unlikely to be
+ * the same across clients, so it still provides good
+ * mixing. We're not trying for perfection here. All we
+ * need is a low probability that multiple clients
+ * won't converge on the same subvolume.
+ */
+ pid = getpid();
+ memcpy(gfid_copy, &pid, sizeof(pid));
+ }
+ child = SuperFastHash((char *)gfid_copy, sizeof(gfid_copy)) %
+ priv->child_count;
+ break;
case 3:
- child = afr_least_pending_reads_child (priv);
- break;
- }
+ child = afr_least_pending_reads_child(priv);
+ break;
+ }
- return child;
+ return child;
}
int
-afr_read_subvol_select_by_policy (inode_t *inode, xlator_t *this,
- unsigned char *readable,
- afr_read_subvol_args_t *args)
-{
- int i = 0;
- int read_subvol = -1;
- afr_private_t *priv = NULL;
- afr_read_subvol_args_t local_args = {0,};
-
- priv = this->private;
-
- /* first preference - explicitly specified or local subvolume */
- if (priv->read_child >= 0 && readable[priv->read_child])
- return priv->read_child;
-
- if (inode_is_linked (inode)) {
- gf_uuid_copy (local_args.gfid, inode->gfid);
- local_args.ia_type = inode->ia_type;
- } else if (args) {
- local_args = *args;
- }
+afr_read_subvol_select_by_policy(inode_t *inode, xlator_t *this,
+ unsigned char *readable,
+ afr_read_subvol_args_t *args)
+{
+ int i = 0;
+ int read_subvol = -1;
+ afr_private_t *priv = NULL;
+ afr_read_subvol_args_t local_args = {
+ 0,
+ };
- /* second preference - use hashed mode */
- read_subvol = afr_hash_child (&local_args, priv);
- if (read_subvol >= 0 && readable[read_subvol])
- return read_subvol;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (readable[i])
- return i;
- }
+ /* first preference - explicitly specified or local subvolume */
+ if (priv->read_child >= 0 && readable[priv->read_child])
+ return priv->read_child;
- /* no readable subvolumes, either split brain or all subvols down */
+ if (inode_is_linked(inode)) {
+ gf_uuid_copy(local_args.gfid, inode->gfid);
+ local_args.ia_type = inode->ia_type;
+ } else if (args) {
+ local_args = *args;
+ }
- return -1;
-}
+ /* second preference - use hashed mode */
+ read_subvol = afr_hash_child(&local_args, priv);
+ if (read_subvol >= 0 && readable[read_subvol])
+ return read_subvol;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (readable[i])
+ return i;
+ }
+ /* no readable subvolumes, either split brain or all subvols down */
+
+ return -1;
+}
int
-afr_inode_read_subvol_type_get (inode_t *inode, xlator_t *this,
- unsigned char *readable, int *event_p,
- int type)
+afr_inode_read_subvol_type_get(inode_t *inode, xlator_t *this,
+ unsigned char *readable, int *event_p, int type)
{
- int ret = -1;
+ int ret = -1;
- if (type == AFR_METADATA_TRANSACTION)
- ret = afr_inode_read_subvol_get (inode, this, 0, readable,
- event_p);
- else
- ret = afr_inode_read_subvol_get (inode, this, readable, 0,
- event_p);
- return ret;
+ if (type == AFR_METADATA_TRANSACTION)
+ ret = afr_inode_read_subvol_get(inode, this, 0, readable, event_p);
+ else
+ ret = afr_inode_read_subvol_get(inode, this, readable, 0, event_p);
+ return ret;
}
void
-afr_readables_intersect_get (inode_t *inode, xlator_t *this, int *event,
- unsigned char *intersection)
+afr_readables_intersect_get(inode_t *inode, xlator_t *this, int *event,
+ unsigned char *intersection)
{
- afr_private_t *priv = NULL;
- unsigned char *data_readable = NULL;
- unsigned char *metadata_readable = NULL;
- unsigned char *intersect = NULL;
+ afr_private_t *priv = NULL;
+ unsigned char *data_readable = NULL;
+ unsigned char *metadata_readable = NULL;
+ unsigned char *intersect = NULL;
- priv = this->private;
- data_readable = alloca0 (priv->child_count);
- metadata_readable = alloca0 (priv->child_count);
- intersect = alloca0 (priv->child_count);
+ priv = this->private;
+ data_readable = alloca0(priv->child_count);
+ metadata_readable = alloca0(priv->child_count);
+ intersect = alloca0(priv->child_count);
- afr_inode_read_subvol_get (inode, this, data_readable,
- metadata_readable, event);
+ afr_inode_read_subvol_get(inode, this, data_readable, metadata_readable,
+ event);
- AFR_INTERSECT (intersect, data_readable, metadata_readable,
- priv->child_count);
- if (intersection)
- memcpy (intersection, intersect,
- sizeof (*intersection) * priv->child_count);
+ AFR_INTERSECT(intersect, data_readable, metadata_readable,
+ priv->child_count);
+ if (intersection)
+ memcpy(intersection, intersect,
+ sizeof(*intersection) * priv->child_count);
}
int
-afr_read_subvol_get (inode_t *inode, xlator_t *this, int *subvol_p,
- unsigned char *readables,
- int *event_p, afr_transaction_type type,
- afr_read_subvol_args_t *args)
+afr_read_subvol_get(inode_t *inode, xlator_t *this, int *subvol_p,
+ unsigned char *readables, int *event_p,
+ afr_transaction_type type, afr_read_subvol_args_t *args)
{
- afr_private_t *priv = NULL;
- unsigned char *readable = NULL;
- unsigned char *intersection = NULL;
- int subvol = -1;
- int event = 0;
+ afr_private_t *priv = NULL;
+ unsigned char *readable = NULL;
+ unsigned char *intersection = NULL;
+ int subvol = -1;
+ int event = 0;
- priv = this->private;
+ priv = this->private;
- readable = alloca0 (priv->child_count);
- intersection = alloca0 (priv->child_count);
+ readable = alloca0(priv->child_count);
+ intersection = alloca0(priv->child_count);
- afr_inode_read_subvol_type_get (inode, this, readable, &event, type);
+ afr_inode_read_subvol_type_get(inode, this, readable, &event, type);
- afr_readables_intersect_get (inode, this, &event, intersection);
+ afr_readables_intersect_get(inode, this, &event, intersection);
- if (AFR_COUNT (intersection, priv->child_count) > 0)
- subvol = afr_read_subvol_select_by_policy (inode, this,
- intersection, args);
- else
- subvol = afr_read_subvol_select_by_policy (inode, this,
- readable, args);
- if (subvol_p)
- *subvol_p = subvol;
- if (event_p)
- *event_p = event;
- if (readables)
- memcpy (readables, readable,
- sizeof (*readables) * priv->child_count);
- return subvol;
+ if (AFR_COUNT(intersection, priv->child_count) > 0)
+ subvol = afr_read_subvol_select_by_policy(inode, this, intersection,
+ args);
+ else
+ subvol = afr_read_subvol_select_by_policy(inode, this, readable, args);
+ if (subvol_p)
+ *subvol_p = subvol;
+ if (event_p)
+ *event_p = event;
+ if (readables)
+ memcpy(readables, readable, sizeof(*readables) * priv->child_count);
+ return subvol;
}
-
void
-afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
+afr_local_transaction_cleanup(afr_local_t *local, xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
- priv = this->private;
+ priv = this->private;
- afr_matrix_cleanup (local->pending, priv->child_count);
+ afr_matrix_cleanup(local->pending, priv->child_count);
- GF_FREE (local->internal_lock.locked_nodes);
+ GF_FREE(local->internal_lock.locked_nodes);
- GF_FREE (local->internal_lock.lower_locked_nodes);
+ GF_FREE(local->internal_lock.lower_locked_nodes);
- afr_entry_lockee_cleanup (&local->internal_lock);
+ afr_entry_lockee_cleanup(&local->internal_lock);
- GF_FREE (local->transaction.pre_op);
+ GF_FREE(local->transaction.pre_op);
- GF_FREE (local->transaction.pre_op_sources);
- if (local->transaction.changelog_xdata) {
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.changelog_xdata[i])
- continue;
- dict_unref (local->transaction.changelog_xdata[i]);
- }
- GF_FREE (local->transaction.changelog_xdata);
+ GF_FREE(local->transaction.pre_op_sources);
+ if (local->transaction.changelog_xdata) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.changelog_xdata[i])
+ continue;
+ dict_unref(local->transaction.changelog_xdata[i]);
}
+ GF_FREE(local->transaction.changelog_xdata);
+ }
- GF_FREE (local->transaction.failed_subvols);
+ GF_FREE(local->transaction.failed_subvols);
- GF_FREE (local->transaction.basename);
- GF_FREE (local->transaction.new_basename);
-
- loc_wipe (&local->transaction.parent_loc);
- loc_wipe (&local->transaction.new_parent_loc);
+ GF_FREE(local->transaction.basename);
+ GF_FREE(local->transaction.new_basename);
+ loc_wipe(&local->transaction.parent_loc);
+ loc_wipe(&local->transaction.new_parent_loc);
}
void
-afr_reply_wipe (struct afr_reply *reply)
+afr_reply_wipe(struct afr_reply *reply)
{
- if (reply->xdata) {
- dict_unref (reply->xdata);
- reply->xdata = NULL;
- }
+ if (reply->xdata) {
+ dict_unref(reply->xdata);
+ reply->xdata = NULL;
+ }
- if (reply->xattr) {
- dict_unref (reply->xattr);
- reply->xattr = NULL;
- }
+ if (reply->xattr) {
+ dict_unref(reply->xattr);
+ reply->xattr = NULL;
+ }
}
void
-afr_replies_wipe (struct afr_reply *replies, int count)
+afr_replies_wipe(struct afr_reply *replies, int count)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < count; i++) {
- afr_reply_wipe (&replies[i]);
- }
+ for (i = 0; i < count; i++) {
+ afr_reply_wipe(&replies[i]);
+ }
}
void
-afr_local_replies_wipe (afr_local_t *local, afr_private_t *priv)
+afr_local_replies_wipe(afr_local_t *local, afr_private_t *priv)
{
+ if (!local->replies)
+ return;
- if (!local->replies)
- return;
-
- afr_replies_wipe (local->replies, priv->child_count);
+ afr_replies_wipe(local->replies, priv->child_count);
- memset (local->replies, 0, sizeof(*local->replies) * priv->child_count);
+ memset(local->replies, 0, sizeof(*local->replies) * priv->child_count);
}
static gf_boolean_t
-afr_fop_lock_is_unlock (call_frame_t *frame)
+afr_fop_lock_is_unlock(call_frame_t *frame)
{
- afr_local_t *local = frame->local;
- switch (local->op) {
+ afr_local_t *local = frame->local;
+ switch (local->op) {
case GF_FOP_INODELK:
case GF_FOP_FINODELK:
- if ((F_UNLCK == local->cont.inodelk.in_flock.l_type) &&
- (local->cont.inodelk.in_cmd == F_SETLKW ||
- local->cont.inodelk.in_cmd == F_SETLK))
- return _gf_true;
- break;
+ if ((F_UNLCK == local->cont.inodelk.in_flock.l_type) &&
+ (local->cont.inodelk.in_cmd == F_SETLKW ||
+ local->cont.inodelk.in_cmd == F_SETLK))
+ return _gf_true;
+ break;
case GF_FOP_ENTRYLK:
case GF_FOP_FENTRYLK:
- if (ENTRYLK_UNLOCK == local->cont.entrylk.in_cmd)
- return _gf_true;
- break;
+ if (ENTRYLK_UNLOCK == local->cont.entrylk.in_cmd)
+ return _gf_true;
+ break;
default:
- return _gf_false;
- }
- return _gf_false;
+ return _gf_false;
+ }
+ return _gf_false;
}
static gf_boolean_t
-afr_lk_is_unlock (int32_t cmd, struct gf_flock *flock)
+afr_lk_is_unlock(int32_t cmd, struct gf_flock *flock)
{
- switch (cmd) {
+ switch (cmd) {
case F_RESLK_UNLCK:
- return _gf_true;
- break;
+ return _gf_true;
+ break;
#if F_SETLKW != F_SETLKW64
case F_SETLKW64:
@@ -1953,571 +1898,565 @@ afr_lk_is_unlock (int32_t cmd, struct gf_flock *flock)
case F_SETLK64:
#endif
case F_SETLK:
- if (F_UNLCK == flock->l_type)
- return _gf_true;
- break;
+ if (F_UNLCK == flock->l_type)
+ return _gf_true;
+ break;
default:
- return _gf_false;
- }
- return _gf_false;
+ return _gf_false;
+ }
+ return _gf_false;
}
void
-afr_handle_inconsistent_fop (call_frame_t *frame, int32_t *op_ret,
- int32_t *op_errno)
+afr_handle_inconsistent_fop(call_frame_t *frame, int32_t *op_ret,
+ int32_t *op_errno)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- if (!frame || !frame->this || !frame->local || !frame->this->private)
- return;
+ if (!frame || !frame->this || !frame->local || !frame->this->private)
+ return;
- if (*op_ret < 0)
- return;
+ if (*op_ret < 0)
+ return;
- /* Failing inodelk/entrylk/lk here is not a good idea because we
- * need to cleanup the locks on the other bricks if we choose to fail
- * the fop here. The brick may go down just after unwind happens as well
- * so anyways the fop will fail when the next fop is sent so leaving
- * it like this for now.*/
- local = frame->local;
- switch (local->op) {
+ /* Failing inodelk/entrylk/lk here is not a good idea because we
+ * need to cleanup the locks on the other bricks if we choose to fail
+ * the fop here. The brick may go down just after unwind happens as well
+ * so anyways the fop will fail when the next fop is sent so leaving
+ * it like this for now.*/
+ local = frame->local;
+ switch (local->op) {
case GF_FOP_LOOKUP:
case GF_FOP_INODELK:
case GF_FOP_FINODELK:
case GF_FOP_ENTRYLK:
case GF_FOP_FENTRYLK:
case GF_FOP_LK:
- return;
+ return;
default:
- break;
- }
+ break;
+ }
- priv = frame->this->private;
- if (!priv->consistent_io)
- return;
+ priv = frame->this->private;
+ if (!priv->consistent_io)
+ return;
- if (local->event_generation &&
- (local->event_generation != priv->event_generation))
- goto inconsistent;
+ if (local->event_generation &&
+ (local->event_generation != priv->event_generation))
+ goto inconsistent;
- return;
+ return;
inconsistent:
- *op_ret = -1;
- *op_errno = ENOTCONN;
+ *op_ret = -1;
+ *op_errno = ENOTCONN;
}
void
-afr_local_cleanup (afr_local_t *local, xlator_t *this)
+afr_local_cleanup(afr_local_t *local, xlator_t *this)
{
- afr_private_t * priv = NULL;
-
- if (!local)
- return;
-
- syncbarrier_destroy (&local->barrier);
-
- afr_local_transaction_cleanup (local, this);
-
- priv = this->private;
+ afr_private_t *priv = NULL;
- loc_wipe (&local->loc);
- loc_wipe (&local->newloc);
+ if (!local)
+ return;
- if (local->fd)
- fd_unref (local->fd);
+ syncbarrier_destroy(&local->barrier);
- if (local->xattr_req)
- dict_unref (local->xattr_req);
+ afr_local_transaction_cleanup(local, this);
- if (local->xattr_rsp)
- dict_unref (local->xattr_rsp);
+ priv = this->private;
- if (local->dict)
- dict_unref (local->dict);
+ loc_wipe(&local->loc);
+ loc_wipe(&local->newloc);
- afr_local_replies_wipe (local, priv);
- GF_FREE(local->replies);
+ if (local->fd)
+ fd_unref(local->fd);
- GF_FREE (local->child_up);
+ if (local->xattr_req)
+ dict_unref(local->xattr_req);
- GF_FREE (local->read_attempted);
+ if (local->xattr_rsp)
+ dict_unref(local->xattr_rsp);
- GF_FREE (local->readable);
- GF_FREE (local->readable2);
+ if (local->dict)
+ dict_unref(local->dict);
- if (local->inode)
- inode_unref (local->inode);
+ afr_local_replies_wipe(local, priv);
+ GF_FREE(local->replies);
- if (local->parent)
- inode_unref (local->parent);
+ GF_FREE(local->child_up);
- if (local->parent2)
- inode_unref (local->parent2);
+ GF_FREE(local->read_attempted);
- if (local->refreshinode)
- inode_unref (local->refreshinode);
+ GF_FREE(local->readable);
+ GF_FREE(local->readable2);
- { /* getxattr */
- GF_FREE (local->cont.getxattr.name);
- }
+ if (local->inode)
+ inode_unref(local->inode);
- { /* lk */
- GF_FREE (local->cont.lk.locked_nodes);
- }
+ if (local->parent)
+ inode_unref(local->parent);
- { /* create */
- if (local->cont.create.fd)
- fd_unref (local->cont.create.fd);
- if (local->cont.create.params)
- dict_unref (local->cont.create.params);
- }
+ if (local->parent2)
+ inode_unref(local->parent2);
- { /* mknod */
- if (local->cont.mknod.params)
- dict_unref (local->cont.mknod.params);
- }
+ if (local->refreshinode)
+ inode_unref(local->refreshinode);
- { /* mkdir */
- if (local->cont.mkdir.params)
- dict_unref (local->cont.mkdir.params);
- }
+ { /* getxattr */
+ GF_FREE(local->cont.getxattr.name);
+ }
- { /* symlink */
- if (local->cont.symlink.params)
- dict_unref (local->cont.symlink.params);
- }
+ { /* lk */
+ GF_FREE(local->cont.lk.locked_nodes);
+ }
- { /* writev */
- GF_FREE (local->cont.writev.vector);
- if (local->cont.writev.iobref)
- iobref_unref (local->cont.writev.iobref);
- }
+ { /* create */
+ if (local->cont.create.fd)
+ fd_unref(local->cont.create.fd);
+ if (local->cont.create.params)
+ dict_unref(local->cont.create.params);
+ }
- { /* setxattr */
- if (local->cont.setxattr.dict)
- dict_unref (local->cont.setxattr.dict);
- }
+ { /* mknod */
+ if (local->cont.mknod.params)
+ dict_unref(local->cont.mknod.params);
+ }
- { /* fsetxattr */
- if (local->cont.fsetxattr.dict)
- dict_unref (local->cont.fsetxattr.dict);
- }
+ { /* mkdir */
+ if (local->cont.mkdir.params)
+ dict_unref(local->cont.mkdir.params);
+ }
- { /* removexattr */
- GF_FREE (local->cont.removexattr.name);
- }
- { /* xattrop */
- if (local->cont.xattrop.xattr)
- dict_unref (local->cont.xattrop.xattr);
- }
- { /* symlink */
- GF_FREE (local->cont.symlink.linkpath);
- }
+ { /* symlink */
+ if (local->cont.symlink.params)
+ dict_unref(local->cont.symlink.params);
+ }
- { /* opendir */
- GF_FREE (local->cont.opendir.checksum);
- }
+ { /* writev */
+ GF_FREE(local->cont.writev.vector);
+ if (local->cont.writev.iobref)
+ iobref_unref(local->cont.writev.iobref);
+ }
- { /* open */
- if (local->cont.open.fd)
- fd_unref (local->cont.open.fd);
- }
+ { /* setxattr */
+ if (local->cont.setxattr.dict)
+ dict_unref(local->cont.setxattr.dict);
+ }
- { /* readdirp */
- if (local->cont.readdir.dict)
- dict_unref (local->cont.readdir.dict);
- }
+ { /* fsetxattr */
+ if (local->cont.fsetxattr.dict)
+ dict_unref(local->cont.fsetxattr.dict);
+ }
- { /* inodelk */
- GF_FREE (local->cont.inodelk.volume);
- if (local->cont.inodelk.xdata)
- dict_unref (local->cont.inodelk.xdata);
- }
+ { /* removexattr */
+ GF_FREE(local->cont.removexattr.name);
+ }
+ { /* xattrop */
+ if (local->cont.xattrop.xattr)
+ dict_unref(local->cont.xattrop.xattr);
+ }
+ { /* symlink */
+ GF_FREE(local->cont.symlink.linkpath);
+ }
- { /* entrylk */
- GF_FREE (local->cont.entrylk.volume);
- GF_FREE (local->cont.entrylk.basename);
- if (local->cont.entrylk.xdata)
- dict_unref (local->cont.entrylk.xdata);
- }
+ { /* opendir */
+ GF_FREE(local->cont.opendir.checksum);
+ }
- if (local->xdata_req)
- dict_unref (local->xdata_req);
+ { /* open */
+ if (local->cont.open.fd)
+ fd_unref(local->cont.open.fd);
+ }
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
+ { /* readdirp */
+ if (local->cont.readdir.dict)
+ dict_unref(local->cont.readdir.dict);
+ }
+
+ { /* inodelk */
+ GF_FREE(local->cont.inodelk.volume);
+ if (local->cont.inodelk.xdata)
+ dict_unref(local->cont.inodelk.xdata);
+ }
+
+ { /* entrylk */
+ GF_FREE(local->cont.entrylk.volume);
+ GF_FREE(local->cont.entrylk.basename);
+ if (local->cont.entrylk.xdata)
+ dict_unref(local->cont.entrylk.xdata);
+ }
+
+ if (local->xdata_req)
+ dict_unref(local->xdata_req);
+
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
}
-
int
-afr_frame_return (call_frame_t *frame)
+afr_frame_return(call_frame_t *frame)
{
- afr_local_t *local = NULL;
- int call_count = 0;
+ afr_local_t *local = NULL;
+ int call_count = 0;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- call_count = --local->call_count;
- }
- UNLOCK (&frame->lock);
+ LOCK(&frame->lock);
+ {
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
- return call_count;
+ return call_count;
}
-static char *afr_ignore_xattrs[] = {
- GF_SELINUX_XATTR_KEY,
- QUOTA_SIZE_KEY,
- NULL
-};
+static char *afr_ignore_xattrs[] = {GF_SELINUX_XATTR_KEY, QUOTA_SIZE_KEY, NULL};
gf_boolean_t
-afr_is_xattr_ignorable (char *key)
+afr_is_xattr_ignorable(char *key)
{
- int i = 0;
+ int i = 0;
- if (!strncmp (key, AFR_XATTR_PREFIX, SLEN (AFR_XATTR_PREFIX)))
- return _gf_true;
- for (i = 0; afr_ignore_xattrs[i]; i++) {
- if (!strcmp (key, afr_ignore_xattrs[i]))
- return _gf_true;
- }
- return _gf_false;
+ if (!strncmp(key, AFR_XATTR_PREFIX, SLEN(AFR_XATTR_PREFIX)))
+ return _gf_true;
+ for (i = 0; afr_ignore_xattrs[i]; i++) {
+ if (!strcmp(key, afr_ignore_xattrs[i]))
+ return _gf_true;
+ }
+ return _gf_false;
}
static gf_boolean_t
-afr_xattr_match_needed (dict_t *this, char *key1, data_t *value1, void *data)
+afr_xattr_match_needed(dict_t *this, char *key1, data_t *value1, void *data)
{
- /* Ignore all non-disk (i.e. virtual) xattrs right away. */
- if (!gf_is_valid_xattr_namespace (key1))
- return _gf_false;
+ /* Ignore all non-disk (i.e. virtual) xattrs right away. */
+ if (!gf_is_valid_xattr_namespace(key1))
+ return _gf_false;
- /* Ignore on-disk xattrs that AFR doesn't need to heal. */
- if (!afr_is_xattr_ignorable (key1))
- return _gf_true;
+ /* Ignore on-disk xattrs that AFR doesn't need to heal. */
+ if (!afr_is_xattr_ignorable(key1))
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
gf_boolean_t
-afr_xattrs_are_equal (dict_t *dict1, dict_t *dict2)
+afr_xattrs_are_equal(dict_t *dict1, dict_t *dict2)
{
- return are_dicts_equal (dict1, dict2, afr_xattr_match_needed, NULL);
+ return are_dicts_equal(dict1, dict2, afr_xattr_match_needed, NULL);
}
static int
-afr_get_parent_read_subvol (xlator_t *this, inode_t *parent,
- struct afr_reply *replies, unsigned char *readable)
+afr_get_parent_read_subvol(xlator_t *this, inode_t *parent,
+ struct afr_reply *replies, unsigned char *readable)
{
- int i = 0;
- int par_read_subvol = -1;
- int par_read_subvol_iter = -1;
- afr_private_t *priv = NULL;
-
- priv = this->private;
+ int i = 0;
+ int par_read_subvol = -1;
+ int par_read_subvol_iter = -1;
+ afr_private_t *priv = NULL;
- if (parent)
- par_read_subvol = afr_data_subvol_get (parent, this, NULL, NULL,
- NULL, NULL);
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
+ if (parent)
+ par_read_subvol = afr_data_subvol_get(parent, this, NULL, NULL, NULL,
+ NULL);
- if (replies[i].op_ret < 0)
- continue;
-
- if (par_read_subvol_iter == -1) {
- par_read_subvol_iter = i;
- continue;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- if ((par_read_subvol_iter != par_read_subvol) && readable[i])
- par_read_subvol_iter = i;
+ if (replies[i].op_ret < 0)
+ continue;
- if (i == par_read_subvol)
- par_read_subvol_iter = i;
+ if (par_read_subvol_iter == -1) {
+ par_read_subvol_iter = i;
+ continue;
}
- /* At the end of the for-loop, the only reason why @par_read_subvol_iter
- * could be -1 is when this LOOKUP has failed on all sub-volumes.
- * So it is okay to send an arbitrary subvolume (0 in this case)
- * as parent read subvol.
- */
- if (par_read_subvol_iter == -1)
- par_read_subvol_iter = 0;
- return par_read_subvol_iter;
+ if ((par_read_subvol_iter != par_read_subvol) && readable[i])
+ par_read_subvol_iter = i;
+
+ if (i == par_read_subvol)
+ par_read_subvol_iter = i;
+ }
+ /* At the end of the for-loop, the only reason why @par_read_subvol_iter
+ * could be -1 is when this LOOKUP has failed on all sub-volumes.
+ * So it is okay to send an arbitrary subvolume (0 in this case)
+ * as parent read subvol.
+ */
+ if (par_read_subvol_iter == -1)
+ par_read_subvol_iter = 0;
+ return par_read_subvol_iter;
}
int
-afr_read_subvol_decide (inode_t *inode, xlator_t *this,
- afr_read_subvol_args_t *args, unsigned char *readable)
+afr_read_subvol_decide(inode_t *inode, xlator_t *this,
+ afr_read_subvol_args_t *args, unsigned char *readable)
{
- int event = 0;
- afr_private_t *priv = NULL;
- unsigned char *intersection = NULL;
+ int event = 0;
+ afr_private_t *priv = NULL;
+ unsigned char *intersection = NULL;
- priv = this->private;
- intersection = alloca0 (priv->child_count);
+ priv = this->private;
+ intersection = alloca0(priv->child_count);
- afr_readables_intersect_get (inode, this, &event, intersection);
+ afr_readables_intersect_get(inode, this, &event, intersection);
- if (AFR_COUNT (intersection, priv->child_count) <= 0) {
- /* TODO: If we have one brick with valid data_readable and
- * another with metadata_readable, try to send an iatt with
- * valid bits from both.*/
- return -1;
- }
+ if (AFR_COUNT(intersection, priv->child_count) <= 0) {
+ /* TODO: If we have one brick with valid data_readable and
+ * another with metadata_readable, try to send an iatt with
+ * valid bits from both.*/
+ return -1;
+ }
- memcpy (readable, intersection, sizeof (*readable) * priv->child_count);
+ memcpy(readable, intersection, sizeof(*readable) * priv->child_count);
- return afr_read_subvol_select_by_policy (inode, this, intersection,
- args);
+ return afr_read_subvol_select_by_policy(inode, this, intersection, args);
}
static inline int
-afr_first_up_child (call_frame_t *frame, xlator_t *this)
+afr_first_up_child(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++)
- if (local->replies[i].valid &&
- local->replies[i].op_ret == 0)
- return i;
- return -1;
+ for (i = 0; i < priv->child_count; i++)
+ if (local->replies[i].valid && local->replies[i].op_ret == 0)
+ return i;
+ return -1;
}
static void
-afr_attempt_readsubvol_set (call_frame_t *frame, xlator_t *this,
- unsigned char *success_replies,
- unsigned char *data_readable, int *read_subvol)
+afr_attempt_readsubvol_set(call_frame_t *frame, xlator_t *this,
+ unsigned char *success_replies,
+ unsigned char *data_readable, int *read_subvol)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int spb_choice = -1;
- int child_count = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int spb_choice = -1;
+ int child_count = -1;
- if (*read_subvol != -1)
- return;
+ if (*read_subvol != -1)
+ return;
- priv = this->private;
- local = frame->local;
- child_count = priv->child_count;
-
- afr_inode_split_brain_choice_get (local->inode, this,
- &spb_choice);
- if ((spb_choice >= 0) &&
- (AFR_COUNT(success_replies, child_count) == child_count)) {
- *read_subvol = spb_choice;
- } else if (!priv->quorum_count) {
- *read_subvol = afr_first_up_child (frame, this);
- } else if (priv->quorum_count &&
- afr_has_quorum (data_readable, this)) {
- /* read_subvol is guaranteed to be valid if we hit this path. */
- *read_subvol = afr_first_up_child (frame, this);
- } else {
- /* If quorum is enabled and we do not have a
- readable yet, it means all good copies are down.
- */
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_READ_SUBVOL_ERROR, "no read "
- "subvols for %s", local->loc.path);
- }
- if (*read_subvol >= 0)
- dict_del (local->replies[*read_subvol].xdata, GF_CONTENT_KEY);
+ priv = this->private;
+ local = frame->local;
+ child_count = priv->child_count;
+
+ afr_inode_split_brain_choice_get(local->inode, this, &spb_choice);
+ if ((spb_choice >= 0) &&
+ (AFR_COUNT(success_replies, child_count) == child_count)) {
+ *read_subvol = spb_choice;
+ } else if (!priv->quorum_count) {
+ *read_subvol = afr_first_up_child(frame, this);
+ } else if (priv->quorum_count && afr_has_quorum(data_readable, this)) {
+ /* read_subvol is guaranteed to be valid if we hit this path. */
+ *read_subvol = afr_first_up_child(frame, this);
+ } else {
+ /* If quorum is enabled and we do not have a
+ readable yet, it means all good copies are down.
+ */
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_READ_SUBVOL_ERROR,
+ "no read "
+ "subvols for %s",
+ local->loc.path);
+ }
+ if (*read_subvol >= 0)
+ dict_del(local->replies[*read_subvol].xdata, GF_CONTENT_KEY);
}
static void
-afr_lookup_done (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = -1;
- int op_errno = 0;
- int read_subvol = 0;
- int par_read_subvol = 0;
- int ret = -1;
- unsigned char *readable = NULL;
- unsigned char *success_replies = NULL;
- int event = 0;
- struct afr_reply *replies = NULL;
- uuid_t read_gfid = {0, };
- gf_boolean_t locked_entry = _gf_false;
- gf_boolean_t can_interpret = _gf_true;
- inode_t *parent = NULL;
- ia_type_t ia_type = IA_INVAL;
- afr_read_subvol_args_t args = {0,};
- char *gfid_heal_msg = NULL;
-
- priv = this->private;
- local = frame->local;
- replies = local->replies;
- parent = local->loc.parent;
-
- locked_entry = afr_is_possibly_under_txn (AFR_ENTRY_TRANSACTION, local,
- this);
-
- readable = alloca0 (priv->child_count);
- success_replies = alloca0 (priv->child_count);
-
- afr_inode_read_subvol_get (parent, this, readable, NULL, &event);
- par_read_subvol = afr_get_parent_read_subvol (this, parent, replies,
- readable);
-
- /* First, check if we have a gfid-change from somewhere,
- If so, propagate that so that a fresh lookup can be
- issued
- */
- if (local->cont.lookup.needs_fresh_lookup) {
- local->op_ret = -1;
- local->op_errno = ESTALE;
- goto error;
+afr_lookup_done(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = -1;
+ int op_errno = 0;
+ int read_subvol = 0;
+ int par_read_subvol = 0;
+ int ret = -1;
+ unsigned char *readable = NULL;
+ unsigned char *success_replies = NULL;
+ int event = 0;
+ struct afr_reply *replies = NULL;
+ uuid_t read_gfid = {
+ 0,
+ };
+ gf_boolean_t locked_entry = _gf_false;
+ gf_boolean_t can_interpret = _gf_true;
+ inode_t *parent = NULL;
+ ia_type_t ia_type = IA_INVAL;
+ afr_read_subvol_args_t args = {
+ 0,
+ };
+ char *gfid_heal_msg = NULL;
+
+ priv = this->private;
+ local = frame->local;
+ replies = local->replies;
+ parent = local->loc.parent;
+
+ locked_entry = afr_is_possibly_under_txn(AFR_ENTRY_TRANSACTION, local,
+ this);
+
+ readable = alloca0(priv->child_count);
+ success_replies = alloca0(priv->child_count);
+
+ afr_inode_read_subvol_get(parent, this, readable, NULL, &event);
+ par_read_subvol = afr_get_parent_read_subvol(this, parent, replies,
+ readable);
+
+ /* First, check if we have a gfid-change from somewhere,
+ If so, propagate that so that a fresh lookup can be
+ issued
+ */
+ if (local->cont.lookup.needs_fresh_lookup) {
+ local->op_ret = -1;
+ local->op_errno = ESTALE;
+ goto error;
+ }
+
+ op_errno = afr_final_errno(frame->local, this->private);
+ local->op_errno = op_errno;
+
+ read_subvol = -1;
+ afr_fill_success_replies(local, priv, success_replies);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+
+ if (locked_entry && replies[i].op_ret == -1 &&
+ replies[i].op_errno == ENOENT) {
+ /* Second, check entry is still
+ "underway" in creation */
+ local->op_ret = -1;
+ local->op_errno = ENOENT;
+ goto error;
}
- op_errno = afr_final_errno (frame->local, this->private);
- local->op_errno = op_errno;
+ if (replies[i].op_ret == -1)
+ continue;
- read_subvol = -1;
- afr_fill_success_replies (local, priv, success_replies);
+ if (read_subvol == -1 || !readable[read_subvol]) {
+ read_subvol = i;
+ gf_uuid_copy(read_gfid, replies[i].poststat.ia_gfid);
+ ia_type = replies[i].poststat.ia_type;
+ local->op_ret = 0;
+ }
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
+ if (read_subvol == -1)
+ goto error;
+ /* We now have a read_subvol, which is readable[] (if there
+ were any). Next we look for GFID mismatches. We don't
+ consider a GFID mismatch as an error if read_subvol is
+ readable[] but the mismatching GFID subvol is not.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1) {
+ continue;
+ }
- if (locked_entry && replies[i].op_ret == -1 &&
- replies[i].op_errno == ENOENT) {
- /* Second, check entry is still
- "underway" in creation */
- local->op_ret = -1;
- local->op_errno = ENOENT;
- goto error;
- }
+ if (!gf_uuid_compare(replies[i].poststat.ia_gfid, read_gfid))
+ continue;
- if (replies[i].op_ret == -1)
- continue;
+ can_interpret = _gf_false;
- if (read_subvol == -1 || !readable[read_subvol]) {
- read_subvol = i;
- gf_uuid_copy (read_gfid, replies[i].poststat.ia_gfid);
- ia_type = replies[i].poststat.ia_type;
- local->op_ret = 0;
- }
- }
+ if (locked_entry)
+ continue;
- if (read_subvol == -1)
- goto error;
- /* We now have a read_subvol, which is readable[] (if there
- were any). Next we look for GFID mismatches. We don't
- consider a GFID mismatch as an error if read_subvol is
- readable[] but the mismatching GFID subvol is not.
- */
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1) {
- continue;
- }
-
- if (!gf_uuid_compare (replies[i].poststat.ia_gfid, read_gfid))
- continue;
-
- can_interpret = _gf_false;
-
- if (locked_entry)
- continue;
-
- /* Now GFIDs mismatch. It's OK as long as this subvol
- is not readable[] but read_subvol is */
- if (readable[read_subvol] && !readable[i])
- continue;
-
- /* If we were called from glfsheal and there is still a gfid
- * mismatch, succeed the lookup and let glfsheal print the
- * response via gfid-heal-msg.*/
- if (!dict_get_str (local->xattr_req, "gfid-heal-msg",
- &gfid_heal_msg))
- goto cant_interpret;
-
- /* LOG ERROR */
- local->op_ret = -1;
- local->op_errno = EIO;
- goto error;
- }
-
- /* Forth, for the finalized GFID, pick the best subvolume
- to return stats from.
- */
- read_subvol = -1;
- memset (readable, 0, sizeof (*readable) * priv->child_count);
- if (can_interpret) {
- if (!afr_has_quorum (success_replies, this))
- goto cant_interpret;
- /* It is safe to call afr_replies_interpret() because we have
- a response from all the UP subvolumes and all of them resolved
- to the same GFID
- */
- gf_uuid_copy (args.gfid, read_gfid);
- args.ia_type = ia_type;
- ret = afr_replies_interpret (frame, this, local->inode, NULL);
- read_subvol = afr_read_subvol_decide (local->inode, this, &args,
- readable);
- if (read_subvol == -1)
- goto cant_interpret;
- if (ret) {
- afr_inode_event_gen_reset (local->inode, this);
- dict_del (local->replies[read_subvol].xdata,
- GF_CONTENT_KEY);
- }
- } else {
- cant_interpret:
- afr_attempt_readsubvol_set (frame, this, success_replies,
- readable, &read_subvol);
- if (read_subvol == -1) {
- goto error;
- }
- }
+ /* Now GFIDs mismatch. It's OK as long as this subvol
+ is not readable[] but read_subvol is */
+ if (readable[read_subvol] && !readable[i])
+ continue;
- afr_handle_quota_size (frame, this);
+ /* If we were called from glfsheal and there is still a gfid
+ * mismatch, succeed the lookup and let glfsheal print the
+ * response via gfid-heal-msg.*/
+ if (!dict_get_str(local->xattr_req, "gfid-heal-msg", &gfid_heal_msg))
+ goto cant_interpret;
- afr_set_need_heal (this, local);
- if (AFR_IS_ARBITER_BRICK (priv, read_subvol) && local->op_ret == 0) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- gf_msg_debug(this->name, 0, "Arbiter cannot be a read subvol "
- "for %s", local->loc.path);
- goto error;
+ /* LOG ERROR */
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ goto error;
+ }
+
+ /* Forth, for the finalized GFID, pick the best subvolume
+ to return stats from.
+ */
+ read_subvol = -1;
+ memset(readable, 0, sizeof(*readable) * priv->child_count);
+ if (can_interpret) {
+ if (!afr_has_quorum(success_replies, this))
+ goto cant_interpret;
+ /* It is safe to call afr_replies_interpret() because we have
+ a response from all the UP subvolumes and all of them resolved
+ to the same GFID
+ */
+ gf_uuid_copy(args.gfid, read_gfid);
+ args.ia_type = ia_type;
+ ret = afr_replies_interpret(frame, this, local->inode, NULL);
+ read_subvol = afr_read_subvol_decide(local->inode, this, &args,
+ readable);
+ if (read_subvol == -1)
+ goto cant_interpret;
+ if (ret) {
+ afr_inode_event_gen_reset(local->inode, this);
+ dict_del(local->replies[read_subvol].xdata, GF_CONTENT_KEY);
+ }
+ } else {
+ cant_interpret:
+ afr_attempt_readsubvol_set(frame, this, success_replies, readable,
+ &read_subvol);
+ if (read_subvol == -1) {
+ goto error;
}
+ }
- ret = dict_get_str (local->xattr_req, "gfid-heal-msg", &gfid_heal_msg);
- if (!ret) {
- ret = dict_set_str (local->replies[read_subvol].xdata,
- "gfid-heal-msg", gfid_heal_msg);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED,
- "Error setting gfid-heal-msg dict");
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- }
+ afr_handle_quota_size(frame, this);
+
+ afr_set_need_heal(this, local);
+ if (AFR_IS_ARBITER_BRICK(priv, read_subvol) && local->op_ret == 0) {
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ gf_msg_debug(this->name, 0,
+ "Arbiter cannot be a read subvol "
+ "for %s",
+ local->loc.path);
+ goto error;
+ }
+
+ ret = dict_get_str(local->xattr_req, "gfid-heal-msg", &gfid_heal_msg);
+ if (!ret) {
+ ret = dict_set_str(local->replies[read_subvol].xdata, "gfid-heal-msg",
+ gfid_heal_msg);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_SET_FAILED,
+ "Error setting gfid-heal-msg dict");
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
}
+ }
- AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->replies[read_subvol].poststat,
- local->replies[read_subvol].xdata,
- &local->replies[par_read_subvol].postparent);
- return;
+ AFR_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->replies[read_subvol].poststat,
+ local->replies[read_subvol].xdata,
+ &local->replies[par_read_subvol].postparent);
+ return;
error:
- AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno, NULL,
- NULL, NULL, NULL);
+ AFR_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno, NULL, NULL,
+ NULL, NULL);
}
/*
@@ -2529,843 +2468,834 @@ error:
*/
int
-afr_higher_errno (int32_t old_errno, int32_t new_errno)
+afr_higher_errno(int32_t old_errno, int32_t new_errno)
{
- if (old_errno == ENODATA || new_errno == ENODATA)
- return ENODATA;
- if (old_errno == ENOENT || new_errno == ENOENT)
- return ENOENT;
- if (old_errno == ESTALE || new_errno == ESTALE)
- return ESTALE;
+ if (old_errno == ENODATA || new_errno == ENODATA)
+ return ENODATA;
+ if (old_errno == ENOENT || new_errno == ENOENT)
+ return ENOENT;
+ if (old_errno == ESTALE || new_errno == ESTALE)
+ return ESTALE;
- return new_errno;
+ return new_errno;
}
-
int
-afr_final_errno (afr_local_t *local, afr_private_t *priv)
+afr_final_errno(afr_local_t *local, afr_private_t *priv)
{
- int i = 0;
- int op_errno = 0;
- int tmp_errno = 0;
+ int i = 0;
+ int op_errno = 0;
+ int tmp_errno = 0;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret >= 0)
- continue;
- tmp_errno = local->replies[i].op_errno;
- op_errno = afr_higher_errno (op_errno, tmp_errno);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret >= 0)
+ continue;
+ tmp_errno = local->replies[i].op_errno;
+ op_errno = afr_higher_errno(op_errno, tmp_errno);
+ }
- return op_errno;
+ return op_errno;
}
static int32_t
-afr_local_discovery_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
-{
- int ret = 0;
- char *pathinfo = NULL;
- gf_boolean_t is_local = _gf_false;
- afr_private_t *priv = NULL;
- int32_t child_index = -1;
-
- if (op_ret != 0) {
- goto out;
- }
-
- priv = this->private;
- child_index = (int32_t)(long)cookie;
-
- ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
- if (ret != 0) {
- goto out;
- }
-
- ret = glusterfs_is_local_pathinfo (pathinfo, &is_local);
- if (ret) {
- goto out;
- }
-
- /*
- * Note that one local subvolume will override another here. The only
- * way to avoid that would be to retain extra information about whether
- * the previous read_child is local, and it's just not worth it. Even
- * the slowest local subvolume is far preferable to a remote one.
- */
- if (is_local) {
- priv->local[child_index] = 1;
- /* Don't set arbiter as read child. */
- if (AFR_IS_ARBITER_BRICK(priv, child_index))
- goto out;
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_LOCAL_CHILD, "selecting local read_child %s",
- priv->children[child_index]->name);
-
- priv->read_child = child_index;
- }
+afr_local_discovery_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
+{
+ int ret = 0;
+ char *pathinfo = NULL;
+ gf_boolean_t is_local = _gf_false;
+ afr_private_t *priv = NULL;
+ int32_t child_index = -1;
+
+ if (op_ret != 0) {
+ goto out;
+ }
+
+ priv = this->private;
+ child_index = (int32_t)(long)cookie;
+
+ ret = dict_get_str(dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
+ if (ret != 0) {
+ goto out;
+ }
+
+ ret = glusterfs_is_local_pathinfo(pathinfo, &is_local);
+ if (ret) {
+ goto out;
+ }
+
+ /*
+ * Note that one local subvolume will override another here. The only
+ * way to avoid that would be to retain extra information about whether
+ * the previous read_child is local, and it's just not worth it. Even
+ * the slowest local subvolume is far preferable to a remote one.
+ */
+ if (is_local) {
+ priv->local[child_index] = 1;
+ /* Don't set arbiter as read child. */
+ if (AFR_IS_ARBITER_BRICK(priv, child_index))
+ goto out;
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_LOCAL_CHILD,
+ "selecting local read_child %s",
+ priv->children[child_index]->name);
+
+ priv->read_child = child_index;
+ }
out:
- STACK_DESTROY(frame->root);
- return 0;
+ STACK_DESTROY(frame->root);
+ return 0;
}
static void
-afr_attempt_local_discovery (xlator_t *this, int32_t child_index)
+afr_attempt_local_discovery(xlator_t *this, int32_t child_index)
{
- call_frame_t *newframe = NULL;
- loc_t tmploc = {0,};
- afr_private_t *priv = this->private;
+ call_frame_t *newframe = NULL;
+ loc_t tmploc = {
+ 0,
+ };
+ afr_private_t *priv = this->private;
- newframe = create_frame(this,this->ctx->pool);
- if (!newframe) {
- return;
- }
+ newframe = create_frame(this, this->ctx->pool);
+ if (!newframe) {
+ return;
+ }
- tmploc.gfid[sizeof(tmploc.gfid)-1] = 1;
- STACK_WIND_COOKIE (newframe, afr_local_discovery_cbk,
- (void *)(long)child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->getxattr,
- &tmploc, GF_XATTR_PATHINFO_KEY, NULL);
+ tmploc.gfid[sizeof(tmploc.gfid) - 1] = 1;
+ STACK_WIND_COOKIE(newframe, afr_local_discovery_cbk,
+ (void *)(long)child_index, priv->children[child_index],
+ priv->children[child_index]->fops->getxattr, &tmploc,
+ GF_XATTR_PATHINFO_KEY, NULL);
}
int
-afr_lookup_sh_metadata_wrap (void *opaque)
-{
- call_frame_t *frame = opaque;
- afr_local_t *local = NULL;
- xlator_t *this = NULL;
- inode_t *inode = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- int i= 0, first = -1;
- int ret = -1;
- dict_t *dict = NULL;
-
- local = frame->local;
- this = frame->this;
- priv = this->private;
- replies = local->replies;
-
- for (i =0; i < priv->child_count; i++) {
- if(!replies[i].valid || replies[i].op_ret == -1)
- continue;
- first = i;
- break;
- }
- if (first == -1)
- goto out;
-
- if (afr_selfheal_metadata_by_stbuf (this, &replies[first].poststat))
- goto out;
-
- afr_local_replies_wipe (local, this->private);
-
- dict = dict_new ();
- if (!dict)
- goto out;
- ret = dict_set_str (dict, "link-count", GF_XATTROP_INDEX_COUNT);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set link-count in dict ");
- }
-
- if (loc_is_nameless (&local->loc)) {
- ret = afr_selfheal_unlocked_discover_on (frame, local->inode,
- local->loc.gfid,
- local->replies,
- local->child_up);
- } else {
- inode = afr_selfheal_unlocked_lookup_on (frame,
- local->loc.parent,
- local->loc.name,
- local->replies,
- local->child_up, dict);
- }
- if (inode)
- inode_unref (inode);
+afr_lookup_sh_metadata_wrap(void *opaque)
+{
+ call_frame_t *frame = opaque;
+ afr_local_t *local = NULL;
+ xlator_t *this = NULL;
+ inode_t *inode = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ int i = 0, first = -1;
+ int ret = -1;
+ dict_t *dict = NULL;
+
+ local = frame->local;
+ this = frame->this;
+ priv = this->private;
+ replies = local->replies;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ first = i;
+ break;
+ }
+ if (first == -1)
+ goto out;
+
+ if (afr_selfheal_metadata_by_stbuf(this, &replies[first].poststat))
+ goto out;
+
+ afr_local_replies_wipe(local, this->private);
+
+ dict = dict_new();
+ if (!dict)
+ goto out;
+ ret = dict_set_str(dict, "link-count", GF_XATTROP_INDEX_COUNT);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to set link-count in dict ");
+ }
+
+ if (loc_is_nameless(&local->loc)) {
+ ret = afr_selfheal_unlocked_discover_on(frame, local->inode,
+ local->loc.gfid, local->replies,
+ local->child_up);
+ } else {
+ inode = afr_selfheal_unlocked_lookup_on(frame, local->loc.parent,
+ local->loc.name, local->replies,
+ local->child_up, dict);
+ }
+ if (inode)
+ inode_unref(inode);
out:
- if (loc_is_nameless (&local->loc))
- afr_discover_done (frame, this);
- else
- afr_lookup_done (frame, this);
+ if (loc_is_nameless(&local->loc))
+ afr_discover_done(frame, this);
+ else
+ afr_lookup_done(frame, this);
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return 0;
+ return 0;
}
gf_boolean_t
-afr_is_pending_set (xlator_t *this, dict_t *xdata, int type)
+afr_is_pending_set(xlator_t *this, dict_t *xdata, int type)
{
- int idx = -1;
- afr_private_t *priv = NULL;
- void *pending_raw = NULL;
- int *pending_int = NULL;
- int i = 0;
+ int idx = -1;
+ afr_private_t *priv = NULL;
+ void *pending_raw = NULL;
+ int *pending_int = NULL;
+ int i = 0;
- priv = this->private;
- idx = afr_index_for_transaction_type (type);
+ priv = this->private;
+ idx = afr_index_for_transaction_type(type);
- if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw) == 0) {
- if (pending_raw) {
- pending_int = pending_raw;
+ if (dict_get_ptr(xdata, AFR_DIRTY, &pending_raw) == 0) {
+ if (pending_raw) {
+ pending_int = pending_raw;
- if (ntoh32 (pending_int[idx]))
- return _gf_true;
- }
+ if (ntoh32(pending_int[idx]))
+ return _gf_true;
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get_ptr (xdata, priv->pending_key[i],
- &pending_raw))
- continue;
- if (!pending_raw)
- continue;
- pending_int = pending_raw;
-
- if (ntoh32 (pending_int[idx]))
- return _gf_true;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_get_ptr(xdata, priv->pending_key[i], &pending_raw))
+ continue;
+ if (!pending_raw)
+ continue;
+ pending_int = pending_raw;
- return _gf_false;
+ if (ntoh32(pending_int[idx]))
+ return _gf_true;
+ }
+
+ return _gf_false;
}
static gf_boolean_t
afr_can_start_metadata_self_heal(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- int i = 0, first = -1;
- gf_boolean_t start = _gf_false;
- struct iatt stbuf = {0, };
-
- local = frame->local;
- replies = local->replies;
- priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ int i = 0, first = -1;
+ gf_boolean_t start = _gf_false;
+ struct iatt stbuf = {
+ 0,
+ };
- if (!priv->metadata_self_heal)
- return _gf_false;
+ local = frame->local;
+ replies = local->replies;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if(!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (first == -1) {
- first = i;
- stbuf = replies[i].poststat;
- continue;
- }
+ if (!priv->metadata_self_heal)
+ return _gf_false;
- if (afr_is_pending_set (this, replies[i].xdata,
- AFR_METADATA_TRANSACTION)) {
- /* Let shd do the heal so that lookup is not blocked
- * on getting metadata lock/doing the heal */
- start = _gf_false;
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (first == -1) {
+ first = i;
+ stbuf = replies[i].poststat;
+ continue;
+ }
- if (gf_uuid_compare (stbuf.ia_gfid, replies[i].poststat.ia_gfid)) {
- start = _gf_false;
- break;
- }
- if (!IA_EQUAL (stbuf, replies[i].poststat, type)) {
- start = _gf_false;
- break;
- }
+ if (afr_is_pending_set(this, replies[i].xdata,
+ AFR_METADATA_TRANSACTION)) {
+ /* Let shd do the heal so that lookup is not blocked
+ * on getting metadata lock/doing the heal */
+ start = _gf_false;
+ break;
+ }
- /*Check if iattrs need heal*/
- if ((!IA_EQUAL (stbuf, replies[i].poststat, uid)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, gid)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, prot))) {
- start = _gf_true;
- continue;
- }
+ if (gf_uuid_compare(stbuf.ia_gfid, replies[i].poststat.ia_gfid)) {
+ start = _gf_false;
+ break;
+ }
+ if (!IA_EQUAL(stbuf, replies[i].poststat, type)) {
+ start = _gf_false;
+ break;
+ }
- /*Check if xattrs need heal*/
- if (!afr_xattrs_are_equal (replies[first].xdata,
- replies[i].xdata))
- start = _gf_true;
+ /*Check if iattrs need heal*/
+ if ((!IA_EQUAL(stbuf, replies[i].poststat, uid)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, gid)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, prot))) {
+ start = _gf_true;
+ continue;
}
- return start;
+ /*Check if xattrs need heal*/
+ if (!afr_xattrs_are_equal(replies[first].xdata, replies[i].xdata))
+ start = _gf_true;
+ }
+
+ return start;
}
int
-afr_lookup_metadata_heal_check (call_frame_t *frame, xlator_t *this)
+afr_lookup_metadata_heal_check(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *heal = NULL;
- afr_local_t *local = NULL;
- int ret = 0;
+ call_frame_t *heal = NULL;
+ afr_local_t *local = NULL;
+ int ret = 0;
- local = frame->local;
- if (!afr_can_start_metadata_self_heal (frame, this))
- goto out;
+ local = frame->local;
+ if (!afr_can_start_metadata_self_heal(frame, this))
+ goto out;
- heal = afr_frame_create (this, &ret);
- if (!heal) {
- ret = -ret;
- goto out;
- }
+ heal = afr_frame_create(this, &ret);
+ if (!heal) {
+ ret = -ret;
+ goto out;
+ }
- ret = synctask_new (this->ctx->env, afr_lookup_sh_metadata_wrap,
- afr_refresh_selfheal_done, heal, frame);
- if (ret)
- goto out;
- return ret;
+ ret = synctask_new(this->ctx->env, afr_lookup_sh_metadata_wrap,
+ afr_refresh_selfheal_done, heal, frame);
+ if (ret)
+ goto out;
+ return ret;
out:
- if (loc_is_nameless (&local->loc))
- afr_discover_done (frame, this);
- else
- afr_lookup_done (frame, this);
- if (heal)
- AFR_STACK_DESTROY (heal);
- return ret;
+ if (loc_is_nameless(&local->loc))
+ afr_discover_done(frame, this);
+ else
+ afr_lookup_done(frame, this);
+ if (heal)
+ AFR_STACK_DESTROY(heal);
+ return ret;
}
int
-afr_lookup_selfheal_wrap (void *opaque)
+afr_lookup_selfheal_wrap(void *opaque)
{
- int ret = 0;
- call_frame_t *frame = opaque;
- afr_local_t *local = NULL;
- xlator_t *this = NULL;
- inode_t *inode = NULL;
- uuid_t pargfid = {0,};
+ int ret = 0;
+ call_frame_t *frame = opaque;
+ afr_local_t *local = NULL;
+ xlator_t *this = NULL;
+ inode_t *inode = NULL;
+ uuid_t pargfid = {
+ 0,
+ };
- local = frame->local;
- this = frame->this;
- loc_pargfid (&local->loc, pargfid);
+ local = frame->local;
+ this = frame->this;
+ loc_pargfid(&local->loc, pargfid);
- ret = afr_selfheal_name (frame->this, pargfid, local->loc.name,
- &local->cont.lookup.gfid_req, local->xattr_req);
- if (ret == -EIO)
- goto unwind;
+ ret = afr_selfheal_name(frame->this, pargfid, local->loc.name,
+ &local->cont.lookup.gfid_req, local->xattr_req);
+ if (ret == -EIO)
+ goto unwind;
- afr_local_replies_wipe (local, this->private);
+ afr_local_replies_wipe(local, this->private);
- inode = afr_selfheal_unlocked_lookup_on (frame, local->loc.parent,
- local->loc.name, local->replies,
- local->child_up, NULL);
- if (inode)
- inode_unref (inode);
+ inode = afr_selfheal_unlocked_lookup_on(frame, local->loc.parent,
+ local->loc.name, local->replies,
+ local->child_up, NULL);
+ if (inode)
+ inode_unref(inode);
- afr_lookup_metadata_heal_check(frame, this);
- return 0;
+ afr_lookup_metadata_heal_check(frame, this);
+ return 0;
unwind:
- AFR_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(lookup, frame, -1, EIO, NULL, NULL, NULL, NULL);
+ return 0;
}
int
-afr_lookup_entry_heal (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- call_frame_t *heal = NULL;
- int i = 0, first = -1;
- gf_boolean_t name_state_mismatch = _gf_false;
- struct afr_reply *replies = NULL;
- int ret = 0;
- unsigned char *par_readables = NULL;
- unsigned char *success = NULL;
- int32_t op_errno = 0;
- uuid_t gfid = {0};
-
- local = frame->local;
- replies = local->replies;
- priv = this->private;
- par_readables = alloca0(priv->child_count);
- success = alloca0(priv->child_count);
-
- ret = afr_inode_read_subvol_get (local->loc.parent, this, par_readables,
- NULL, NULL);
- if (ret < 0 || AFR_COUNT (par_readables, priv->child_count) == 0) {
- /* In this case set par_readables to all 1 so that name_heal
- * need checks at the end of this function will flag missing
- * entry when name state mismatches*/
- memset (par_readables, 1, priv->child_count);
+afr_lookup_entry_heal(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ call_frame_t *heal = NULL;
+ int i = 0, first = -1;
+ gf_boolean_t name_state_mismatch = _gf_false;
+ struct afr_reply *replies = NULL;
+ int ret = 0;
+ unsigned char *par_readables = NULL;
+ unsigned char *success = NULL;
+ int32_t op_errno = 0;
+ uuid_t gfid = {0};
+
+ local = frame->local;
+ replies = local->replies;
+ priv = this->private;
+ par_readables = alloca0(priv->child_count);
+ success = alloca0(priv->child_count);
+
+ ret = afr_inode_read_subvol_get(local->loc.parent, this, par_readables,
+ NULL, NULL);
+ if (ret < 0 || AFR_COUNT(par_readables, priv->child_count) == 0) {
+ /* In this case set par_readables to all 1 so that name_heal
+ * need checks at the end of this function will flag missing
+ * entry when name state mismatches*/
+ memset(par_readables, 1, priv->child_count);
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret == 0) {
+ if (uuid_is_null(gfid)) {
+ gf_uuid_copy(gfid, replies[i].poststat.ia_gfid);
+ }
+ success[i] = 1;
+ } else {
+ if ((replies[i].op_errno != ENOTCONN) &&
+ (replies[i].op_errno != ENOENT) &&
+ (replies[i].op_errno != ESTALE)) {
+ op_errno = replies[i].op_errno;
+ }
}
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret == 0) {
- if (uuid_is_null (gfid)) {
- gf_uuid_copy (gfid,
- replies[i].poststat.ia_gfid);
- }
- success[i] = 1;
- } else {
- if ((replies[i].op_errno != ENOTCONN) &&
- (replies[i].op_errno != ENOENT) &&
- (replies[i].op_errno != ESTALE)) {
- op_errno = replies[i].op_errno;
- }
- }
+ /*gfid is missing, needs heal*/
+ if ((replies[i].op_ret == -1) && (replies[i].op_errno == ENODATA)) {
+ goto name_heal;
+ }
- /*gfid is missing, needs heal*/
- if ((replies[i].op_ret == -1) &&
- (replies[i].op_errno == ENODATA)) {
- goto name_heal;
- }
+ if (first == -1) {
+ first = i;
+ continue;
+ }
- if (first == -1) {
- first = i;
- continue;
- }
-
- if (replies[i].op_ret != replies[first].op_ret) {
- name_state_mismatch = _gf_true;
- }
-
- if (replies[i].op_ret == 0) {
- /* Rename after this lookup may succeed if we don't do
- * a name-heal and the destination may not have pending xattrs
- * to indicate which name is good and which is bad so always do
- * this heal*/
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- gfid)) {
- goto name_heal;
- }
- }
- }
-
- if (name_state_mismatch) {
- if (!priv->quorum_count)
- goto name_heal;
- if (!afr_has_quorum (success, this))
- goto name_heal;
- if (op_errno)
- goto name_heal;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
- if (par_readables[i] && replies[i].op_ret < 0 &&
- replies[i].op_errno != ENOTCONN) {
- goto name_heal;
- }
- }
+ if (replies[i].op_ret != replies[first].op_ret) {
+ name_state_mismatch = _gf_true;
}
- goto metadata_heal;
+ if (replies[i].op_ret == 0) {
+ /* Rename after this lookup may succeed if we don't do
+ * a name-heal and the destination may not have pending xattrs
+ * to indicate which name is good and which is bad so always do
+ * this heal*/
+ if (gf_uuid_compare(replies[i].poststat.ia_gfid, gfid)) {
+ goto name_heal;
+ }
+ }
+ }
+
+ if (name_state_mismatch) {
+ if (!priv->quorum_count)
+ goto name_heal;
+ if (!afr_has_quorum(success, this))
+ goto name_heal;
+ if (op_errno)
+ goto name_heal;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (par_readables[i] && replies[i].op_ret < 0 &&
+ replies[i].op_errno != ENOTCONN) {
+ goto name_heal;
+ }
+ }
+ }
+
+ goto metadata_heal;
name_heal:
- heal = afr_frame_create (this, NULL);
- if (!heal)
- goto metadata_heal;
+ heal = afr_frame_create(this, NULL);
+ if (!heal)
+ goto metadata_heal;
- ret = synctask_new (this->ctx->env, afr_lookup_selfheal_wrap,
- afr_refresh_selfheal_done, heal, frame);
- if (ret) {
- AFR_STACK_DESTROY (heal);
- goto metadata_heal;
- }
- return ret;
+ ret = synctask_new(this->ctx->env, afr_lookup_selfheal_wrap,
+ afr_refresh_selfheal_done, heal, frame);
+ if (ret) {
+ AFR_STACK_DESTROY(heal);
+ goto metadata_heal;
+ }
+ return ret;
metadata_heal:
- ret = afr_lookup_metadata_heal_check (frame, this);
+ ret = afr_lookup_metadata_heal_check(frame, this);
- return ret;
+ return ret;
}
-
int
-afr_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
- dict_t *xdata, struct iatt *postparent)
-{
- afr_local_t * local = NULL;
- int call_count = -1;
- int child_index = -1;
- GF_UNUSED int ret = 0;
- int8_t need_heal = 1;
-
- child_index = (long) cookie;
-
- local = frame->local;
-
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- /*
- * On revalidate lookup if the gfid-changed, afr should unwind the fop
- * with ESTALE so that a fresh lookup will be sent by the top xlator.
- * So remember it.
- */
- if (xdata && dict_get (xdata, "gfid-changed"))
- local->cont.lookup.needs_fresh_lookup = _gf_true;
+afr_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int child_index = -1;
+ GF_UNUSED int ret = 0;
+ int8_t need_heal = 1;
+
+ child_index = (long)cookie;
+
+ local = frame->local;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ /*
+ * On revalidate lookup if the gfid-changed, afr should unwind the fop
+ * with ESTALE so that a fresh lookup will be sent by the top xlator.
+ * So remember it.
+ */
+ if (xdata && dict_get(xdata, "gfid-changed"))
+ local->cont.lookup.needs_fresh_lookup = _gf_true;
+
+ if (xdata) {
+ ret = dict_get_int8(xdata, "link-count", &need_heal);
+ local->replies[child_index].need_heal = need_heal;
+ } else {
+ local->replies[child_index].need_heal = need_heal;
+ }
+ if (op_ret != -1) {
+ local->replies[child_index].poststat = *buf;
+ local->replies[child_index].postparent = *postparent;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
+ }
- if (xdata) {
- ret = dict_get_int8 (xdata, "link-count", &need_heal);
- local->replies[child_index].need_heal = need_heal;
- } else {
- local->replies[child_index].need_heal = need_heal;
- }
- if (op_ret != -1) {
- local->replies[child_index].poststat = *buf;
- local->replies[child_index].postparent = *postparent;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
- }
-
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_need_heal (this, local);
- afr_lookup_entry_heal (frame, this);
- }
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ afr_set_need_heal(this, local);
+ afr_lookup_entry_heal(frame, this);
+ }
- return 0;
+ return 0;
}
static void
-afr_discover_unwind (call_frame_t *frame, xlator_t *this)
+afr_discover_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int op_errno = 0;
- int read_subvol = -1;
- unsigned char *data_readable = NULL;
- unsigned char *success_replies = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int op_errno = 0;
+ int read_subvol = -1;
+ unsigned char *data_readable = NULL;
+ unsigned char *success_replies = NULL;
- priv = this->private;
- local = frame->local;
- data_readable = alloca0 (priv->child_count);
- success_replies = alloca0 (priv->child_count);
+ priv = this->private;
+ local = frame->local;
+ data_readable = alloca0(priv->child_count);
+ success_replies = alloca0(priv->child_count);
- afr_fill_success_replies (local, priv, success_replies);
- if (AFR_COUNT (success_replies, priv->child_count) > 0)
- local->op_ret = 0;
+ afr_fill_success_replies(local, priv, success_replies);
+ if (AFR_COUNT(success_replies, priv->child_count) > 0)
+ local->op_ret = 0;
- op_errno = afr_final_errno (frame->local, this->private);
+ op_errno = afr_final_errno(frame->local, this->private);
- if (local->op_ret < 0) {
- AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL,
- NULL, NULL);
- return;
- }
+ if (local->op_ret < 0) {
+ AFR_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ return;
+ }
- if (!afr_has_quorum (success_replies, this))
- goto unwind;
+ if (!afr_has_quorum(success_replies, this))
+ goto unwind;
- afr_replies_interpret (frame, this, local->inode, NULL);
+ afr_replies_interpret(frame, this, local->inode, NULL);
- read_subvol = afr_read_subvol_decide (local->inode, this, NULL,
- data_readable);
+ read_subvol = afr_read_subvol_decide(local->inode, this, NULL,
+ data_readable);
unwind:
- afr_attempt_readsubvol_set (frame, this, success_replies, data_readable,
- &read_subvol);
- if (read_subvol == -1) {
- AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- NULL, NULL, NULL, NULL);
- return;
- }
+ afr_attempt_readsubvol_set(frame, this, success_replies, data_readable,
+ &read_subvol);
+ if (read_subvol == -1) {
+ AFR_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno, NULL,
+ NULL, NULL, NULL);
+ return;
+ }
- if (AFR_IS_ARBITER_BRICK (priv, read_subvol) && local->op_ret == 0) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- gf_msg_debug (this->name, 0, "Arbiter cannot be a read subvol "
- "for %s", local->loc.path);
- }
+ if (AFR_IS_ARBITER_BRICK(priv, read_subvol) && local->op_ret == 0) {
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ gf_msg_debug(this->name, 0,
+ "Arbiter cannot be a read subvol "
+ "for %s",
+ local->loc.path);
+ }
- AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->replies[read_subvol].poststat,
- local->replies[read_subvol].xdata,
- &local->replies[read_subvol].postparent);
+ AFR_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->replies[read_subvol].poststat,
+ local->replies[read_subvol].xdata,
+ &local->replies[read_subvol].postparent);
}
static int
-afr_ta_id_file_check (void *opaque)
-{
- afr_private_t *priv = NULL;
- xlator_t *this = NULL;
- loc_t loc = {0, };
- struct iatt stbuf = {0,};
- dict_t *dict = NULL;
- uuid_t gfid = {0,};
- fd_t *fd = NULL;
- int ret = 0;
-
- this = opaque;
- priv = this->private;
-
- ret = afr_fill_ta_loc (this, &loc);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed to populate thin-arbiter loc for: %s.",
- loc.name);
- goto out;
- }
-
- ret = syncop_lookup (priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
- &stbuf, 0, 0, 0);
- if (ret == 0) {
- goto out;
- } else if (ret == -ENOENT) {
- fd = fd_create (loc.inode, getpid());
- if (!fd)
- goto out;
- dict = dict_new ();
- if (!dict)
- goto out;
- gf_uuid_generate (gfid);
- ret = dict_set_gfuuid (dict, "gfid-req", gfid, true);
- ret = syncop_create (priv->children[THIN_ARBITER_BRICK_INDEX],
- &loc, O_RDWR, 0664, fd, &stbuf, dict,
- NULL);
- }
+afr_ta_id_file_check(void *opaque)
+{
+ afr_private_t *priv = NULL;
+ xlator_t *this = NULL;
+ loc_t loc = {
+ 0,
+ };
+ struct iatt stbuf = {
+ 0,
+ };
+ dict_t *dict = NULL;
+ uuid_t gfid = {
+ 0,
+ };
+ fd_t *fd = NULL;
+ int ret = 0;
+
+ this = opaque;
+ priv = this->private;
+
+ ret = afr_fill_ta_loc(this, &loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate thin-arbiter loc for: %s.", loc.name);
+ goto out;
+ }
+
+ ret = syncop_lookup(priv->children[THIN_ARBITER_BRICK_INDEX], &loc, &stbuf,
+ 0, 0, 0);
+ if (ret == 0) {
+ goto out;
+ } else if (ret == -ENOENT) {
+ fd = fd_create(loc.inode, getpid());
+ if (!fd)
+ goto out;
+ dict = dict_new();
+ if (!dict)
+ goto out;
+ gf_uuid_generate(gfid);
+ ret = dict_set_gfuuid(dict, "gfid-req", gfid, true);
+ ret = syncop_create(priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
+ O_RDWR, 0664, fd, &stbuf, dict, NULL);
+ }
out:
- if (ret == 0) {
- gf_uuid_copy (priv->ta_gfid, stbuf.ia_gfid);
- } else {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed to lookup/create thin-arbiter id file.");
- }
- if (dict)
- dict_unref (dict);
- if (fd)
- fd_unref (fd);
- loc_wipe (&loc);
+ if (ret == 0) {
+ gf_uuid_copy(priv->ta_gfid, stbuf.ia_gfid);
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to lookup/create thin-arbiter id file.");
+ }
+ if (dict)
+ dict_unref(dict);
+ if (fd)
+ fd_unref(fd);
+ loc_wipe(&loc);
- return 0;
+ return 0;
}
static int
-afr_ta_id_file_check_cbk (int ret, call_frame_t *ta_frame, void *opaque)
+afr_ta_id_file_check_cbk(int ret, call_frame_t *ta_frame, void *opaque)
{
- return 0;
+ return 0;
}
static void
-afr_discover_done (call_frame_t *frame, xlator_t *this)
+afr_discover_done(call_frame_t *frame, xlator_t *this)
{
- int ret = 0;
- afr_private_t *priv = NULL;
+ int ret = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
- if (!priv->thin_arbiter_count)
- goto unwind;
- if (!gf_uuid_is_null(priv->ta_gfid))
- goto unwind;
+ priv = this->private;
+ if (!priv->thin_arbiter_count)
+ goto unwind;
+ if (!gf_uuid_is_null(priv->ta_gfid))
+ goto unwind;
- ret = synctask_new (this->ctx->env, afr_ta_id_file_check,
- afr_ta_id_file_check_cbk, NULL, this);
- if (ret)
- goto unwind;
+ ret = synctask_new(this->ctx->env, afr_ta_id_file_check,
+ afr_ta_id_file_check_cbk, NULL, this);
+ if (ret)
+ goto unwind;
unwind:
- afr_discover_unwind (frame, this);
+ afr_discover_unwind(frame, this);
}
int
-afr_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
- dict_t *xdata, struct iatt *postparent)
-{
- afr_local_t * local = NULL;
- int call_count = -1;
- int child_index = -1;
- GF_UNUSED int ret = 0;
- int8_t need_heal = 1;
-
- child_index = (long) cookie;
-
- local = frame->local;
-
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (op_ret != -1) {
- local->replies[child_index].poststat = *buf;
- local->replies[child_index].postparent = *postparent;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
- }
-
- if (local->do_discovery && (op_ret == 0))
- afr_attempt_local_discovery (this, child_index);
-
- if (xdata) {
- ret = dict_get_int8 (xdata, "link-count", &need_heal);
- local->replies[child_index].need_heal = need_heal;
- } else {
- local->replies[child_index].need_heal = need_heal;
- }
-
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_need_heal (this, local);
- afr_lookup_metadata_heal_check (frame, this);
- }
+afr_discover_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int child_index = -1;
+ GF_UNUSED int ret = 0;
+ int8_t need_heal = 1;
+
+ child_index = (long)cookie;
+
+ local = frame->local;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (op_ret != -1) {
+ local->replies[child_index].poststat = *buf;
+ local->replies[child_index].postparent = *postparent;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
+ }
+
+ if (local->do_discovery && (op_ret == 0))
+ afr_attempt_local_discovery(this, child_index);
+
+ if (xdata) {
+ ret = dict_get_int8(xdata, "link-count", &need_heal);
+ local->replies[child_index].need_heal = need_heal;
+ } else {
+ local->replies[child_index].need_heal = need_heal;
+ }
+
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ afr_set_need_heal(this, local);
+ afr_lookup_metadata_heal_check(frame, this);
+ }
- return 0;
+ return 0;
}
-
int
-afr_discover_do (call_frame_t *frame, xlator_t *this, int err)
-{
- int ret = 0;
- int i = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (err) {
- local->op_errno = err;
- goto out;
- }
-
- call_count = local->call_count = AFR_COUNT (local->child_up,
- priv->child_count);
-
- ret = afr_lookup_xattr_req_prepare (local, this, local->xattr_req,
- &local->loc);
- if (ret) {
- local->op_errno = -ret;
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_discover_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->lookup,
- &local->loc, local->xattr_req);
- if (!--call_count)
- break;
- }
+afr_discover_do(call_frame_t *frame, xlator_t *this, int err)
+{
+ int ret = 0;
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (err) {
+ local->op_errno = err;
+ goto out;
+ }
+
+ call_count = local->call_count = AFR_COUNT(local->child_up,
+ priv->child_count);
+
+ ret = afr_lookup_xattr_req_prepare(local, this, local->xattr_req,
+ &local->loc);
+ if (ret) {
+ local->op_errno = -ret;
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(
+ frame, afr_discover_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->lookup, &local->loc, local->xattr_req);
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
- return 0;
+ AFR_STACK_UNWIND(lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
+ return 0;
}
-
int
-afr_discover (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
+afr_discover(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
{
- int op_errno = ENOMEM;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int event = 0;
+ int op_errno = ENOMEM;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int event = 0;
- priv = this->private;
+ priv = this->private;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- if (!local->call_count) {
- op_errno = ENOTCONN;
- goto out;
- }
+ if (!local->call_count) {
+ op_errno = ENOTCONN;
+ goto out;
+ }
- if (__is_root_gfid (loc->inode->gfid)) {
- if (!this->itable)
- this->itable = loc->inode->table;
- if (!priv->root_inode)
- priv->root_inode = inode_ref (loc->inode);
-
- if (priv->choose_local && !priv->did_discovery) {
- /* Logic to detect which subvolumes of AFR are
- local, in order to prefer them for reads
- */
- local->do_discovery = _gf_true;
- priv->did_discovery = _gf_true;
- }
- }
+ if (__is_root_gfid(loc->inode->gfid)) {
+ if (!this->itable)
+ this->itable = loc->inode->table;
+ if (!priv->root_inode)
+ priv->root_inode = inode_ref(loc->inode);
- local->op = GF_FOP_LOOKUP;
+ if (priv->choose_local && !priv->did_discovery) {
+ /* Logic to detect which subvolumes of AFR are
+ local, in order to prefer them for reads
+ */
+ local->do_discovery = _gf_true;
+ priv->did_discovery = _gf_true;
+ }
+ }
- loc_copy (&local->loc, loc);
+ local->op = GF_FOP_LOOKUP;
- local->inode = inode_ref (loc->inode);
+ loc_copy(&local->loc, loc);
- if (xattr_req)
- /* If xattr_req was null, afr_lookup_xattr_req_prepare() will
- allocate one for us */
- local->xattr_req = dict_ref (xattr_req);
+ local->inode = inode_ref(loc->inode);
- if (gf_uuid_is_null (loc->inode->gfid)) {
- afr_discover_do (frame, this, 0);
- return 0;
- }
+ if (xattr_req)
+ /* If xattr_req was null, afr_lookup_xattr_req_prepare() will
+ allocate one for us */
+ local->xattr_req = dict_ref(xattr_req);
- afr_read_subvol_get (loc->inode, this, NULL, NULL, &event,
- AFR_DATA_TRANSACTION, NULL);
+ if (gf_uuid_is_null(loc->inode->gfid)) {
+ afr_discover_do(frame, this, 0);
+ return 0;
+ }
- if (afr_is_inode_refresh_reqd (loc->inode, this, event,
- local->event_generation))
- afr_inode_refresh (frame, this, loc->inode, NULL,
- afr_discover_do);
- else
- afr_discover_do (frame, this, 0);
+ afr_read_subvol_get(loc->inode, this, NULL, NULL, &event,
+ AFR_DATA_TRANSACTION, NULL);
- return 0;
+ if (afr_is_inode_refresh_reqd(loc->inode, this, event,
+ local->event_generation))
+ afr_inode_refresh(frame, this, loc->inode, NULL, afr_discover_do);
+ else
+ afr_discover_do(frame, this, 0);
+
+ return 0;
out:
- AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ return 0;
}
-
int
-afr_lookup_do (call_frame_t *frame, xlator_t *this, int err)
-{
- int ret = 0;
- int i = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (err < 0) {
- local->op_errno = err;
- ret = -1;
- goto out;
- }
-
- call_count = local->call_count = AFR_COUNT (local->child_up,
- priv->child_count);
-
- ret = afr_lookup_xattr_req_prepare (local, this, local->xattr_req,
- &local->loc);
- if (ret) {
- local->op_errno = -ret;
- ret = -1;
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_lookup_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->lookup,
- &local->loc, local->xattr_req);
- if (!--call_count)
- break;
- }
+afr_lookup_do(call_frame_t *frame, xlator_t *this, int err)
+{
+ int ret = 0;
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (err < 0) {
+ local->op_errno = err;
+ ret = -1;
+ goto out;
+ }
+
+ call_count = local->call_count = AFR_COUNT(local->child_up,
+ priv->child_count);
+
+ ret = afr_lookup_xattr_req_prepare(local, this, local->xattr_req,
+ &local->loc);
+ if (ret) {
+ local->op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(
+ frame, afr_lookup_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->lookup, &local->loc, local->xattr_req);
+ if (!--call_count)
+ break;
}
- return 0;
+ }
+ return 0;
out:
- AFR_STACK_UNWIND (lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
- return 0;
+ AFR_STACK_UNWIND(lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
+ return 0;
}
/*
@@ -3405,1619 +3335,1568 @@ out:
*/
int
-afr_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
-{
- afr_local_t *local = NULL;
- int32_t op_errno = 0;
- int event = 0;
- int ret = 0;
-
- if (loc_is_nameless (loc)) {
- if (xattr_req)
- dict_del (xattr_req, "gfid-req");
- afr_discover (frame, this, loc, xattr_req);
- return 0;
- }
-
- if (__is_root_gfid (loc->parent->gfid)) {
- if (!strcmp (loc->name, GF_REPLICATE_TRASH_DIR)) {
- op_errno = EPERM;
- goto out;
- }
- }
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- if (!local->call_count) {
- op_errno = ENOTCONN;
- goto out;
+afr_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
+{
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
+ int event = 0;
+ int ret = 0;
+
+ if (loc_is_nameless(loc)) {
+ if (xattr_req)
+ dict_del(xattr_req, "gfid-req");
+ afr_discover(frame, this, loc, xattr_req);
+ return 0;
+ }
+
+ if (__is_root_gfid(loc->parent->gfid)) {
+ if (!strcmp(loc->name, GF_REPLICATE_TRASH_DIR)) {
+ op_errno = EPERM;
+ goto out;
}
+ }
- local->op = GF_FOP_LOOKUP;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- loc_copy (&local->loc, loc);
+ if (!local->call_count) {
+ op_errno = ENOTCONN;
+ goto out;
+ }
- local->inode = inode_ref (loc->inode);
+ local->op = GF_FOP_LOOKUP;
- if (xattr_req) {
- /* If xattr_req was null, afr_lookup_xattr_req_prepare() will
- allocate one for us */
- local->xattr_req = dict_copy_with_ref (xattr_req, NULL);
- if (!local->xattr_req) {
- op_errno = ENOMEM;
- goto out;
- }
- ret = dict_get_gfuuid (local->xattr_req, "gfid-req",
- &local->cont.lookup.gfid_req);
- if (ret == 0) {
- dict_del (local->xattr_req, "gfid-req");
- }
+ loc_copy(&local->loc, loc);
+
+ local->inode = inode_ref(loc->inode);
+
+ if (xattr_req) {
+ /* If xattr_req was null, afr_lookup_xattr_req_prepare() will
+ allocate one for us */
+ local->xattr_req = dict_copy_with_ref(xattr_req, NULL);
+ if (!local->xattr_req) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ ret = dict_get_gfuuid(local->xattr_req, "gfid-req",
+ &local->cont.lookup.gfid_req);
+ if (ret == 0) {
+ dict_del(local->xattr_req, "gfid-req");
}
+ }
- afr_read_subvol_get (loc->parent, this, NULL, NULL, &event,
- AFR_DATA_TRANSACTION, NULL);
+ afr_read_subvol_get(loc->parent, this, NULL, NULL, &event,
+ AFR_DATA_TRANSACTION, NULL);
- if (afr_is_inode_refresh_reqd (loc->inode, this, event,
- local->event_generation))
- afr_inode_refresh (frame, this, loc->parent, NULL,
- afr_lookup_do);
- else
- afr_lookup_do (frame, this, 0);
+ if (afr_is_inode_refresh_reqd(loc->inode, this, event,
+ local->event_generation))
+ afr_inode_refresh(frame, this, loc->parent, NULL, afr_lookup_do);
+ else
+ afr_lookup_do(frame, this, 0);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ AFR_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
void
-_afr_cleanup_fd_ctx (afr_fd_ctx_t *fd_ctx)
+_afr_cleanup_fd_ctx(afr_fd_ctx_t *fd_ctx)
{
- GF_FREE (fd_ctx->opened_on);
- GF_FREE (fd_ctx);
- return;
+ GF_FREE(fd_ctx->opened_on);
+ GF_FREE(fd_ctx);
+ return;
}
int
-afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
+afr_cleanup_fd_ctx(xlator_t *this, fd_t *fd)
{
- uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- int ret = 0;
+ uint64_t ctx = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int ret = 0;
- ret = fd_ctx_get (fd, this, &ctx);
- if (ret < 0)
- goto out;
+ ret = fd_ctx_get(fd, this, &ctx);
+ if (ret < 0)
+ goto out;
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ fd_ctx = (afr_fd_ctx_t *)(long)ctx;
- if (fd_ctx) {
- _afr_cleanup_fd_ctx (fd_ctx);
- }
+ if (fd_ctx) {
+ _afr_cleanup_fd_ctx(fd_ctx);
+ }
out:
- return 0;
+ return 0;
}
int
-afr_release (xlator_t *this, fd_t *fd)
+afr_release(xlator_t *this, fd_t *fd)
{
- afr_cleanup_fd_ctx (this, fd);
+ afr_cleanup_fd_ctx(this, fd);
- return 0;
+ return 0;
}
afr_fd_ctx_t *
-__afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+__afr_fd_ctx_get(fd_t *fd, xlator_t *this)
{
- uint64_t ctx = 0;
- int ret = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
+ uint64_t ctx = 0;
+ int ret = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
- ret = __fd_ctx_get (fd, this, &ctx);
+ ret = __fd_ctx_get(fd, this, &ctx);
- if (ret < 0) {
- ret = __afr_fd_ctx_set (this, fd);
- if (ret < 0)
- goto out;
+ if (ret < 0) {
+ ret = __afr_fd_ctx_set(this, fd);
+ if (ret < 0)
+ goto out;
- ret = __fd_ctx_get (fd, this, &ctx);
- if (ret < 0)
- goto out;
- }
+ ret = __fd_ctx_get(fd, this, &ctx);
+ if (ret < 0)
+ goto out;
+ }
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ fd_ctx = (afr_fd_ctx_t *)(long)ctx;
out:
- return fd_ctx;
+ return fd_ctx;
}
-
afr_fd_ctx_t *
-afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+afr_fd_ctx_get(fd_t *fd, xlator_t *this)
{
- afr_fd_ctx_t *fd_ctx = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
- LOCK(&fd->lock);
- {
- fd_ctx = __afr_fd_ctx_get (fd, this);
- }
- UNLOCK(&fd->lock);
+ LOCK(&fd->lock);
+ {
+ fd_ctx = __afr_fd_ctx_get(fd, this);
+ }
+ UNLOCK(&fd->lock);
- return fd_ctx;
+ return fd_ctx;
}
-
int
-__afr_fd_ctx_set (xlator_t *this, fd_t *fd)
+__afr_fd_ctx_set(xlator_t *this, fd_t *fd)
{
- afr_private_t * priv = NULL;
- int ret = -1;
- uint64_t ctx = 0;
- afr_fd_ctx_t * fd_ctx = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ uint64_t ctx = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int i = 0;
- VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (fd, out);
+ VALIDATE_OR_GOTO(this->private, out);
+ VALIDATE_OR_GOTO(fd, out);
- priv = this->private;
+ priv = this->private;
- ret = __fd_ctx_get (fd, this, &ctx);
+ ret = __fd_ctx_get(fd, this, &ctx);
- if (ret == 0)
- goto out;
+ if (ret == 0)
+ goto out;
- fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
- gf_afr_mt_afr_fd_ctx_t);
- if (!fd_ctx) {
- ret = -ENOMEM;
- goto out;
- }
+ fd_ctx = GF_CALLOC(1, sizeof(afr_fd_ctx_t), gf_afr_mt_afr_fd_ctx_t);
+ if (!fd_ctx) {
+ ret = -ENOMEM;
+ goto out;
+ }
- fd_ctx->opened_on = GF_CALLOC (sizeof (*fd_ctx->opened_on),
- priv->child_count,
- gf_afr_mt_int32_t);
- if (!fd_ctx->opened_on) {
- ret = -ENOMEM;
- goto out;
- }
+ fd_ctx->opened_on = GF_CALLOC(sizeof(*fd_ctx->opened_on), priv->child_count,
+ gf_afr_mt_int32_t);
+ if (!fd_ctx->opened_on) {
+ ret = -ENOMEM;
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (fd_is_anonymous (fd))
- fd_ctx->opened_on[i] = AFR_FD_OPENED;
- else
- fd_ctx->opened_on[i] = AFR_FD_NOT_OPENED;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (fd_is_anonymous(fd))
+ fd_ctx->opened_on[i] = AFR_FD_OPENED;
+ else
+ fd_ctx->opened_on[i] = AFR_FD_NOT_OPENED;
+ }
- fd_ctx->readdir_subvol = -1;
+ fd_ctx->readdir_subvol = -1;
- ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
- if (ret)
- gf_msg_debug (this->name, 0,
- "failed to set fd ctx (%p)", fd);
+ ret = __fd_ctx_set(fd, this, (uint64_t)(long)fd_ctx);
+ if (ret)
+ gf_msg_debug(this->name, 0, "failed to set fd ctx (%p)", fd);
out:
- if (ret && fd_ctx)
- _afr_cleanup_fd_ctx (fd_ctx);
- return ret;
+ if (ret && fd_ctx)
+ _afr_cleanup_fd_ctx(fd_ctx);
+ return ret;
}
-
/* {{{ flush */
int
-afr_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
+ afr_local_t *local = NULL;
+ int call_count = -1;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- if (op_ret != -1) {
- local->op_ret = op_ret;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- } else {
- local->op_errno = op_errno;
- }
+ LOCK(&frame->lock);
+ {
+ if (op_ret != -1) {
+ local->op_ret = op_ret;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ } else {
+ local->op_errno = op_errno;
}
- UNLOCK (&frame->lock);
+ }
+ UNLOCK(&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- AFR_STACK_UNWIND (flush, frame, local->op_ret,
- local->op_errno, local->xdata_rsp);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(flush, frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
- return 0;
+ return 0;
}
static int
-afr_flush_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
-{
- int i = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = -1;
-
- priv = this->private;
- local = frame->local;
- call_count = local->call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_flush_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->flush,
- local->fd, xdata);
- if (!--call_count)
- break;
-
- }
+afr_flush_wrapper(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+
+ priv = this->private;
+ local = frame->local;
+ call_count = local->call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, afr_flush_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->flush,
+ local->fd, xdata);
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
}
-afr_local_t*
-afr_wakeup_same_fd_delayed_op (xlator_t *this, afr_lock_t *lock, fd_t *fd)
+afr_local_t *
+afr_wakeup_same_fd_delayed_op(xlator_t *this, afr_lock_t *lock, fd_t *fd)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- if (lock->delay_timer) {
- local = list_entry(lock->post_op.next, afr_local_t,
- transaction.owner_list);
- if (fd == local->fd) {
- if (gf_timer_call_cancel (this->ctx,
- lock->delay_timer)) {
- local = NULL;
- } else {
- lock->delay_timer = NULL;
- }
- } else {
- local = NULL;
- }
+ if (lock->delay_timer) {
+ local = list_entry(lock->post_op.next, afr_local_t,
+ transaction.owner_list);
+ if (fd == local->fd) {
+ if (gf_timer_call_cancel(this->ctx, lock->delay_timer)) {
+ local = NULL;
+ } else {
+ lock->delay_timer = NULL;
+ }
+ } else {
+ local = NULL;
}
+ }
- return local;
+ return local;
}
void
-afr_delayed_changelog_wake_resume (xlator_t *this, inode_t *inode,
- call_stub_t *stub)
-{
- afr_inode_ctx_t *ctx = NULL;
- afr_lock_t *lock = NULL;
- afr_local_t *metadata_local = NULL;
- afr_local_t *data_local = NULL;
- LOCK (&inode->lock);
- {
- (void)__afr_inode_ctx_get (this, inode, &ctx);
- lock = &ctx->lock[AFR_DATA_TRANSACTION];
- data_local = afr_wakeup_same_fd_delayed_op (this, lock,
- stub->args.fd);
- lock = &ctx->lock[AFR_METADATA_TRANSACTION];
- metadata_local = afr_wakeup_same_fd_delayed_op (this, lock,
- stub->args.fd);
- }
- UNLOCK (&inode->lock);
-
- if (data_local) {
- data_local->transaction.resume_stub = stub;
- } else if (metadata_local) {
- metadata_local->transaction.resume_stub = stub;
- } else {
- call_resume (stub);
- }
- if (data_local) {
- afr_delayed_changelog_wake_up_cbk (data_local);
- }
- if (metadata_local) {
- afr_delayed_changelog_wake_up_cbk (metadata_local);
- }
+afr_delayed_changelog_wake_resume(xlator_t *this, inode_t *inode,
+ call_stub_t *stub)
+{
+ afr_inode_ctx_t *ctx = NULL;
+ afr_lock_t *lock = NULL;
+ afr_local_t *metadata_local = NULL;
+ afr_local_t *data_local = NULL;
+ LOCK(&inode->lock);
+ {
+ (void)__afr_inode_ctx_get(this, inode, &ctx);
+ lock = &ctx->lock[AFR_DATA_TRANSACTION];
+ data_local = afr_wakeup_same_fd_delayed_op(this, lock, stub->args.fd);
+ lock = &ctx->lock[AFR_METADATA_TRANSACTION];
+ metadata_local = afr_wakeup_same_fd_delayed_op(this, lock,
+ stub->args.fd);
+ }
+ UNLOCK(&inode->lock);
+
+ if (data_local) {
+ data_local->transaction.resume_stub = stub;
+ } else if (metadata_local) {
+ metadata_local->transaction.resume_stub = stub;
+ } else {
+ call_resume(stub);
+ }
+ if (data_local) {
+ afr_delayed_changelog_wake_up_cbk(data_local);
+ }
+ if (metadata_local) {
+ afr_delayed_changelog_wake_up_cbk(metadata_local);
+ }
}
int
-afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+afr_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_stub_t *stub = NULL;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = ENOMEM;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_FLUSH;
- if (!afr_is_consistent_io_possible (local, this->private, &op_errno))
- goto out;
+ local->op = GF_FOP_FLUSH;
+ if (!afr_is_consistent_io_possible(local, this->private, &op_errno))
+ goto out;
- local->fd = fd_ref(fd);
+ local->fd = fd_ref(fd);
- stub = fop_flush_stub (frame, afr_flush_wrapper, fd, xdata);
- if (!stub)
- goto out;
+ stub = fop_flush_stub(frame, afr_flush_wrapper, fd, xdata);
+ if (!stub)
+ goto out;
- afr_delayed_changelog_wake_resume (this, fd->inode, stub);
+ afr_delayed_changelog_wake_resume(this, fd->inode, stub);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
- return 0;
+ AFR_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
+ return 0;
}
int
-afr_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_fsyncdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
+ afr_local_t *local = NULL;
+ int call_count = -1;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- if (op_ret == 0) {
- local->op_ret = 0;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- } else {
- local->op_errno = op_errno;
- }
+ LOCK(&frame->lock);
+ {
+ if (op_ret == 0) {
+ local->op_ret = 0;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ } else {
+ local->op_errno = op_errno;
}
- UNLOCK (&frame->lock);
+ }
+ UNLOCK(&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
- local->op_errno, local->xdata_rsp);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(fsyncdir, frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
- return 0;
+ return 0;
}
-
int
-afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
- dict_t *xdata)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int32_t call_count = 0;
- int32_t op_errno = ENOMEM;
-
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_FSYNCDIR;
- if (!afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
-
- call_count = local->call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND (frame, afr_fsyncdir_cbk,
- priv->children[i],
- priv->children[i]->fops->fsyncdir,
- fd, datasync, xdata);
- if (!--call_count)
- break;
- }
+afr_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int32_t call_count = 0;
+ int32_t op_errno = ENOMEM;
+
+ priv = this->private;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_FSYNCDIR;
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
+
+ call_count = local->call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND(frame, afr_fsyncdir_cbk, priv->children[i],
+ priv->children[i]->fops->fsyncdir, fd, datasync, xdata);
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND(fsyncdir, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
/* }}} */
static int
-afr_serialized_lock_wind (call_frame_t *frame, xlator_t *this);
+afr_serialized_lock_wind(call_frame_t *frame, xlator_t *this);
static gf_boolean_t
-afr_is_conflicting_lock_present (int32_t op_ret, int32_t op_errno)
+afr_is_conflicting_lock_present(int32_t op_ret, int32_t op_errno)
{
- if (op_ret == -1 && op_errno == EAGAIN)
- return _gf_true;
- return _gf_false;
+ if (op_ret == -1 && op_errno == EAGAIN)
+ return _gf_true;
+ return _gf_false;
}
static void
-afr_fop_lock_unwind (call_frame_t *frame, glusterfs_fop_t op, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+afr_fop_lock_unwind(call_frame_t *frame, glusterfs_fop_t op, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- switch (op) {
+ switch (op) {
case GF_FOP_INODELK:
- AFR_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
- break;
+ AFR_STACK_UNWIND(inodelk, frame, op_ret, op_errno, xdata);
+ break;
case GF_FOP_FINODELK:
- AFR_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
- break;
+ AFR_STACK_UNWIND(finodelk, frame, op_ret, op_errno, xdata);
+ break;
case GF_FOP_ENTRYLK:
- AFR_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
- break;
+ AFR_STACK_UNWIND(entrylk, frame, op_ret, op_errno, xdata);
+ break;
case GF_FOP_FENTRYLK:
- AFR_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, xdata);
- break;
+ AFR_STACK_UNWIND(fentrylk, frame, op_ret, op_errno, xdata);
+ break;
default:
- break;
- }
+ break;
+ }
}
static void
-afr_fop_lock_wind (call_frame_t *frame, xlator_t *this, int child_index,
- int32_t (*lock_cbk) (call_frame_t *, void *, xlator_t *,
- int32_t, int32_t, dict_t *))
+afr_fop_lock_wind(call_frame_t *frame, xlator_t *this, int child_index,
+ int32_t (*lock_cbk)(call_frame_t *, void *, xlator_t *,
+ int32_t, int32_t, dict_t *))
{
- afr_local_t *local = frame->local;
- afr_private_t *priv = this->private;
- int i = child_index;
+ afr_local_t *local = frame->local;
+ afr_private_t *priv = this->private;
+ int i = child_index;
- switch (local->op) {
+ switch (local->op) {
case GF_FOP_INODELK:
- STACK_WIND_COOKIE (frame, lock_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->inodelk,
- (const char *)local->cont.inodelk.volume,
- &local->loc, local->cont.inodelk.cmd,
- &local->cont.inodelk.flock,
- local->cont.inodelk.xdata);
- break;
+ STACK_WIND_COOKIE(
+ frame, lock_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->inodelk,
+ (const char *)local->cont.inodelk.volume, &local->loc,
+ local->cont.inodelk.cmd, &local->cont.inodelk.flock,
+ local->cont.inodelk.xdata);
+ break;
case GF_FOP_FINODELK:
- STACK_WIND_COOKIE (frame, lock_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->finodelk,
- (const char *)local->cont.inodelk.volume,
- local->fd, local->cont.inodelk.cmd,
- &local->cont.inodelk.flock,
- local->cont.inodelk.xdata);
- break;
+ STACK_WIND_COOKIE(
+ frame, lock_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->finodelk,
+ (const char *)local->cont.inodelk.volume, local->fd,
+ local->cont.inodelk.cmd, &local->cont.inodelk.flock,
+ local->cont.inodelk.xdata);
+ break;
case GF_FOP_ENTRYLK:
- STACK_WIND_COOKIE (frame, lock_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- local->cont.entrylk.volume, &local->loc,
- local->cont.entrylk.basename,
- local->cont.entrylk.cmd,
- local->cont.entrylk.type,
- local->cont.entrylk.xdata);
- break;
+ STACK_WIND_COOKIE(
+ frame, lock_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->entrylk, local->cont.entrylk.volume,
+ &local->loc, local->cont.entrylk.basename,
+ local->cont.entrylk.cmd, local->cont.entrylk.type,
+ local->cont.entrylk.xdata);
+ break;
case GF_FOP_FENTRYLK:
- STACK_WIND_COOKIE (frame, lock_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fentrylk,
- local->cont.entrylk.volume, local->fd,
- local->cont.entrylk.basename,
- local->cont.entrylk.cmd,
- local->cont.entrylk.type,
- local->cont.entrylk.xdata);
- break;
+ STACK_WIND_COOKIE(
+ frame, lock_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->fentrylk, local->cont.entrylk.volume,
+ local->fd, local->cont.entrylk.basename,
+ local->cont.entrylk.cmd, local->cont.entrylk.type,
+ local->cont.entrylk.xdata);
+ break;
default:
- break;
- }
+ break;
+ }
}
void
-afr_fop_lock_proceed (call_frame_t *frame)
+afr_fop_lock_proceed(call_frame_t *frame)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = frame->this->private;
+ local = frame->local;
+ priv = frame->this->private;
- if (local->fop_lock_state != AFR_FOP_LOCK_PARALLEL) {
- afr_fop_lock_unwind (frame, local->op, local->op_ret,
- local->op_errno, local->xdata_rsp);
- return;
- }
- /* At least one child is up */
- /*
- * Non-blocking locks also need to be serialized. Otherwise there is
- * a chance that both the mounts which issued same non-blocking inodelk
- * may endup not acquiring the lock on any-brick.
- * Ex: Mount1 and Mount2
- * request for full length lock on file f1. Mount1 afr may acquire the
- * partial lock on brick-1 and may not acquire the lock on brick-2
- * because Mount2 already got the lock on brick-2, vice versa. Since
- * both the mounts only got partial locks, afr treats them as failure in
- * gaining the locks and unwinds with EAGAIN errno.
- */
- local->op_ret = -1;
- local->op_errno = EUCLEAN;
- local->fop_lock_state = AFR_FOP_LOCK_SERIAL;
- afr_local_replies_wipe (local, priv);
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = NULL;
- switch (local->op) {
+ if (local->fop_lock_state != AFR_FOP_LOCK_PARALLEL) {
+ afr_fop_lock_unwind(frame, local->op, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return;
+ }
+ /* At least one child is up */
+ /*
+ * Non-blocking locks also need to be serialized. Otherwise there is
+ * a chance that both the mounts which issued same non-blocking inodelk
+ * may endup not acquiring the lock on any-brick.
+ * Ex: Mount1 and Mount2
+ * request for full length lock on file f1. Mount1 afr may acquire the
+ * partial lock on brick-1 and may not acquire the lock on brick-2
+ * because Mount2 already got the lock on brick-2, vice versa. Since
+ * both the mounts only got partial locks, afr treats them as failure in
+ * gaining the locks and unwinds with EAGAIN errno.
+ */
+ local->op_ret = -1;
+ local->op_errno = EUCLEAN;
+ local->fop_lock_state = AFR_FOP_LOCK_SERIAL;
+ afr_local_replies_wipe(local, priv);
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = NULL;
+ switch (local->op) {
case GF_FOP_INODELK:
case GF_FOP_FINODELK:
- local->cont.inodelk.cmd = local->cont.inodelk.in_cmd;
- local->cont.inodelk.flock = local->cont.inodelk.in_flock;
- if (local->cont.inodelk.xdata)
- dict_unref (local->cont.inodelk.xdata);
- local->cont.inodelk.xdata = NULL;
- if (local->xdata_req)
- local->cont.inodelk.xdata = dict_ref (local->xdata_req);
- break;
+ local->cont.inodelk.cmd = local->cont.inodelk.in_cmd;
+ local->cont.inodelk.flock = local->cont.inodelk.in_flock;
+ if (local->cont.inodelk.xdata)
+ dict_unref(local->cont.inodelk.xdata);
+ local->cont.inodelk.xdata = NULL;
+ if (local->xdata_req)
+ local->cont.inodelk.xdata = dict_ref(local->xdata_req);
+ break;
case GF_FOP_ENTRYLK:
case GF_FOP_FENTRYLK:
- local->cont.entrylk.cmd = local->cont.entrylk.in_cmd;
- if (local->cont.entrylk.xdata)
- dict_unref (local->cont.entrylk.xdata);
- local->cont.entrylk.xdata = NULL;
- if (local->xdata_req)
- local->cont.entrylk.xdata = dict_ref (local->xdata_req);
- break;
+ local->cont.entrylk.cmd = local->cont.entrylk.in_cmd;
+ if (local->cont.entrylk.xdata)
+ dict_unref(local->cont.entrylk.xdata);
+ local->cont.entrylk.xdata = NULL;
+ if (local->xdata_req)
+ local->cont.entrylk.xdata = dict_ref(local->xdata_req);
+ break;
default:
- break;
- }
- afr_serialized_lock_wind (frame, frame->this);
+ break;
+ }
+ afr_serialized_lock_wind(frame, frame->this);
}
static int32_t
-afr_unlock_partial_lock_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+afr_unlock_partial_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = -1;
- int child_index = (long)cookie;
- uuid_t gfid = {0};
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int child_index = (long)cookie;
+ uuid_t gfid = {0};
- local = frame->local;
- priv = this->private;
-
- if (op_ret < 0 && op_errno != ENOTCONN) {
- if (local->fd)
- gf_uuid_copy (gfid, local->fd->inode->gfid);
- else
- loc_gfid (&local->loc, gfid);
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- AFR_MSG_UNLOCK_FAIL,
- "%s: Failed to unlock %s on %s "
- "with lk_owner: %s", uuid_utoa (gfid),
- gf_fop_list[local->op],
- priv->children[child_index]->name,
- lkowner_utoa (&frame->root->lk_owner));
- }
+ local = frame->local;
+ priv = this->private;
+
+ if (op_ret < 0 && op_errno != ENOTCONN) {
+ if (local->fd)
+ gf_uuid_copy(gfid, local->fd->inode->gfid);
+ else
+ loc_gfid(&local->loc, gfid);
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_UNLOCK_FAIL,
+ "%s: Failed to unlock %s on %s "
+ "with lk_owner: %s",
+ uuid_utoa(gfid), gf_fop_list[local->op],
+ priv->children[child_index]->name,
+ lkowner_utoa(&frame->root->lk_owner));
+ }
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_fop_lock_proceed (frame);
+ call_count = afr_frame_return(frame);
+ if (call_count == 0)
+ afr_fop_lock_proceed(frame);
- return 0;
+ return 0;
}
static int32_t
-afr_unlock_locks_and_proceed (call_frame_t *frame, xlator_t *this,
+afr_unlock_locks_and_proceed(call_frame_t *frame, xlator_t *this,
int call_count)
{
- int i = 0;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- if (call_count == 0) {
- afr_fop_lock_proceed (frame);
- goto out;
- }
+ if (call_count == 0) {
+ afr_fop_lock_proceed(frame);
+ goto out;
+ }
- local = frame->local;
- priv = this->private;
- local->call_count = call_count;
- switch (local->op) {
+ local = frame->local;
+ priv = this->private;
+ local->call_count = call_count;
+ switch (local->op) {
case GF_FOP_INODELK:
case GF_FOP_FINODELK:
- local->cont.inodelk.flock.l_type = F_UNLCK;
- local->cont.inodelk.cmd = F_SETLK;
- if (local->cont.inodelk.xdata)
- dict_unref (local->cont.inodelk.xdata);
- local->cont.inodelk.xdata = NULL;
- break;
+ local->cont.inodelk.flock.l_type = F_UNLCK;
+ local->cont.inodelk.cmd = F_SETLK;
+ if (local->cont.inodelk.xdata)
+ dict_unref(local->cont.inodelk.xdata);
+ local->cont.inodelk.xdata = NULL;
+ break;
case GF_FOP_ENTRYLK:
case GF_FOP_FENTRYLK:
- local->cont.entrylk.cmd = ENTRYLK_UNLOCK;
- if (local->cont.entrylk.xdata)
- dict_unref (local->cont.entrylk.xdata);
- local->cont.entrylk.xdata = NULL;
- break;
+ local->cont.entrylk.cmd = ENTRYLK_UNLOCK;
+ if (local->cont.entrylk.xdata)
+ dict_unref(local->cont.entrylk.xdata);
+ local->cont.entrylk.xdata = NULL;
+ break;
default:
- break;
- }
+ break;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
- if (local->replies[i].op_ret == -1)
- continue;
+ if (local->replies[i].op_ret == -1)
+ continue;
- afr_fop_lock_wind (frame, this, i, afr_unlock_partial_lock_cbk);
+ afr_fop_lock_wind(frame, this, i, afr_unlock_partial_lock_cbk);
- if (!--call_count)
- break;
- }
+ if (!--call_count)
+ break;
+ }
out:
- return 0;
+ return 0;
}
int32_t
-afr_fop_lock_done (call_frame_t *frame, xlator_t *this)
+afr_fop_lock_done(call_frame_t *frame, xlator_t *this)
{
- int i = 0;
- int lock_count = 0;
- unsigned char *success = NULL;
+ int i = 0;
+ int lock_count = 0;
+ unsigned char *success = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
- success = alloca0(priv->child_count);
+ local = frame->local;
+ priv = this->private;
+ success = alloca0(priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
- if (local->replies[i].op_ret == 0) {
- lock_count++;
- success[i] = 1;
- }
+ if (local->replies[i].op_ret == 0) {
+ lock_count++;
+ success[i] = 1;
+ }
- if (local->op_ret == -1 && local->op_errno == EAGAIN)
- continue;
+ if (local->op_ret == -1 && local->op_errno == EAGAIN)
+ continue;
- if ((local->replies[i].op_ret == -1) &&
- (local->replies[i].op_errno == EAGAIN)) {
- local->op_ret = -1;
- local->op_errno = EAGAIN;
- continue;
- }
+ if ((local->replies[i].op_ret == -1) &&
+ (local->replies[i].op_errno == EAGAIN)) {
+ local->op_ret = -1;
+ local->op_errno = EAGAIN;
+ continue;
+ }
- if (local->replies[i].op_ret == 0)
- local->op_ret = 0;
+ if (local->replies[i].op_ret == 0)
+ local->op_ret = 0;
- local->op_errno = local->replies[i].op_errno;
- }
+ local->op_errno = local->replies[i].op_errno;
+ }
- if (afr_fop_lock_is_unlock (frame))
- goto unwind;
-
- if (afr_is_conflicting_lock_present (local->op_ret, local->op_errno)) {
- afr_unlock_locks_and_proceed (frame, this, lock_count);
- } else if (priv->quorum_count && !afr_has_quorum (success, this)) {
- local->fop_lock_state = AFR_FOP_LOCK_QUORUM_FAILED;
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
- if (local->op_errno == 0)
- local->op_errno = afr_quorum_errno (priv);
- afr_unlock_locks_and_proceed (frame, this, lock_count);
- } else {
- goto unwind;
- }
+ if (afr_fop_lock_is_unlock(frame))
+ goto unwind;
- return 0;
+ if (afr_is_conflicting_lock_present(local->op_ret, local->op_errno)) {
+ afr_unlock_locks_and_proceed(frame, this, lock_count);
+ } else if (priv->quorum_count && !afr_has_quorum(success, this)) {
+ local->fop_lock_state = AFR_FOP_LOCK_QUORUM_FAILED;
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ if (local->op_errno == 0)
+ local->op_errno = afr_quorum_errno(priv);
+ afr_unlock_locks_and_proceed(frame, this, lock_count);
+ } else {
+ goto unwind;
+ }
+
+ return 0;
unwind:
- afr_fop_lock_unwind (frame, local->op, local->op_ret,
- local->op_errno, local->xdata_rsp);
- return 0;
+ afr_fop_lock_unwind(frame, local->op, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
}
static int
-afr_common_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_common_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int child_index = (long)cookie;
+ afr_local_t *local = NULL;
+ int child_index = (long)cookie;
- local = frame->local;
+ local = frame->local;
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (op_ret == 0 && xdata) {
- local->replies[child_index].xdata = dict_ref (xdata);
- LOCK (&frame->lock);
- {
- if (!local->xdata_rsp)
- local->xdata_rsp = dict_ref (xdata);
- }
- UNLOCK (&frame->lock);
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (op_ret == 0 && xdata) {
+ local->replies[child_index].xdata = dict_ref(xdata);
+ LOCK(&frame->lock);
+ {
+ if (!local->xdata_rsp)
+ local->xdata_rsp = dict_ref(xdata);
}
- return 0;
+ UNLOCK(&frame->lock);
+ }
+ return 0;
}
static int32_t
-afr_serialized_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_serialized_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int child_index = (long)cookie;
- int next_child = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = (long)cookie;
+ int next_child = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- afr_common_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
- for (next_child = child_index + 1; next_child < priv->child_count;
- next_child++) {
- if (local->child_up[next_child])
- break;
- }
+ for (next_child = child_index + 1; next_child < priv->child_count;
+ next_child++) {
+ if (local->child_up[next_child])
+ break;
+ }
- if (afr_is_conflicting_lock_present (op_ret, op_errno) ||
- (next_child == priv->child_count)) {
- afr_fop_lock_done (frame, this);
- } else {
- afr_fop_lock_wind (frame, this, next_child,
- afr_serialized_lock_cbk);
- }
+ if (afr_is_conflicting_lock_present(op_ret, op_errno) ||
+ (next_child == priv->child_count)) {
+ afr_fop_lock_done(frame, this);
+ } else {
+ afr_fop_lock_wind(frame, this, next_child, afr_serialized_lock_cbk);
+ }
- return 0;
+ return 0;
}
static int
-afr_serialized_lock_wind (call_frame_t *frame, xlator_t *this)
+afr_serialized_lock_wind(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- afr_fop_lock_wind (frame, this, i,
- afr_serialized_lock_cbk);
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ afr_fop_lock_wind(frame, this, i, afr_serialized_lock_cbk);
+ break;
}
- return 0;
+ }
+ return 0;
}
static int32_t
-afr_parallel_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_parallel_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- int call_count = 0;
+ int call_count = 0;
- afr_common_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_fop_lock_done (frame, this);
+ call_count = afr_frame_return(frame);
+ if (call_count == 0)
+ afr_fop_lock_done(frame, this);
- return 0;
+ return 0;
}
static int
-afr_parallel_lock_wind (call_frame_t *frame, xlator_t *this)
+afr_parallel_lock_wind(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int call_count = 0;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int call_count = 0;
+ int i = 0;
- priv = this->private;
- local = frame->local;
- call_count = local->call_count;
+ priv = this->private;
+ local = frame->local;
+ call_count = local->call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
- continue;
- afr_fop_lock_wind (frame, this, i, afr_parallel_lock_cbk);
- if (!--call_count)
- break;
- }
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->child_up[i])
+ continue;
+ afr_fop_lock_wind(frame, this, i, afr_parallel_lock_cbk);
+ if (!--call_count)
+ break;
+ }
+ return 0;
}
static int
-afr_fop_handle_lock (call_frame_t *frame, xlator_t *this)
+afr_fop_handle_lock(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = frame->local;
- int op_errno = 0;
+ afr_local_t *local = frame->local;
+ int op_errno = 0;
- if (!afr_fop_lock_is_unlock (frame)) {
- if (!afr_is_consistent_io_possible (local, this->private,
- &op_errno))
- goto out;
+ if (!afr_fop_lock_is_unlock(frame)) {
+ if (!afr_is_consistent_io_possible(local, this->private, &op_errno))
+ goto out;
- switch (local->op) {
- case GF_FOP_INODELK:
- case GF_FOP_FINODELK:
- local->cont.inodelk.cmd = F_SETLK;
- break;
- case GF_FOP_ENTRYLK:
- case GF_FOP_FENTRYLK:
- local->cont.entrylk.cmd = ENTRYLK_LOCK_NB;
- break;
- default:
- break;
- }
+ switch (local->op) {
+ case GF_FOP_INODELK:
+ case GF_FOP_FINODELK:
+ local->cont.inodelk.cmd = F_SETLK;
+ break;
+ case GF_FOP_ENTRYLK:
+ case GF_FOP_FENTRYLK:
+ local->cont.entrylk.cmd = ENTRYLK_LOCK_NB;
+ break;
+ default:
+ break;
}
+ }
- if (local->xdata_req) {
- switch (local->op) {
- case GF_FOP_INODELK:
- case GF_FOP_FINODELK:
- local->cont.inodelk.xdata = dict_ref (local->xdata_req);
- break;
- case GF_FOP_ENTRYLK:
- case GF_FOP_FENTRYLK:
- local->cont.entrylk.xdata = dict_ref (local->xdata_req);
- break;
- default:
- break;
- }
+ if (local->xdata_req) {
+ switch (local->op) {
+ case GF_FOP_INODELK:
+ case GF_FOP_FINODELK:
+ local->cont.inodelk.xdata = dict_ref(local->xdata_req);
+ break;
+ case GF_FOP_ENTRYLK:
+ case GF_FOP_FENTRYLK:
+ local->cont.entrylk.xdata = dict_ref(local->xdata_req);
+ break;
+ default:
+ break;
}
+ }
- local->fop_lock_state = AFR_FOP_LOCK_PARALLEL;
- afr_parallel_lock_wind (frame, this);
+ local->fop_lock_state = AFR_FOP_LOCK_PARALLEL;
+ afr_parallel_lock_wind(frame, this);
out:
- return -op_errno;
+ return -op_errno;
}
static int32_t
-afr_handle_inodelk (call_frame_t *frame, glusterfs_fop_t fop,
- const char *volume, loc_t *loc, fd_t *fd, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata)
+afr_handle_inodelk(call_frame_t *frame, glusterfs_fop_t fop, const char *volume,
+ loc_t *loc, fd_t *fd, int32_t cmd, struct gf_flock *flock,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t op_errno = ENOMEM;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = fop;
- if (loc)
- loc_copy (&local->loc, loc);
- if (fd)
- local->fd = fd_ref (fd);
-
- local->cont.inodelk.volume = gf_strdup (volume);
- if (!local->cont.inodelk.volume) {
- op_errno = ENOMEM;
- goto out;
- }
-
- local->cont.inodelk.in_cmd = cmd;
- local->cont.inodelk.cmd = cmd;
- local->cont.inodelk.in_flock = *flock;
- local->cont.inodelk.flock = *flock;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
-
- op_errno = -afr_fop_handle_lock (frame, frame->this);
- if (op_errno)
- goto out;
- return 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = ENOMEM;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = fop;
+ if (loc)
+ loc_copy(&local->loc, loc);
+ if (fd)
+ local->fd = fd_ref(fd);
+
+ local->cont.inodelk.volume = gf_strdup(volume);
+ if (!local->cont.inodelk.volume) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->cont.inodelk.in_cmd = cmd;
+ local->cont.inodelk.cmd = cmd;
+ local->cont.inodelk.in_flock = *flock;
+ local->cont.inodelk.flock = *flock;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+
+ op_errno = -afr_fop_handle_lock(frame, frame->this);
+ if (op_errno)
+ goto out;
+ return 0;
out:
- afr_fop_lock_unwind (frame, fop, -1, op_errno, NULL);
+ afr_fop_lock_unwind(frame, fop, -1, op_errno, NULL);
- return 0;
+ return 0;
}
int32_t
-afr_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata)
+afr_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- afr_handle_inodelk (frame, GF_FOP_INODELK, volume, loc, NULL, cmd,
- flock, xdata);
- return 0;
+ afr_handle_inodelk(frame, GF_FOP_INODELK, volume, loc, NULL, cmd, flock,
+ xdata);
+ return 0;
}
int32_t
-afr_finodelk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
- int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+afr_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- afr_handle_inodelk (frame, GF_FOP_FINODELK, volume, NULL, fd, cmd,
- flock, xdata);
- return 0;
+ afr_handle_inodelk(frame, GF_FOP_FINODELK, volume, NULL, fd, cmd, flock,
+ xdata);
+ return 0;
}
static int
-afr_handle_entrylk (call_frame_t *frame, glusterfs_fop_t fop,
- const char *volume, loc_t *loc, fd_t *fd,
- const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata)
-{
- afr_local_t *local = NULL;
- int32_t op_errno = ENOMEM;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = fop;
- if (loc)
- loc_copy (&local->loc, loc);
- if (fd)
- local->fd = fd_ref (fd);
- local->cont.entrylk.cmd = cmd;
- local->cont.entrylk.in_cmd = cmd;
- local->cont.entrylk.type = type;
- local->cont.entrylk.volume = gf_strdup (volume);
- local->cont.entrylk.basename = gf_strdup (basename);
- if (!local->cont.entrylk.volume || !local->cont.entrylk.basename) {
- op_errno = ENOMEM;
- goto out;
- }
- if (xdata)
- local->xdata_req = dict_ref (xdata);
- op_errno = -afr_fop_handle_lock (frame, frame->this);
- if (op_errno)
- goto out;
-
- return 0;
+afr_handle_entrylk(call_frame_t *frame, glusterfs_fop_t fop, const char *volume,
+ loc_t *loc, fd_t *fd, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int32_t op_errno = ENOMEM;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = fop;
+ if (loc)
+ loc_copy(&local->loc, loc);
+ if (fd)
+ local->fd = fd_ref(fd);
+ local->cont.entrylk.cmd = cmd;
+ local->cont.entrylk.in_cmd = cmd;
+ local->cont.entrylk.type = type;
+ local->cont.entrylk.volume = gf_strdup(volume);
+ local->cont.entrylk.basename = gf_strdup(basename);
+ if (!local->cont.entrylk.volume || !local->cont.entrylk.basename) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+ op_errno = -afr_fop_handle_lock(frame, frame->this);
+ if (op_errno)
+ goto out;
+
+ return 0;
out:
- afr_fop_lock_unwind (frame, fop, -1, op_errno, NULL);
- return 0;
+ afr_fop_lock_unwind(frame, fop, -1, op_errno, NULL);
+ return 0;
}
int
-afr_entrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata)
+afr_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- afr_handle_entrylk (frame, GF_FOP_ENTRYLK, volume, loc, NULL, basename,
- cmd, type, xdata);
- return 0;
+ afr_handle_entrylk(frame, GF_FOP_ENTRYLK, volume, loc, NULL, basename, cmd,
+ type, xdata);
+ return 0;
}
int
-afr_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
- const char *basename, entrylk_cmd cmd, entrylk_type type,
- dict_t *xdata)
+afr_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- afr_handle_entrylk (frame, GF_FOP_FENTRYLK, volume, NULL, fd, basename,
- cmd, type, xdata);
- return 0;
+ afr_handle_entrylk(frame, GF_FOP_FENTRYLK, volume, NULL, fd, basename, cmd,
+ type, xdata);
+ return 0;
}
-
int
-afr_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, struct statvfs *statvfs, dict_t *xdata)
+afr_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct statvfs *statvfs, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = 0;
- struct statvfs *buf = NULL;
+ afr_local_t *local = NULL;
+ int call_count = 0;
+ struct statvfs *buf = NULL;
- LOCK (&frame->lock);
- {
- local = frame->local;
-
- if (op_ret != 0) {
- local->op_errno = op_errno;
- goto unlock;
- }
-
- local->op_ret = op_ret;
-
- buf = &local->cont.statfs.buf;
- if (local->cont.statfs.buf_set) {
- if (statvfs->f_bavail < buf->f_bavail) {
- *buf = *statvfs;
- if (xdata) {
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = dict_ref (xdata);
- }
- }
- } else {
- *buf = *statvfs;
- local->cont.statfs.buf_set = 1;
- if (xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
+ LOCK(&frame->lock);
+ {
+ local = frame->local;
+
+ if (op_ret != 0) {
+ local->op_errno = op_errno;
+ goto unlock;
}
+
+ local->op_ret = op_ret;
+
+ buf = &local->cont.statfs.buf;
+ if (local->cont.statfs.buf_set) {
+ if (statvfs->f_bavail < buf->f_bavail) {
+ *buf = *statvfs;
+ if (xdata) {
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = dict_ref(xdata);
+ }
+ }
+ } else {
+ *buf = *statvfs;
+ local->cont.statfs.buf_set = 1;
+ if (xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ }
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- AFR_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
- &local->cont.statfs.buf, local->xdata_rsp);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(statfs, frame, local->op_ret, local->op_errno,
+ &local->cont.statfs.buf, local->xdata_rsp);
- return 0;
+ return 0;
}
-
int
-afr_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
-{
- afr_local_t * local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int call_count = 0;
- int32_t op_errno = ENOMEM;
-
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_STATFS;
- if (!afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
-
- if (priv->arbiter_count == 1 && local->child_up[ARBITER_BRICK_INDEX])
- local->call_count--;
- call_count = local->call_count;
- if (!call_count) {
- op_errno = ENOTCONN;
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- if (AFR_IS_ARBITER_BRICK(priv, i))
- continue;
- STACK_WIND (frame, afr_statfs_cbk,
- priv->children[i],
- priv->children[i]->fops->statfs,
- loc, xdata);
- if (!--call_count)
- break;
- }
+afr_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int call_count = 0;
+ int32_t op_errno = ENOMEM;
+
+ priv = this->private;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_STATFS;
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
+
+ if (priv->arbiter_count == 1 && local->child_up[ARBITER_BRICK_INDEX])
+ local->call_count--;
+ call_count = local->call_count;
+ if (!call_count) {
+ op_errno = ENOTCONN;
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ if (AFR_IS_ARBITER_BRICK(priv, i))
+ continue;
+ STACK_WIND(frame, afr_statfs_cbk, priv->children[i],
+ priv->children[i]->fops->statfs, loc, xdata);
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(statfs, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
- dict_t *xdata)
+afr_lk_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = this->private;
- int call_count = -1;
- int child_index = (long)cookie;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = this->private;
+ int call_count = -1;
+ int child_index = (long)cookie;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- AFR_MSG_UNLOCK_FAIL,
- "gfid=%s: unlock failed on subvolume %s "
- "with lock owner %s",
- uuid_utoa (local->fd->inode->gfid),
- priv->children[child_index]->name,
- lkowner_utoa (&frame->root->lk_owner));
- }
+ if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_UNLOCK_FAIL,
+ "gfid=%s: unlock failed on subvolume %s "
+ "with lock owner %s",
+ uuid_utoa(local->fd->inode->gfid),
+ priv->children[child_index]->name,
+ lkowner_utoa(&frame->root->lk_owner));
+ }
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- NULL, local->xdata_rsp);
+ call_count = afr_frame_return(frame);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(lk, frame, local->op_ret, local->op_errno, NULL,
+ local->xdata_rsp);
- return 0;
+ return 0;
}
-
int32_t
-afr_lk_unlock (call_frame_t *frame, xlator_t *this)
+afr_lk_unlock(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
- int i = 0;
- int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int call_count = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- call_count = afr_locked_nodes_count (local->cont.lk.locked_nodes,
- priv->child_count);
+ call_count = afr_locked_nodes_count(local->cont.lk.locked_nodes,
+ priv->child_count);
- if (call_count == 0) {
- AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- NULL, local->xdata_rsp);
- return 0;
- }
+ if (call_count == 0) {
+ AFR_STACK_UNWIND(lk, frame, local->op_ret, local->op_errno, NULL,
+ local->xdata_rsp);
+ return 0;
+ }
- local->call_count = call_count;
+ local->call_count = call_count;
- local->cont.lk.user_flock.l_type = F_UNLCK;
+ local->cont.lk.user_flock.l_type = F_UNLCK;
- for (i = 0; i < priv->child_count; i++) {
- if (local->cont.lk.locked_nodes[i]) {
- STACK_WIND_COOKIE (frame, afr_lk_unlock_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->lk,
- local->fd, F_SETLK,
- &local->cont.lk.user_flock, NULL);
-
- if (!--call_count)
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->cont.lk.locked_nodes[i]) {
+ STACK_WIND_COOKIE(frame, afr_lk_unlock_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->lk,
+ local->fd, F_SETLK, &local->cont.lk.user_flock,
+ NULL);
+
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
}
-
int32_t
-afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
- dict_t *xdata)
+afr_lk_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int child_index = -1;
-
-
- local = frame->local;
- priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = -1;
- child_index = (long) cookie;
+ local = frame->local;
+ priv = this->private;
- afr_common_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- if (op_ret < 0 && op_errno == EAGAIN) {
- local->op_ret = -1;
- local->op_errno = EAGAIN;
+ child_index = (long)cookie;
- afr_lk_unlock (frame, this);
- return 0;
- }
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ if (op_ret < 0 && op_errno == EAGAIN) {
+ local->op_ret = -1;
+ local->op_errno = EAGAIN;
- if (op_ret == 0) {
- local->op_ret = 0;
- local->op_errno = 0;
- local->cont.lk.locked_nodes[child_index] = 1;
- local->cont.lk.ret_flock = *lock;
- }
+ afr_lk_unlock(frame, this);
+ return 0;
+ }
+
+ if (op_ret == 0) {
+ local->op_ret = 0;
+ local->op_errno = 0;
+ local->cont.lk.locked_nodes[child_index] = 1;
+ local->cont.lk.ret_flock = *lock;
+ }
+
+ child_index++;
+
+ if (child_index < priv->child_count) {
+ STACK_WIND_COOKIE(frame, afr_lk_cbk, (void *)(long)child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->lk, local->fd,
+ local->cont.lk.cmd, &local->cont.lk.user_flock,
+ local->xdata_req);
+ } else if (priv->quorum_count &&
+ !afr_has_quorum(local->cont.lk.locked_nodes, this)) {
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
- child_index++;
-
- if (child_index < priv->child_count) {
- STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->lk,
- local->fd, local->cont.lk.cmd,
- &local->cont.lk.user_flock,
- local->xdata_req);
- } else if (priv->quorum_count &&
- !afr_has_quorum (local->cont.lk.locked_nodes, this)) {
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
-
- afr_lk_unlock (frame, this);
- } else {
- if (local->op_ret < 0)
- local->op_errno = afr_final_errno (local, priv);
+ afr_lk_unlock(frame, this);
+ } else {
+ if (local->op_ret < 0)
+ local->op_errno = afr_final_errno(local, priv);
- AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- &local->cont.lk.ret_flock, local->xdata_rsp);
- }
+ AFR_STACK_UNWIND(lk, frame, local->op_ret, local->op_errno,
+ &local->cont.lk.ret_flock, local->xdata_rsp);
+ }
- return 0;
+ return 0;
}
int
-afr_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int32_t op_errno = ENOMEM;
-
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_LK;
- if (!afr_lk_is_unlock (cmd, flock) &&
- !afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
-
- local->cont.lk.locked_nodes = GF_CALLOC (priv->child_count,
- sizeof (*local->cont.lk.locked_nodes),
- gf_afr_mt_char);
-
- if (!local->cont.lk.locked_nodes) {
- op_errno = ENOMEM;
- goto out;
- }
-
- local->fd = fd_ref (fd);
- local->cont.lk.cmd = cmd;
- local->cont.lk.user_flock = *flock;
- local->cont.lk.ret_flock = *flock;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
-
- STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
- priv->children[i],
- priv->children[i]->fops->lk,
- fd, cmd, flock, local->xdata_req);
-
- return 0;
+afr_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int32_t op_errno = ENOMEM;
+
+ priv = this->private;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_LK;
+ if (!afr_lk_is_unlock(cmd, flock) &&
+ !afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
+
+ local->cont.lk.locked_nodes = GF_CALLOC(
+ priv->child_count, sizeof(*local->cont.lk.locked_nodes),
+ gf_afr_mt_char);
+
+ if (!local->cont.lk.locked_nodes) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->fd = fd_ref(fd);
+ local->cont.lk.cmd = cmd;
+ local->cont.lk.user_flock = *flock;
+ local->cont.lk.ret_flock = *flock;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+
+ STACK_WIND_COOKIE(frame, afr_lk_cbk, (void *)(long)0, priv->children[i],
+ priv->children[i]->fops->lk, fd, cmd, flock,
+ local->xdata_req);
+
+ return 0;
out:
- AFR_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(lk, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
int32_t
-afr_lease_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_lease *lease,
- dict_t *xdata)
+afr_lease_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_lease *lease,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
+ afr_local_t *local = NULL;
+ int call_count = -1;
- local = frame->local;
- call_count = afr_frame_return (frame);
+ local = frame->local;
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- AFR_STACK_UNWIND (lease, frame, local->op_ret, local->op_errno,
- lease, xdata);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(lease, frame, local->op_ret, local->op_errno, lease,
+ xdata);
- return 0;
+ return 0;
}
int32_t
-afr_lease_unlock (call_frame_t *frame, xlator_t *this)
+afr_lease_unlock(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int call_count = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- call_count = afr_locked_nodes_count (local->cont.lease.locked_nodes,
- priv->child_count);
+ call_count = afr_locked_nodes_count(local->cont.lease.locked_nodes,
+ priv->child_count);
- if (call_count == 0) {
- AFR_STACK_UNWIND (lease, frame, local->op_ret, local->op_errno,
- &local->cont.lease.ret_lease, NULL);
- return 0;
- }
+ if (call_count == 0) {
+ AFR_STACK_UNWIND(lease, frame, local->op_ret, local->op_errno,
+ &local->cont.lease.ret_lease, NULL);
+ return 0;
+ }
- local->call_count = call_count;
+ local->call_count = call_count;
- local->cont.lease.user_lease.cmd = GF_UNLK_LEASE;
+ local->cont.lease.user_lease.cmd = GF_UNLK_LEASE;
- for (i = 0; i < priv->child_count; i++) {
- if (local->cont.lease.locked_nodes[i]) {
- STACK_WIND (frame, afr_lease_unlock_cbk,
- priv->children[i],
- priv->children[i]->fops->lease,
- &local->loc, &local->cont.lease.user_lease, NULL);
-
- if (!--call_count)
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->cont.lease.locked_nodes[i]) {
+ STACK_WIND(frame, afr_lease_unlock_cbk, priv->children[i],
+ priv->children[i]->fops->lease, &local->loc,
+ &local->cont.lease.user_lease, NULL);
+
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
}
int32_t
-afr_lease_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_lease *lease,
- dict_t *xdata)
+afr_lease_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct gf_lease *lease, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int child_index = -1;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = -1;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- child_index = (long) cookie;
+ child_index = (long)cookie;
- afr_common_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- if (op_ret < 0 && op_errno == EAGAIN) {
- local->op_ret = -1;
- local->op_errno = EAGAIN;
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ if (op_ret < 0 && op_errno == EAGAIN) {
+ local->op_ret = -1;
+ local->op_errno = EAGAIN;
- afr_lease_unlock (frame, this);
- return 0;
- }
+ afr_lease_unlock(frame, this);
+ return 0;
+ }
- if (op_ret == 0) {
- local->op_ret = 0;
- local->op_errno = 0;
- local->cont.lease.locked_nodes[child_index] = 1;
- local->cont.lease.ret_lease = *lease;
- }
+ if (op_ret == 0) {
+ local->op_ret = 0;
+ local->op_errno = 0;
+ local->cont.lease.locked_nodes[child_index] = 1;
+ local->cont.lease.ret_lease = *lease;
+ }
+
+ child_index++;
+ if (child_index < priv->child_count) {
+ STACK_WIND_COOKIE(frame, afr_lease_cbk, (void *)(long)child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->lease, &local->loc,
+ &local->cont.lease.user_lease, xdata);
+ } else if (priv->quorum_count &&
+ !afr_has_quorum(local->cont.lk.locked_nodes, this)) {
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
- child_index++;
- if (child_index < priv->child_count) {
- STACK_WIND_COOKIE (frame, afr_lease_cbk, (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->lease,
- &local->loc, &local->cont.lease.user_lease, xdata);
- } else if (priv->quorum_count &&
- !afr_has_quorum (local->cont.lk.locked_nodes, this)) {
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
-
- afr_lease_unlock (frame, this);
- } else {
- if (local->op_ret < 0)
- local->op_errno = afr_final_errno (local, priv);
- AFR_STACK_UNWIND (lease, frame, local->op_ret, local->op_errno,
- &local->cont.lease.ret_lease, NULL);
- }
+ afr_lease_unlock(frame, this);
+ } else {
+ if (local->op_ret < 0)
+ local->op_errno = afr_final_errno(local, priv);
+ AFR_STACK_UNWIND(lease, frame, local->op_ret, local->op_errno,
+ &local->cont.lease.ret_lease, NULL);
+ }
- return 0;
+ return 0;
}
int
-afr_lease (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct gf_lease *lease, dict_t *xdata)
+afr_lease(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct gf_lease *lease, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int32_t op_errno = ENOMEM;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int32_t op_errno = ENOMEM;
- priv = this->private;
+ priv = this->private;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_LEASE;
- local->cont.lease.locked_nodes = GF_CALLOC (priv->child_count,
- sizeof (*local->cont.lease.locked_nodes),
- gf_afr_mt_char);
+ local->op = GF_FOP_LEASE;
+ local->cont.lease.locked_nodes = GF_CALLOC(
+ priv->child_count, sizeof(*local->cont.lease.locked_nodes),
+ gf_afr_mt_char);
- if (!local->cont.lease.locked_nodes) {
- op_errno = ENOMEM;
- goto out;
- }
+ if (!local->cont.lease.locked_nodes) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- loc_copy (&local->loc, loc);
- local->cont.lease.user_lease = *lease;
- local->cont.lease.ret_lease = *lease;
+ loc_copy(&local->loc, loc);
+ local->cont.lease.user_lease = *lease;
+ local->cont.lease.ret_lease = *lease;
- STACK_WIND_COOKIE (frame, afr_lease_cbk, (void *) (long) 0,
- priv->children[0],
- priv->children[0]->fops->lease,
- loc, lease, xdata);
+ STACK_WIND_COOKIE(frame, afr_lease_cbk, (void *)(long)0, priv->children[0],
+ priv->children[0]->fops->lease, loc, lease, xdata);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (lease, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(lease, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
int
-afr_ipc_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- afr_local_t *local = NULL;
- int child_index = (long)cookie;
- int call_count = 0;
- gf_boolean_t failed = _gf_false;
- gf_boolean_t succeeded = _gf_false;
- int i = 0;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
-
- call_count = afr_frame_return (frame);
- if (call_count)
- goto out;
- /* If any of the subvolumes failed with other than ENOTCONN
- * return error else return success unless all the subvolumes
- * failed.
- * TODO: In case of failure, we need to unregister the xattrs
- * from the other subvolumes where it succeeded (once upcall
- * fixes the Bz-1371622)*/
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret < 0 &&
- local->replies[i].op_errno != ENOTCONN) {
- local->op_ret = local->replies[i].op_ret;
- local->op_errno = local->replies[i].op_errno;
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = NULL;
- if (local->replies[i].xdata) {
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- }
- failed = _gf_true;
- break;
- }
- if (local->replies[i].op_ret == 0) {
- succeeded = _gf_true;
- local->op_ret = 0;
- local->op_errno = 0;
- if (!local->xdata_rsp && local->replies[i].xdata) {
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- }
- }
+afr_ipc_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int child_index = (long)cookie;
+ int call_count = 0;
+ gf_boolean_t failed = _gf_false;
+ gf_boolean_t succeeded = _gf_false;
+ int i = 0;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
+
+ call_count = afr_frame_return(frame);
+ if (call_count)
+ goto out;
+ /* If any of the subvolumes failed with other than ENOTCONN
+ * return error else return success unless all the subvolumes
+ * failed.
+ * TODO: In case of failure, we need to unregister the xattrs
+ * from the other subvolumes where it succeeded (once upcall
+ * fixes the Bz-1371622)*/
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret < 0 &&
+ local->replies[i].op_errno != ENOTCONN) {
+ local->op_ret = local->replies[i].op_ret;
+ local->op_errno = local->replies[i].op_errno;
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = NULL;
+ if (local->replies[i].xdata) {
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ }
+ failed = _gf_true;
+ break;
+ }
+ if (local->replies[i].op_ret == 0) {
+ succeeded = _gf_true;
+ local->op_ret = 0;
+ local->op_errno = 0;
+ if (!local->xdata_rsp && local->replies[i].xdata) {
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ }
}
+ }
- if (!succeeded && !failed) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- }
+ if (!succeeded && !failed) {
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ }
- AFR_STACK_UNWIND (ipc, frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ AFR_STACK_UNWIND(ipc, frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
out:
- return 0;
+ return 0;
}
int
-afr_ipc (call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
+afr_ipc(call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t op_errno = -1;
- afr_private_t *priv = NULL;
- int i = 0;
- int call_cnt = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
+ afr_local_t *local = NULL;
+ int32_t op_errno = -1;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int call_cnt = -1;
- if (op != GF_IPC_TARGET_UPCALL)
- goto wind_default;
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
- VALIDATE_OR_GOTO (this->private, err);
- priv = this->private;
+ if (op != GF_IPC_TARGET_UPCALL)
+ goto wind_default;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto err;
+ VALIDATE_OR_GOTO(this->private, err);
+ priv = this->private;
- call_cnt = local->call_count;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto err;
- if (xdata) {
- for (i = 0; i < priv->child_count; i++) {
- if (dict_set_int8 (xdata, priv->pending_key[i], 0) < 0)
- goto err;
- }
- }
+ call_cnt = local->call_count;
+ if (xdata) {
for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
- continue;
-
- STACK_WIND_COOKIE (frame, afr_ipc_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->ipc,
- op, xdata);
- if (!--call_cnt)
- break;
+ if (dict_set_int8(xdata, priv->pending_key[i], 0) < 0)
+ goto err;
}
- return 0;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->child_up[i])
+ continue;
+
+ STACK_WIND_COOKIE(frame, afr_ipc_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->ipc, op,
+ xdata);
+ if (!--call_cnt)
+ break;
+ }
+ return 0;
err:
- if (op_errno == -1)
- op_errno = errno;
- AFR_STACK_UNWIND (ipc, frame, -1, op_errno, NULL);
+ if (op_errno == -1)
+ op_errno = errno;
+ AFR_STACK_UNWIND(ipc, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
wind_default:
- STACK_WIND (frame, default_ipc_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ipc, op, xdata);
- return 0;
+ STACK_WIND(frame, default_ipc_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ipc, op, xdata);
+ return 0;
}
int
-afr_forget (xlator_t *this, inode_t *inode)
+afr_forget(xlator_t *this, inode_t *inode)
{
- uint64_t ctx_int = 0;
- afr_inode_ctx_t *ctx = NULL;
-
- afr_spb_choice_timeout_cancel (this, inode);
- inode_ctx_del (inode, this, &ctx_int);
- if (!ctx_int)
- return 0;
+ uint64_t ctx_int = 0;
+ afr_inode_ctx_t *ctx = NULL;
- ctx = (afr_inode_ctx_t *)ctx_int;
- afr_inode_ctx_destroy (ctx);
+ afr_spb_choice_timeout_cancel(this, inode);
+ inode_ctx_del(inode, this, &ctx_int);
+ if (!ctx_int)
return 0;
+
+ ctx = (afr_inode_ctx_t *)ctx_int;
+ afr_inode_ctx_destroy(ctx);
+ return 0;
}
int
-afr_priv_dump (xlator_t *this)
-{
- afr_private_t *priv = NULL;
- char key_prefix[GF_DUMP_MAX_BUF_LEN];
- char key[GF_DUMP_MAX_BUF_LEN];
- int i = 0;
-
-
- GF_ASSERT (this);
- priv = this->private;
-
- GF_ASSERT (priv);
- snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
- gf_proc_dump_add_section(key_prefix);
- gf_proc_dump_write("child_count", "%u", priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- sprintf (key, "child_up[%d]", i);
- gf_proc_dump_write(key, "%d", priv->child_up[i]);
- sprintf (key, "pending_key[%d]", i);
- gf_proc_dump_write(key, "%s", priv->pending_key[i]);
- sprintf (key, "pending_reads[%d]", i);
- gf_proc_dump_write(key, "%"PRId64, GF_ATOMIC_GET(priv->pending_reads[i]));
- sprintf (key, "child_latency[%d]", i);
- gf_proc_dump_write(key, "%"PRId64, priv->child_latency[i]);
- }
- gf_proc_dump_write("data_self_heal", "%s", priv->data_self_heal);
- gf_proc_dump_write("metadata_self_heal", "%d", priv->metadata_self_heal);
- gf_proc_dump_write("entry_self_heal", "%d", priv->entry_self_heal);
- gf_proc_dump_write("read_child", "%d", priv->read_child);
- gf_proc_dump_write("wait_count", "%u", priv->wait_count);
- gf_proc_dump_write("heal-wait-queue-length", "%d",
- priv->heal_wait_qlen);
- gf_proc_dump_write("heal-waiters", "%d", priv->heal_waiters);
- gf_proc_dump_write("background-self-heal-count", "%d",
- priv->background_self_heal_count);
- gf_proc_dump_write("healers", "%d", priv->healers);
- gf_proc_dump_write("read-hash-mode", "%d", priv->hash_mode);
- if (priv->quorum_count == AFR_QUORUM_AUTO) {
- gf_proc_dump_write ("quorum-type", "auto");
- } else if (priv->quorum_count == 0) {
- gf_proc_dump_write ("quorum-type", "none");
- } else {
- gf_proc_dump_write("quorum-type", "fixed");
- gf_proc_dump_write("quorum-count", "%d", priv->quorum_count);
- }
- gf_proc_dump_write("up", "%u", afr_has_quorum (priv->child_up, this));
-
- return 0;
+afr_priv_dump(xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ char key_prefix[GF_DUMP_MAX_BUF_LEN];
+ char key[GF_DUMP_MAX_BUF_LEN];
+ int i = 0;
+
+ GF_ASSERT(this);
+ priv = this->private;
+
+ GF_ASSERT(priv);
+ snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
+ gf_proc_dump_add_section(key_prefix);
+ gf_proc_dump_write("child_count", "%u", priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ sprintf(key, "child_up[%d]", i);
+ gf_proc_dump_write(key, "%d", priv->child_up[i]);
+ sprintf(key, "pending_key[%d]", i);
+ gf_proc_dump_write(key, "%s", priv->pending_key[i]);
+ sprintf(key, "pending_reads[%d]", i);
+ gf_proc_dump_write(key, "%" PRId64,
+ GF_ATOMIC_GET(priv->pending_reads[i]));
+ sprintf(key, "child_latency[%d]", i);
+ gf_proc_dump_write(key, "%" PRId64, priv->child_latency[i]);
+ }
+ gf_proc_dump_write("data_self_heal", "%s", priv->data_self_heal);
+ gf_proc_dump_write("metadata_self_heal", "%d", priv->metadata_self_heal);
+ gf_proc_dump_write("entry_self_heal", "%d", priv->entry_self_heal);
+ gf_proc_dump_write("read_child", "%d", priv->read_child);
+ gf_proc_dump_write("wait_count", "%u", priv->wait_count);
+ gf_proc_dump_write("heal-wait-queue-length", "%d", priv->heal_wait_qlen);
+ gf_proc_dump_write("heal-waiters", "%d", priv->heal_waiters);
+ gf_proc_dump_write("background-self-heal-count", "%d",
+ priv->background_self_heal_count);
+ gf_proc_dump_write("healers", "%d", priv->healers);
+ gf_proc_dump_write("read-hash-mode", "%d", priv->hash_mode);
+ if (priv->quorum_count == AFR_QUORUM_AUTO) {
+ gf_proc_dump_write("quorum-type", "auto");
+ } else if (priv->quorum_count == 0) {
+ gf_proc_dump_write("quorum-type", "none");
+ } else {
+ gf_proc_dump_write("quorum-type", "fixed");
+ gf_proc_dump_write("quorum-count", "%d", priv->quorum_count);
+ }
+ gf_proc_dump_write("up", "%u", afr_has_quorum(priv->child_up, this));
+
+ return 0;
}
-
/**
* find_child_index - find the child's index in the array of subvolumes
* @this: AFR
@@ -5025,1974 +4904,1924 @@ afr_priv_dump (xlator_t *this)
*/
static int
-find_child_index (xlator_t *this, xlator_t *child)
+find_child_index(xlator_t *this, xlator_t *child)
{
- afr_private_t *priv = NULL;
- int i = -1;
+ afr_private_t *priv = NULL;
+ int i = -1;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if ((xlator_t *) child == priv->children[i])
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if ((xlator_t *)child == priv->children[i])
+ break;
+ }
- return i;
+ return i;
}
static int
-__afr_get_up_children_count (afr_private_t *priv)
+__afr_get_up_children_count(afr_private_t *priv)
{
- int up_children = 0;
- int i = 0;
+ int up_children = 0;
+ int i = 0;
- for (i = 0; i < priv->child_count; i++)
- if (priv->child_up[i] == 1)
- up_children++;
+ for (i = 0; i < priv->child_count; i++)
+ if (priv->child_up[i] == 1)
+ up_children++;
- return up_children;
+ return up_children;
}
glusterfs_event_t
-__afr_transform_event_from_state (afr_private_t *priv)
+__afr_transform_event_from_state(afr_private_t *priv)
{
- int i = 0;
- int up_children = 0;
+ int i = 0;
+ int up_children = 0;
- if (AFR_COUNT (priv->last_event, priv->child_count) ==
- priv->child_count)
- /* have_heard_from_all. Let afr_notify() do the propagation. */
- return GF_EVENT_MAXVAL;
+ if (AFR_COUNT(priv->last_event, priv->child_count) == priv->child_count)
+ /* have_heard_from_all. Let afr_notify() do the propagation. */
+ return GF_EVENT_MAXVAL;
- up_children = __afr_get_up_children_count (priv);
- /* Treat the children with pending notification, as having sent a
- * GF_EVENT_CHILD_DOWN. i.e. set the event as GF_EVENT_SOME_DESCENDENT_DOWN,
- * as done in afr_notify() */
- for (i = 0; i < priv->child_count; i++) {
- if (priv->last_event[i])
- continue;
- priv->last_event[i] = GF_EVENT_SOME_DESCENDENT_DOWN;
- priv->child_up[i] = 0;
- }
+ up_children = __afr_get_up_children_count(priv);
+ /* Treat the children with pending notification, as having sent a
+ * GF_EVENT_CHILD_DOWN. i.e. set the event as GF_EVENT_SOME_DESCENDENT_DOWN,
+ * as done in afr_notify() */
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->last_event[i])
+ continue;
+ priv->last_event[i] = GF_EVENT_SOME_DESCENDENT_DOWN;
+ priv->child_up[i] = 0;
+ }
- if (up_children)
- /* We received at least one child up */
- return GF_EVENT_CHILD_UP;
- else
- return GF_EVENT_CHILD_DOWN;
+ if (up_children)
+ /* We received at least one child up */
+ return GF_EVENT_CHILD_UP;
+ else
+ return GF_EVENT_CHILD_DOWN;
- return GF_EVENT_MAXVAL;
+ return GF_EVENT_MAXVAL;
}
static void
-afr_notify_cbk (void *data)
-{
- xlator_t *this = data;
- afr_private_t *priv = this->private;
- glusterfs_event_t event = GF_EVENT_MAXVAL;
- gf_boolean_t propagate = _gf_false;
-
- LOCK (&priv->lock);
- {
- if (!priv->timer) {
- /*
- * Either child_up/child_down is already sent to parent.
- * This is a spurious wake up.
- */
- goto unlock;
- }
- priv->timer = NULL;
- event = __afr_transform_event_from_state (priv);
- if (event != GF_EVENT_MAXVAL)
- propagate = _gf_true;
- }
+afr_notify_cbk(void *data)
+{
+ xlator_t *this = data;
+ afr_private_t *priv = this->private;
+ glusterfs_event_t event = GF_EVENT_MAXVAL;
+ gf_boolean_t propagate = _gf_false;
+
+ LOCK(&priv->lock);
+ {
+ if (!priv->timer) {
+ /*
+ * Either child_up/child_down is already sent to parent.
+ * This is a spurious wake up.
+ */
+ goto unlock;
+ }
+ priv->timer = NULL;
+ event = __afr_transform_event_from_state(priv);
+ if (event != GF_EVENT_MAXVAL)
+ propagate = _gf_true;
+ }
unlock:
- UNLOCK (&priv->lock);
- if (propagate)
- default_notify (this, event, NULL);
+ UNLOCK(&priv->lock);
+ if (propagate)
+ default_notify(this, event, NULL);
}
static void
-__afr_launch_notify_timer (xlator_t *this, afr_private_t *priv)
+__afr_launch_notify_timer(xlator_t *this, afr_private_t *priv)
{
+ struct timespec delay = {
+ 0,
+ };
- struct timespec delay = {0, };
-
- gf_msg_debug (this->name, 0, "Initiating child-down timer");
- delay.tv_sec = 10;
- delay.tv_nsec = 0;
- priv->timer = gf_timer_call_after (this->ctx, delay,
- afr_notify_cbk, this);
- if (priv->timer == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_TIMER_CREATE_FAIL,
- "Cannot create timer for delayed initialization");
- }
+ gf_msg_debug(this->name, 0, "Initiating child-down timer");
+ delay.tv_sec = 10;
+ delay.tv_nsec = 0;
+ priv->timer = gf_timer_call_after(this->ctx, delay, afr_notify_cbk, this);
+ if (priv->timer == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_TIMER_CREATE_FAIL,
+ "Cannot create timer for delayed initialization");
+ }
}
int
-__get_heard_from_all_status (xlator_t *this)
+__get_heard_from_all_status(xlator_t *this)
{
- afr_private_t *priv = this->private;
- int heard_from_all = 1;
- int i = 0;
+ afr_private_t *priv = this->private;
+ int heard_from_all = 1;
+ int i = 0;
- for (i = 0; i < priv->child_count; i++) {
- if (!priv->last_event[i]) {
- heard_from_all = 0;
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!priv->last_event[i]) {
+ heard_from_all = 0;
+ break;
}
- return heard_from_all;
+ }
+ return heard_from_all;
}
static int
-find_best_down_child (xlator_t *this)
+find_best_down_child(xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = -1;
- int32_t best_child = -1;
- int64_t best_latency = INT64_MAX;
+ afr_private_t *priv = NULL;
+ int i = -1;
+ int32_t best_child = -1;
+ int64_t best_latency = INT64_MAX;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (priv->child_up[i] &&
- priv->child_latency[i] >= 0 &&
- priv->child_latency[i] < best_latency) {
- best_child = i;
- best_latency = priv->child_latency[i];
- }
- }
- if (best_child >= 0) {
- gf_msg_debug (this->name, 0, "Found best down child (%d) "
- "@ %ld ms latency", best_child, best_latency);
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->child_up[i] && priv->child_latency[i] >= 0 &&
+ priv->child_latency[i] < best_latency) {
+ best_child = i;
+ best_latency = priv->child_latency[i];
}
- return best_child;
+ }
+ if (best_child >= 0) {
+ gf_msg_debug(this->name, 0,
+ "Found best down child (%d) "
+ "@ %ld ms latency",
+ best_child, best_latency);
+ }
+ return best_child;
}
int
-find_worst_up_child (xlator_t *this)
+find_worst_up_child(xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = -1;
- int32_t worst_child = -1;
- int64_t worst_latency = INT64_MIN;
+ afr_private_t *priv = NULL;
+ int i = -1;
+ int32_t worst_child = -1;
+ int64_t worst_latency = INT64_MIN;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (priv->child_up[i] &&
- priv->child_latency[i] >= 0 &&
- priv->child_latency[i] > worst_latency) {
- worst_child = i;
- worst_latency = priv->child_latency[i];
- }
- }
- if (worst_child >= 0) {
- gf_msg_debug (this->name, 0, "Found worst up child (%d)"
- " @ %ld ms latency", worst_child, worst_latency);
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->child_up[i] && priv->child_latency[i] >= 0 &&
+ priv->child_latency[i] > worst_latency) {
+ worst_child = i;
+ worst_latency = priv->child_latency[i];
}
- return worst_child;
+ }
+ if (worst_child >= 0) {
+ gf_msg_debug(this->name, 0,
+ "Found worst up child (%d)"
+ " @ %ld ms latency",
+ worst_child, worst_latency);
+ }
+ return worst_child;
}
void
-__afr_handle_ping_event (xlator_t *this, xlator_t *child_xlator,
- const int idx, int64_t halo_max_latency_msec, int32_t *event,
- int64_t child_latency_msec)
+__afr_handle_ping_event(xlator_t *this, xlator_t *child_xlator, const int idx,
+ int64_t halo_max_latency_msec, int32_t *event,
+ int64_t child_latency_msec)
{
- afr_private_t *priv = NULL;
- int up_children = 0;
+ afr_private_t *priv = NULL;
+ int up_children = 0;
- priv = this->private;
+ priv = this->private;
- priv->child_latency[idx] = child_latency_msec;
- gf_msg_debug (child_xlator->name, 0, "Client ping @ %ld ms",
- child_latency_msec);
- if (priv->shd.iamshd)
- return;
-
- up_children = __afr_get_up_children_count (priv);
-
- if (child_latency_msec > halo_max_latency_msec &&
- priv->child_up[idx] == 1 &&
- up_children > priv->halo_min_replicas) {
- if ((up_children - 1) <
- priv->halo_min_replicas) {
- gf_log (child_xlator->name, GF_LOG_INFO,
- "Overriding halo threshold, "
- "min replicas: %d",
- priv->halo_min_replicas);
- } else {
- gf_log (child_xlator->name, GF_LOG_INFO,
- "Child latency (%ld ms) "
- "exceeds halo threshold (%ld), "
- "marking child down.",
- child_latency_msec,
- halo_max_latency_msec);
- *event = GF_EVENT_CHILD_DOWN;
- }
- } else if (child_latency_msec < halo_max_latency_msec &&
- priv->child_up[idx] == 0) {
- if (up_children < priv->halo_max_replicas) {
- gf_log (child_xlator->name, GF_LOG_INFO,
- "Child latency (%ld ms) "
- "below halo threshold (%ld), "
- "marking child up.",
- child_latency_msec,
- halo_max_latency_msec);
- *event = GF_EVENT_CHILD_UP;
- } else {
- gf_log (child_xlator->name, GF_LOG_INFO,
- "Not marking child %d up, "
- "max replicas (%d) reached.", idx,
- priv->halo_max_replicas);
- }
+ priv->child_latency[idx] = child_latency_msec;
+ gf_msg_debug(child_xlator->name, 0, "Client ping @ %ld ms",
+ child_latency_msec);
+ if (priv->shd.iamshd)
+ return;
+
+ up_children = __afr_get_up_children_count(priv);
+
+ if (child_latency_msec > halo_max_latency_msec &&
+ priv->child_up[idx] == 1 && up_children > priv->halo_min_replicas) {
+ if ((up_children - 1) < priv->halo_min_replicas) {
+ gf_log(child_xlator->name, GF_LOG_INFO,
+ "Overriding halo threshold, "
+ "min replicas: %d",
+ priv->halo_min_replicas);
+ } else {
+ gf_log(child_xlator->name, GF_LOG_INFO,
+ "Child latency (%ld ms) "
+ "exceeds halo threshold (%ld), "
+ "marking child down.",
+ child_latency_msec, halo_max_latency_msec);
+ *event = GF_EVENT_CHILD_DOWN;
+ }
+ } else if (child_latency_msec < halo_max_latency_msec &&
+ priv->child_up[idx] == 0) {
+ if (up_children < priv->halo_max_replicas) {
+ gf_log(child_xlator->name, GF_LOG_INFO,
+ "Child latency (%ld ms) "
+ "below halo threshold (%ld), "
+ "marking child up.",
+ child_latency_msec, halo_max_latency_msec);
+ *event = GF_EVENT_CHILD_UP;
+ } else {
+ gf_log(child_xlator->name, GF_LOG_INFO,
+ "Not marking child %d up, "
+ "max replicas (%d) reached.",
+ idx, priv->halo_max_replicas);
}
+ }
}
static int64_t
-afr_get_halo_latency (xlator_t *this)
+afr_get_halo_latency(xlator_t *this)
{
- afr_private_t *priv = NULL;
- int64_t halo_max_latency_msec = 0;
+ afr_private_t *priv = NULL;
+ int64_t halo_max_latency_msec = 0;
- priv = this->private;
+ priv = this->private;
- if (priv->shd.iamshd) {
- halo_max_latency_msec = priv->shd.halo_max_latency_msec;
- } else if (priv->nfsd.iamnfsd) {
- halo_max_latency_msec =
- priv->nfsd.halo_max_latency_msec;
- } else {
- halo_max_latency_msec = priv->halo_max_latency_msec;
- }
- gf_msg_debug (this->name, 0, "Using halo latency %ld",
- halo_max_latency_msec);
- return halo_max_latency_msec;
+ if (priv->shd.iamshd) {
+ halo_max_latency_msec = priv->shd.halo_max_latency_msec;
+ } else if (priv->nfsd.iamnfsd) {
+ halo_max_latency_msec = priv->nfsd.halo_max_latency_msec;
+ } else {
+ halo_max_latency_msec = priv->halo_max_latency_msec;
+ }
+ gf_msg_debug(this->name, 0, "Using halo latency %ld",
+ halo_max_latency_msec);
+ return halo_max_latency_msec;
}
void
-__afr_handle_child_up_event (xlator_t *this, xlator_t *child_xlator,
- const int idx, int64_t child_latency_msec,
- int32_t *event, int32_t *call_psh, int32_t *up_child)
-{
- afr_private_t *priv = NULL;
- int up_children = 0;
- int worst_up_child = -1;
- int64_t halo_max_latency_msec = afr_get_halo_latency (this);
-
- priv = this->private;
-
- /*
- * This only really counts if the child was never up
- * (value = -1) or had been down (value = 0). See
- * comment at GF_EVENT_CHILD_DOWN for a more detailed
- * explanation.
- */
- if (priv->child_up[idx] != 1) {
- priv->event_generation++;
- }
- priv->child_up[idx] = 1;
-
- *call_psh = 1;
- *up_child = idx;
- up_children = __afr_get_up_children_count (priv);
- /*
- * If this is an _actual_ CHILD_UP event, we
- * want to set the child_latency to MAX to indicate
- * the child needs ping data to be available before doing child-up
- */
- if (child_latency_msec < 0 && priv->halo_enabled) {
- /*set to INT64_MAX-1 so that it is found for best_down_child*/
- priv->child_latency[idx] = AFR_HALO_MAX_LATENCY;
- }
-
- /*
- * Handle the edge case where we exceed
- * halo_min_replicas and we've got a child which is
- * marked up as it was helping to satisfy the
- * halo_min_replicas even though it's latency exceeds
- * halo_max_latency_msec.
- */
- if (up_children > priv->halo_min_replicas) {
- worst_up_child = find_worst_up_child (this);
- if (worst_up_child >= 0 &&
- priv->child_latency[worst_up_child] >
- halo_max_latency_msec) {
- gf_msg_debug (this->name, 0, "Marking child %d down, "
- "doesn't meet halo threshold (%ld), and > "
- "halo_min_replicas (%d)",
- worst_up_child, halo_max_latency_msec,
- priv->halo_min_replicas);
- priv->child_up[worst_up_child] = 0;
- up_children--;
- }
- }
-
- if (up_children > priv->halo_max_replicas &&
- !priv->shd.iamshd) {
- worst_up_child = find_worst_up_child (this);
- if (worst_up_child < 0) {
- worst_up_child = idx;
- }
- priv->child_up[worst_up_child] = 0;
- up_children--;
- gf_msg_debug (this->name, 0, "Marking child %d down, "
- "up_children (%d) > halo_max_replicas (%d)",
- worst_up_child, up_children, priv->halo_max_replicas);
- }
-
- if (up_children == 1) {
- gf_msg (this->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOL_UP,
- "Subvolume '%s' came back up; "
- "going online.",
- child_xlator->name);
- gf_event (EVENT_AFR_SUBVOL_UP, "subvol=%s", this->name);
- } else {
- *event = GF_EVENT_SOME_DESCENDENT_UP;
- }
-
- priv->last_event[idx] = *event;
+__afr_handle_child_up_event(xlator_t *this, xlator_t *child_xlator,
+ const int idx, int64_t child_latency_msec,
+ int32_t *event, int32_t *call_psh,
+ int32_t *up_child)
+{
+ afr_private_t *priv = NULL;
+ int up_children = 0;
+ int worst_up_child = -1;
+ int64_t halo_max_latency_msec = afr_get_halo_latency(this);
+
+ priv = this->private;
+
+ /*
+ * This only really counts if the child was never up
+ * (value = -1) or had been down (value = 0). See
+ * comment at GF_EVENT_CHILD_DOWN for a more detailed
+ * explanation.
+ */
+ if (priv->child_up[idx] != 1) {
+ priv->event_generation++;
+ }
+ priv->child_up[idx] = 1;
+
+ *call_psh = 1;
+ *up_child = idx;
+ up_children = __afr_get_up_children_count(priv);
+ /*
+ * If this is an _actual_ CHILD_UP event, we
+ * want to set the child_latency to MAX to indicate
+ * the child needs ping data to be available before doing child-up
+ */
+ if (child_latency_msec < 0 && priv->halo_enabled) {
+ /*set to INT64_MAX-1 so that it is found for best_down_child*/
+ priv->child_latency[idx] = AFR_HALO_MAX_LATENCY;
+ }
+
+ /*
+ * Handle the edge case where we exceed
+ * halo_min_replicas and we've got a child which is
+ * marked up as it was helping to satisfy the
+ * halo_min_replicas even though it's latency exceeds
+ * halo_max_latency_msec.
+ */
+ if (up_children > priv->halo_min_replicas) {
+ worst_up_child = find_worst_up_child(this);
+ if (worst_up_child >= 0 &&
+ priv->child_latency[worst_up_child] > halo_max_latency_msec) {
+ gf_msg_debug(this->name, 0,
+ "Marking child %d down, "
+ "doesn't meet halo threshold (%ld), and > "
+ "halo_min_replicas (%d)",
+ worst_up_child, halo_max_latency_msec,
+ priv->halo_min_replicas);
+ priv->child_up[worst_up_child] = 0;
+ up_children--;
+ }
+ }
+
+ if (up_children > priv->halo_max_replicas && !priv->shd.iamshd) {
+ worst_up_child = find_worst_up_child(this);
+ if (worst_up_child < 0) {
+ worst_up_child = idx;
+ }
+ priv->child_up[worst_up_child] = 0;
+ up_children--;
+ gf_msg_debug(this->name, 0,
+ "Marking child %d down, "
+ "up_children (%d) > halo_max_replicas (%d)",
+ worst_up_child, up_children, priv->halo_max_replicas);
+ }
+
+ if (up_children == 1) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOL_UP,
+ "Subvolume '%s' came back up; "
+ "going online.",
+ child_xlator->name);
+ gf_event(EVENT_AFR_SUBVOL_UP, "subvol=%s", this->name);
+ } else {
+ *event = GF_EVENT_SOME_DESCENDENT_UP;
+ }
+
+ priv->last_event[idx] = *event;
}
void
-__afr_handle_child_down_event (xlator_t *this, xlator_t *child_xlator,
- int idx, int64_t child_latency_msec, int32_t *event,
- int32_t *call_psh, int32_t *up_child)
-{
- afr_private_t *priv = NULL;
- int i = 0;
- int up_children = 0;
- int down_children = 0;
- int best_down_child = -1;
-
- priv = this->private;
-
- /*
- * If a brick is down when we start, we'll get a
- * CHILD_DOWN to indicate its initial state. There
- * was never a CHILD_UP in this case, so if we
- * increment "down_count" the difference between than
- * and "up_count" will no longer be the number of
- * children that are currently up. This has serious
- * implications e.g. for quorum enforcement, so we
- * don't increment these values unless the event
- * represents an actual state transition between "up"
- * (value = 1) and anything else.
- */
- if (priv->child_up[idx] == 1) {
- priv->event_generation++;
- }
-
- /*
- * If this is an _actual_ CHILD_DOWN event, we
- * want to set the child_latency to < 0 to indicate
- * the child is really disconnected.
- */
- if (child_latency_msec < 0) {
- priv->child_latency[idx] = child_latency_msec;
- }
- priv->child_up[idx] = 0;
-
- up_children = __afr_get_up_children_count (priv);
- /*
- * Handle the edge case where we need to find the
- * next best child (to mark up) as marking this child
- * down would cause us to fall below halo_min_replicas.
- * We will also force the SHD to heal this child _now_
- * as we want it to be up to date if we are going to
- * begin using it synchronously.
- */
- if (up_children < priv->halo_min_replicas) {
- best_down_child = find_best_down_child (this);
- if (best_down_child >= 0) {
- gf_msg_debug (this->name, 0,
- "Swapping out child %d for "
- "child %d to satisfy halo_min_replicas (%d).",
- idx, best_down_child, priv->halo_min_replicas);
- priv->child_up[best_down_child] = 1;
- *call_psh = 1;
- *up_child = best_down_child;
- }
- }
-
- for (i = 0; i < priv->child_count; i++)
- if (priv->child_up[i] == 0)
- down_children++;
- if (down_children == priv->child_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_SUBVOLS_DOWN,
- "All subvolumes are down. Going "
- "offline until at least one of them "
- "comes back up.");
- gf_event (EVENT_AFR_SUBVOLS_DOWN, "subvol=%s", this->name);
- } else {
- *event = GF_EVENT_SOME_DESCENDENT_DOWN;
- }
- priv->last_event[idx] = *event;
+__afr_handle_child_down_event(xlator_t *this, xlator_t *child_xlator, int idx,
+ int64_t child_latency_msec, int32_t *event,
+ int32_t *call_psh, int32_t *up_child)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int up_children = 0;
+ int down_children = 0;
+ int best_down_child = -1;
+
+ priv = this->private;
+
+ /*
+ * If a brick is down when we start, we'll get a
+ * CHILD_DOWN to indicate its initial state. There
+ * was never a CHILD_UP in this case, so if we
+ * increment "down_count" the difference between than
+ * and "up_count" will no longer be the number of
+ * children that are currently up. This has serious
+ * implications e.g. for quorum enforcement, so we
+ * don't increment these values unless the event
+ * represents an actual state transition between "up"
+ * (value = 1) and anything else.
+ */
+ if (priv->child_up[idx] == 1) {
+ priv->event_generation++;
+ }
+
+ /*
+ * If this is an _actual_ CHILD_DOWN event, we
+ * want to set the child_latency to < 0 to indicate
+ * the child is really disconnected.
+ */
+ if (child_latency_msec < 0) {
+ priv->child_latency[idx] = child_latency_msec;
+ }
+ priv->child_up[idx] = 0;
+
+ up_children = __afr_get_up_children_count(priv);
+ /*
+ * Handle the edge case where we need to find the
+ * next best child (to mark up) as marking this child
+ * down would cause us to fall below halo_min_replicas.
+ * We will also force the SHD to heal this child _now_
+ * as we want it to be up to date if we are going to
+ * begin using it synchronously.
+ */
+ if (up_children < priv->halo_min_replicas) {
+ best_down_child = find_best_down_child(this);
+ if (best_down_child >= 0) {
+ gf_msg_debug(this->name, 0,
+ "Swapping out child %d for "
+ "child %d to satisfy halo_min_replicas (%d).",
+ idx, best_down_child, priv->halo_min_replicas);
+ priv->child_up[best_down_child] = 1;
+ *call_psh = 1;
+ *up_child = best_down_child;
+ }
+ }
+
+ for (i = 0; i < priv->child_count; i++)
+ if (priv->child_up[i] == 0)
+ down_children++;
+ if (down_children == priv->child_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SUBVOLS_DOWN,
+ "All subvolumes are down. Going "
+ "offline until at least one of them "
+ "comes back up.");
+ gf_event(EVENT_AFR_SUBVOLS_DOWN, "subvol=%s", this->name);
+ } else {
+ *event = GF_EVENT_SOME_DESCENDENT_DOWN;
+ }
+ priv->last_event[idx] = *event;
}
int32_t
-afr_notify (xlator_t *this, int32_t event,
- void *data, void *data2)
-{
- afr_private_t *priv = NULL;
- xlator_t *child_xlator = NULL;
- int i = -1;
- int propagate = 0;
- int had_heard_from_all = 0;
- int have_heard_from_all = 0;
- int idx = -1;
- int ret = -1;
- int call_psh = 0;
- int up_child = -1;
- dict_t *input = NULL;
- dict_t *output = NULL;
- gf_boolean_t had_quorum = _gf_false;
- gf_boolean_t has_quorum = _gf_false;
- struct gf_upcall *up_data = NULL;
- struct gf_upcall_cache_invalidation *up_ci = NULL;
- inode_table_t *itable = NULL;
- inode_t *inode = NULL;
- int64_t halo_max_latency_msec = 0;
- int64_t child_latency_msec = -1;
-
- child_xlator = (xlator_t *)data;
-
- priv = this->private;
-
- if (!priv)
- return 0;
-
- /*
- * We need to reset this in case children come up in "staggered"
- * fashion, so that we discover a late-arriving local subvolume. Note
- * that we could end up issuing N lookups to the first subvolume, and
- * O(N^2) overall, but N is small for AFR so it shouldn't be an issue.
- */
- priv->did_discovery = _gf_false;
-
+afr_notify(xlator_t *this, int32_t event, void *data, void *data2)
+{
+ afr_private_t *priv = NULL;
+ xlator_t *child_xlator = NULL;
+ int i = -1;
+ int propagate = 0;
+ int had_heard_from_all = 0;
+ int have_heard_from_all = 0;
+ int idx = -1;
+ int ret = -1;
+ int call_psh = 0;
+ int up_child = -1;
+ dict_t *input = NULL;
+ dict_t *output = NULL;
+ gf_boolean_t had_quorum = _gf_false;
+ gf_boolean_t has_quorum = _gf_false;
+ struct gf_upcall *up_data = NULL;
+ struct gf_upcall_cache_invalidation *up_ci = NULL;
+ inode_table_t *itable = NULL;
+ inode_t *inode = NULL;
+ int64_t halo_max_latency_msec = 0;
+ int64_t child_latency_msec = -1;
+
+ child_xlator = (xlator_t *)data;
+
+ priv = this->private;
+
+ if (!priv)
+ return 0;
- /* parent xlators don't need to know about every child_up, child_down
- * because of afr ha. If all subvolumes go down, child_down has
- * to be triggered. In that state when 1 subvolume comes up child_up
- * needs to be triggered. dht optimizes revalidate lookup by sending
- * it only to one of its subvolumes. When child up/down happens
- * for afr's subvolumes dht should be notified by child_modified. The
- * subsequent revalidate lookup happens on all the dht's subvolumes
- * which triggers afr self-heals if any.
- */
- idx = find_child_index (this, child_xlator);
- if (idx < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_CHILD_UP,
- "Received child_up from invalid subvolume");
- goto out;
- }
+ /*
+ * We need to reset this in case children come up in "staggered"
+ * fashion, so that we discover a late-arriving local subvolume. Note
+ * that we could end up issuing N lookups to the first subvolume, and
+ * O(N^2) overall, but N is small for AFR so it shouldn't be an issue.
+ */
+ priv->did_discovery = _gf_false;
+
+ /* parent xlators don't need to know about every child_up, child_down
+ * because of afr ha. If all subvolumes go down, child_down has
+ * to be triggered. In that state when 1 subvolume comes up child_up
+ * needs to be triggered. dht optimizes revalidate lookup by sending
+ * it only to one of its subvolumes. When child up/down happens
+ * for afr's subvolumes dht should be notified by child_modified. The
+ * subsequent revalidate lookup happens on all the dht's subvolumes
+ * which triggers afr self-heals if any.
+ */
+ idx = find_child_index(this, child_xlator);
+ if (idx < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_CHILD_UP,
+ "Received child_up from invalid subvolume");
+ goto out;
+ }
+
+ had_quorum = priv->quorum_count && afr_has_quorum(priv->child_up, this);
+ if (priv->halo_enabled) {
+ halo_max_latency_msec = afr_get_halo_latency(this);
- had_quorum = priv->quorum_count && afr_has_quorum (priv->child_up,
- this);
- if (priv->halo_enabled) {
- halo_max_latency_msec = afr_get_halo_latency (this);
-
- if (event == GF_EVENT_CHILD_PING) {
- /* Calculates the child latency and sets event
- */
- child_latency_msec = (int64_t)(uintptr_t)data2;
- LOCK (&priv->lock);
- {
- __afr_handle_ping_event (this, child_xlator,
- idx, halo_max_latency_msec, &event,
+ if (event == GF_EVENT_CHILD_PING) {
+ /* Calculates the child latency and sets event
+ */
+ child_latency_msec = (int64_t)(uintptr_t)data2;
+ LOCK(&priv->lock);
+ {
+ __afr_handle_ping_event(this, child_xlator, idx,
+ halo_max_latency_msec, &event,
child_latency_msec);
- }
- UNLOCK (&priv->lock);
- }
+ }
+ UNLOCK(&priv->lock);
}
+ }
- if (event == GF_EVENT_CHILD_PING) {
- /* This is the only xlator that handles PING, no reason to
- * propagate.
- */
- goto out;
- }
+ if (event == GF_EVENT_CHILD_PING) {
+ /* This is the only xlator that handles PING, no reason to
+ * propagate.
+ */
+ goto out;
+ }
- if (event == GF_EVENT_TRANSLATOR_OP) {
- LOCK (&priv->lock);
- {
- had_heard_from_all = __get_heard_from_all_status (this);
- }
- UNLOCK (&priv->lock);
-
- if (!had_heard_from_all) {
- ret = -1;
- } else {
- input = data;
- output = data2;
- ret = afr_xl_op (this, input, output);
- }
- goto out;
+ if (event == GF_EVENT_TRANSLATOR_OP) {
+ LOCK(&priv->lock);
+ {
+ had_heard_from_all = __get_heard_from_all_status(this);
}
+ UNLOCK(&priv->lock);
- LOCK (&priv->lock);
- {
- had_heard_from_all = __get_heard_from_all_status (this);
- switch (event) {
- case GF_EVENT_PARENT_UP:
- __afr_launch_notify_timer (this, priv);
- propagate = 1;
- break;
- case GF_EVENT_CHILD_UP:
- __afr_handle_child_up_event (this, child_xlator,
- idx, child_latency_msec, &event, &call_psh,
- &up_child);
- break;
+ if (!had_heard_from_all) {
+ ret = -1;
+ } else {
+ input = data;
+ output = data2;
+ ret = afr_xl_op(this, input, output);
+ }
+ goto out;
+ }
+
+ LOCK(&priv->lock);
+ {
+ had_heard_from_all = __get_heard_from_all_status(this);
+ switch (event) {
+ case GF_EVENT_PARENT_UP:
+ __afr_launch_notify_timer(this, priv);
+ propagate = 1;
+ break;
+ case GF_EVENT_CHILD_UP:
+ __afr_handle_child_up_event(this, child_xlator, idx,
+ child_latency_msec, &event,
+ &call_psh, &up_child);
+ break;
- case GF_EVENT_CHILD_DOWN:
- __afr_handle_child_down_event (this, child_xlator, idx,
- child_latency_msec, &event, &call_psh,
- &up_child);
- break;
+ case GF_EVENT_CHILD_DOWN:
+ __afr_handle_child_down_event(this, child_xlator, idx,
+ child_latency_msec, &event,
+ &call_psh, &up_child);
+ break;
- case GF_EVENT_CHILD_CONNECTING:
- priv->last_event[idx] = event;
+ case GF_EVENT_CHILD_CONNECTING:
+ priv->last_event[idx] = event;
- break;
+ break;
+
+ case GF_EVENT_SOME_DESCENDENT_DOWN:
+ priv->last_event[idx] = event;
+ break;
+ case GF_EVENT_UPCALL:
+ up_data = (struct gf_upcall *)data;
+ if (up_data->event_type != GF_UPCALL_CACHE_INVALIDATION)
+ break;
+ up_ci = (struct gf_upcall_cache_invalidation *)up_data->data;
+
+ /* Since md-cache will be aggressively filtering
+ * lookups, the stale read issue will be more
+ * pronounced. Hence when a pending xattr is set notify
+ * all the md-cache clients to invalidate the existing
+ * stat cache and send the lookup next time */
+ if (!up_ci->dict)
+ break;
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_get(up_ci->dict, priv->pending_key[i])) {
+ up_ci->flags |= UP_INVAL_ATTR;
+ itable = ((xlator_t *)this->graph->top)->itable;
+ /*Internal processes may not have itable for top
+ * xlator*/
+ if (itable)
+ inode = inode_find(itable, up_data->gfid);
+ if (inode)
+ afr_inode_need_refresh_set(inode, this);
- case GF_EVENT_SOME_DESCENDENT_DOWN:
- priv->last_event[idx] = event;
- break;
- case GF_EVENT_UPCALL:
- up_data = (struct gf_upcall *)data;
- if (up_data->event_type != GF_UPCALL_CACHE_INVALIDATION)
- break;
- up_ci = (struct gf_upcall_cache_invalidation *)up_data->data;
-
- /* Since md-cache will be aggressively filtering
- * lookups, the stale read issue will be more
- * pronounced. Hence when a pending xattr is set notify
- * all the md-cache clients to invalidate the existing
- * stat cache and send the lookup next time */
- if (!up_ci->dict)
- break;
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get (up_ci->dict, priv->pending_key[i])) {
- up_ci->flags |= UP_INVAL_ATTR;
- itable = ((xlator_t *)this->graph->top)->itable;
- /*Internal processes may not have itable for top xlator*/
- if (itable)
- inode = inode_find (itable, up_data->gfid);
- if (inode)
- afr_inode_need_refresh_set (inode, this);
-
- break;
- }
- }
- break;
- default:
- propagate = 1;
break;
+ }
}
- have_heard_from_all = __get_heard_from_all_status (this);
- if (!had_heard_from_all && have_heard_from_all) {
- if (priv->timer) {
- gf_timer_call_cancel (this->ctx, priv->timer);
- priv->timer = NULL;
- }
- /* This is the first event which completes aggregation
- of events from all subvolumes. If at least one subvol
- had come up, propagate CHILD_UP, but only this time
- */
- event = GF_EVENT_CHILD_DOWN;
- for (i = 0; i < priv->child_count; i++) {
- if (priv->last_event[i] == GF_EVENT_CHILD_UP) {
- event = GF_EVENT_CHILD_UP;
- break;
- }
-
- if (priv->last_event[i] ==
- GF_EVENT_CHILD_CONNECTING) {
- event = GF_EVENT_CHILD_CONNECTING;
- /* continue to check other events for CHILD_UP */
- }
- }
- }
+ break;
+ default:
+ propagate = 1;
+ break;
}
- UNLOCK (&priv->lock);
-
- if (priv->quorum_count) {
- has_quorum = afr_has_quorum (priv->child_up, this);
- if (!had_quorum && has_quorum) {
- gf_msg (this->name, GF_LOG_INFO, 0, AFR_MSG_QUORUM_MET,
- "Client-quorum is met");
- gf_event (EVENT_AFR_QUORUM_MET,
- "subvol=%s", this->name);
- }
- if (had_quorum && !has_quorum) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_QUORUM_FAIL,
- "Client-quorum is not met");
- gf_event (EVENT_AFR_QUORUM_FAIL, "subvol=%s",
- this->name);
+ have_heard_from_all = __get_heard_from_all_status(this);
+ if (!had_heard_from_all && have_heard_from_all) {
+ if (priv->timer) {
+ gf_timer_call_cancel(this->ctx, priv->timer);
+ priv->timer = NULL;
+ }
+ /* This is the first event which completes aggregation
+ of events from all subvolumes. If at least one subvol
+ had come up, propagate CHILD_UP, but only this time
+ */
+ event = GF_EVENT_CHILD_DOWN;
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->last_event[i] == GF_EVENT_CHILD_UP) {
+ event = GF_EVENT_CHILD_UP;
+ break;
}
- }
-
- /* if all subvols have reported status, no need to hide anything
- or wait for anything else. Just propagate blindly */
- if (have_heard_from_all)
- propagate = 1;
- ret = 0;
- if (propagate)
- ret = default_notify (this, event, data);
-
- if ((!had_heard_from_all) || call_psh) {
- /* Launch self-heal on all local subvolumes if:
- * a) We have_heard_from_all for the first time
- * b) Already heard from everyone, but we now got a child-up
- * event.
- */
- if (have_heard_from_all && priv->shd.iamshd) {
- for (i = 0; i < priv->child_count; i++)
- if (priv->child_up[i])
- afr_selfheal_childup (this, i);
+ if (priv->last_event[i] == GF_EVENT_CHILD_CONNECTING) {
+ event = GF_EVENT_CHILD_CONNECTING;
+ /* continue to check other events for CHILD_UP */
}
+ }
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (priv->quorum_count) {
+ has_quorum = afr_has_quorum(priv->child_up, this);
+ if (!had_quorum && has_quorum) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_QUORUM_MET,
+ "Client-quorum is met");
+ gf_event(EVENT_AFR_QUORUM_MET, "subvol=%s", this->name);
+ }
+ if (had_quorum && !has_quorum) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_FAIL,
+ "Client-quorum is not met");
+ gf_event(EVENT_AFR_QUORUM_FAIL, "subvol=%s", this->name);
+ }
+ }
+
+ /* if all subvols have reported status, no need to hide anything
+ or wait for anything else. Just propagate blindly */
+ if (have_heard_from_all)
+ propagate = 1;
+
+ ret = 0;
+ if (propagate)
+ ret = default_notify(this, event, data);
+
+ if ((!had_heard_from_all) || call_psh) {
+ /* Launch self-heal on all local subvolumes if:
+ * a) We have_heard_from_all for the first time
+ * b) Already heard from everyone, but we now got a child-up
+ * event.
+ */
+ if (have_heard_from_all && priv->shd.iamshd) {
+ for (i = 0; i < priv->child_count; i++)
+ if (priv->child_up[i])
+ afr_selfheal_childup(this, i);
}
+ }
out:
- return ret;
+ return ret;
}
int
-afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
+afr_local_init(afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
{
- int __ret = -1;
- local->op_ret = -1;
- local->op_errno = EUCLEAN;
+ int __ret = -1;
+ local->op_ret = -1;
+ local->op_errno = EUCLEAN;
- __ret = syncbarrier_init (&local->barrier);
- if (__ret) {
- if (op_errno)
- *op_errno = __ret;
- goto out;
- }
+ __ret = syncbarrier_init(&local->barrier);
+ if (__ret) {
+ if (op_errno)
+ *op_errno = __ret;
+ goto out;
+ }
- local->child_up = GF_CALLOC (priv->child_count,
- sizeof (*local->child_up),
- gf_afr_mt_char);
- if (!local->child_up) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
+ local->child_up = GF_CALLOC(priv->child_count, sizeof(*local->child_up),
+ gf_afr_mt_char);
+ if (!local->child_up) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ memcpy(local->child_up, priv->child_up,
+ sizeof(*local->child_up) * priv->child_count);
+ local->call_count = AFR_COUNT(local->child_up, priv->child_count);
+ if (local->call_count == 0) {
+ gf_msg(THIS->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOLS_DOWN,
+ "no subvolumes up");
+ if (op_errno)
+ *op_errno = ENOTCONN;
+ goto out;
+ }
- memcpy (local->child_up, priv->child_up,
- sizeof (*local->child_up) * priv->child_count);
- local->call_count = AFR_COUNT (local->child_up, priv->child_count);
- if (local->call_count == 0) {
- gf_msg (THIS->name, GF_LOG_INFO, 0,
- AFR_MSG_SUBVOLS_DOWN, "no subvolumes up");
- if (op_errno)
- *op_errno = ENOTCONN;
- goto out;
- }
+ local->event_generation = priv->event_generation;
- local->event_generation = priv->event_generation;
-
- local->read_attempted = GF_CALLOC (priv->child_count, sizeof (char),
- gf_afr_mt_char);
- if (!local->read_attempted) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
-
- local->readable = GF_CALLOC (priv->child_count, sizeof (char),
- gf_afr_mt_char);
- if (!local->readable) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
-
- local->readable2 = GF_CALLOC (priv->child_count, sizeof (char),
+ local->read_attempted = GF_CALLOC(priv->child_count, sizeof(char),
gf_afr_mt_char);
- if (!local->readable2) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
+ if (!local->read_attempted) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
- local->read_subvol = -1;
+ local->readable = GF_CALLOC(priv->child_count, sizeof(char),
+ gf_afr_mt_char);
+ if (!local->readable) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
- local->replies = GF_CALLOC(priv->child_count, sizeof(*local->replies),
- gf_afr_mt_reply_t);
- if (!local->replies) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
+ local->readable2 = GF_CALLOC(priv->child_count, sizeof(char),
+ gf_afr_mt_char);
+ if (!local->readable2) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
- local->need_full_crawl = _gf_false;
+ local->read_subvol = -1;
- INIT_LIST_HEAD (&local->healer);
- return 0;
+ local->replies = GF_CALLOC(priv->child_count, sizeof(*local->replies),
+ gf_afr_mt_reply_t);
+ if (!local->replies) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->need_full_crawl = _gf_false;
+
+ INIT_LIST_HEAD(&local->healer);
+ return 0;
out:
- return -1;
+ return -1;
}
int
-afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count)
+afr_internal_lock_init(afr_internal_lock_t *lk, size_t child_count)
{
- int ret = -ENOMEM;
+ int ret = -ENOMEM;
- lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
- child_count, gf_afr_mt_char);
- if (NULL == lk->locked_nodes)
- goto out;
+ lk->locked_nodes = GF_CALLOC(sizeof(*lk->locked_nodes), child_count,
+ gf_afr_mt_char);
+ if (NULL == lk->locked_nodes)
+ goto out;
- lk->lower_locked_nodes = GF_CALLOC (sizeof (*lk->lower_locked_nodes),
- child_count, gf_afr_mt_char);
- if (NULL == lk->lower_locked_nodes)
- goto out;
+ lk->lower_locked_nodes = GF_CALLOC(sizeof(*lk->lower_locked_nodes),
+ child_count, gf_afr_mt_char);
+ if (NULL == lk->lower_locked_nodes)
+ goto out;
- lk->lock_op_ret = -1;
- lk->lock_op_errno = EUCLEAN;
+ lk->lock_op_ret = -1;
+ lk->lock_op_errno = EUCLEAN;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
void
-afr_matrix_cleanup (int32_t **matrix, unsigned int m)
+afr_matrix_cleanup(int32_t **matrix, unsigned int m)
{
- int i = 0;
+ int i = 0;
- if (!matrix)
- goto out;
- for (i = 0; i < m; i++) {
- GF_FREE (matrix[i]);
- }
+ if (!matrix)
+ goto out;
+ for (i = 0; i < m; i++) {
+ GF_FREE(matrix[i]);
+ }
- GF_FREE (matrix);
+ GF_FREE(matrix);
out:
- return;
+ return;
}
-int32_t**
-afr_matrix_create (unsigned int m, unsigned int n)
+int32_t **
+afr_matrix_create(unsigned int m, unsigned int n)
{
- int32_t **matrix = NULL;
- int i = 0;
+ int32_t **matrix = NULL;
+ int i = 0;
- matrix = GF_CALLOC (sizeof (*matrix), m, gf_afr_mt_int32_t);
- if (!matrix)
- goto out;
+ matrix = GF_CALLOC(sizeof(*matrix), m, gf_afr_mt_int32_t);
+ if (!matrix)
+ goto out;
- for (i = 0; i < m; i++) {
- matrix[i] = GF_CALLOC (sizeof (*matrix[i]), n,
- gf_afr_mt_int32_t);
- if (!matrix[i])
- goto out;
- }
- return matrix;
+ for (i = 0; i < m; i++) {
+ matrix[i] = GF_CALLOC(sizeof(*matrix[i]), n, gf_afr_mt_int32_t);
+ if (!matrix[i])
+ goto out;
+ }
+ return matrix;
out:
- afr_matrix_cleanup (matrix, m);
- return NULL;
+ afr_matrix_cleanup(matrix, m);
+ return NULL;
}
int
-afr_transaction_local_init (afr_local_t *local, xlator_t *this)
-{
- int ret = -ENOMEM;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- ret = afr_internal_lock_init (&local->internal_lock, priv->child_count);
- if (ret < 0)
- goto out;
-
- ret = -ENOMEM;
- local->pre_op_compat = priv->pre_op_compat;
-
- local->transaction.pre_op = GF_CALLOC (sizeof (*local->transaction.pre_op),
- priv->child_count,
- gf_afr_mt_char);
- if (!local->transaction.pre_op)
- goto out;
-
- local->transaction.changelog_xdata =
- GF_CALLOC (sizeof (*local->transaction.changelog_xdata),
- priv->child_count, gf_afr_mt_dict_t);
- if (!local->transaction.changelog_xdata)
- goto out;
-
- if (priv->arbiter_count == 1) {
- local->transaction.pre_op_sources =
- GF_CALLOC (sizeof (*local->transaction.pre_op_sources),
- priv->child_count, gf_afr_mt_char);
- if (!local->transaction.pre_op_sources)
- goto out;
- }
-
- local->transaction.failed_subvols = GF_CALLOC (sizeof (*local->transaction.failed_subvols),
- priv->child_count,
- gf_afr_mt_char);
- if (!local->transaction.failed_subvols)
- goto out;
-
- local->pending = afr_matrix_create (priv->child_count,
- AFR_NUM_CHANGE_LOGS);
- if (!local->pending)
- goto out;
-
- ret = 0;
- INIT_LIST_HEAD (&local->transaction.wait_list);
- INIT_LIST_HEAD (&local->transaction.owner_list);
+afr_transaction_local_init(afr_local_t *local, xlator_t *this)
+{
+ int ret = -ENOMEM;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ ret = afr_internal_lock_init(&local->internal_lock, priv->child_count);
+ if (ret < 0)
+ goto out;
+
+ ret = -ENOMEM;
+ local->pre_op_compat = priv->pre_op_compat;
+
+ local->transaction.pre_op = GF_CALLOC(sizeof(*local->transaction.pre_op),
+ priv->child_count, gf_afr_mt_char);
+ if (!local->transaction.pre_op)
+ goto out;
+
+ local->transaction.changelog_xdata = GF_CALLOC(
+ sizeof(*local->transaction.changelog_xdata), priv->child_count,
+ gf_afr_mt_dict_t);
+ if (!local->transaction.changelog_xdata)
+ goto out;
+
+ if (priv->arbiter_count == 1) {
+ local->transaction.pre_op_sources = GF_CALLOC(
+ sizeof(*local->transaction.pre_op_sources), priv->child_count,
+ gf_afr_mt_char);
+ if (!local->transaction.pre_op_sources)
+ goto out;
+ }
+
+ local->transaction.failed_subvols = GF_CALLOC(
+ sizeof(*local->transaction.failed_subvols), priv->child_count,
+ gf_afr_mt_char);
+ if (!local->transaction.failed_subvols)
+ goto out;
+
+ local->pending = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!local->pending)
+ goto out;
+
+ ret = 0;
+ INIT_LIST_HEAD(&local->transaction.wait_list);
+ INIT_LIST_HEAD(&local->transaction.owner_list);
out:
- return ret;
+ return ret;
}
-
void
-afr_set_low_priority (call_frame_t *frame)
+afr_set_low_priority(call_frame_t *frame)
{
- frame->root->pid = LOW_PRIO_PROC_PID;
+ frame->root->pid = LOW_PRIO_PROC_PID;
}
-
void
-afr_priv_destroy (afr_private_t *priv)
-{
- int i = 0;
- int child_count = -1;
-
- if (!priv)
- goto out;
- GF_FREE (priv->last_event);
-
- child_count = priv->child_count;
- if (priv->thin_arbiter_count) {
- child_count++;
- }
- if (priv->pending_key) {
- for (i = 0; i < child_count; i++)
- GF_FREE (priv->pending_key[i]);
- }
-
- GF_FREE (priv->pending_reads);
- GF_FREE (priv->local);
- GF_FREE (priv->pending_key);
- GF_FREE (priv->children);
- GF_FREE (priv->child_up);
- GF_FREE (priv->child_latency);
- LOCK_DESTROY (&priv->lock);
-
- GF_FREE (priv);
+afr_priv_destroy(afr_private_t *priv)
+{
+ int i = 0;
+ int child_count = -1;
+
+ if (!priv)
+ goto out;
+ GF_FREE(priv->last_event);
+
+ child_count = priv->child_count;
+ if (priv->thin_arbiter_count) {
+ child_count++;
+ }
+ if (priv->pending_key) {
+ for (i = 0; i < child_count; i++)
+ GF_FREE(priv->pending_key[i]);
+ }
+
+ GF_FREE(priv->pending_reads);
+ GF_FREE(priv->local);
+ GF_FREE(priv->pending_key);
+ GF_FREE(priv->children);
+ GF_FREE(priv->child_up);
+ GF_FREE(priv->child_latency);
+ LOCK_DESTROY(&priv->lock);
+
+ GF_FREE(priv);
out:
- return;
+ return;
}
-int**
-afr_mark_pending_changelog (afr_private_t *priv, unsigned char *pending,
- dict_t *xattr, ia_type_t iat)
+int **
+afr_mark_pending_changelog(afr_private_t *priv, unsigned char *pending,
+ dict_t *xattr, ia_type_t iat)
{
- int i = 0;
- int **changelog = NULL;
- int idx = -1;
- int m_idx = 0;
- int d_idx = 0;
- int ret = 0;
+ int i = 0;
+ int **changelog = NULL;
+ int idx = -1;
+ int m_idx = 0;
+ int d_idx = 0;
+ int ret = 0;
- m_idx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
- d_idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+ m_idx = afr_index_for_transaction_type(AFR_METADATA_TRANSACTION);
+ d_idx = afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
- idx = afr_index_from_ia_type (iat);
+ idx = afr_index_from_ia_type(iat);
- changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
- if (!changelog)
- goto out;
+ changelog = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!changelog)
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- if (!pending[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!pending[i])
+ continue;
- changelog[i][m_idx] = hton32(1);
- if (idx != -1)
- changelog[i][idx] = hton32(1);
- /* If the newentry marking is on a newly created directory,
- * then mark it with the full-heal indicator.
- */
- if ((IA_ISDIR (iat)) && (priv->esh_granular))
- changelog[i][d_idx] = hton32(1);
- }
- ret = afr_set_pending_dict (priv, xattr, changelog);
- if (ret < 0) {
- afr_matrix_cleanup (changelog, priv->child_count);
- return NULL;
- }
+ changelog[i][m_idx] = hton32(1);
+ if (idx != -1)
+ changelog[i][idx] = hton32(1);
+ /* If the newentry marking is on a newly created directory,
+ * then mark it with the full-heal indicator.
+ */
+ if ((IA_ISDIR(iat)) && (priv->esh_granular))
+ changelog[i][d_idx] = hton32(1);
+ }
+ ret = afr_set_pending_dict(priv, xattr, changelog);
+ if (ret < 0) {
+ afr_matrix_cleanup(changelog, priv->child_count);
+ return NULL;
+ }
out:
- return changelog;
+ return changelog;
}
gf_boolean_t
-afr_decide_heal_info (afr_private_t *priv, unsigned char *sources, int source)
+afr_decide_heal_info(afr_private_t *priv, unsigned char *sources, int source)
{
- int sources_count = 0;
+ int sources_count = 0;
- if (source < 0)
- goto out;
+ if (source < 0)
+ goto out;
- sources_count = AFR_COUNT (sources, priv->child_count);
- if (sources_count == priv->child_count)
- return _gf_false;
+ sources_count = AFR_COUNT(sources, priv->child_count);
+ if (sources_count == priv->child_count)
+ return _gf_false;
out:
- return _gf_true;
+ return _gf_true;
}
int
-afr_selfheal_locked_metadata_inspect (call_frame_t *frame, xlator_t *this,
- inode_t *inode, gf_boolean_t *msh,
- gf_boolean_t *pending)
-{
- int ret = -1;
- unsigned char *locked_on = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
-
- afr_private_t *priv = this->private;
-
- locked_on = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, inode, this->name,
- LLONG_MAX - 1, 0, locked_on);
- {
- if (ret == 0) {
- /* Not a single lock */
- ret = -afr_final_errno (frame->local, priv);
- if (ret == 0)
- ret = -ENOTCONN;/* all invalid responses */
- goto out;
- }
- ret = __afr_selfheal_metadata_prepare (frame, this, inode,
- locked_on, sources,
- sinks, healed_sinks,
- undid_pending,
- locked_replies,
- pending);
- *msh = afr_decide_heal_info (priv, sources, ret);
- }
- afr_selfheal_uninodelk (frame, this, inode, this->name,
- LLONG_MAX - 1, 0, locked_on);
+afr_selfheal_locked_metadata_inspect(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, gf_boolean_t *msh,
+ gf_boolean_t *pending)
+{
+ int ret = -1;
+ unsigned char *locked_on = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+
+ afr_private_t *priv = this->private;
+
+ locked_on = alloca0(priv->child_count);
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = afr_selfheal_inodelk(frame, this, inode, this->name, LLONG_MAX - 1, 0,
+ locked_on);
+ {
+ if (ret == 0) {
+ /* Not a single lock */
+ ret = -afr_final_errno(frame->local, priv);
+ if (ret == 0)
+ ret = -ENOTCONN; /* all invalid responses */
+ goto out;
+ }
+ ret = __afr_selfheal_metadata_prepare(
+ frame, this, inode, locked_on, sources, sinks, healed_sinks,
+ undid_pending, locked_replies, pending);
+ *msh = afr_decide_heal_info(priv, sources, ret);
+ }
+ afr_selfheal_uninodelk(frame, this, inode, this->name, LLONG_MAX - 1, 0,
+ locked_on);
out:
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
+ return ret;
}
int
-afr_selfheal_locked_data_inspect (call_frame_t *frame, xlator_t *this,
- fd_t *fd, gf_boolean_t *dsh,
- gf_boolean_t *pflag)
-{
- int ret = -1;
- unsigned char *data_lock = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *locked_replies = NULL;
- inode_t *inode = fd->inode;
-
- priv = this->private;
- data_lock = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, inode, this->name,
- 0, 0, data_lock);
- {
- if (ret == 0) {
- ret = -afr_final_errno (frame->local, priv);
- if (ret == 0)
- ret = -ENOTCONN; /* all invalid responses */
- goto out;
- }
- ret = __afr_selfheal_data_prepare (frame, this, inode,
- data_lock, sources, sinks,
- healed_sinks, undid_pending,
- locked_replies, pflag);
- *dsh = afr_decide_heal_info (priv, sources, ret);
- }
- afr_selfheal_uninodelk (frame, this, inode, this->name, 0, 0,
- data_lock);
+afr_selfheal_locked_data_inspect(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_boolean_t *dsh, gf_boolean_t *pflag)
+{
+ int ret = -1;
+ unsigned char *data_lock = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *locked_replies = NULL;
+ inode_t *inode = fd->inode;
+
+ priv = this->private;
+ data_lock = alloca0(priv->child_count);
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = afr_selfheal_inodelk(frame, this, inode, this->name, 0, 0, data_lock);
+ {
+ if (ret == 0) {
+ ret = -afr_final_errno(frame->local, priv);
+ if (ret == 0)
+ ret = -ENOTCONN; /* all invalid responses */
+ goto out;
+ }
+ ret = __afr_selfheal_data_prepare(frame, this, inode, data_lock,
+ sources, sinks, healed_sinks,
+ undid_pending, locked_replies, pflag);
+ *dsh = afr_decide_heal_info(priv, sources, ret);
+ }
+ afr_selfheal_uninodelk(frame, this, inode, this->name, 0, 0, data_lock);
out:
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
+ return ret;
}
int
-afr_selfheal_locked_entry_inspect (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- gf_boolean_t *esh, gf_boolean_t *pflag)
-{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- unsigned char *locked_on = NULL;
- unsigned char *data_lock = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- struct afr_reply *locked_replies = NULL;
- gf_boolean_t granular_locks = _gf_false;
-
- priv = this->private;
- if (strcmp ("granular", priv->locking_scheme) == 0)
- granular_locks = _gf_true;
- locked_on = alloca0 (priv->child_count);
- data_lock = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- if (!granular_locks) {
- ret = afr_selfheal_tryentrylk (frame, this, inode,
- priv->sh_domain, NULL, locked_on);
- }
+afr_selfheal_locked_entry_inspect(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, gf_boolean_t *esh,
+ gf_boolean_t *pflag)
+{
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ unsigned char *locked_on = NULL;
+ unsigned char *data_lock = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ struct afr_reply *locked_replies = NULL;
+ gf_boolean_t granular_locks = _gf_false;
+
+ priv = this->private;
+ if (strcmp("granular", priv->locking_scheme) == 0)
+ granular_locks = _gf_true;
+ locked_on = alloca0(priv->child_count);
+ data_lock = alloca0(priv->child_count);
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ if (!granular_locks) {
+ ret = afr_selfheal_tryentrylk(frame, this, inode, priv->sh_domain, NULL,
+ locked_on);
+ }
+ {
+ if (!granular_locks && ret == 0) {
+ ret = -afr_final_errno(frame->local, priv);
+ if (ret == 0)
+ ret = -ENOTCONN; /* all invalid responses */
+ goto out;
+ }
+
+ ret = afr_selfheal_entrylk(frame, this, inode, this->name, NULL,
+ data_lock);
{
- if (!granular_locks && ret == 0) {
- ret = -afr_final_errno (frame->local, priv);
- if (ret == 0)
- ret = -ENOTCONN;/* all invalid responses */
- goto out;
- }
-
- ret = afr_selfheal_entrylk (frame, this, inode, this->name,
- NULL, data_lock);
- {
- if (ret == 0) {
- ret = -afr_final_errno (frame->local, priv);
- if (ret == 0)
- ret = -ENOTCONN;
- /* all invalid responses */
- goto unlock;
- }
- ret = __afr_selfheal_entry_prepare (frame, this, inode,
- data_lock, sources,
- sinks, healed_sinks,
- locked_replies,
- &source, pflag);
- if ((ret == 0) && source < 0)
- ret = -EIO;
- *esh = afr_decide_heal_info (priv, sources, ret);
- }
- afr_selfheal_unentrylk (frame, this, inode, this->name, NULL,
- data_lock, NULL);
- }
+ if (ret == 0) {
+ ret = -afr_final_errno(frame->local, priv);
+ if (ret == 0)
+ ret = -ENOTCONN;
+ /* all invalid responses */
+ goto unlock;
+ }
+ ret = __afr_selfheal_entry_prepare(frame, this, inode, data_lock,
+ sources, sinks, healed_sinks,
+ locked_replies, &source, pflag);
+ if ((ret == 0) && source < 0)
+ ret = -EIO;
+ *esh = afr_decide_heal_info(priv, sources, ret);
+ }
+ afr_selfheal_unentrylk(frame, this, inode, this->name, NULL, data_lock,
+ NULL);
+ }
unlock:
- if (!granular_locks)
- afr_selfheal_unentrylk (frame, this, inode, priv->sh_domain,
- NULL, locked_on, NULL);
+ if (!granular_locks)
+ afr_selfheal_unentrylk(frame, this, inode, priv->sh_domain, NULL,
+ locked_on, NULL);
out:
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
+ return ret;
}
int
-afr_selfheal_locked_inspect (call_frame_t *frame, xlator_t *this, uuid_t gfid,
- inode_t **inode,
- gf_boolean_t *entry_selfheal,
- gf_boolean_t *data_selfheal,
- gf_boolean_t *metadata_selfheal,
- gf_boolean_t *pending)
-
-{
- int ret = -1;
- fd_t *fd = NULL;
- gf_boolean_t dsh = _gf_false;
- gf_boolean_t msh = _gf_false;
- gf_boolean_t esh = _gf_false;
-
- ret = afr_selfheal_unlocked_inspect (frame, this, gfid, inode,
- &dsh, &msh, &esh);
- if (ret)
- goto out;
-
- /* For every heal type hold locks and check if it indeed needs heal */
-
-
- /* Heal-info does an open() on the file being examined so that the
- * current eager-lock holding client, if present, at some point sees
- * open-fd count being > 1 and releases the eager-lock so that heal-info
- * doesn't remain blocked forever until IO completes.
- */
- if ((*inode)->ia_type == IA_IFREG) {
- ret = afr_selfheal_data_open (this, *inode, &fd);
- if (ret < 0) {
- gf_msg_debug (this->name, -ret, "%s: Failed to open",
- uuid_utoa ((*inode)->gfid));
- goto out;
- }
+afr_selfheal_locked_inspect(call_frame_t *frame, xlator_t *this, uuid_t gfid,
+ inode_t **inode, gf_boolean_t *entry_selfheal,
+ gf_boolean_t *data_selfheal,
+ gf_boolean_t *metadata_selfheal,
+ gf_boolean_t *pending)
+
+{
+ int ret = -1;
+ fd_t *fd = NULL;
+ gf_boolean_t dsh = _gf_false;
+ gf_boolean_t msh = _gf_false;
+ gf_boolean_t esh = _gf_false;
+
+ ret = afr_selfheal_unlocked_inspect(frame, this, gfid, inode, &dsh, &msh,
+ &esh);
+ if (ret)
+ goto out;
+
+ /* For every heal type hold locks and check if it indeed needs heal */
+
+ /* Heal-info does an open() on the file being examined so that the
+ * current eager-lock holding client, if present, at some point sees
+ * open-fd count being > 1 and releases the eager-lock so that heal-info
+ * doesn't remain blocked forever until IO completes.
+ */
+ if ((*inode)->ia_type == IA_IFREG) {
+ ret = afr_selfheal_data_open(this, *inode, &fd);
+ if (ret < 0) {
+ gf_msg_debug(this->name, -ret, "%s: Failed to open",
+ uuid_utoa((*inode)->gfid));
+ goto out;
}
+ }
- if (msh) {
- ret = afr_selfheal_locked_metadata_inspect (frame, this,
- *inode, &msh,
- pending);
- if (ret == -EIO)
- goto out;
- }
+ if (msh) {
+ ret = afr_selfheal_locked_metadata_inspect(frame, this, *inode, &msh,
+ pending);
+ if (ret == -EIO)
+ goto out;
+ }
- if (dsh) {
- ret = afr_selfheal_locked_data_inspect (frame, this, fd,
- &dsh, pending);
- if (ret == -EIO || (ret == -EAGAIN))
- goto out;
- }
+ if (dsh) {
+ ret = afr_selfheal_locked_data_inspect(frame, this, fd, &dsh, pending);
+ if (ret == -EIO || (ret == -EAGAIN))
+ goto out;
+ }
- if (esh) {
- ret = afr_selfheal_locked_entry_inspect (frame, this, *inode,
- &esh, pending);
- }
+ if (esh) {
+ ret = afr_selfheal_locked_entry_inspect(frame, this, *inode, &esh,
+ pending);
+ }
out:
- *data_selfheal = dsh;
- *entry_selfheal = esh;
- *metadata_selfheal = msh;
- if (fd)
- fd_unref (fd);
- return ret;
+ *data_selfheal = dsh;
+ *entry_selfheal = esh;
+ *metadata_selfheal = msh;
+ if (fd)
+ fd_unref(fd);
+ return ret;
}
-dict_t*
-afr_set_heal_info (char *status)
+dict_t *
+afr_set_heal_info(char *status)
{
- dict_t *dict = NULL;
- int ret = -1;
-
- dict = dict_new ();
- if (!dict) {
- ret = -ENOMEM;
- goto out;
- }
+ dict_t *dict = NULL;
+ int ret = -1;
- ret = dict_set_str (dict, "heal-info", status);
- if (ret)
- gf_msg ("", GF_LOG_WARNING, -ret,
- AFR_MSG_DICT_SET_FAILED,
- "Failed to set heal-info key to "
- "%s", status);
+ dict = dict_new();
+ if (!dict) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ ret = dict_set_str(dict, "heal-info", status);
+ if (ret)
+ gf_msg("", GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Failed to set heal-info key to "
+ "%s",
+ status);
out:
- return dict;
+ return dict;
}
int
-afr_get_heal_info (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
- gf_boolean_t data_selfheal = _gf_false;
- gf_boolean_t metadata_selfheal = _gf_false;
- gf_boolean_t entry_selfheal = _gf_false;
- gf_boolean_t pending = _gf_false;
- dict_t *dict = NULL;
- int ret = -1;
- int op_errno = 0;
- inode_t *inode = NULL;
- char *substr = NULL;
- char *status = NULL;
-
- ret = afr_selfheal_locked_inspect (frame, this, loc->gfid, &inode,
- &entry_selfheal,
- &data_selfheal, &metadata_selfheal,
- &pending);
-
- if (ret == -ENOMEM) {
- op_errno = -ret;
- ret = -1;
+afr_get_heal_info(call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+ gf_boolean_t data_selfheal = _gf_false;
+ gf_boolean_t metadata_selfheal = _gf_false;
+ gf_boolean_t entry_selfheal = _gf_false;
+ gf_boolean_t pending = _gf_false;
+ dict_t *dict = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ inode_t *inode = NULL;
+ char *substr = NULL;
+ char *status = NULL;
+
+ ret = afr_selfheal_locked_inspect(frame, this, loc->gfid, &inode,
+ &entry_selfheal, &data_selfheal,
+ &metadata_selfheal, &pending);
+
+ if (ret == -ENOMEM) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ if (pending) {
+ gf_asprintf(&substr, "-pending");
+ if (!substr)
+ goto out;
+ }
+
+ if (ret == -EIO) {
+ ret = gf_asprintf(&status, "split-brain%s", substr ? substr : "");
+ if (ret < 0)
+ goto out;
+ dict = afr_set_heal_info(status);
+ } else if (ret == -EAGAIN) {
+ ret = gf_asprintf(&status, "possibly-healing%s", substr ? substr : "");
+ if (ret < 0)
+ goto out;
+ dict = afr_set_heal_info(status);
+ } else if (ret >= 0) {
+ /* value of ret = source index
+ * so ret >= 0 and at least one of the 3 booleans set to
+ * true means a source is identified; heal is required.
+ */
+ if (!data_selfheal && !entry_selfheal && !metadata_selfheal) {
+ dict = afr_set_heal_info("no-heal");
+ } else {
+ ret = gf_asprintf(&status, "heal%s", substr ? substr : "");
+ if (ret < 0)
goto out;
+ dict = afr_set_heal_info(status);
+ }
+ } else if (ret < 0) {
+ /* Apart from above checked -ve ret values, there are
+ * other possible ret values like ENOTCONN
+ * (returned when number of valid replies received are
+ * less than 2)
+ * in which case heal is required when one of the
+ * selfheal booleans is set.
+ */
+ if (data_selfheal || entry_selfheal || metadata_selfheal) {
+ ret = gf_asprintf(&status, "heal%s", substr ? substr : "");
+ if (ret < 0)
+ goto out;
+ dict = afr_set_heal_info(status);
}
-
- if (pending) {
- gf_asprintf (&substr, "-pending");
- if (!substr)
- goto out;
- }
-
- if (ret == -EIO) {
- ret = gf_asprintf (&status, "split-brain%s",
- substr? substr : "");
- if (ret < 0)
- goto out;
- dict = afr_set_heal_info (status);
- } else if (ret == -EAGAIN) {
- ret = gf_asprintf (&status, "possibly-healing%s",
- substr? substr : "");
- if (ret < 0)
- goto out;
- dict = afr_set_heal_info (status);
- } else if (ret >= 0) {
- /* value of ret = source index
- * so ret >= 0 and at least one of the 3 booleans set to
- * true means a source is identified; heal is required.
- */
- if (!data_selfheal && !entry_selfheal &&
- !metadata_selfheal) {
- dict = afr_set_heal_info ("no-heal");
- } else {
- ret = gf_asprintf (&status, "heal%s",
- substr? substr : "");
- if (ret < 0)
- goto out;
- dict = afr_set_heal_info (status);
- }
- } else if (ret < 0) {
- /* Apart from above checked -ve ret values, there are
- * other possible ret values like ENOTCONN
- * (returned when number of valid replies received are
- * less than 2)
- * in which case heal is required when one of the
- * selfheal booleans is set.
- */
- if (data_selfheal || entry_selfheal ||
- metadata_selfheal) {
- ret = gf_asprintf (&status, "heal%s",
- substr? substr : "");
- if (ret < 0)
- goto out;
- dict = afr_set_heal_info (status);
- }
- }
- ret = 0;
+ }
+ ret = 0;
out:
- AFR_STACK_UNWIND (getxattr, frame, ret, op_errno, dict, NULL);
- if (dict)
- dict_unref (dict);
- if (inode)
- inode_unref (inode);
- GF_FREE (substr);
- return ret;
+ AFR_STACK_UNWIND(getxattr, frame, ret, op_errno, dict, NULL);
+ if (dict)
+ dict_unref(dict);
+ if (inode)
+ inode_unref(inode);
+ GF_FREE(substr);
+ return ret;
}
int
-_afr_is_split_brain (call_frame_t *frame, xlator_t *this,
- struct afr_reply *replies,
- afr_transaction_type type,
- gf_boolean_t *spb)
-{
- afr_private_t *priv = NULL;
- uint64_t *witness = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- int sources_count = 0;
- int ret = 0;
-
- priv = this->private;
-
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- witness = alloca0(priv->child_count * sizeof (*witness));
-
- ret = afr_selfheal_find_direction (frame, this, replies,
- type, priv->child_up, sources,
- sinks, witness, NULL);
- if (ret)
- return ret;
+_afr_is_split_brain(call_frame_t *frame, xlator_t *this,
+ struct afr_reply *replies, afr_transaction_type type,
+ gf_boolean_t *spb)
+{
+ afr_private_t *priv = NULL;
+ uint64_t *witness = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ int sources_count = 0;
+ int ret = 0;
+
+ priv = this->private;
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ witness = alloca0(priv->child_count * sizeof(*witness));
+
+ ret = afr_selfheal_find_direction(frame, this, replies, type,
+ priv->child_up, sources, sinks, witness,
+ NULL);
+ if (ret)
+ return ret;
- sources_count = AFR_COUNT (sources, priv->child_count);
- if (!sources_count)
- *spb = _gf_true;
+ sources_count = AFR_COUNT(sources, priv->child_count);
+ if (!sources_count)
+ *spb = _gf_true;
- return ret;
+ return ret;
}
int
-afr_is_split_brain (call_frame_t *frame, xlator_t *this, inode_t *inode,
- uuid_t gfid, gf_boolean_t *d_spb, gf_boolean_t *m_spb)
+afr_is_split_brain(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ uuid_t gfid, gf_boolean_t *d_spb, gf_boolean_t *m_spb)
{
- int ret = -1;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
+ int ret = -1;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
- priv = this->private;
+ priv = this->private;
- replies = alloca0 (sizeof (*replies) * priv->child_count);
+ replies = alloca0(sizeof(*replies) * priv->child_count);
- ret = afr_selfheal_unlocked_discover (frame, inode, gfid, replies);
- if (ret)
- goto out;
+ ret = afr_selfheal_unlocked_discover(frame, inode, gfid, replies);
+ if (ret)
+ goto out;
- if (!afr_can_decide_split_brain_source_sinks (replies,
- priv->child_count)) {
- ret = -EAGAIN;
- goto out;
- }
+ if (!afr_can_decide_split_brain_source_sinks(replies, priv->child_count)) {
+ ret = -EAGAIN;
+ goto out;
+ }
- ret = _afr_is_split_brain (frame, this, replies,
- AFR_DATA_TRANSACTION, d_spb);
- if (ret)
- goto out;
+ ret = _afr_is_split_brain(frame, this, replies, AFR_DATA_TRANSACTION,
+ d_spb);
+ if (ret)
+ goto out;
- ret = _afr_is_split_brain (frame, this, replies,
- AFR_METADATA_TRANSACTION, m_spb);
+ ret = _afr_is_split_brain(frame, this, replies, AFR_METADATA_TRANSACTION,
+ m_spb);
out:
- if (replies) {
- afr_replies_wipe (replies, priv->child_count);
- replies = NULL;
- }
- return ret;
+ if (replies) {
+ afr_replies_wipe(replies, priv->child_count);
+ replies = NULL;
+ }
+ return ret;
}
int
-afr_get_split_brain_status_cbk (int ret, call_frame_t *frame, void *opaque)
+afr_get_split_brain_status_cbk(int ret, call_frame_t *frame, void *opaque)
{
- GF_FREE (opaque);
- return 0;
+ GF_FREE(opaque);
+ return 0;
}
int
-afr_get_split_brain_status (void *opaque)
-{
- gf_boolean_t d_spb = _gf_false;
- gf_boolean_t m_spb = _gf_false;
- int ret = -1;
- int op_errno = 0;
- int i = 0;
- char *choices = NULL;
- char *status = NULL;
- dict_t *dict = NULL;
- inode_t *inode = NULL;
- afr_private_t *priv = NULL;
- xlator_t **children = NULL;
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- loc_t *loc = NULL;
- afr_spb_status_t *data = NULL;
-
- data = opaque;
- frame = data->frame;
- this = frame->this;
- loc = data->loc;
- priv = this->private;
- children = priv->children;
-
- inode = afr_inode_find (this, loc->gfid);
- if (!inode)
- goto out;
-
- dict = dict_new ();
- if (!dict) {
- op_errno = ENOMEM;
- ret = -1;
- goto out;
+afr_get_split_brain_status(void *opaque)
+{
+ gf_boolean_t d_spb = _gf_false;
+ gf_boolean_t m_spb = _gf_false;
+ int ret = -1;
+ int op_errno = 0;
+ int i = 0;
+ char *choices = NULL;
+ char *status = NULL;
+ dict_t *dict = NULL;
+ inode_t *inode = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ loc_t *loc = NULL;
+ afr_spb_status_t *data = NULL;
+
+ data = opaque;
+ frame = data->frame;
+ this = frame->this;
+ loc = data->loc;
+ priv = this->private;
+ children = priv->children;
+
+ inode = afr_inode_find(this, loc->gfid);
+ if (!inode)
+ goto out;
+
+ dict = dict_new();
+ if (!dict) {
+ op_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ /* Calculation for string length :
+ * (child_count X length of child-name) + SLEN (" Choices :")
+ * child-name consists of :
+ * a) 251 = max characters for volname according to GD_VOLUME_NAME_MAX
+ * b) strlen ("-client-00,") assuming 16 replicas
+ */
+ choices = alloca0(priv->child_count * (256 + SLEN("-client-00,")) +
+ SLEN(" Choices:"));
+
+ ret = afr_is_split_brain(frame, this, inode, loc->gfid, &d_spb, &m_spb);
+ if (ret) {
+ op_errno = -ret;
+ if (ret == -EAGAIN) {
+ ret = dict_set_str(dict, GF_AFR_SBRAIN_STATUS,
+ SBRAIN_HEAL_NO_GO_MSG);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ AFR_MSG_DICT_SET_FAILED,
+ "Failed to set GF_AFR_SBRAIN_STATUS in dict");
+ }
}
+ ret = -1;
+ goto out;
+ }
- /* Calculation for string length :
- * (child_count X length of child-name) + SLEN (" Choices :")
- * child-name consists of :
- * a) 251 = max characters for volname according to GD_VOLUME_NAME_MAX
- * b) strlen ("-client-00,") assuming 16 replicas
- */
- choices = alloca0 (priv->child_count * (256 + SLEN ("-client-00,")) +
- SLEN (" Choices:"));
-
- ret = afr_is_split_brain (frame, this, inode, loc->gfid, &d_spb,
- &m_spb);
- if (ret) {
- op_errno = -ret;
- if (ret == -EAGAIN) {
- ret = dict_set_str (dict, GF_AFR_SBRAIN_STATUS,
- SBRAIN_HEAL_NO_GO_MSG);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Failed to set GF_AFR_SBRAIN_STATUS in dict");
- }
- }
- ret = -1;
- goto out;
+ if (d_spb || m_spb) {
+ sprintf(choices, " Choices:");
+ for (i = 0; i < priv->child_count; i++) {
+ strcat(choices, children[i]->name);
+ strcat(choices, ",");
}
+ choices[strlen(choices) - 1] = '\0';
- if (d_spb || m_spb) {
- sprintf (choices, " Choices:");
- for (i = 0; i < priv->child_count; i++) {
- strcat (choices, children[i]->name);
- strcat (choices, ",");
- }
- choices[strlen (choices) - 1] = '\0';
-
- ret = gf_asprintf (&status, "data-split-brain:%s "
- "metadata-split-brain:%s%s",
- (d_spb) ? "yes" : "no",
- (m_spb) ? "yes" : "no", choices);
+ ret = gf_asprintf(&status,
+ "data-split-brain:%s "
+ "metadata-split-brain:%s%s",
+ (d_spb) ? "yes" : "no", (m_spb) ? "yes" : "no",
+ choices);
- if (-1 == ret) {
- op_errno = ENOMEM;
- goto out;
- }
- ret = dict_set_dynstr (dict, GF_AFR_SBRAIN_STATUS, status);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
- } else {
- ret = dict_set_str (dict, GF_AFR_SBRAIN_STATUS,
- "The file is not under data or"
- " metadata split-brain");
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
+ if (-1 == ret) {
+ op_errno = ENOMEM;
+ goto out;
}
+ ret = dict_set_dynstr(dict, GF_AFR_SBRAIN_STATUS, status);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ } else {
+ ret = dict_set_str(dict, GF_AFR_SBRAIN_STATUS,
+ "The file is not under data or"
+ " metadata split-brain");
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ }
- ret = 0;
+ ret = 0;
out:
- AFR_STACK_UNWIND (getxattr, frame, ret, op_errno, dict, NULL);
- if (dict)
- dict_unref (dict);
- if (inode)
- inode_unref (inode);
- return ret;
+ AFR_STACK_UNWIND(getxattr, frame, ret, op_errno, dict, NULL);
+ if (dict)
+ dict_unref(dict);
+ if (inode)
+ inode_unref(inode);
+ return ret;
}
int32_t
afr_heal_splitbrain_file(call_frame_t *frame, xlator_t *this, loc_t *loc)
{
- int ret = 0;
- int op_errno = 0;
- dict_t *dict = NULL;
- afr_local_t *local = NULL;
+ int ret = 0;
+ int op_errno = 0;
+ dict_t *dict = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- dict = dict_new ();
- if (!dict) {
- op_errno = ENOMEM;
- ret = -1;
- goto out;
- }
+ local = frame->local;
+ dict = dict_new();
+ if (!dict) {
+ op_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
- ret = afr_selfheal_do (frame, this, loc->gfid);
+ ret = afr_selfheal_do(frame, this, loc->gfid);
- if (ret == 1 || ret == 2) {
- ret = dict_set_str (dict, "sh-fail-msg",
- "File not in split-brain");
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Failed to set sh-fail-msg in dict");
- ret = 0;
- goto out;
- } else {
- if (local->xdata_rsp) {
- /* 'sh-fail-msg' has been set in the dict during self-heal.*/
- dict_copy (local->xdata_rsp, dict);
- ret = 0;
- } else if (ret < 0) {
- op_errno = -ret;
- ret = -1;
- }
+ if (ret == 1 || ret == 2) {
+ ret = dict_set_str(dict, "sh-fail-msg", "File not in split-brain");
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Failed to set sh-fail-msg in dict");
+ ret = 0;
+ goto out;
+ } else {
+ if (local->xdata_rsp) {
+ /* 'sh-fail-msg' has been set in the dict during self-heal.*/
+ dict_copy(local->xdata_rsp, dict);
+ ret = 0;
+ } else if (ret < 0) {
+ op_errno = -ret;
+ ret = -1;
}
+ }
out:
- if (local->op == GF_FOP_GETXATTR)
- AFR_STACK_UNWIND (getxattr, frame, ret, op_errno, dict, NULL);
- else if (local->op == GF_FOP_SETXATTR)
- AFR_STACK_UNWIND (setxattr, frame, ret, op_errno, NULL);
- if (dict)
- dict_unref(dict);
- return ret;
+ if (local->op == GF_FOP_GETXATTR)
+ AFR_STACK_UNWIND(getxattr, frame, ret, op_errno, dict, NULL);
+ else if (local->op == GF_FOP_SETXATTR)
+ AFR_STACK_UNWIND(setxattr, frame, ret, op_errno, NULL);
+ if (dict)
+ dict_unref(dict);
+ return ret;
}
int
-afr_get_child_index_from_name (xlator_t *this, char *name)
+afr_get_child_index_from_name(xlator_t *this, char *name)
{
- afr_private_t *priv = this->private;
- int index = -1;
+ afr_private_t *priv = this->private;
+ int index = -1;
- for (index = 0; index < priv->child_count; index++) {
- if (!strcmp (priv->children[index]->name, name))
- goto out;
- }
- index = -1;
+ for (index = 0; index < priv->child_count; index++) {
+ if (!strcmp(priv->children[index]->name, name))
+ goto out;
+ }
+ index = -1;
out:
- return index;
+ return index;
}
void
-afr_priv_need_heal_set (afr_private_t *priv, gf_boolean_t need_heal)
+afr_priv_need_heal_set(afr_private_t *priv, gf_boolean_t need_heal)
{
- LOCK (&priv->lock);
- {
- priv->need_heal = need_heal;
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ priv->need_heal = need_heal;
+ }
+ UNLOCK(&priv->lock);
}
void
-afr_set_need_heal (xlator_t *this, afr_local_t *local)
+afr_set_need_heal(xlator_t *this, afr_local_t *local)
{
- int i = 0;
- afr_private_t *priv = this->private;
- gf_boolean_t need_heal = _gf_false;
+ int i = 0;
+ afr_private_t *priv = this->private;
+ gf_boolean_t need_heal = _gf_false;
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].valid && local->replies[i].need_heal) {
- need_heal = _gf_true;
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].valid && local->replies[i].need_heal) {
+ need_heal = _gf_true;
+ break;
}
- afr_priv_need_heal_set (priv, need_heal);
- return;
+ }
+ afr_priv_need_heal_set(priv, need_heal);
+ return;
}
gf_boolean_t
-afr_get_need_heal (xlator_t *this)
+afr_get_need_heal(xlator_t *this)
{
- afr_private_t *priv = this->private;
- gf_boolean_t need_heal = _gf_true;
+ afr_private_t *priv = this->private;
+ gf_boolean_t need_heal = _gf_true;
- LOCK (&priv->lock);
- {
- need_heal = priv->need_heal;
- }
- UNLOCK (&priv->lock);
- return need_heal;
+ LOCK(&priv->lock);
+ {
+ need_heal = priv->need_heal;
+ }
+ UNLOCK(&priv->lock);
+ return need_heal;
}
int
-afr_get_msg_id (char *op_type)
+afr_get_msg_id(char *op_type)
{
-
- if (!strcmp (op_type, GF_AFR_REPLACE_BRICK))
- return AFR_MSG_REPLACE_BRICK_STATUS;
- else if (!strcmp (op_type, GF_AFR_ADD_BRICK))
- return AFR_MSG_ADD_BRICK_STATUS;
- return -1;
+ if (!strcmp(op_type, GF_AFR_REPLACE_BRICK))
+ return AFR_MSG_REPLACE_BRICK_STATUS;
+ else if (!strcmp(op_type, GF_AFR_ADD_BRICK))
+ return AFR_MSG_ADD_BRICK_STATUS;
+ return -1;
}
int
-afr_fav_child_reset_sink_xattrs_cbk (int ret, call_frame_t *heal_frame,
- void *opaque)
+afr_fav_child_reset_sink_xattrs_cbk(int ret, call_frame_t *heal_frame,
+ void *opaque)
{
+ call_frame_t *txn_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *heal_local = NULL;
+ xlator_t *this = NULL;
- call_frame_t *txn_frame = NULL;
- afr_local_t *local = NULL;
- afr_local_t *heal_local = NULL;
- xlator_t *this = NULL;
-
- heal_local = heal_frame->local;
- txn_frame = heal_local->heal_frame;
- local = txn_frame->local;
- this = txn_frame->this;
+ heal_local = heal_frame->local;
+ txn_frame = heal_local->heal_frame;
+ local = txn_frame->local;
+ this = txn_frame->this;
- /* Refresh the inode agan and proceed with the transaction.*/
- afr_inode_refresh (txn_frame, this, local->inode, NULL,
- local->refreshfn);
+ /* Refresh the inode agan and proceed with the transaction.*/
+ afr_inode_refresh(txn_frame, this, local->inode, NULL, local->refreshfn);
- AFR_STACK_DESTROY (heal_frame);
+ AFR_STACK_DESTROY(heal_frame);
- return 0;
+ return 0;
}
int
-afr_fav_child_reset_sink_xattrs (void *opaque)
-{
- call_frame_t *heal_frame = NULL;
- call_frame_t *txn_frame = NULL;
- xlator_t *this = NULL;
- gf_boolean_t d_spb = _gf_false;
- gf_boolean_t m_spb = _gf_false;
- afr_local_t *heal_local = NULL;
- afr_local_t *txn_local = NULL;
- afr_private_t *priv = NULL;
- inode_t *inode = NULL;
- unsigned char *locked_on = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
- int ret = 0;
-
- heal_frame = (call_frame_t *) opaque;
- heal_local = heal_frame->local;
- txn_frame = heal_local->heal_frame;
- txn_local = txn_frame->local;
- this = txn_frame->this;
- inode = txn_local->inode;
- priv = this->private;
- locked_on = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = _afr_is_split_brain (txn_frame, this, txn_local->replies,
- AFR_DATA_TRANSACTION, &d_spb);
-
- ret = _afr_is_split_brain (txn_frame, this, txn_local->replies,
- AFR_METADATA_TRANSACTION, &m_spb);
-
- /* Take appropriate locks and reset sink xattrs. */
- if (d_spb) {
- ret = afr_selfheal_inodelk (heal_frame, this, inode, this->name,
- 0, 0, locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS)
- goto data_unlock;
- ret = __afr_selfheal_data_prepare (heal_frame, this,
- inode, locked_on,
- sources, sinks,
- healed_sinks,
- undid_pending,
- locked_replies,
- NULL);
- }
-data_unlock:
- afr_selfheal_uninodelk (heal_frame, this, inode, this->name,
- 0, 0, locked_on);
- }
-
- if (m_spb) {
- memset (locked_on, 0, sizeof (*locked_on) * priv->child_count);
- memset (undid_pending, 0,
- sizeof (*undid_pending) * priv->child_count);
- ret = afr_selfheal_inodelk (heal_frame, this, inode, this->name,
- LLONG_MAX-1, 0, locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS)
- goto mdata_unlock;
- ret = __afr_selfheal_metadata_prepare (heal_frame, this,
- inode, locked_on,
- sources, sinks,
- healed_sinks,
- undid_pending,
- locked_replies,
- NULL);
-
- }
-mdata_unlock:
- afr_selfheal_uninodelk (heal_frame, this, inode, this->name,
- LLONG_MAX-1, 0, locked_on);
+afr_fav_child_reset_sink_xattrs(void *opaque)
+{
+ call_frame_t *heal_frame = NULL;
+ call_frame_t *txn_frame = NULL;
+ xlator_t *this = NULL;
+ gf_boolean_t d_spb = _gf_false;
+ gf_boolean_t m_spb = _gf_false;
+ afr_local_t *heal_local = NULL;
+ afr_local_t *txn_local = NULL;
+ afr_private_t *priv = NULL;
+ inode_t *inode = NULL;
+ unsigned char *locked_on = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+ int ret = 0;
+
+ heal_frame = (call_frame_t *)opaque;
+ heal_local = heal_frame->local;
+ txn_frame = heal_local->heal_frame;
+ txn_local = txn_frame->local;
+ this = txn_frame->this;
+ inode = txn_local->inode;
+ priv = this->private;
+ locked_on = alloca0(priv->child_count);
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = _afr_is_split_brain(txn_frame, this, txn_local->replies,
+ AFR_DATA_TRANSACTION, &d_spb);
+
+ ret = _afr_is_split_brain(txn_frame, this, txn_local->replies,
+ AFR_METADATA_TRANSACTION, &m_spb);
+
+ /* Take appropriate locks and reset sink xattrs. */
+ if (d_spb) {
+ ret = afr_selfheal_inodelk(heal_frame, this, inode, this->name, 0, 0,
+ locked_on);
+ {
+ if (ret < AFR_SH_MIN_PARTICIPANTS)
+ goto data_unlock;
+ ret = __afr_selfheal_data_prepare(
+ heal_frame, this, inode, locked_on, sources, sinks,
+ healed_sinks, undid_pending, locked_replies, NULL);
+ }
+ data_unlock:
+ afr_selfheal_uninodelk(heal_frame, this, inode, this->name, 0, 0,
+ locked_on);
+ }
+
+ if (m_spb) {
+ memset(locked_on, 0, sizeof(*locked_on) * priv->child_count);
+ memset(undid_pending, 0, sizeof(*undid_pending) * priv->child_count);
+ ret = afr_selfheal_inodelk(heal_frame, this, inode, this->name,
+ LLONG_MAX - 1, 0, locked_on);
+ {
+ if (ret < AFR_SH_MIN_PARTICIPANTS)
+ goto mdata_unlock;
+ ret = __afr_selfheal_metadata_prepare(
+ heal_frame, this, inode, locked_on, sources, sinks,
+ healed_sinks, undid_pending, locked_replies, NULL);
}
+ mdata_unlock:
+ afr_selfheal_uninodelk(heal_frame, this, inode, this->name,
+ LLONG_MAX - 1, 0, locked_on);
+ }
- return ret;
-
+ return ret;
}
/*
* Concatenates the xattrs in local->replies separated by a delimiter.
*/
int
-afr_serialize_xattrs_with_delimiter (call_frame_t *frame, xlator_t *this,
- char *buf, const char *default_str,
- int32_t *serz_len, char delimiter)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- char *xattr = NULL;
- int i = 0;
- int len = 0;
- size_t str_len = 0;
- int ret = -1;
-
- priv = this->private;
- local = frame->local;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid || local->replies[i].op_ret) {
- str_len = strlen (default_str);
- buf = strncat (buf, default_str, str_len);
- len += str_len;
- buf[len++] = delimiter;
- buf[len] = '\0';
- } else {
- ret = dict_get_str (local->replies[i].xattr,
- local->cont.getxattr.name, &xattr);
- if (ret) {
- gf_msg ("TEST", GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_GET_FAILED,
- "Failed to get the node_uuid of brick "
- "%d", i);
- goto out;
- }
- str_len = strlen (xattr);
- buf = strncat (buf, xattr, str_len);
- len += str_len;
- buf[len++] = delimiter;
- buf[len] = '\0';
- }
- }
- buf[--len] = '\0'; /*remove the last delimiter*/
- if (serz_len)
- *serz_len = ++len;
- ret = 0;
+afr_serialize_xattrs_with_delimiter(call_frame_t *frame, xlator_t *this,
+ char *buf, const char *default_str,
+ int32_t *serz_len, char delimiter)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ char *xattr = NULL;
+ int i = 0;
+ int len = 0;
+ size_t str_len = 0;
+ int ret = -1;
+
+ priv = this->private;
+ local = frame->local;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid || local->replies[i].op_ret) {
+ str_len = strlen(default_str);
+ buf = strncat(buf, default_str, str_len);
+ len += str_len;
+ buf[len++] = delimiter;
+ buf[len] = '\0';
+ } else {
+ ret = dict_get_str(local->replies[i].xattr,
+ local->cont.getxattr.name, &xattr);
+ if (ret) {
+ gf_msg("TEST", GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "Failed to get the node_uuid of brick "
+ "%d",
+ i);
+ goto out;
+ }
+ str_len = strlen(xattr);
+ buf = strncat(buf, xattr, str_len);
+ len += str_len;
+ buf[len++] = delimiter;
+ buf[len] = '\0';
+ }
+ }
+ buf[--len] = '\0'; /*remove the last delimiter*/
+ if (serz_len)
+ *serz_len = ++len;
+ ret = 0;
out:
- return ret;
+ return ret;
}
uint64_t
-afr_write_subvol_get (call_frame_t *frame, xlator_t *this)
+afr_write_subvol_get(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- uint64_t write_subvol = 0;
+ afr_local_t *local = NULL;
+ uint64_t write_subvol = 0;
- local = frame->local;
- LOCK(&local->inode->lock);
- write_subvol = local->inode_ctx->write_subvol;
- UNLOCK (&local->inode->lock);
+ local = frame->local;
+ LOCK(&local->inode->lock);
+ write_subvol = local->inode_ctx->write_subvol;
+ UNLOCK(&local->inode->lock);
- return write_subvol;
+ return write_subvol;
}
int
-afr_write_subvol_set (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- unsigned char *data_accused = NULL;
- unsigned char *metadata_accused = NULL;
- unsigned char *data_readable = NULL;
- unsigned char *metadata_readable = NULL;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint64_t val = 0;
- int event = 0;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
- data_accused = alloca0 (priv->child_count);
- metadata_accused = alloca0 (priv->child_count);
- data_readable = alloca0 (priv->child_count);
- metadata_readable = alloca0 (priv->child_count);
- event = local->event_generation;
-
- afr_readables_fill (frame, this, local->inode, data_accused,
- metadata_accused, data_readable, metadata_readable,
- NULL);
-
- for (i = 0; i < priv->child_count; i++) {
- if (data_readable[i])
- datamap |= (1 << i);
- if (metadata_readable[i])
- metadatamap |= (1 << i);
- }
-
- val = ((uint64_t) metadatamap) |
- (((uint64_t) datamap) << 16) |
- (((uint64_t) event) << 32);
-
- LOCK(&local->inode->lock);
- {
- if (local->inode_ctx->write_subvol == 0 &&
- local->transaction.type == AFR_DATA_TRANSACTION) {
- local->inode_ctx->write_subvol = val;
- }
- }
- UNLOCK (&local->inode->lock);
-
- return 0;
+afr_write_subvol_set(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ unsigned char *data_accused = NULL;
+ unsigned char *metadata_accused = NULL;
+ unsigned char *data_readable = NULL;
+ unsigned char *metadata_readable = NULL;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint64_t val = 0;
+ int event = 0;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+ data_accused = alloca0(priv->child_count);
+ metadata_accused = alloca0(priv->child_count);
+ data_readable = alloca0(priv->child_count);
+ metadata_readable = alloca0(priv->child_count);
+ event = local->event_generation;
+
+ afr_readables_fill(frame, this, local->inode, data_accused,
+ metadata_accused, data_readable, metadata_readable,
+ NULL);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (data_readable[i])
+ datamap |= (1 << i);
+ if (metadata_readable[i])
+ metadatamap |= (1 << i);
+ }
+
+ val = ((uint64_t)metadatamap) | (((uint64_t)datamap) << 16) |
+ (((uint64_t)event) << 32);
+
+ LOCK(&local->inode->lock);
+ {
+ if (local->inode_ctx->write_subvol == 0 &&
+ local->transaction.type == AFR_DATA_TRANSACTION) {
+ local->inode_ctx->write_subvol = val;
+ }
+ }
+ UNLOCK(&local->inode->lock);
+
+ return 0;
}
int
-afr_write_subvol_reset (call_frame_t *frame, xlator_t *this)
+afr_write_subvol_reset(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- LOCK(&local->inode->lock);
- {
- GF_ASSERT (local->inode_ctx->lock_count > 0);
- local->inode_ctx->lock_count--;
+ local = frame->local;
+ LOCK(&local->inode->lock);
+ {
+ GF_ASSERT(local->inode_ctx->lock_count > 0);
+ local->inode_ctx->lock_count--;
- if (!local->inode_ctx->lock_count)
- local->inode_ctx->write_subvol = 0;
- }
- UNLOCK(&local->inode->lock);
+ if (!local->inode_ctx->lock_count)
+ local->inode_ctx->write_subvol = 0;
+ }
+ UNLOCK(&local->inode->lock);
- return 0;
+ return 0;
}
int
-afr_set_inode_local (xlator_t *this, afr_local_t *local, inode_t *inode)
+afr_set_inode_local(xlator_t *this, afr_local_t *local, inode_t *inode)
{
- int ret = 0;
+ int ret = 0;
- local->inode = inode_ref (inode);
- LOCK(&local->inode->lock);
- {
- ret = __afr_inode_ctx_get (this, local->inode,
- &local->inode_ctx);
- }
- UNLOCK (&local->inode->lock);
- if (ret < 0) {
- gf_msg_callingfn (this->name, GF_LOG_ERROR, ENOMEM,
- AFR_MSG_INODE_CTX_GET_FAILED,
- "Error getting inode ctx %s",
- uuid_utoa (local->inode->gfid));
- }
- return ret;
+ local->inode = inode_ref(inode);
+ LOCK(&local->inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, local->inode, &local->inode_ctx);
+ }
+ UNLOCK(&local->inode->lock);
+ if (ret < 0) {
+ gf_msg_callingfn(
+ this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_INODE_CTX_GET_FAILED,
+ "Error getting inode ctx %s", uuid_utoa(local->inode->gfid));
+ }
+ return ret;
}
gf_boolean_t
-afr_ta_is_fop_called_from_synctask (xlator_t *this)
+afr_ta_is_fop_called_from_synctask(xlator_t *this)
{
- struct synctask *task = NULL;
- gf_lkowner_t tmp_owner = {0,};
+ struct synctask *task = NULL;
+ gf_lkowner_t tmp_owner = {
+ 0,
+ };
- task = synctask_get ();
- if (!task)
- return _gf_false;
+ task = synctask_get();
+ if (!task)
+ return _gf_false;
- set_lk_owner_from_ptr(&tmp_owner, (void *)this);
+ set_lk_owner_from_ptr(&tmp_owner, (void *)this);
- if (!is_same_lkowner (&tmp_owner, &task->frame->root->lk_owner))
- return _gf_false;
+ if (!is_same_lkowner(&tmp_owner, &task->frame->root->lk_owner))
+ return _gf_false;
- return _gf_true;
+ return _gf_true;
}
int
-afr_ta_post_op_lock (xlator_t *this, loc_t *loc)
-{
- /*Note: At any given time, only one instance of this function must
- * be in progress.*/
-
- int ret = 0;
- uuid_t gfid = {0,};
- afr_private_t *priv = this->private;
- gf_boolean_t locked = _gf_false;
- struct gf_flock flock1 = {0, };
- struct gf_flock flock2 = {0, };
- int32_t cmd = 0;
-
- GF_ASSERT (afr_ta_is_fop_called_from_synctask (this));
- flock1.l_type = F_WRLCK;
-
- while (!locked) {
- if (priv->shd.iamshd) {
- cmd = F_SETLKW;
- flock1.l_start = 0;
- flock1.l_len = 0;
-
- } else {
- cmd = F_SETLK;
- if (priv->ta_notify_dom_lock_offset) {
- flock1.l_start =
- priv->ta_notify_dom_lock_offset;
- } else {
- gf_uuid_generate (gfid);
- flock1.l_start = gfid_to_ino (gfid);
- if (flock1.l_start < 0)
- flock1.l_start = -flock1.l_start;
- }
- flock1.l_len = 1;
- }
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_NOTIFY, loc, cmd, &flock1,
- NULL, NULL);
- if (!ret) {
- locked = _gf_true;
- priv->ta_notify_dom_lock_offset = flock1.l_start;
- } else if (ret == -EAGAIN) {
- continue;
- } else {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_THIN_ARB, "Failed to get "
- "AFR_TA_DOM_NOTIFY lock on %s.", loc->name);
- goto out;
- }
- }
+afr_ta_post_op_lock(xlator_t *this, loc_t *loc)
+{
+ /*Note: At any given time, only one instance of this function must
+ * be in progress.*/
+
+ int ret = 0;
+ uuid_t gfid = {
+ 0,
+ };
+ afr_private_t *priv = this->private;
+ gf_boolean_t locked = _gf_false;
+ struct gf_flock flock1 = {
+ 0,
+ };
+ struct gf_flock flock2 = {
+ 0,
+ };
+ int32_t cmd = 0;
+
+ GF_ASSERT(afr_ta_is_fop_called_from_synctask(this));
+ flock1.l_type = F_WRLCK;
+
+ while (!locked) {
+ if (priv->shd.iamshd) {
+ cmd = F_SETLKW;
+ flock1.l_start = 0;
+ flock1.l_len = 0;
- flock2.l_type = F_WRLCK;
- flock2.l_start = 0;
- flock2.l_len = 0;
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_MODIFY, loc, F_SETLKW, &flock2,
- NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed to get AFR_TA_DOM_MODIFY lock.");
- flock1.l_type = F_UNLCK;
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_NOTIFY, loc, F_SETLK, &flock1,
- NULL, NULL);
+ } else {
+ cmd = F_SETLK;
+ if (priv->ta_notify_dom_lock_offset) {
+ flock1.l_start = priv->ta_notify_dom_lock_offset;
+ } else {
+ gf_uuid_generate(gfid);
+ flock1.l_start = gfid_to_ino(gfid);
+ if (flock1.l_start < 0)
+ flock1.l_start = -flock1.l_start;
+ }
+ flock1.l_len = 1;
}
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, loc, cmd, &flock1, NULL, NULL);
+ if (!ret) {
+ locked = _gf_true;
+ priv->ta_notify_dom_lock_offset = flock1.l_start;
+ } else if (ret == -EAGAIN) {
+ continue;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to get "
+ "AFR_TA_DOM_NOTIFY lock on %s.",
+ loc->name);
+ goto out;
+ }
+ }
+
+ flock2.l_type = F_WRLCK;
+ flock2.l_start = 0;
+ flock2.l_len = 0;
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_MODIFY, loc, F_SETLKW, &flock2, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to get AFR_TA_DOM_MODIFY lock.");
+ flock1.l_type = F_UNLCK;
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, loc, F_SETLK, &flock1, NULL,
+ NULL);
+ }
out:
- return ret;
+ return ret;
}
int
-afr_ta_post_op_unlock (xlator_t *this, loc_t *loc)
-{
- afr_private_t *priv = this->private;
- struct gf_flock flock = {0, };
- int ret = 0;
-
- GF_ASSERT (afr_ta_is_fop_called_from_synctask (this));
- flock.l_type = F_UNLCK;
- flock.l_start = 0;
- flock.l_len = 0;
-
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_MODIFY, loc, F_SETLK, &flock, NULL,
- NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed to unlock AFR_TA_DOM_MODIFY lock.");
- goto out;
- }
-
- if (!priv->shd.iamshd)
- /* Mounts (clients) will not release the AFR_TA_DOM_NOTIFY lock
- * in post-op as they use it as a notification mechanism. When
- * shd sends a lock request on TA during heal, the clients will
- * receive a lock-contention upcall notification upon which they
- * will release the AFR_TA_DOM_NOTIFY lock after completing the
- * in flight I/O.*/
- goto out;
-
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_NOTIFY, loc, F_SETLK, &flock,
- NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed to unlock AFR_TA_DOM_NOTIFY lock.");
- }
+afr_ta_post_op_unlock(xlator_t *this, loc_t *loc)
+{
+ afr_private_t *priv = this->private;
+ struct gf_flock flock = {
+ 0,
+ };
+ int ret = 0;
+
+ GF_ASSERT(afr_ta_is_fop_called_from_synctask(this));
+ flock.l_type = F_UNLCK;
+ flock.l_start = 0;
+ flock.l_len = 0;
+
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_MODIFY, loc, F_SETLK, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to unlock AFR_TA_DOM_MODIFY lock.");
+ goto out;
+ }
+
+ if (!priv->shd.iamshd)
+ /* Mounts (clients) will not release the AFR_TA_DOM_NOTIFY lock
+ * in post-op as they use it as a notification mechanism. When
+ * shd sends a lock request on TA during heal, the clients will
+ * receive a lock-contention upcall notification upon which they
+ * will release the AFR_TA_DOM_NOTIFY lock after completing the
+ * in flight I/O.*/
+ goto out;
+
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, loc, F_SETLK, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to unlock AFR_TA_DOM_NOTIFY lock.");
+ }
out:
- return ret;
+ return ret;
}
call_frame_t *
-afr_ta_frame_create (xlator_t *this)
+afr_ta_frame_create(xlator_t *this)
{
- call_frame_t *frame = NULL;
- void *lk_owner = NULL;
+ call_frame_t *frame = NULL;
+ void *lk_owner = NULL;
- frame = create_frame (this, this->ctx->pool);
- if (!frame)
- return NULL;
- lk_owner = (void *)this;
- afr_set_lk_owner (frame, this, lk_owner);
- return frame;
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame)
+ return NULL;
+ lk_owner = (void *)this;
+ afr_set_lk_owner(frame, this, lk_owner);
+ return frame;
}
diff --git a/xlators/cluster/afr/src/afr-dir-read.c b/xlators/cluster/afr/src/afr-dir-read.c
index 76d0cdfa826..4c40e85f393 100644
--- a/xlators/cluster/afr/src/afr-dir-read.c
+++ b/xlators/cluster/afr/src/afr-dir-read.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include <libgen.h>
#include <unistd.h>
#include <fnmatch.h>
@@ -33,327 +32,309 @@
#include "afr.h"
#include "afr-transaction.h"
-
int32_t
-afr_opendir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd, dict_t *xdata)
+afr_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
- int32_t child_index = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = frame->local;
- fd_ctx = local->fd_ctx;
- child_index = (long) cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
- } else {
- local->op_ret = op_ret;
- fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int32_t child_index = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ local = frame->local;
+ fd_ctx = local->fd_ctx;
+ child_index = (long)cookie;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+ } else {
+ local->op_ret = op_ret;
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
}
- UNLOCK (&frame->lock);
+ }
+ UNLOCK(&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd, NULL);
- return 0;
+ if (call_count == 0)
+ AFR_STACK_UNWIND(opendir, frame, local->op_ret, local->op_errno,
+ local->fd, NULL);
+ return 0;
}
-
int
-afr_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+afr_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- int i = 0;
- int call_count = -1;
- int32_t op_errno = ENOMEM;
- afr_fd_ctx_t *fd_ctx = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = -1;
+ int32_t op_errno = ENOMEM;
+ afr_fd_ctx_t *fd_ctx = NULL;
- priv = this->private;
+ priv = this->private;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_OPENDIR;
- if (!afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
+ local->op = GF_FOP_OPENDIR;
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx)
- goto out;
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx)
+ goto out;
- loc_copy (&local->loc, loc);
+ loc_copy(&local->loc, loc);
- local->fd = fd_ref (fd);
- local->fd_ctx = fd_ctx;
+ local->fd = fd_ref(fd);
+ local->fd_ctx = fd_ctx;
- call_count = local->call_count;
+ call_count = local->call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_opendir_cbk,
- (void*) (long) i,
- priv->children[i],
- priv->children[i]->fops->opendir,
- loc, fd, NULL);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, afr_opendir_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->opendir, loc, fd, NULL);
- if (!--call_count)
- break;
- }
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (opendir, frame, -1, op_errno, fd, NULL);
- return 0;
+ AFR_STACK_UNWIND(opendir, frame, -1, op_errno, fd, NULL);
+ return 0;
}
static int
-afr_validate_read_subvol (inode_t *inode, xlator_t *this, int par_read_subvol)
+afr_validate_read_subvol(inode_t *inode, xlator_t *this, int par_read_subvol)
{
- int gen = 0;
- int entry_read_subvol = 0;
- unsigned char *data_readable = NULL;
- unsigned char *metadata_readable = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- data_readable = alloca0 (priv->child_count);
- metadata_readable = alloca0 (priv->child_count);
-
- afr_inode_read_subvol_get (inode, this, data_readable,
- metadata_readable, &gen);
-
- if (gen != priv->event_generation ||
- !data_readable[par_read_subvol] ||
- !metadata_readable[par_read_subvol])
- return -1;
-
- /* Once the control reaches the following statement, it means that the
- * parent's read subvol is perfectly readable. So calling
- * either afr_data_subvol_get() or afr_metadata_subvol_get() would
- * yield the same result. Hence, choosing afr_data_subvol_get() below.
- */
-
- if (!priv->consistent_metadata)
- return 0;
-
- /* For an inode fetched through readdirp which is yet to be linked,
- * inode ctx would not be initialised (yet). So this function returns
- * -1 above due to gen being 0, which is why it is OK to pass NULL for
- * read_subvol_args here.
- */
- entry_read_subvol = afr_data_subvol_get (inode, this, NULL, NULL,
- NULL, NULL);
- if (entry_read_subvol != par_read_subvol)
- return -1;
-
+ int gen = 0;
+ int entry_read_subvol = 0;
+ unsigned char *data_readable = NULL;
+ unsigned char *metadata_readable = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ data_readable = alloca0(priv->child_count);
+ metadata_readable = alloca0(priv->child_count);
+
+ afr_inode_read_subvol_get(inode, this, data_readable, metadata_readable,
+ &gen);
+
+ if (gen != priv->event_generation || !data_readable[par_read_subvol] ||
+ !metadata_readable[par_read_subvol])
+ return -1;
+
+ /* Once the control reaches the following statement, it means that the
+ * parent's read subvol is perfectly readable. So calling
+ * either afr_data_subvol_get() or afr_metadata_subvol_get() would
+ * yield the same result. Hence, choosing afr_data_subvol_get() below.
+ */
+
+ if (!priv->consistent_metadata)
return 0;
+ /* For an inode fetched through readdirp which is yet to be linked,
+ * inode ctx would not be initialised (yet). So this function returns
+ * -1 above due to gen being 0, which is why it is OK to pass NULL for
+ * read_subvol_args here.
+ */
+ entry_read_subvol = afr_data_subvol_get(inode, this, NULL, NULL, NULL,
+ NULL);
+ if (entry_read_subvol != par_read_subvol)
+ return -1;
+
+ return 0;
}
static void
-afr_readdir_transform_entries (gf_dirent_t *subvol_entries, int subvol,
- gf_dirent_t *entries, fd_t *fd)
+afr_readdir_transform_entries(gf_dirent_t *subvol_entries, int subvol,
+ gf_dirent_t *entries, fd_t *fd)
{
- int ret = -1;
- gf_dirent_t *entry = NULL;
- gf_dirent_t *tmp = NULL;
- xlator_t *this = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t need_heal = _gf_false;
- gf_boolean_t validate_subvol = _gf_false;
-
- this = THIS;
- priv = this->private;
-
- need_heal = afr_get_need_heal (this);
- validate_subvol = need_heal | priv->consistent_metadata;
-
- list_for_each_entry_safe (entry, tmp, &subvol_entries->list, list) {
- if (__is_root_gfid (fd->inode->gfid) &&
- !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
- continue;
- }
-
- list_del_init (&entry->list);
- list_add_tail (&entry->list, &entries->list);
-
- if (!validate_subvol)
- continue;
-
- if (entry->inode) {
- ret = afr_validate_read_subvol (entry->inode, this,
- subvol);
- if (ret == -1) {
- inode_unref (entry->inode);
- entry->inode = NULL;
- continue;
- }
- }
+ int ret = -1;
+ gf_dirent_t *entry = NULL;
+ gf_dirent_t *tmp = NULL;
+ xlator_t *this = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t need_heal = _gf_false;
+ gf_boolean_t validate_subvol = _gf_false;
+
+ this = THIS;
+ priv = this->private;
+
+ need_heal = afr_get_need_heal(this);
+ validate_subvol = need_heal | priv->consistent_metadata;
+
+ list_for_each_entry_safe(entry, tmp, &subvol_entries->list, list)
+ {
+ if (__is_root_gfid(fd->inode->gfid) &&
+ !strcmp(entry->d_name, GF_REPLICATE_TRASH_DIR)) {
+ continue;
}
-}
+ list_del_init(&entry->list);
+ list_add_tail(&entry->list, &entries->list);
+
+ if (!validate_subvol)
+ continue;
+
+ if (entry->inode) {
+ ret = afr_validate_read_subvol(entry->inode, this, subvol);
+ if (ret == -1) {
+ inode_unref(entry->inode);
+ entry->inode = NULL;
+ continue;
+ }
+ }
+ }
+}
int32_t
-afr_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *subvol_entries,
- dict_t *xdata)
+afr_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *subvol_entries,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- gf_dirent_t entries;
+ afr_local_t *local = NULL;
+ gf_dirent_t entries;
- INIT_LIST_HEAD (&entries.list);
+ INIT_LIST_HEAD(&entries.list);
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0 && !local->cont.readdir.offset) {
- /* failover only if this was first readdir, detected
- by offset == 0 */
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0 && !local->cont.readdir.offset) {
+ /* failover only if this was first readdir, detected
+ by offset == 0 */
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- if (op_ret >= 0)
- afr_readdir_transform_entries (subvol_entries, (long) cookie,
- &entries, local->fd);
+ if (op_ret >= 0)
+ afr_readdir_transform_entries(subvol_entries, (long)cookie, &entries,
+ local->fd);
- AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries, xdata);
+ AFR_STACK_UNWIND(readdir, frame, op_ret, op_errno, &entries, xdata);
- gf_dirent_free (&entries);
+ gf_dirent_free(&entries);
- return 0;
+ return 0;
}
-
int
-afr_readdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_readdir_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- priv = this->private;
- local = frame->local;
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- if (!fd_ctx) {
- local->op_errno = EINVAL;
- local->op_ret = -1;
- }
-
- if (subvol == -1 || !fd_ctx) {
- AFR_STACK_UNWIND (readdir, frame, local->op_ret,
- local->op_errno, 0, 0);
- return 0;
- }
-
- fd_ctx->readdir_subvol = subvol;
-
- if (local->op == GF_FOP_READDIR)
- STACK_WIND_COOKIE (frame, afr_readdir_cbk,
- (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->readdir,
- local->fd, local->cont.readdir.size,
- local->cont.readdir.offset,
- local->xdata_req);
- else
- STACK_WIND_COOKIE (frame, afr_readdir_cbk,
- (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->readdirp,
- local->fd, local->cont.readdir.size,
- local->cont.readdir.offset,
- local->xdata_req);
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ priv = this->private;
+ local = frame->local;
+ fd_ctx = afr_fd_ctx_get(local->fd, this);
+ if (!fd_ctx) {
+ local->op_errno = EINVAL;
+ local->op_ret = -1;
+ }
+
+ if (subvol == -1 || !fd_ctx) {
+ AFR_STACK_UNWIND(readdir, frame, local->op_ret, local->op_errno, 0, 0);
+ return 0;
+ }
+
+ fd_ctx->readdir_subvol = subvol;
+
+ if (local->op == GF_FOP_READDIR)
+ STACK_WIND_COOKIE(frame, afr_readdir_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->readdir, local->fd,
+ local->cont.readdir.size, local->cont.readdir.offset,
+ local->xdata_req);
+ else
+ STACK_WIND_COOKIE(frame, afr_readdir_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->readdirp, local->fd,
+ local->cont.readdir.size, local->cont.readdir.offset,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, int whichop, dict_t *dict)
+afr_do_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, int whichop, dict_t *dict)
{
- afr_local_t *local = NULL;
- int32_t op_errno = 0;
- int subvol = -1;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx) {
- op_errno = EINVAL;
- goto out;
- }
-
- local->op = whichop;
- local->fd = fd_ref (fd);
- local->cont.readdir.size = size;
- local->cont.readdir.offset = offset;
- local->xdata_req = (dict)? dict_ref (dict) : NULL;
-
- subvol = fd_ctx->readdir_subvol;
-
- if (offset == 0 || subvol == -1) {
- /* First readdir has option of failing over and selecting
- an appropriate read subvolume */
- afr_read_txn (frame, this, fd->inode, afr_readdir_wind,
- AFR_DATA_TRANSACTION);
- } else {
- /* But continued readdirs MUST stick to the same subvolume
- without an option to failover */
- afr_readdir_wind (frame, this, subvol);
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
+ int subvol = -1;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx) {
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ local->op = whichop;
+ local->fd = fd_ref(fd);
+ local->cont.readdir.size = size;
+ local->cont.readdir.offset = offset;
+ local->xdata_req = (dict) ? dict_ref(dict) : NULL;
+
+ subvol = fd_ctx->readdir_subvol;
+
+ if (offset == 0 || subvol == -1) {
+ /* First readdir has option of failing over and selecting
+ an appropriate read subvolume */
+ afr_read_txn(frame, this, fd->inode, afr_readdir_wind,
+ AFR_DATA_TRANSACTION);
+ } else {
+ /* But continued readdirs MUST stick to the same subvolume
+ without an option to failover */
+ afr_readdir_wind(frame, this, subvol);
+ }
+
+ return 0;
out:
- AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(readdir, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
-
int32_t
-afr_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *xdata)
+afr_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
+ afr_do_readdir(frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
- return 0;
+ return 0;
}
-
int32_t
-afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *dict)
+afr_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict)
{
- afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP, dict);
+ afr_do_readdir(frame, this, fd, size, offset, GF_FOP_READDIRP, dict);
- return 0;
+ return 0;
}
-
int32_t
-afr_releasedir (xlator_t *this, fd_t *fd)
+afr_releasedir(xlator_t *this, fd_t *fd)
{
- afr_cleanup_fd_ctx (this, fd);
+ afr_cleanup_fd_ctx(this, fd);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index 0e3ae18b46a..5725b1c5cb3 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include <libgen.h>
#include <unistd.h>
#include <fnmatch.h>
@@ -34,501 +33,470 @@
#include "afr-transaction.h"
void
-afr_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this);
+afr_mark_entry_pending_changelog(call_frame_t *frame, xlator_t *this);
int
-afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno)
+afr_build_parent_loc(loc_t *parent, loc_t *child, int32_t *op_errno)
{
- int ret = -1;
- char *child_path = NULL;
-
- if (!child->parent) {
- if (op_errno)
- *op_errno = EINVAL;
- goto out;
- }
-
- child_path = gf_strdup (child->path);
- if (!child_path) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
-
- parent->path = gf_strdup (dirname (child_path));
- if (!parent->path) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
-
- parent->inode = inode_ref (child->parent);
- gf_uuid_copy (parent->gfid, child->pargfid);
-
- ret = 0;
+ int ret = -1;
+ char *child_path = NULL;
+
+ if (!child->parent) {
+ if (op_errno)
+ *op_errno = EINVAL;
+ goto out;
+ }
+
+ child_path = gf_strdup(child->path);
+ if (!child_path) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ parent->path = gf_strdup(dirname(child_path));
+ if (!parent->path) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ parent->inode = inode_ref(child->parent);
+ gf_uuid_copy(parent->gfid, child->pargfid);
+
+ ret = 0;
out:
- GF_FREE (child_path);
+ GF_FREE(child_path);
- return ret;
+ return ret;
}
-
static void
-__afr_dir_write_finalize (call_frame_t *frame, xlator_t *this)
+__afr_dir_write_finalize(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int inode_read_subvol = -1;
- int parent_read_subvol = -1;
- int parent2_read_subvol = -1;
- int i = 0;
- afr_read_subvol_args_t args = {0,};
-
- local = frame->local;
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret == -1)
- continue;
- gf_uuid_copy (args.gfid, local->replies[i].poststat.ia_gfid);
- args.ia_type = local->replies[i].poststat.ia_type;
- break;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int inode_read_subvol = -1;
+ int parent_read_subvol = -1;
+ int parent2_read_subvol = -1;
+ int i = 0;
+ afr_read_subvol_args_t args = {
+ 0,
+ };
+
+ local = frame->local;
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret == -1)
+ continue;
+ gf_uuid_copy(args.gfid, local->replies[i].poststat.ia_gfid);
+ args.ia_type = local->replies[i].poststat.ia_type;
+ break;
+ }
+
+ if (local->inode) {
+ afr_replies_interpret(frame, this, local->inode, NULL);
+ inode_read_subvol = afr_data_subvol_get(local->inode, this, NULL, NULL,
+ NULL, &args);
+ }
+
+ if (local->parent)
+ parent_read_subvol = afr_data_subvol_get(local->parent, this, NULL,
+ local->readable, NULL, NULL);
+
+ if (local->parent2)
+ parent2_read_subvol = afr_data_subvol_get(local->parent2, this, NULL,
+ local->readable2, NULL, NULL);
+
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ afr_pick_error_xdata(local, priv, local->parent, local->readable,
+ local->parent2, local->readable2);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret < 0) {
+ if (local->inode)
+ afr_inode_event_gen_reset(local->inode, this);
+ if (local->parent)
+ afr_inode_event_gen_reset(local->parent, this);
+ if (local->parent2)
+ afr_inode_event_gen_reset(local->parent2, this);
+ continue;
}
- if (local->inode) {
- afr_replies_interpret (frame, this, local->inode, NULL);
- inode_read_subvol = afr_data_subvol_get (local->inode, this,
- NULL, NULL, NULL, &args);
- }
-
- if (local->parent)
- parent_read_subvol = afr_data_subvol_get (local->parent, this,
- NULL, local->readable, NULL, NULL);
-
- if (local->parent2)
- parent2_read_subvol = afr_data_subvol_get (local->parent2, this,
- NULL, local->readable2, NULL, NULL);
-
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
- afr_pick_error_xdata (local, priv, local->parent, local->readable,
- local->parent2, local->readable2);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret < 0) {
- if (local->inode)
- afr_inode_event_gen_reset (local->inode, this);
- if (local->parent)
- afr_inode_event_gen_reset (local->parent,
- this);
- if (local->parent2)
- afr_inode_event_gen_reset (local->parent2,
- this);
- continue;
- }
-
- if (local->op_ret == -1) {
- local->op_ret = local->replies[i].op_ret;
- local->op_errno = local->replies[i].op_errno;
-
- local->cont.dir_fop.buf =
- local->replies[i].poststat;
- local->cont.dir_fop.preparent =
- local->replies[i].preparent;
- local->cont.dir_fop.postparent =
- local->replies[i].postparent;
- local->cont.dir_fop.prenewparent =
- local->replies[i].preparent2;
- local->cont.dir_fop.postnewparent =
- local->replies[i].postparent2;
- if (local->xdata_rsp) {
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = NULL;
- }
-
- if (local->replies[i].xdata)
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- continue;
- }
-
- if (i == inode_read_subvol) {
- local->cont.dir_fop.buf =
- local->replies[i].poststat;
- if (local->replies[i].xdata) {
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- }
- }
-
- if (i == parent_read_subvol) {
- local->cont.dir_fop.preparent =
- local->replies[i].preparent;
- local->cont.dir_fop.postparent =
- local->replies[i].postparent;
- }
-
- if (i == parent2_read_subvol) {
- local->cont.dir_fop.prenewparent =
- local->replies[i].preparent2;
- local->cont.dir_fop.postnewparent =
- local->replies[i].postparent2;
- }
- }
+ if (local->op_ret == -1) {
+ local->op_ret = local->replies[i].op_ret;
+ local->op_errno = local->replies[i].op_errno;
+
+ local->cont.dir_fop.buf = local->replies[i].poststat;
+ local->cont.dir_fop.preparent = local->replies[i].preparent;
+ local->cont.dir_fop.postparent = local->replies[i].postparent;
+ local->cont.dir_fop.prenewparent = local->replies[i].preparent2;
+ local->cont.dir_fop.postnewparent = local->replies[i].postparent2;
+ if (local->xdata_rsp) {
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = NULL;
+ }
+
+ if (local->replies[i].xdata)
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ continue;
+ }
-}
+ if (i == inode_read_subvol) {
+ local->cont.dir_fop.buf = local->replies[i].poststat;
+ if (local->replies[i].xdata) {
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ }
+ }
+ if (i == parent_read_subvol) {
+ local->cont.dir_fop.preparent = local->replies[i].preparent;
+ local->cont.dir_fop.postparent = local->replies[i].postparent;
+ }
+
+ if (i == parent2_read_subvol) {
+ local->cont.dir_fop.prenewparent = local->replies[i].preparent2;
+ local->cont.dir_fop.postnewparent = local->replies[i].postparent2;
+ }
+ }
+}
static void
-__afr_dir_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
- int op_ret, int op_errno, struct iatt *poststat,
- struct iatt *preparent, struct iatt *postparent,
- struct iatt *preparent2, struct iatt *postparent2,
- dict_t *xdata)
+__afr_dir_write_fill(call_frame_t *frame, xlator_t *this, int child_index,
+ int op_ret, int op_errno, struct iatt *poststat,
+ struct iatt *preparent, struct iatt *postparent,
+ struct iatt *preparent2, struct iatt *postparent2,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = frame->local;
- fd_ctx = local->fd_ctx;
-
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
-
-
- if (op_ret >= 0) {
- if (poststat)
- local->replies[child_index].poststat = *poststat;
- if (preparent)
- local->replies[child_index].preparent = *preparent;
- if (postparent)
- local->replies[child_index].postparent = *postparent;
- if (preparent2)
- local->replies[child_index].preparent2 = *preparent2;
- if (postparent2)
- local->replies[child_index].postparent2 = *postparent2;
- if (fd_ctx)
- fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- } else {
- if (op_errno != ENOTEMPTY)
- afr_transaction_fop_failed (frame, this, child_index);
- if (fd_ctx)
- fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
- }
-
- return;
+ afr_local_t *local = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ local = frame->local;
+ fd_ctx = local->fd_ctx;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
+
+ if (op_ret >= 0) {
+ if (poststat)
+ local->replies[child_index].poststat = *poststat;
+ if (preparent)
+ local->replies[child_index].preparent = *preparent;
+ if (postparent)
+ local->replies[child_index].postparent = *postparent;
+ if (preparent2)
+ local->replies[child_index].preparent2 = *preparent2;
+ if (postparent2)
+ local->replies[child_index].postparent2 = *postparent2;
+ if (fd_ctx)
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+ } else {
+ if (op_errno != ENOTEMPTY)
+ afr_transaction_fop_failed(frame, this, child_index);
+ if (fd_ctx)
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+ }
+
+ return;
}
-
static int
-__afr_dir_write_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- struct iatt *preparent2, struct iatt *postparent2,
- dict_t *xdata)
+__afr_dir_write_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent,
+ struct iatt *preparent2, struct iatt *postparent2,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int child_index = (long) cookie;
- int call_count = -1;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- __afr_dir_write_fill (frame, this, child_index, op_ret,
- op_errno, buf, preparent, postparent,
- preparent2, postparent2, xdata);
- }
- UNLOCK (&frame->lock);
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- __afr_dir_write_finalize (frame, this);
-
- if (afr_txn_nothing_failed (frame, this)) {
- /*if it did pre-op, it will do post-op changing ctime*/
- if (priv->consistent_metadata &&
- afr_needs_changelog_update (local))
- afr_zero_fill_stat (local);
- local->transaction.unwind (frame, this);
- }
-
- afr_mark_entry_pending_changelog (frame, this);
-
- afr_transaction_resume (frame, this);
+ afr_local_t *local = NULL;
+ int child_index = (long)cookie;
+ int call_count = -1;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ LOCK(&frame->lock);
+ {
+ __afr_dir_write_fill(frame, this, child_index, op_ret, op_errno, buf,
+ preparent, postparent, preparent2, postparent2,
+ xdata);
+ }
+ UNLOCK(&frame->lock);
+ call_count = afr_frame_return(frame);
+
+ if (call_count == 0) {
+ __afr_dir_write_finalize(frame, this);
+
+ if (afr_txn_nothing_failed(frame, this)) {
+ /*if it did pre-op, it will do post-op changing ctime*/
+ if (priv->consistent_metadata && afr_needs_changelog_update(local))
+ afr_zero_fill_stat(local);
+ local->transaction.unwind(frame, this);
}
- return 0;
-}
+ afr_mark_entry_pending_changelog(frame, this);
+ afr_transaction_resume(frame, this);
+ }
+
+ return 0;
+}
int
-afr_mark_new_entry_changelog_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- dict_t *xattr, dict_t *xdata)
+afr_mark_new_entry_changelog_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ dict_t *xattr, dict_t *xdata)
{
- int call_count = 0;
+ int call_count = 0;
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- AFR_STACK_DESTROY (frame);
+ if (call_count == 0)
+ AFR_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
-
void
-afr_mark_new_entry_changelog (call_frame_t *frame, xlator_t *this)
+afr_mark_new_entry_changelog(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *new_frame = NULL;
- afr_local_t *local = NULL;
- afr_local_t *new_local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- int32_t **changelog = NULL;
- int i = 0;
- int op_errno = ENOMEM;
- unsigned char *pending = NULL;
- int call_count = 0;
-
- local = frame->local;
- priv = this->private;
-
- new_frame = copy_frame (frame);
- if (!new_frame)
- goto out;
-
- new_local = AFR_FRAME_INIT (new_frame, op_errno);
- if (!new_local)
- goto out;
-
- xattr = dict_new ();
- if (!xattr)
- goto out;
-
- pending = alloca0 (priv->child_count);
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] &&
- !local->transaction.failed_subvols[i]) {
- call_count ++;
- continue;
- }
- pending[i] = 1;
- }
-
- changelog = afr_mark_pending_changelog (priv, pending, xattr,
- local->cont.dir_fop.buf.ia_type);
- if (!changelog)
- goto out;
-
- new_local->pending = changelog;
- gf_uuid_copy (new_local->loc.gfid, local->cont.dir_fop.buf.ia_gfid);
- new_local->loc.inode = inode_ref (local->inode);
-
- new_local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (pending[i])
- continue;
-
- STACK_WIND_COOKIE (new_frame, afr_mark_new_entry_changelog_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->xattrop,
- &new_local->loc, GF_XATTROP_ADD_ARRAY,
- xattr, NULL);
- if (!--call_count)
- break;
+ call_frame_t *new_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *new_local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int32_t **changelog = NULL;
+ int i = 0;
+ int op_errno = ENOMEM;
+ unsigned char *pending = NULL;
+ int call_count = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ new_frame = copy_frame(frame);
+ if (!new_frame)
+ goto out;
+
+ new_local = AFR_FRAME_INIT(new_frame, op_errno);
+ if (!new_local)
+ goto out;
+
+ xattr = dict_new();
+ if (!xattr)
+ goto out;
+
+ pending = alloca0(priv->child_count);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] &&
+ !local->transaction.failed_subvols[i]) {
+ call_count++;
+ continue;
}
+ pending[i] = 1;
+ }
+
+ changelog = afr_mark_pending_changelog(priv, pending, xattr,
+ local->cont.dir_fop.buf.ia_type);
+ if (!changelog)
+ goto out;
+
+ new_local->pending = changelog;
+ gf_uuid_copy(new_local->loc.gfid, local->cont.dir_fop.buf.ia_gfid);
+ new_local->loc.inode = inode_ref(local->inode);
+
+ new_local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (pending[i])
+ continue;
- new_frame = NULL;
+ STACK_WIND_COOKIE(new_frame, afr_mark_new_entry_changelog_cbk,
+ (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->xattrop, &new_local->loc,
+ GF_XATTROP_ADD_ARRAY, xattr, NULL);
+ if (!--call_count)
+ break;
+ }
+
+ new_frame = NULL;
out:
- if (new_frame)
- AFR_STACK_DESTROY (new_frame);
- if (xattr)
- dict_unref (xattr);
- return;
+ if (new_frame)
+ AFR_STACK_DESTROY(new_frame);
+ if (xattr)
+ dict_unref(xattr);
+ return;
}
-
void
-afr_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this)
+afr_mark_entry_pending_changelog(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int pre_op_count = 0;
- int failed_count = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int pre_op_count = 0;
+ int failed_count = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (local->op_ret < 0)
- return;
+ if (local->op_ret < 0)
+ return;
- if (local->op != GF_FOP_CREATE && local->op != GF_FOP_MKNOD &&
- local->op != GF_FOP_MKDIR)
- return;
+ if (local->op != GF_FOP_CREATE && local->op != GF_FOP_MKNOD &&
+ local->op != GF_FOP_MKDIR)
+ return;
- pre_op_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
- failed_count = AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count);
+ pre_op_count = AFR_COUNT(local->transaction.pre_op, priv->child_count);
+ failed_count = AFR_COUNT(local->transaction.failed_subvols,
+ priv->child_count);
- if (pre_op_count == priv->child_count && !failed_count)
- return;
+ if (pre_op_count == priv->child_count && !failed_count)
+ return;
- afr_mark_new_entry_changelog (frame, this);
+ afr_mark_new_entry_changelog(frame, this);
- return;
+ return;
}
-
/* {{{ create */
int
-afr_create_unwind (call_frame_t *frame, xlator_t *this)
+afr_create_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
+ local = frame->local;
- if (!main_frame)
- return 0;
+ main_frame = afr_transaction_detach_fop_frame(frame);
- AFR_STACK_UNWIND (create, main_frame, local->op_ret, local->op_errno,
- local->cont.create.fd, local->inode,
- &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(create, main_frame, local->op_ret, local->op_errno,
+ local->cont.create.fd, local->inode,
+ &local->cont.dir_fop.buf, &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- fd_t *fd, inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+afr_create_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_create_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_create_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- STACK_WIND_COOKIE (frame, afr_create_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->create,
- &local->loc, local->cont.create.flags,
- local->cont.create.mode, local->umask,
- local->cont.create.fd, local->xdata_req);
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ STACK_WIND_COOKIE(frame, afr_create_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->create, &local->loc,
+ local->cont.create.flags, local->cont.create.mode,
+ local->umask, local->cont.create.fd, local->xdata_req);
+ return 0;
}
-
int
-afr_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+afr_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
-
- local->fd_ctx = afr_fd_ctx_get (fd, this);
- if (!local->fd_ctx)
- goto out;
-
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->op = GF_FOP_CREATE;
- local->cont.create.flags = flags;
- local->fd_ctx->flags = flags;
- local->cont.create.mode = mode;
- local->cont.create.fd = fd_ref (fd);
- local->umask = umask;
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->transaction.wind = afr_create_wind;
- local->transaction.unwind = afr_create_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ priv = this->private;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+
+ local->fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!local->fd_ctx)
+ goto out;
+
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->op = GF_FOP_CREATE;
+ local->cont.create.flags = flags;
+ local->fd_ctx->flags = flags;
+ local->cont.create.mode = mode;
+ local->cont.create.fd = fd_ref(fd);
+ local->umask = umask;
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->transaction.wind = afr_create_wind;
+ local->transaction.unwind = afr_create_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee(&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
/* }}} */
@@ -536,516 +504,492 @@ out:
/* {{{ mknod */
int
-afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
+afr_mknod_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (mknod, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(mknod, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_mknod_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_mknod_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_mknod_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_mknod_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- STACK_WIND_COOKIE (frame, afr_mknod_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->mknod,
- &local->loc, local->cont.mknod.mode,
- local->cont.mknod.dev, local->umask,
- local->xdata_req);
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ STACK_WIND_COOKIE(frame, afr_mknod_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->mknod, &local->loc,
+ local->cont.mknod.mode, local->cont.mknod.dev,
+ local->umask, local->xdata_req);
+ return 0;
}
int
-afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t dev, mode_t umask, dict_t *xdata)
+afr_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->op = GF_FOP_MKNOD;
- local->cont.mknod.mode = mode;
- local->cont.mknod.dev = dev;
- local->umask = umask;
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->transaction.wind = afr_mknod_wind;
- local->transaction.unwind = afr_mknod_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ priv = this->private;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->op = GF_FOP_MKNOD;
+ local->cont.mknod.mode = mode;
+ local->cont.mknod.dev = dev;
+ local->umask = umask;
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->transaction.wind = afr_mknod_wind;
+ local->transaction.unwind = afr_mknod_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee(&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
+ AFR_STACK_UNWIND(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
/* {{{ mkdir */
-
int
-afr_mkdir_unwind (call_frame_t *frame, xlator_t *this)
+afr_mkdir_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (mkdir, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(mkdir, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_mkdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_mkdir_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_mkdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_mkdir_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_mkdir_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->mkdir, &local->loc,
- local->cont.mkdir.mode, local->umask,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_mkdir_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->mkdir, &local->loc,
+ local->cont.mkdir.mode, local->umask, local->xdata_req);
+ return 0;
}
-
int
-afr_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+afr_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->cont.mkdir.mode = mode;
- local->umask = umask;
-
- if (!xdata || !dict_get (xdata, "gfid-req")) {
- op_errno = EPERM;
- gf_msg_callingfn (this->name, GF_LOG_WARNING, op_errno,
- AFR_MSG_GFID_NULL, "mkdir: %s is received "
- "without gfid-req %p", loc->path, xdata);
- goto out;
- }
-
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- if (!local->xdata_req) {
- op_errno = ENOMEM;
- goto out;
- }
-
- local->op = GF_FOP_MKDIR;
- local->transaction.wind = afr_mkdir_wind;
- local->transaction.unwind = afr_mkdir_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ priv = this->private;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->cont.mkdir.mode = mode;
+ local->umask = umask;
+
+ if (!xdata || !dict_get(xdata, "gfid-req")) {
+ op_errno = EPERM;
+ gf_msg_callingfn(this->name, GF_LOG_WARNING, op_errno,
+ AFR_MSG_GFID_NULL,
+ "mkdir: %s is received "
+ "without gfid-req %p",
+ loc->path, xdata);
+ goto out;
+ }
+
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ if (!local->xdata_req) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->op = GF_FOP_MKDIR;
+ local->transaction.wind = afr_mkdir_wind;
+ local->transaction.unwind = afr_mkdir_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee(&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
+ AFR_STACK_UNWIND(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
/* {{{ link */
-
int
-afr_link_unwind (call_frame_t *frame, xlator_t *this)
+afr_link_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (link, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(link, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_link_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_link_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_link_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_link_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->link,
- &local->loc, &local->newloc, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_link_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->link, &local->loc,
+ &local->newloc, local->xdata_req);
+ return 0;
}
-
int
-afr_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+afr_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->newloc, newloc);
-
- local->inode = inode_ref (oldloc->inode);
- local->parent = inode_ref (newloc->parent);
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_LINK;
-
- local->transaction.wind = afr_link_wind;
- local->transaction.unwind = afr_link_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, newloc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (newloc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ priv = this->private;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, oldloc);
+ loc_copy(&local->newloc, newloc);
+
+ local->inode = inode_ref(oldloc->inode);
+ local->parent = inode_ref(newloc->parent);
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_LINK;
+
+ local->transaction.wind = afr_link_wind;
+ local->transaction.unwind = afr_link_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, newloc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(newloc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee(&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
+ AFR_STACK_UNWIND(link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
/* {{{ symlink */
-
int
-afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
+afr_symlink_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (symlink, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(symlink, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_symlink_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_symlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_symlink_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- STACK_WIND_COOKIE (frame, afr_symlink_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->symlink,
- local->cont.symlink.linkpath, &local->loc,
- local->umask, local->xdata_req);
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ STACK_WIND_COOKIE(frame, afr_symlink_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->symlink,
+ local->cont.symlink.linkpath, &local->loc, local->umask,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+afr_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->cont.symlink.linkpath = gf_strdup (linkpath);
- local->umask = umask;
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_SYMLINK;
- local->transaction.wind = afr_symlink_wind;
- local->transaction.unwind = afr_symlink_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ priv = this->private;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->cont.symlink.linkpath = gf_strdup(linkpath);
+ local->umask = umask;
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_SYMLINK;
+ local->transaction.wind = afr_symlink_wind;
+ local->transaction.unwind = afr_symlink_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee(&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (symlink, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(symlink, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
/* }}} */
@@ -1053,159 +997,152 @@ out:
/* {{{ rename */
int
-afr_rename_unwind (call_frame_t *frame, xlator_t *this)
+afr_rename_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (rename, main_frame, local->op_ret, local->op_errno,
- &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent,
- &local->cont.dir_fop.prenewparent,
- &local->cont.dir_fop.postnewparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(rename, main_frame, local->op_ret, local->op_errno,
+ &local->cont.dir_fop.buf, &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ &local->cont.dir_fop.prenewparent,
+ &local->cont.dir_fop.postnewparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+afr_rename_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preoldparent, postoldparent, prenewparent,
- postnewparent, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preoldparent, postoldparent, prenewparent,
+ postnewparent, xdata);
}
-
int
-afr_rename_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_rename_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_rename_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->rename,
- &local->loc, &local->newloc, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_rename_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->rename, &local->loc,
+ &local->newloc, local->xdata_req);
+ return 0;
}
-
int
-afr_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+afr_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
- int nlockee = 0;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- op_errno = ENOMEM;
- goto out;
- }
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->newloc, newloc);
-
- local->inode = inode_ref (oldloc->inode);
- local->parent = inode_ref (oldloc->parent);
- local->parent2 = inode_ref (newloc->parent);
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_RENAME;
- local->transaction.wind = afr_rename_wind;
- local->transaction.unwind = afr_rename_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, oldloc,
- &op_errno);
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+ int nlockee = 0;
+
+ priv = this->private;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, oldloc);
+ loc_copy(&local->newloc, newloc);
+
+ local->inode = inode_ref(oldloc->inode);
+ local->parent = inode_ref(oldloc->parent);
+ local->parent2 = inode_ref(newloc->parent);
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_RENAME;
+ local->transaction.wind = afr_rename_wind;
+ local->transaction.unwind = afr_rename_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, oldloc,
+ &op_errno);
+ if (ret)
+ goto out;
+ ret = afr_build_parent_loc(&local->transaction.new_parent_loc, newloc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(oldloc->path);
+ local->transaction.new_basename = AFR_BASENAME(newloc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = nlockee = 0;
+ ret = afr_init_entry_lockee(
+ &int_lock->lockee[nlockee], local, &local->transaction.new_parent_loc,
+ local->transaction.new_basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ ret = afr_init_entry_lockee(&int_lock->lockee[nlockee], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ if (local->newloc.inode && IA_ISDIR(local->newloc.inode->ia_type)) {
+ ret = afr_init_entry_lockee(&int_lock->lockee[nlockee], local,
+ &local->newloc, NULL, priv->child_count);
if (ret)
- goto out;
- ret = afr_build_parent_loc (&local->transaction.new_parent_loc, newloc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (oldloc->path);
- local->transaction.new_basename = AFR_BASENAME (newloc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = nlockee = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
+ goto out;
nlockee++;
- if (local->newloc.inode && IA_ISDIR (local->newloc.inode->ia_type)) {
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->newloc,
- NULL,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- }
- qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
- afr_entry_lockee_cmp);
- int_lock->lockee_count = nlockee;
-
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ }
+ qsort(int_lock->lockee, nlockee, sizeof(*int_lock->lockee),
+ afr_entry_lockee_cmp);
+ int_lock->lockee_count = nlockee;
+
+ ret = afr_transaction(transaction_frame, this,
+ AFR_ENTRY_RENAME_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
/* }}} */
@@ -1213,259 +1150,244 @@ out:
/* {{{ unlink */
int
-afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
+afr_unlink_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (unlink, main_frame, local->op_ret, local->op_errno,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(unlink, main_frame, local->op_ret, local->op_errno,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_unlink_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, NULL,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_unlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_unlink_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_unlink_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->unlink,
- &local->loc, local->xflag, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_unlink_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->unlink, &local->loc,
+ local->xflag, local->xdata_req);
+ return 0;
}
-
int
-afr_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+afr_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
- local->xflag = xflag;
-
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_UNLINK;
- local->transaction.wind = afr_unlink_wind;
- local->transaction.unwind = afr_unlink_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ priv = this->private;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->xflag = xflag;
+
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_UNLINK;
+ local->transaction.wind = afr_unlink_wind;
+ local->transaction.unwind = afr_unlink_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee(&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(unlink, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
/* {{{ rmdir */
-
-
int
-afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
+afr_rmdir_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (rmdir, main_frame, local->op_ret, local->op_errno,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(rmdir, main_frame, local->op_ret, local->op_errno,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_rmdir_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, NULL,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_rmdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_rmdir_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_rmdir_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->rmdir,
- &local->loc, local->cont.rmdir.flags, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_rmdir_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->rmdir, &local->loc,
+ local->cont.rmdir.flags, local->xdata_req);
+ return 0;
}
-
int
-afr_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+afr_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
- int nlockee = 0;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
-
- loc_copy (&local->loc, loc);
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->cont.rmdir.flags = flags;
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_RMDIR;
- local->transaction.wind = afr_rmdir_wind;
- local->transaction.unwind = afr_rmdir_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = nlockee = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->loc,
- NULL,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
- afr_entry_lockee_cmp);
- int_lock->lockee_count = nlockee;
-
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+ int nlockee = 0;
+
+ priv = this->private;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->cont.rmdir.flags = flags;
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_RMDIR;
+ local->transaction.wind = afr_rmdir_wind;
+ local->transaction.unwind = afr_rmdir_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = nlockee = 0;
+ ret = afr_init_entry_lockee(&int_lock->lockee[nlockee], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename, priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ ret = afr_init_entry_lockee(&int_lock->lockee[nlockee], local, &local->loc,
+ NULL, priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ qsort(int_lock->lockee, nlockee, sizeof(*int_lock->lockee),
+ afr_entry_lockee_cmp);
+ int_lock->lockee_count = nlockee;
+
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(rmdir, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
diff --git a/xlators/cluster/afr/src/afr-inode-read.c b/xlators/cluster/afr/src/afr-inode-read.c
index d0b07e9064f..113e39acfe8 100644
--- a/xlators/cluster/afr/src/afr-inode-read.c
+++ b/xlators/cluster/afr/src/afr-inode-read.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include <libgen.h>
#include <unistd.h>
#include <fnmatch.h>
@@ -44,146 +43,145 @@
* */
int
-afr_handle_quota_size (call_frame_t *frame, xlator_t *this)
+afr_handle_quota_size(call_frame_t *frame, xlator_t *this)
{
- unsigned char *readable = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- int i = 0;
- int ret = 0;
- quota_meta_t size = {0, };
- quota_meta_t max_size = {0, };
- int readable_cnt = 0;
- int read_subvol = -1;
-
- local = frame->local;
- priv = this->private;
- replies = local->replies;
-
- readable = alloca0 (priv->child_count);
-
- afr_inode_read_subvol_get (local->inode, this, readable, 0, 0);
-
- readable_cnt = AFR_COUNT (readable, priv->child_count);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (readable_cnt && !readable[i])
- continue;
- if (!replies[i].xdata)
- continue;
- ret = quota_dict_get_meta (replies[i].xdata, QUOTA_SIZE_KEY,
- &size);
- if (ret == -1)
- continue;
- if (read_subvol == -1)
- read_subvol = i;
- if (size.size > max_size.size ||
- (size.file_count + size.dir_count) >
- (max_size.file_count + max_size.dir_count))
- read_subvol = i;
-
- if (size.size > max_size.size)
- max_size.size = size.size;
- if (size.file_count > max_size.file_count)
- max_size.file_count = size.file_count;
- if (size.dir_count > max_size.dir_count)
- max_size.dir_count = size.dir_count;
- }
-
- if (max_size.size == 0 && max_size.file_count == 0 &&
- max_size.dir_count == 0)
- return read_subvol;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (readable_cnt && !readable[i])
- continue;
- if (!replies[i].xdata)
- continue;
- quota_dict_set_meta (replies[i].xdata, QUOTA_SIZE_KEY,
- &max_size, IA_IFDIR);
- }
-
+ unsigned char *readable = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ int i = 0;
+ int ret = 0;
+ quota_meta_t size = {
+ 0,
+ };
+ quota_meta_t max_size = {
+ 0,
+ };
+ int readable_cnt = 0;
+ int read_subvol = -1;
+
+ local = frame->local;
+ priv = this->private;
+ replies = local->replies;
+
+ readable = alloca0(priv->child_count);
+
+ afr_inode_read_subvol_get(local->inode, this, readable, 0, 0);
+
+ readable_cnt = AFR_COUNT(readable, priv->child_count);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (readable_cnt && !readable[i])
+ continue;
+ if (!replies[i].xdata)
+ continue;
+ ret = quota_dict_get_meta(replies[i].xdata, QUOTA_SIZE_KEY, &size);
+ if (ret == -1)
+ continue;
+ if (read_subvol == -1)
+ read_subvol = i;
+ if (size.size > max_size.size ||
+ (size.file_count + size.dir_count) >
+ (max_size.file_count + max_size.dir_count))
+ read_subvol = i;
+
+ if (size.size > max_size.size)
+ max_size.size = size.size;
+ if (size.file_count > max_size.file_count)
+ max_size.file_count = size.file_count;
+ if (size.dir_count > max_size.dir_count)
+ max_size.dir_count = size.dir_count;
+ }
+
+ if (max_size.size == 0 && max_size.file_count == 0 &&
+ max_size.dir_count == 0)
return read_subvol;
-}
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (readable_cnt && !readable[i])
+ continue;
+ if (!replies[i].xdata)
+ continue;
+ quota_dict_set_meta(replies[i].xdata, QUOTA_SIZE_KEY, &max_size,
+ IA_IFDIR);
+ }
+
+ return read_subvol;
+}
/* {{{ access */
int
-afr_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+afr_access_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
+ AFR_STACK_UNWIND(access, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
-
int
-afr_access_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_access_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (access, frame, local->op_ret,
- local->op_errno, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_access_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->access,
- &local->loc, local->cont.access.mask,
- local->xdata_req);
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(access, frame, local->op_ret, local->op_errno, 0);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(frame, afr_access_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->access, &local->loc,
+ local->cont.access.mask, local->xdata_req);
+ return 0;
}
int
-afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int mask, dict_t *xdata)
+afr_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int mask,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int op_errno = 0;
+ afr_local_t *local = NULL;
+ int op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_ACCESS;
- loc_copy (&local->loc, loc);
- local->cont.access.mask = mask;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_ACCESS;
+ loc_copy(&local->loc, loc);
+ local->cont.access.mask = mask;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_read_txn (frame, this, loc->inode, afr_access_wind,
- AFR_METADATA_TRANSACTION);
+ afr_read_txn(frame, this, loc->inode, afr_access_wind,
+ AFR_METADATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND(access, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
/* }}} */
@@ -191,152 +189,139 @@ out:
/* {{{ stat */
int
-afr_stat_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *buf, dict_t *xdata)
+afr_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
+ AFR_STACK_UNWIND(stat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ return 0;
}
-
int
-afr_stat_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_stat_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
- 0, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->stat,
- &local->loc, local->xdata_req);
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(stat, frame, local->op_ret, local->op_errno, 0, 0);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(
+ frame, afr_stat_cbk, (void *)(long)subvol, priv->children[subvol],
+ priv->children[subvol]->fops->stat, &local->loc, local->xdata_req);
+ return 0;
}
int
-afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+afr_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int op_errno = 0;
+ afr_local_t *local = NULL;
+ int op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_STAT;
- loc_copy (&local->loc, loc);
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_STAT;
+ loc_copy(&local->loc, loc);
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_read_txn (frame, this, loc->inode, afr_stat_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, loc->inode, afr_stat_wind, AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(stat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
/* }}} */
/* {{{ fstat */
int
-afr_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+afr_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
+ AFR_STACK_UNWIND(fstat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ return 0;
}
-
int
-afr_fstat_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fstat_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (fstat, frame, local->op_ret, local->op_errno,
- 0, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fstat,
- local->fd, local->xdata_req);
- return 0;
-}
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(fstat, frame, local->op_ret, local->op_errno, 0, 0);
+ return 0;
+ }
+ STACK_WIND_COOKIE(
+ frame, afr_fstat_cbk, (void *)(long)subvol, priv->children[subvol],
+ priv->children[subvol]->fops->fstat, local->fd, local->xdata_req);
+ return 0;
+}
int32_t
-afr_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xdata)
+afr_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int op_errno = 0;
+ afr_local_t *local = NULL;
+ int op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_FSTAT;
- local->fd = fd_ref (fd);
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_FSTAT;
+ local->fd = fd_ref(fd);
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_read_txn (frame, this, fd->inode, afr_fstat_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, fd->inode, afr_fstat_wind, AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(fstat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
/* }}} */
@@ -344,1565 +329,1484 @@ out:
/* {{{ readlink */
int
-afr_readlink_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- const char *buf, struct iatt *sbuf, dict_t *xdata)
+afr_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, const char *buf,
+ struct iatt *sbuf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno,
- buf, sbuf, xdata);
- return 0;
+ AFR_STACK_UNWIND(readlink, frame, op_ret, op_errno, buf, sbuf, xdata);
+ return 0;
}
int
-afr_readlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_readlink_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (readlink, frame, local->op_ret,
- local->op_errno, 0, 0, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_readlink_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->readlink,
- &local->loc, local->cont.readlink.size,
- local->xdata_req);
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ local = frame->local;
+ priv = this->private;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(readlink, frame, local->op_ret, local->op_errno, 0, 0,
+ 0);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(frame, afr_readlink_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->readlink, &local->loc,
+ local->cont.readlink.size, local->xdata_req);
+ return 0;
+}
int
-afr_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size, dict_t *xdata)
+afr_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- afr_local_t * local = NULL;
- int32_t op_errno = 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_READLINK;
- loc_copy (&local->loc, loc);
- local->cont.readlink.size = size;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_READLINK;
+ loc_copy(&local->loc, loc);
+ local->cont.readlink.size = size;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_read_txn (frame, this, loc->inode, afr_readlink_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, loc->inode, afr_readlink_wind,
+ AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND(readlink, frame, -1, op_errno, 0, 0, 0);
+ AFR_STACK_UNWIND(readlink, frame, -1, op_errno, 0, 0, 0);
- return 0;
+ return 0;
}
-
/* }}} */
/* {{{ getxattr */
struct _xattr_key {
- char *key;
- struct list_head list;
+ char *key;
+ struct list_head list;
};
-
int
-__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
- void *data)
+__gather_xattr_keys(dict_t *dict, char *key, data_t *value, void *data)
{
- struct list_head * list = data;
- struct _xattr_key * xkey = NULL;
-
- if (!strncmp (key, AFR_XATTR_PREFIX,
- SLEN (AFR_XATTR_PREFIX))) {
+ struct list_head *list = data;
+ struct _xattr_key *xkey = NULL;
- xkey = GF_MALLOC (sizeof (*xkey), gf_afr_mt_xattr_key);
- if (!xkey)
- return -1;
+ if (!strncmp(key, AFR_XATTR_PREFIX, SLEN(AFR_XATTR_PREFIX))) {
+ xkey = GF_MALLOC(sizeof(*xkey), gf_afr_mt_xattr_key);
+ if (!xkey)
+ return -1;
- xkey->key = key;
- INIT_LIST_HEAD (&xkey->list);
+ xkey->key = key;
+ INIT_LIST_HEAD(&xkey->list);
- list_add_tail (&xkey->list, list);
- }
- return 0;
+ list_add_tail(&xkey->list, list);
+ }
+ return 0;
}
-
void
-afr_filter_xattrs (dict_t *dict)
+afr_filter_xattrs(dict_t *dict)
{
- struct list_head keys = {0,};
- struct _xattr_key *key = NULL;
- struct _xattr_key *tmp = NULL;
+ struct list_head keys = {
+ 0,
+ };
+ struct _xattr_key *key = NULL;
+ struct _xattr_key *tmp = NULL;
- INIT_LIST_HEAD (&keys);
+ INIT_LIST_HEAD(&keys);
- dict_foreach (dict, __gather_xattr_keys,
- (void *) &keys);
+ dict_foreach(dict, __gather_xattr_keys, (void *)&keys);
- list_for_each_entry_safe (key, tmp, &keys, list) {
- dict_del (dict, key->key);
+ list_for_each_entry_safe(key, tmp, &keys, list)
+ {
+ dict_del(dict, key->key);
- list_del_init (&key->list);
+ list_del_init(&key->list);
- GF_FREE (key);
- }
+ GF_FREE(key);
+ }
}
-static
-gf_boolean_t
-afr_getxattr_ignorable_errnos (int32_t op_errno)
+static gf_boolean_t
+afr_getxattr_ignorable_errnos(int32_t op_errno)
{
- if (op_errno == ENODATA || op_errno == ENOTSUP || op_errno == ERANGE ||
- op_errno == ENAMETOOLONG)
- return _gf_true;
+ if (op_errno == ENODATA || op_errno == ENOTSUP || op_errno == ERANGE ||
+ op_errno == ENAMETOOLONG)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
int
-afr_getxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0 && !afr_getxattr_ignorable_errnos(op_errno)) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0 && !afr_getxattr_ignorable_errnos(op_errno)) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- if (dict)
- afr_filter_xattrs (dict);
+ if (dict)
+ afr_filter_xattrs(dict);
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ return 0;
}
-
int
-afr_getxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_getxattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, NULL, NULL);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_getxattr_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->getxattr,
- &local->loc, local->cont.getxattr.name,
- local->xdata_req);
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ local = frame->local;
+ priv = this->private;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno, NULL,
+ NULL);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(frame, afr_getxattr_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->getxattr, &local->loc,
+ local->cont.getxattr.name, local->xdata_req);
+ return 0;
+}
int32_t
-afr_getxattr_unwind (call_frame_t *frame, int op_ret, int op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_unwind(call_frame_t *frame, int op_ret, int op_errno, dict_t *dict,
+ dict_t *xdata)
{
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
int32_t
-afr_fgetxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_fgetxattr_clrlk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- xlator_t **children = NULL;
- dict_t *xattr = NULL;
- char *tmp_report = NULL;
- char lk_summary[1024] = {0,};
- int serz_len = 0;
- int32_t callcnt = 0;
- long int cky = 0;
- int ret = 0;
-
- priv = this->private;
- children = priv->children;
-
- local = frame->local;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- if (op_ret == -1)
- local->replies[cky].op_errno = op_errno;
-
- if (!local->dict)
- local->dict = dict_new ();
- if (local->dict) {
- ret = dict_get_str (dict, local->cont.getxattr.name,
- &tmp_report);
- if (ret)
- goto unlock;
- ret = dict_set_dynstr (local->dict,
- children[cky]->name,
- gf_strdup (tmp_report));
- if (ret)
- goto unlock;
- }
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ dict_t *xattr = NULL;
+ char *tmp_report = NULL;
+ char lk_summary[1024] = {
+ 0,
+ };
+ int serz_len = 0;
+ int32_t callcnt = 0;
+ long int cky = 0;
+ int ret = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ cky = (long)cookie;
+
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+ if (op_ret == -1)
+ local->replies[cky].op_errno = op_errno;
+
+ if (!local->dict)
+ local->dict = dict_new();
+ if (local->dict) {
+ ret = dict_get_str(dict, local->cont.getxattr.name, &tmp_report);
+ if (ret)
+ goto unlock;
+ ret = dict_set_dynstr(local->dict, children[cky]->name,
+ gf_strdup(tmp_report));
+ if (ret)
+ goto unlock;
}
+ }
unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- xattr = dict_new ();
- if (!xattr) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- ret = dict_serialize_value_with_delim (local->dict,
- lk_summary,
- &serz_len, '\n');
- if (ret) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- if (serz_len == -1)
- snprintf (lk_summary, sizeof (lk_summary),
- "No locks cleared.");
- ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
- gf_strdup (lk_summary));
- if (ret) {
- op_ret = -1;
- op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_ERROR,
- ENOMEM, AFR_MSG_DICT_SET_FAILED,
- "Error setting dictionary");
- goto unwind;
- }
+ UNLOCK(&frame->lock);
+
+ if (!callcnt) {
+ xattr = dict_new();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_serialize_value_with_delim(local->dict, lk_summary,
+ &serz_len, '\n');
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ if (serz_len == -1)
+ snprintf(lk_summary, sizeof(lk_summary), "No locks cleared.");
+ ret = dict_set_dynstr(xattr, local->cont.getxattr.name,
+ gf_strdup(lk_summary));
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_DICT_SET_FAILED,
+ "Error setting dictionary");
+ goto unwind;
+ }
- op_errno = afr_final_errno (local, priv);
+ op_errno = afr_final_errno(local, priv);
-unwind:
- AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, xattr,
- xdata);
- if (xattr)
- dict_unref (xattr);
- }
+ unwind:
+ AFR_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, xattr, xdata);
+ if (xattr)
+ dict_unref(xattr);
+ }
- return ret;
+ return ret;
}
int32_t
-afr_getxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_clrlk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- xlator_t **children = NULL;
- dict_t *xattr = NULL;
- char *tmp_report = NULL;
- char lk_summary[1024] = {0,};
- int serz_len = 0;
- int32_t callcnt = 0;
- long int cky = 0;
- int ret = 0;
-
- priv = this->private;
- children = priv->children;
-
- local = frame->local;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- if (op_ret == -1)
- local->replies[cky].op_errno = op_errno;
-
- if (!local->dict)
- local->dict = dict_new ();
- if (local->dict) {
- ret = dict_get_str (dict, local->cont.getxattr.name,
- &tmp_report);
- if (ret)
- goto unlock;
- ret = dict_set_dynstr (local->dict,
- children[cky]->name,
- gf_strdup (tmp_report));
- if (ret)
- goto unlock;
- }
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ dict_t *xattr = NULL;
+ char *tmp_report = NULL;
+ char lk_summary[1024] = {
+ 0,
+ };
+ int serz_len = 0;
+ int32_t callcnt = 0;
+ long int cky = 0;
+ int ret = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ cky = (long)cookie;
+
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+ if (op_ret == -1)
+ local->replies[cky].op_errno = op_errno;
+
+ if (!local->dict)
+ local->dict = dict_new();
+ if (local->dict) {
+ ret = dict_get_str(dict, local->cont.getxattr.name, &tmp_report);
+ if (ret)
+ goto unlock;
+ ret = dict_set_dynstr(local->dict, children[cky]->name,
+ gf_strdup(tmp_report));
+ if (ret)
+ goto unlock;
}
+ }
unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- xattr = dict_new ();
- if (!xattr) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- ret = dict_serialize_value_with_delim (local->dict,
- lk_summary,
- &serz_len, '\n');
- if (ret) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- if (serz_len == -1)
- snprintf (lk_summary, sizeof (lk_summary),
- "No locks cleared.");
- ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
- gf_strdup (lk_summary));
- if (ret) {
- op_ret = -1;
- op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_ERROR,
- ENOMEM, AFR_MSG_DICT_SET_FAILED,
- "Error setting dictionary");
- goto unwind;
- }
+ UNLOCK(&frame->lock);
+
+ if (!callcnt) {
+ xattr = dict_new();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_serialize_value_with_delim(local->dict, lk_summary,
+ &serz_len, '\n');
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ if (serz_len == -1)
+ snprintf(lk_summary, sizeof(lk_summary), "No locks cleared.");
+ ret = dict_set_dynstr(xattr, local->cont.getxattr.name,
+ gf_strdup(lk_summary));
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_DICT_SET_FAILED,
+ "Error setting dictionary");
+ goto unwind;
+ }
- op_errno = afr_final_errno (local, priv);
+ op_errno = afr_final_errno(local, priv);
-unwind:
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
+ unwind:
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, xattr, xdata);
- if (xattr)
- dict_unref (xattr);
- }
+ if (xattr)
+ dict_unref(xattr);
+ }
- return ret;
+ return ret;
}
/**
* node-uuid cbk uses next child querying mechanism
*/
int32_t
-afr_getxattr_node_uuid_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_node_uuid_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- xlator_t **children = NULL;
- int unwind = 1;
- int curr_call_child = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int unwind = 1;
+ int curr_call_child = 0;
- priv = this->private;
- children = priv->children;
+ priv = this->private;
+ children = priv->children;
- local = frame->local;
+ local = frame->local;
- if (op_ret == -1) { /** query the _next_ child */
-
- /**
- * _current_ becomes _next_
- * If done with all children and yet no success; give up !
- */
- curr_call_child = (int) ((long)cookie);
- if (++curr_call_child == priv->child_count)
- goto unwind;
-
- gf_msg_debug (this->name, op_errno,
- "op_ret (-1): Re-querying afr-child (%d/%d)",
- curr_call_child, priv->child_count);
-
- unwind = 0;
- STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
- (void *) (long) curr_call_child,
- children[curr_call_child],
- children[curr_call_child]->fops->getxattr,
- &local->loc,
- local->cont.getxattr.name,
- local->xdata_req);
- }
+ if (op_ret == -1) { /** query the _next_ child */
- unwind:
- if (unwind)
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
- xdata);
+ /**
+ * _current_ becomes _next_
+ * If done with all children and yet no success; give up !
+ */
+ curr_call_child = (int)((long)cookie);
+ if (++curr_call_child == priv->child_count)
+ goto unwind;
+
+ gf_msg_debug(this->name, op_errno,
+ "op_ret (-1): Re-querying afr-child (%d/%d)",
+ curr_call_child, priv->child_count);
+
+ unwind = 0;
+ STACK_WIND_COOKIE(
+ frame, afr_getxattr_node_uuid_cbk, (void *)(long)curr_call_child,
+ children[curr_call_child],
+ children[curr_call_child]->fops->getxattr, &local->loc,
+ local->cont.getxattr.name, local->xdata_req);
+ }
- return 0;
+unwind:
+ if (unwind)
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+
+ return 0;
}
/**
* list-node-uuids cbk returns the list of node_uuids for the subvolume.
*/
int32_t
-afr_getxattr_list_node_uuids_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *dict, dict_t *xdata)
+afr_getxattr_list_node_uuids_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int32_t callcnt = 0;
- int ret = 0;
- char *xattr_serz = NULL;
- long cky = 0;
- int32_t tlen = 0;
-
- local = frame->local;
- priv = this->private;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- local->replies[cky].valid = 1;
- local->replies[cky].op_ret = op_ret;
- local->replies[cky].op_errno = op_errno;
-
- if (op_ret < 0)
- goto unlock;
-
- local->op_ret = 0;
-
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- local->replies[cky].xattr = dict_ref (dict);
- }
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr_serz = NULL;
+ long cky = 0;
+ int32_t tlen = 0;
+
+ local = frame->local;
+ priv = this->private;
+ cky = (long)cookie;
+
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+ local->replies[cky].valid = 1;
+ local->replies[cky].op_ret = op_ret;
+ local->replies[cky].op_errno = op_errno;
+
+ if (op_ret < 0)
+ goto unlock;
+
+ local->op_ret = 0;
+
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ local->replies[cky].xattr = dict_ref(dict);
+ }
unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
+ UNLOCK(&frame->lock);
- if (local->op_ret != 0) {
- /* All bricks gave an error. */
- local->op_errno = afr_final_errno (local, priv);
- goto unwind;
- }
+ if (!callcnt) {
+ if (local->op_ret != 0) {
+ /* All bricks gave an error. */
+ local->op_errno = afr_final_errno(local, priv);
+ goto unwind;
+ }
- /*Since we store the UUID0_STR as node uuid for down bricks and
- *for non zero op_ret, assigning length to priv->child_count
- *number of uuids*/
- local->cont.getxattr.xattr_len = (SLEN (UUID0_STR) + 2) *
- priv->child_count;
-
- if (!local->dict)
- local->dict = dict_new ();
- if (!local->dict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
+ /*Since we store the UUID0_STR as node uuid for down bricks and
+ *for non zero op_ret, assigning length to priv->child_count
+ *number of uuids*/
+ local->cont.getxattr.xattr_len = (SLEN(UUID0_STR) + 2) *
+ priv->child_count;
+
+ if (!local->dict)
+ local->dict = dict_new();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
- sizeof (char), gf_common_mt_char);
+ xattr_serz = GF_CALLOC(local->cont.getxattr.xattr_len, sizeof(char),
+ gf_common_mt_char);
- if (!xattr_serz) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
-
- ret = afr_serialize_xattrs_with_delimiter (frame, this,
- xattr_serz,
- UUID0_STR, &tlen,
- ' ');
- if (ret) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
- ret = dict_set_dynstr (local->dict,
- GF_XATTR_LIST_NODE_UUIDS_KEY,
- xattr_serz);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Cannot set node_uuid key in dict");
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- } else {
- local->op_ret = local->cont.getxattr.xattr_len - 1;
- local->op_errno = 0;
- }
+ if (!xattr_serz) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
-unwind:
- AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, local->dict,
- local->xdata_rsp);
+ ret = afr_serialize_xattrs_with_delimiter(frame, this, xattr_serz,
+ UUID0_STR, &tlen, ' ');
+ if (ret) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_set_dynstr(local->dict, GF_XATTR_LIST_NODE_UUIDS_KEY,
+ xattr_serz);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set node_uuid key in dict");
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ local->op_ret = local->cont.getxattr.xattr_len - 1;
+ local->op_errno = 0;
}
- return ret;
-}
+ unwind:
+ AFR_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno,
+ local->dict, local->xdata_rsp);
+ }
+ return ret;
+}
int32_t
-afr_getxattr_quota_size_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_quota_size_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- int idx = (long) cookie;
- int call_count = 0;
- afr_local_t *local = frame->local;
- int read_subvol = -1;
-
- local->replies[idx].valid = 1;
- local->replies[idx].op_ret = op_ret;
- local->replies[idx].op_errno = op_errno;
- if (dict)
- local->replies[idx].xdata = dict_ref (dict);
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- local->inode = inode_ref (local->loc.inode);
- read_subvol = afr_handle_quota_size (frame, this);
- if (read_subvol != -1) {
- op_ret = local->replies[read_subvol].op_ret;
- op_errno = local->replies[read_subvol].op_errno;
- dict = local->replies[read_subvol].xdata;
- }
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
- xdata);
+ int idx = (long)cookie;
+ int call_count = 0;
+ afr_local_t *local = frame->local;
+ int read_subvol = -1;
+
+ local->replies[idx].valid = 1;
+ local->replies[idx].op_ret = op_ret;
+ local->replies[idx].op_errno = op_errno;
+ if (dict)
+ local->replies[idx].xdata = dict_ref(dict);
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ local->inode = inode_ref(local->loc.inode);
+ read_subvol = afr_handle_quota_size(frame, this);
+ if (read_subvol != -1) {
+ op_ret = local->replies[read_subvol].op_ret;
+ op_errno = local->replies[read_subvol].op_errno;
+ dict = local->replies[read_subvol].xdata;
}
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+ }
- return 0;
+ return 0;
}
int32_t
-afr_getxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_lockinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- int call_cnt = 0, len = 0;
- char *lockinfo_buf = NULL;
- dict_t *lockinfo = NULL, *newdict = NULL;
- afr_local_t *local = NULL;
+ int call_cnt = 0, len = 0;
+ char *lockinfo_buf = NULL;
+ dict_t *lockinfo = NULL, *newdict = NULL;
+ afr_local_t *local = NULL;
- LOCK (&frame->lock);
- {
- local = frame->local;
-
- call_cnt = --local->call_count;
+ LOCK(&frame->lock);
+ {
+ local = frame->local;
- if ((op_ret < 0) || (!dict && !xdata)) {
- goto unlock;
- }
+ call_cnt = --local->call_count;
- if (xdata) {
- if (!local->xdata_rsp) {
- local->xdata_rsp = dict_new ();
- if (!local->xdata_rsp) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unlock;
- }
- }
- }
+ if ((op_ret < 0) || (!dict && !xdata)) {
+ goto unlock;
+ }
- if (!dict) {
- goto unlock;
+ if (xdata) {
+ if (!local->xdata_rsp) {
+ local->xdata_rsp = dict_new();
+ if (!local->xdata_rsp) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
}
+ }
+ }
- op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
- (void **)&lockinfo_buf, &len);
+ if (!dict) {
+ goto unlock;
+ }
- if (!lockinfo_buf) {
- goto unlock;
- }
+ op_ret = dict_get_ptr_and_len(dict, GF_XATTR_LOCKINFO_KEY,
+ (void **)&lockinfo_buf, &len);
- if (!local->dict) {
- local->dict = dict_new ();
- if (!local->dict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unlock;
- }
- }
- }
-unlock:
- UNLOCK (&frame->lock);
-
- if (lockinfo_buf != NULL) {
- lockinfo = dict_new ();
- if (lockinfo == NULL) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- } else {
- op_ret = dict_unserialize (lockinfo_buf, len,
- &lockinfo);
-
- if (lockinfo && local->dict) {
- dict_copy (lockinfo, local->dict);
- }
- }
+ if (!lockinfo_buf) {
+ goto unlock;
}
- if (xdata && local->xdata_rsp) {
- dict_copy (xdata, local->xdata_rsp);
+ if (!local->dict) {
+ local->dict = dict_new();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
}
+ }
+unlock:
+ UNLOCK(&frame->lock);
- if (!call_cnt) {
- newdict = dict_new ();
- if (!newdict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
+ if (lockinfo_buf != NULL) {
+ lockinfo = dict_new();
+ if (lockinfo == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ op_ret = dict_unserialize(lockinfo_buf, len, &lockinfo);
- len = dict_serialized_length (local->dict);
- if (len <= 0) {
- goto unwind;
- }
+ if (lockinfo && local->dict) {
+ dict_copy(lockinfo, local->dict);
+ }
+ }
+ }
+
+ if (xdata && local->xdata_rsp) {
+ dict_copy(xdata, local->xdata_rsp);
+ }
+
+ if (!call_cnt) {
+ newdict = dict_new();
+ if (!newdict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
- if (!lockinfo_buf) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
+ len = dict_serialized_length(local->dict);
+ if (len <= 0) {
+ goto unwind;
+ }
- op_ret = dict_serialize (local->dict, lockinfo_buf);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- }
+ lockinfo_buf = GF_CALLOC(1, len, gf_common_mt_char);
+ if (!lockinfo_buf) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
- (void *)lockinfo_buf, len);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- goto unwind;
- }
+ op_ret = dict_serialize(local->dict, lockinfo_buf);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ }
- unwind:
- AFR_STACK_UNWIND (getxattr, frame, op_ret,
- op_errno, newdict,
- local->xdata_rsp);
+ op_ret = dict_set_dynptr(newdict, GF_XATTR_LOCKINFO_KEY,
+ (void *)lockinfo_buf, len);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto unwind;
}
- dict_unref (lockinfo);
+ unwind:
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, newdict,
+ local->xdata_rsp);
+ }
- return 0;
+ dict_unref(lockinfo);
+
+ return 0;
}
int32_t
-afr_fgetxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_fgetxattr_lockinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- int call_cnt = 0, len = 0;
- char *lockinfo_buf = NULL;
- dict_t *lockinfo = NULL, *newdict = NULL;
- afr_local_t *local = NULL;
+ int call_cnt = 0, len = 0;
+ char *lockinfo_buf = NULL;
+ dict_t *lockinfo = NULL, *newdict = NULL;
+ afr_local_t *local = NULL;
- LOCK (&frame->lock);
- {
- local = frame->local;
-
- call_cnt = --local->call_count;
+ LOCK(&frame->lock);
+ {
+ local = frame->local;
- if ((op_ret < 0) || (!dict && !xdata)) {
- goto unlock;
- }
+ call_cnt = --local->call_count;
- if (xdata) {
- if (!local->xdata_rsp) {
- local->xdata_rsp = dict_new ();
- if (!local->xdata_rsp) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unlock;
- }
- }
- }
+ if ((op_ret < 0) || (!dict && !xdata)) {
+ goto unlock;
+ }
- if (!dict) {
- goto unlock;
+ if (xdata) {
+ if (!local->xdata_rsp) {
+ local->xdata_rsp = dict_new();
+ if (!local->xdata_rsp) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
}
+ }
+ }
- op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
- (void **)&lockinfo_buf, &len);
+ if (!dict) {
+ goto unlock;
+ }
- if (!lockinfo_buf) {
- goto unlock;
- }
+ op_ret = dict_get_ptr_and_len(dict, GF_XATTR_LOCKINFO_KEY,
+ (void **)&lockinfo_buf, &len);
- if (!local->dict) {
- local->dict = dict_new ();
- if (!local->dict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unlock;
- }
- }
- }
-unlock:
- UNLOCK (&frame->lock);
-
- if (lockinfo_buf != NULL) {
- lockinfo = dict_new ();
- if (lockinfo == NULL) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- } else {
- op_ret = dict_unserialize (lockinfo_buf, len,
- &lockinfo);
-
- if (lockinfo && local->dict) {
- dict_copy (lockinfo, local->dict);
- }
- }
+ if (!lockinfo_buf) {
+ goto unlock;
}
- if (xdata && local->xdata_rsp) {
- dict_copy (xdata, local->xdata_rsp);
+ if (!local->dict) {
+ local->dict = dict_new();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
}
+ }
+unlock:
+ UNLOCK(&frame->lock);
- if (!call_cnt) {
- newdict = dict_new ();
- if (!newdict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
+ if (lockinfo_buf != NULL) {
+ lockinfo = dict_new();
+ if (lockinfo == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ op_ret = dict_unserialize(lockinfo_buf, len, &lockinfo);
- len = dict_serialized_length (local->dict);
- if (len <= 0) {
- goto unwind;
- }
+ if (lockinfo && local->dict) {
+ dict_copy(lockinfo, local->dict);
+ }
+ }
+ }
+
+ if (xdata && local->xdata_rsp) {
+ dict_copy(xdata, local->xdata_rsp);
+ }
+
+ if (!call_cnt) {
+ newdict = dict_new();
+ if (!newdict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
- if (!lockinfo_buf) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
+ len = dict_serialized_length(local->dict);
+ if (len <= 0) {
+ goto unwind;
+ }
- op_ret = dict_serialize (local->dict, lockinfo_buf);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- }
+ lockinfo_buf = GF_CALLOC(1, len, gf_common_mt_char);
+ if (!lockinfo_buf) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
- (void *)lockinfo_buf, len);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- goto unwind;
- }
+ op_ret = dict_serialize(local->dict, lockinfo_buf);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ }
- unwind:
- AFR_STACK_UNWIND (fgetxattr, frame, op_ret,
- op_errno, newdict,
- local->xdata_rsp);
+ op_ret = dict_set_dynptr(newdict, GF_XATTR_LOCKINFO_KEY,
+ (void *)lockinfo_buf, len);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto unwind;
}
- dict_unref (lockinfo);
+ unwind:
+ AFR_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, newdict,
+ local->xdata_rsp);
+ }
- return 0;
+ dict_unref(lockinfo);
+
+ return 0;
}
int32_t
-afr_fgetxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_fgetxattr_pathinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t callcnt = 0;
- int ret = 0;
- char *xattr = NULL;
- char *xattr_serz = NULL;
- char xattr_cky[1024] = {0,};
- dict_t *nxattr = NULL;
- long cky = 0;
- int32_t padding = 0;
- int32_t tlen = 0;
-
- if (!frame || !frame->local || !this) {
- gf_msg ("", GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_ARG, "possible NULL deref");
- goto out;
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr = NULL;
+ char *xattr_serz = NULL;
+ char xattr_cky[1024] = {
+ 0,
+ };
+ dict_t *nxattr = NULL;
+ long cky = 0;
+ int32_t padding = 0;
+ int32_t tlen = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_msg("", GF_LOG_ERROR, 0, AFR_MSG_INVALID_ARG, "possible NULL deref");
+ goto out;
+ }
+
+ local = frame->local;
+ cky = (long)cookie;
+
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+
+ if (op_ret < 0) {
+ local->op_errno = op_errno;
+ } else {
+ local->op_ret = op_ret;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
}
- local = frame->local;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret < 0) {
- local->op_errno = op_errno;
- } else {
- local->op_ret = op_ret;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
+ if (!dict || (op_ret < 0))
+ goto unlock;
- if (!dict || (op_ret < 0))
- goto unlock;
-
- if (!local->dict)
- local->dict = dict_new ();
-
- if (local->dict) {
- ret = dict_get_str (dict,
- local->cont.getxattr.name,
- &xattr);
- if (ret)
- goto unlock;
-
- xattr = gf_strdup (xattr);
-
- (void)snprintf (xattr_cky, sizeof(xattr_cky), "%s-%ld",
- local->cont.getxattr.name, cky);
- ret = dict_set_dynstr (local->dict,
- xattr_cky, xattr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Cannot set xattr cookie key");
- goto unlock;
- }
-
- local->cont.getxattr.xattr_len
- += strlen (xattr) + 1;
- }
+ if (!local->dict)
+ local->dict = dict_new();
+
+ if (local->dict) {
+ ret = dict_get_str(dict, local->cont.getxattr.name, &xattr);
+ if (ret)
+ goto unlock;
+
+ xattr = gf_strdup(xattr);
+
+ (void)snprintf(xattr_cky, sizeof(xattr_cky), "%s-%ld",
+ local->cont.getxattr.name, cky);
+ ret = dict_set_dynstr(local->dict, xattr_cky, xattr);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set xattr cookie key");
+ goto unlock;
+ }
+
+ local->cont.getxattr.xattr_len += strlen(xattr) + 1;
}
+ }
unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (!local->cont.getxattr.xattr_len)
- goto unwind;
-
- nxattr = dict_new ();
- if (!nxattr)
- goto unwind;
-
- /* extra bytes for decorations (brackets and <>'s) */
- padding += strlen (this->name)
- + SLEN (AFR_PATHINFO_HEADER) + 4;
- local->cont.getxattr.xattr_len += (padding + 2);
-
- xattr_serz = GF_MALLOC (local->cont.getxattr.xattr_len,
- gf_common_mt_char);
-
- if (!xattr_serz)
- goto unwind;
-
- /* the xlator info */
- int xattr_serz_len = sprintf (xattr_serz,
- "(<"AFR_PATHINFO_HEADER"%s> ",
- this->name);
-
- /* actual series of pathinfo */
- ret = dict_serialize_value_with_delim (local->dict,
- xattr_serz
- + xattr_serz_len,
- &tlen, ' ');
- if (ret) {
- goto unwind;
- }
+ UNLOCK(&frame->lock);
+
+ if (!callcnt) {
+ if (!local->cont.getxattr.xattr_len)
+ goto unwind;
- /* closing part */
- *(xattr_serz + padding + tlen) = ')';
- *(xattr_serz + padding + tlen + 1) = '\0';
+ nxattr = dict_new();
+ if (!nxattr)
+ goto unwind;
- ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
- xattr_serz);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Cannot set pathinfo key in dict");
+ /* extra bytes for decorations (brackets and <>'s) */
+ padding += strlen(this->name) + SLEN(AFR_PATHINFO_HEADER) + 4;
+ local->cont.getxattr.xattr_len += (padding + 2);
- unwind:
- AFR_STACK_UNWIND (fgetxattr, frame, local->op_ret,
- local->op_errno, nxattr, local->xdata_rsp);
+ xattr_serz = GF_MALLOC(local->cont.getxattr.xattr_len,
+ gf_common_mt_char);
- if (nxattr)
- dict_unref (nxattr);
+ if (!xattr_serz)
+ goto unwind;
+
+ /* the xlator info */
+ int xattr_serz_len = sprintf(
+ xattr_serz, "(<" AFR_PATHINFO_HEADER "%s> ", this->name);
+
+ /* actual series of pathinfo */
+ ret = dict_serialize_value_with_delim(
+ local->dict, xattr_serz + xattr_serz_len, &tlen, ' ');
+ if (ret) {
+ goto unwind;
}
+ /* closing part */
+ *(xattr_serz + padding + tlen) = ')';
+ *(xattr_serz + padding + tlen + 1) = '\0';
+
+ ret = dict_set_dynstr(nxattr, local->cont.getxattr.name, xattr_serz);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set pathinfo key in dict");
+
+ unwind:
+ AFR_STACK_UNWIND(fgetxattr, frame, local->op_ret, local->op_errno,
+ nxattr, local->xdata_rsp);
+
+ if (nxattr)
+ dict_unref(nxattr);
+ }
+
out:
- return ret;
+ return ret;
}
int32_t
-afr_getxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_pathinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t callcnt = 0;
- int ret = 0;
- char *xattr = NULL;
- char *xattr_serz = NULL;
- char xattr_cky[1024] = {0,};
- dict_t *nxattr = NULL;
- long cky = 0;
- int32_t padding = 0;
- int32_t tlen = 0;
-
- if (!frame || !frame->local || !this) {
- gf_msg ("", GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_ARG, "possible NULL deref");
- goto out;
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr = NULL;
+ char *xattr_serz = NULL;
+ char xattr_cky[1024] = {
+ 0,
+ };
+ dict_t *nxattr = NULL;
+ long cky = 0;
+ int32_t padding = 0;
+ int32_t tlen = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_msg("", GF_LOG_ERROR, 0, AFR_MSG_INVALID_ARG, "possible NULL deref");
+ goto out;
+ }
+
+ local = frame->local;
+ cky = (long)cookie;
+
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+
+ if (op_ret < 0) {
+ local->op_errno = op_errno;
+ } else {
+ local->op_ret = op_ret;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
}
- local = frame->local;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret < 0) {
- local->op_errno = op_errno;
- } else {
- local->op_ret = op_ret;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
-
- if (!dict || (op_ret < 0))
- goto unlock;
-
- if (!local->dict)
- local->dict = dict_new ();
-
- if (local->dict) {
- ret = dict_get_str (dict,
- local->cont.getxattr.name,
- &xattr);
- if (ret)
- goto unlock;
-
- xattr = gf_strdup (xattr);
-
- (void)snprintf (xattr_cky, 1024, "%s-%ld",
- local->cont.getxattr.name, cky);
- ret = dict_set_dynstr (local->dict,
- xattr_cky, xattr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret,
- AFR_MSG_DICT_SET_FAILED,
- "Cannot set xattr "
- "cookie key");
- goto unlock;
- }
-
- local->cont.getxattr.xattr_len += strlen (xattr) + 1;
- }
- }
-unlock:
- UNLOCK (&frame->lock);
+ if (!dict || (op_ret < 0))
+ goto unlock;
+
+ if (!local->dict)
+ local->dict = dict_new();
- if (!callcnt) {
- if (!local->cont.getxattr.xattr_len)
- goto unwind;
+ if (local->dict) {
+ ret = dict_get_str(dict, local->cont.getxattr.name, &xattr);
+ if (ret)
+ goto unlock;
- nxattr = dict_new ();
- if (!nxattr)
- goto unwind;
+ xattr = gf_strdup(xattr);
- /* extra bytes for decorations (brackets and <>'s) */
- padding += strlen (this->name) + SLEN (AFR_PATHINFO_HEADER) + 4;
- local->cont.getxattr.xattr_len += (padding + 2);
+ (void)snprintf(xattr_cky, 1024, "%s-%ld", local->cont.getxattr.name,
+ cky);
+ ret = dict_set_dynstr(local->dict, xattr_cky, xattr);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set xattr "
+ "cookie key");
+ goto unlock;
+ }
- xattr_serz = GF_MALLOC (local->cont.getxattr.xattr_len,
- gf_common_mt_char);
+ local->cont.getxattr.xattr_len += strlen(xattr) + 1;
+ }
+ }
+unlock:
+ UNLOCK(&frame->lock);
- if (!xattr_serz)
- goto unwind;
+ if (!callcnt) {
+ if (!local->cont.getxattr.xattr_len)
+ goto unwind;
- /* the xlator info */
- int xattr_serz_len = sprintf (xattr_serz,
- "(<"AFR_PATHINFO_HEADER"%s> ",
- this->name);
+ nxattr = dict_new();
+ if (!nxattr)
+ goto unwind;
- /* actual series of pathinfo */
- ret = dict_serialize_value_with_delim (local->dict,
- xattr_serz + xattr_serz_len,
- &tlen, ' ');
- if (ret) {
- goto unwind;
- }
+ /* extra bytes for decorations (brackets and <>'s) */
+ padding += strlen(this->name) + SLEN(AFR_PATHINFO_HEADER) + 4;
+ local->cont.getxattr.xattr_len += (padding + 2);
- /* closing part */
- *(xattr_serz + padding + tlen) = ')';
- *(xattr_serz + padding + tlen + 1) = '\0';
+ xattr_serz = GF_MALLOC(local->cont.getxattr.xattr_len,
+ gf_common_mt_char);
- ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
- xattr_serz);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Cannot set pathinfo key in dict");
+ if (!xattr_serz)
+ goto unwind;
- unwind:
- AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, nxattr, local->xdata_rsp);
+ /* the xlator info */
+ int xattr_serz_len = sprintf(
+ xattr_serz, "(<" AFR_PATHINFO_HEADER "%s> ", this->name);
- if (nxattr)
- dict_unref (nxattr);
+ /* actual series of pathinfo */
+ ret = dict_serialize_value_with_delim(
+ local->dict, xattr_serz + xattr_serz_len, &tlen, ' ');
+ if (ret) {
+ goto unwind;
}
+ /* closing part */
+ *(xattr_serz + padding + tlen) = ')';
+ *(xattr_serz + padding + tlen + 1) = '\0';
+
+ ret = dict_set_dynstr(nxattr, local->cont.getxattr.name, xattr_serz);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set pathinfo key in dict");
+
+ unwind:
+ AFR_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno,
+ nxattr, local->xdata_rsp);
+
+ if (nxattr)
+ dict_unref(nxattr);
+ }
+
out:
- return ret;
+ return ret;
}
static int
-afr_aggregate_stime_xattr (dict_t *this, char *key, data_t *value, void *data)
+afr_aggregate_stime_xattr(dict_t *this, char *key, data_t *value, void *data)
{
- int ret = 0;
+ int ret = 0;
- if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
- ret = gf_get_max_stime (THIS, data, key, value);
+ if (fnmatch(GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
+ ret = gf_get_max_stime(THIS, data, key, value);
- return ret;
+ return ret;
}
int32_t
-afr_common_getxattr_stime_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_common_getxattr_stime_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t callcnt = 0;
-
- if (!frame || !frame->local || !this) {
- gf_msg ("", GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_ARG, "possible NULL deref");
- goto out;
- }
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
- local = frame->local;
+ if (!frame || !frame->local || !this) {
+ gf_msg("", GF_LOG_ERROR, 0, AFR_MSG_INVALID_ARG, "possible NULL deref");
+ goto out;
+ }
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
+ local = frame->local;
- if (!dict || (op_ret < 0)) {
- local->op_errno = op_errno;
- goto cleanup;
- }
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
- if (!local->dict)
- local->dict = dict_copy_with_ref (dict, NULL);
- else
- dict_foreach (dict, afr_aggregate_stime_xattr,
- local->dict);
- local->op_ret = 0;
+ if (!dict || (op_ret < 0)) {
+ local->op_errno = op_errno;
+ goto cleanup;
}
+ if (!local->dict)
+ local->dict = dict_copy_with_ref(dict, NULL);
+ else
+ dict_foreach(dict, afr_aggregate_stime_xattr, local->dict);
+ local->op_ret = 0;
+ }
+
cleanup:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
- if (!callcnt) {
- AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, local->dict, xdata);
- }
+ if (!callcnt) {
+ AFR_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno,
+ local->dict, xdata);
+ }
out:
- return 0;
+ return 0;
}
-
static gf_boolean_t
-afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk,
- gf_boolean_t is_fgetxattr)
+afr_is_special_xattr(const char *name, fop_getxattr_cbk_t *cbk,
+ gf_boolean_t is_fgetxattr)
{
- gf_boolean_t is_spl = _gf_true;
-
- GF_ASSERT (cbk);
- if (!cbk || !name) {
- is_spl = _gf_false;
- goto out;
+ gf_boolean_t is_spl = _gf_true;
+
+ GF_ASSERT(cbk);
+ if (!cbk || !name) {
+ is_spl = _gf_false;
+ goto out;
+ }
+
+ if (!strcmp(name, GF_XATTR_PATHINFO_KEY) ||
+ !strcmp(name, GF_XATTR_USER_PATHINFO_KEY)) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_pathinfo_cbk;
+ } else {
+ *cbk = afr_getxattr_pathinfo_cbk;
}
-
- if (!strcmp (name, GF_XATTR_PATHINFO_KEY) ||
- !strcmp (name, GF_XATTR_USER_PATHINFO_KEY)) {
- if (is_fgetxattr) {
- *cbk = afr_fgetxattr_pathinfo_cbk;
- } else {
- *cbk = afr_getxattr_pathinfo_cbk;
- }
- } else if (!strncmp (name, GF_XATTR_CLRLK_CMD,
- SLEN (GF_XATTR_CLRLK_CMD))) {
- if (is_fgetxattr) {
- *cbk = afr_fgetxattr_clrlk_cbk;
- } else {
- *cbk = afr_getxattr_clrlk_cbk;
- }
- } else if (!strncmp (name, GF_XATTR_LOCKINFO_KEY,
- SLEN (GF_XATTR_LOCKINFO_KEY))) {
- if (is_fgetxattr) {
- *cbk = afr_fgetxattr_lockinfo_cbk;
- } else {
- *cbk = afr_getxattr_lockinfo_cbk;
- }
- } else if (fnmatch (GF_XATTR_STIME_PATTERN, name, FNM_NOESCAPE) == 0) {
- *cbk = afr_common_getxattr_stime_cbk;
- } else if (strcmp (name, QUOTA_SIZE_KEY) == 0) {
- *cbk = afr_getxattr_quota_size_cbk;
- } else if (!strcmp (name, GF_XATTR_LIST_NODE_UUIDS_KEY)) {
- *cbk = afr_getxattr_list_node_uuids_cbk;
+ } else if (!strncmp(name, GF_XATTR_CLRLK_CMD, SLEN(GF_XATTR_CLRLK_CMD))) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_clrlk_cbk;
+ } else {
+ *cbk = afr_getxattr_clrlk_cbk;
+ }
+ } else if (!strncmp(name, GF_XATTR_LOCKINFO_KEY,
+ SLEN(GF_XATTR_LOCKINFO_KEY))) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_lockinfo_cbk;
} else {
- is_spl = _gf_false;
+ *cbk = afr_getxattr_lockinfo_cbk;
}
+ } else if (fnmatch(GF_XATTR_STIME_PATTERN, name, FNM_NOESCAPE) == 0) {
+ *cbk = afr_common_getxattr_stime_cbk;
+ } else if (strcmp(name, QUOTA_SIZE_KEY) == 0) {
+ *cbk = afr_getxattr_quota_size_cbk;
+ } else if (!strcmp(name, GF_XATTR_LIST_NODE_UUIDS_KEY)) {
+ *cbk = afr_getxattr_list_node_uuids_cbk;
+ } else {
+ is_spl = _gf_false;
+ }
out:
- return is_spl;
+ return is_spl;
}
static void
-afr_getxattr_all_subvols (xlator_t *this, call_frame_t *frame,
- const char *name, loc_t *loc,
- fop_getxattr_cbk_t cbk)
+afr_getxattr_all_subvols(xlator_t *this, call_frame_t *frame, const char *name,
+ loc_t *loc, fop_getxattr_cbk_t cbk)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int call_count = 0;
-
- priv = this->private;
-
- local = frame->local;
- //local->call_count set in afr_local_init
- call_count = local->call_count;
-
- if (!strcmp (name, GF_XATTR_LIST_NODE_UUIDS_KEY)) {
- GF_FREE (local->cont.getxattr.name);
- local->cont.getxattr.name = gf_strdup (GF_XATTR_NODE_UUID_KEY);
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+
+ priv = this->private;
+
+ local = frame->local;
+ // local->call_count set in afr_local_init
+ call_count = local->call_count;
+
+ if (!strcmp(name, GF_XATTR_LIST_NODE_UUIDS_KEY)) {
+ GF_FREE(local->cont.getxattr.name);
+ local->cont.getxattr.name = gf_strdup(GF_XATTR_NODE_UUID_KEY);
+ }
+
+ // If up-children count is 0, afr_local_init would have failed already
+ // and the call would have unwound so not handling it here.
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->getxattr, loc,
+ local->cont.getxattr.name, NULL);
+ if (!--call_count)
+ break;
}
-
- //If up-children count is 0, afr_local_init would have failed already
- //and the call would have unwound so not handling it here.
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->getxattr,
- loc, local->cont.getxattr.name,
- NULL);
- if (!--call_count)
- break;
- }
- }
- return;
+ }
+ return;
}
int
-afr_marker_populate_args (call_frame_t *frame, int type, int *gauge,
- xlator_t **subvols)
+afr_marker_populate_args(call_frame_t *frame, int type, int *gauge,
+ xlator_t **subvols)
{
- xlator_t *this = frame->this;
- afr_private_t *priv = this->private;
+ xlator_t *this = frame->this;
+ afr_private_t *priv = this->private;
- memcpy (subvols, priv->children, sizeof (*subvols) * priv->child_count);
+ memcpy(subvols, priv->children, sizeof(*subvols) * priv->child_count);
- if (type == MARKER_XTIME_TYPE) {
- /*Don't error out on ENOENT/ENOTCONN */
- gauge[MCNT_NOTFOUND] = 0;
- gauge[MCNT_ENOTCONN] = 0;
- }
- return priv->child_count;
+ if (type == MARKER_XTIME_TYPE) {
+ /*Don't error out on ENOENT/ENOTCONN */
+ gauge[MCNT_NOTFOUND] = 0;
+ gauge[MCNT_ENOTCONN] = 0;
+ }
+ return priv->child_count;
}
static int
-afr_handle_heal_xattrs (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *heal_op)
+afr_handle_heal_xattrs(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *heal_op)
{
- int ret = -1;
- afr_spb_status_t *data = NULL;
+ int ret = -1;
+ afr_spb_status_t *data = NULL;
- if (!strcmp (heal_op, GF_HEAL_INFO)) {
- afr_get_heal_info (frame, this, loc);
- ret = 0;
- goto out;
- }
+ if (!strcmp(heal_op, GF_HEAL_INFO)) {
+ afr_get_heal_info(frame, this, loc);
+ ret = 0;
+ goto out;
+ }
- if (!strcmp (heal_op, GF_AFR_HEAL_SBRAIN)) {
- afr_heal_splitbrain_file (frame, this, loc);
- ret = 0;
- goto out;
+ if (!strcmp(heal_op, GF_AFR_HEAL_SBRAIN)) {
+ afr_heal_splitbrain_file(frame, this, loc);
+ ret = 0;
+ goto out;
+ }
+
+ if (!strcmp(heal_op, GF_AFR_SBRAIN_STATUS)) {
+ data = GF_CALLOC(1, sizeof(*data), gf_afr_mt_spb_status_t);
+ if (!data) {
+ ret = 1;
+ goto out;
}
-
- if (!strcmp (heal_op, GF_AFR_SBRAIN_STATUS)) {
- data = GF_CALLOC (1, sizeof (*data), gf_afr_mt_spb_status_t);
- if (!data) {
- ret = 1;
- goto out;
- }
- data->frame = frame;
- data->loc = loc;
- ret = synctask_new (this->ctx->env,
- afr_get_split_brain_status,
- afr_get_split_brain_status_cbk,
- NULL, data);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN_STATUS,
- "Failed to create"
- " synctask. Unable to fetch split-brain status"
- " for %s.", loc->name);
- ret = 1;
- goto out;
- }
- goto out;
+ data->frame = frame;
+ data->loc = loc;
+ ret = synctask_new(this->ctx->env, afr_get_split_brain_status,
+ afr_get_split_brain_status_cbk, NULL, data);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN_STATUS,
+ "Failed to create"
+ " synctask. Unable to fetch split-brain status"
+ " for %s.",
+ loc->name);
+ ret = 1;
+ goto out;
}
+ goto out;
+ }
out:
- if (ret == 1) {
- AFR_STACK_UNWIND (getxattr, frame, -1, ENOMEM, NULL, NULL);
- if (data)
- GF_FREE (data);
- ret = 0;
- }
- return ret;
+ if (ret == 1) {
+ AFR_STACK_UNWIND(getxattr, frame, -1, ENOMEM, NULL, NULL);
+ if (data)
+ GF_FREE(data);
+ ret = 0;
+ }
+ return ret;
}
int32_t
-afr_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+afr_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- xlator_t **children = NULL;
- int i = 0;
- int32_t op_errno = 0;
- int ret = -1;
- fop_getxattr_cbk_t cbk = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int i = 0;
+ int32_t op_errno = 0;
+ int ret = -1;
+ fop_getxattr_cbk_t cbk = NULL;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ priv = this->private;
- priv = this->private;
+ children = priv->children;
- children = priv->children;
+ loc_copy(&local->loc, loc);
- loc_copy (&local->loc, loc);
+ local->op = GF_FOP_GETXATTR;
- local->op = GF_FOP_GETXATTR;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ if (!name)
+ goto no_name;
- if (!name)
- goto no_name;
+ local->cont.getxattr.name = gf_strdup(name);
- local->cont.getxattr.name = gf_strdup (name);
+ if (!local->cont.getxattr.name) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- if (!local->cont.getxattr.name) {
- op_errno = ENOMEM;
- goto out;
- }
-
- if (!strncmp (name, AFR_XATTR_PREFIX,
- SLEN (AFR_XATTR_PREFIX))) {
- op_errno = ENODATA;
- goto out;
- }
+ if (!strncmp(name, AFR_XATTR_PREFIX, SLEN(AFR_XATTR_PREFIX))) {
+ op_errno = ENODATA;
+ goto out;
+ }
- if (cluster_handle_marker_getxattr (frame, loc, name, priv->vol_uuid,
- afr_getxattr_unwind,
- afr_marker_populate_args) == 0)
- return 0;
-
- ret = afr_handle_heal_xattrs (frame, this, &local->loc, name);
- if (ret == 0)
- return 0;
+ if (cluster_handle_marker_getxattr(frame, loc, name, priv->vol_uuid,
+ afr_getxattr_unwind,
+ afr_marker_populate_args) == 0)
+ return 0;
- /*
- * Heal daemons don't have IO threads ... and as a result they
- * send this getxattr down and eventually crash :(
- */
- op_errno = -1;
- GF_CHECK_XATTR_KEY_AND_GOTO(name, IO_THREADS_QUEUE_SIZE_KEY, op_errno,
- out);
+ ret = afr_handle_heal_xattrs(frame, this, &local->loc, name);
+ if (ret == 0)
+ return 0;
- /*
- * Special xattrs which need responses from all subvols
- */
- if (afr_is_special_xattr (name, &cbk, 0)) {
- afr_getxattr_all_subvols (this, frame, name, loc, cbk);
- return 0;
- }
+ /*
+ * Heal daemons don't have IO threads ... and as a result they
+ * send this getxattr down and eventually crash :(
+ */
+ op_errno = -1;
+ GF_CHECK_XATTR_KEY_AND_GOTO(name, IO_THREADS_QUEUE_SIZE_KEY, op_errno, out);
+
+ /*
+ * Special xattrs which need responses from all subvols
+ */
+ if (afr_is_special_xattr(name, &cbk, 0)) {
+ afr_getxattr_all_subvols(this, frame, name, loc, cbk);
+ return 0;
+ }
- if (XATTR_IS_NODE_UUID (name)) {
- i = 0;
- STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
- (void *) (long) i,
- children[i],
- children[i]->fops->getxattr,
- loc, name, xdata);
- return 0;
- }
+ if (XATTR_IS_NODE_UUID(name)) {
+ i = 0;
+ STACK_WIND_COOKIE(frame, afr_getxattr_node_uuid_cbk, (void *)(long)i,
+ children[i], children[i]->fops->getxattr, loc, name,
+ xdata);
+ return 0;
+ }
no_name:
- afr_read_txn (frame, this, local->loc.inode, afr_getxattr_wind,
- AFR_METADATA_TRANSACTION);
+ afr_read_txn(frame, this, local->loc.inode, afr_getxattr_wind,
+ AFR_METADATA_TRANSACTION);
- ret = 0;
+ ret = 0;
out:
- if (ret < 0)
- AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
- return 0;
+ if (ret < 0)
+ AFR_STACK_UNWIND(getxattr, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
/* {{{ fgetxattr */
-
int32_t
-afr_fgetxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- if (dict)
- afr_filter_xattrs (dict);
+ if (dict)
+ afr_filter_xattrs(dict);
- AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ AFR_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ return 0;
}
int
-afr_fgetxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fgetxattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (fgetxattr, frame, local->op_ret,
- local->op_errno, NULL, NULL);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_fgetxattr_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fgetxattr,
- local->fd, local->cont.getxattr.name,
- local->xdata_req);
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ local = frame->local;
+ priv = this->private;
-static void
-afr_fgetxattr_all_subvols (xlator_t *this, call_frame_t *frame,
- fop_fgetxattr_cbk_t cbk)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int call_count = 0;
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(fgetxattr, frame, local->op_ret, local->op_errno, NULL,
+ NULL);
+ return 0;
+ }
- priv = this->private;
+ STACK_WIND_COOKIE(frame, afr_fgetxattr_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fgetxattr, local->fd,
+ local->cont.getxattr.name, local->xdata_req);
+ return 0;
+}
- local = frame->local;
- //local->call_count set in afr_local_init
- call_count = local->call_count;
-
- //If up-children count is 0, afr_local_init would have failed already
- //and the call would have unwound so not handling it here.
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fgetxattr,
- local->fd, local->cont.getxattr.name,
- NULL);
- if (!--call_count)
- break;
- }
+static void
+afr_fgetxattr_all_subvols(xlator_t *this, call_frame_t *frame,
+ fop_fgetxattr_cbk_t cbk)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+
+ priv = this->private;
+
+ local = frame->local;
+ // local->call_count set in afr_local_init
+ call_count = local->call_count;
+
+ // If up-children count is 0, afr_local_init would have failed already
+ // and the call would have unwound so not handling it here.
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->fgetxattr, local->fd,
+ local->cont.getxattr.name, NULL);
+ if (!--call_count)
+ break;
}
+ }
- return;
+ return;
}
-
int
-afr_fgetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
+afr_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t op_errno = 0;
- fop_fgetxattr_cbk_t cbk = NULL;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_FGETXATTR;
- local->fd = fd_ref (fd);
- if (name) {
- local->cont.getxattr.name = gf_strdup (name);
- if (!local->cont.getxattr.name) {
- op_errno = ENOMEM;
- goto out;
- }
- }
- if (xdata)
- local->xdata_req = dict_ref (xdata);
-
- /* pathinfo gets handled only in getxattr(), but we need to handle
- * lockinfo.
- * If we are doing fgetxattr with lockinfo as the key then we
- * collect information from all children.
- */
- if (afr_is_special_xattr (name, &cbk, 1)) {
- afr_fgetxattr_all_subvols (this, frame, cbk);
- return 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
+ fop_fgetxattr_cbk_t cbk = NULL;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_FGETXATTR;
+ local->fd = fd_ref(fd);
+ if (name) {
+ local->cont.getxattr.name = gf_strdup(name);
+ if (!local->cont.getxattr.name) {
+ op_errno = ENOMEM;
+ goto out;
}
+ }
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+
+ /* pathinfo gets handled only in getxattr(), but we need to handle
+ * lockinfo.
+ * If we are doing fgetxattr with lockinfo as the key then we
+ * collect information from all children.
+ */
+ if (afr_is_special_xattr(name, &cbk, 1)) {
+ afr_fgetxattr_all_subvols(this, frame, cbk);
+ return 0;
+ }
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_read_txn (frame, this, fd->inode, afr_fgetxattr_wind,
- AFR_METADATA_TRANSACTION);
+ afr_read_txn(frame, this, fd->inode, afr_fgetxattr_wind,
+ AFR_METADATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(fgetxattr, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
/* }}} */
/* {{{ readv */
int
-afr_readv_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iovec *vector, int32_t count, struct iatt *buf,
- struct iobref *iobref, dict_t *xdata)
+afr_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iovec *vector, int32_t count,
+ struct iatt *buf, struct iobref *iobref, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
- vector, count, buf, iobref, xdata);
- return 0;
+ AFR_STACK_UNWIND(readv, frame, op_ret, op_errno, vector, count, buf, iobref,
+ xdata);
+ return 0;
}
-
int
-afr_readv_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_readv_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (readv, frame, local->op_ret, local->op_errno,
- 0, 0, 0, 0, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_readv_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->readv,
- local->fd, local->cont.readv.size,
- local->cont.readv.offset, local->cont.readv.flags,
- local->xdata_req);
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(readv, frame, local->op_ret, local->op_errno, 0, 0, 0,
+ 0, 0);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(
+ frame, afr_readv_cbk, (void *)(long)subvol, priv->children[subvol],
+ priv->children[subvol]->fops->readv, local->fd, local->cont.readv.size,
+ local->cont.readv.offset, local->cont.readv.flags, local->xdata_req);
+ return 0;
+}
int
-afr_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+afr_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- afr_local_t * local = NULL;
- int32_t op_errno = 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_READ;
- local->fd = fd_ref (fd);
- local->cont.readv.size = size;
- local->cont.readv.offset = offset;
- local->cont.readv.flags = flags;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_READ;
+ local->fd = fd_ref(fd);
+ local->cont.readv.size = size;
+ local->cont.readv.offset = offset;
+ local->cont.readv.flags = flags;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_read_txn (frame, this, fd->inode, afr_readv_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, fd->inode, afr_readv_wind, AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND(readv, frame, -1, op_errno, 0, 0, 0, 0, 0);
+ AFR_STACK_UNWIND(readv, frame, -1, op_errno, 0, 0, 0, 0, 0);
- return 0;
+ return 0;
}
/* }}} */
@@ -1910,77 +1814,72 @@ out:
/* {{{ seek */
int
-afr_seek_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, off_t offset, dict_t *xdata)
+afr_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, off_t offset, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
-
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- AFR_STACK_UNWIND (seek, frame, op_ret, op_errno, offset, xdata);
+ afr_read_txn_continue(frame, this, (long)cookie);
return 0;
-}
+ }
+ AFR_STACK_UNWIND(seek, frame, op_ret, op_errno, offset, xdata);
+ return 0;
+}
int
-afr_seek_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_seek_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (seek, frame, local->op_ret, local->op_errno,
- 0, NULL);
- return 0;
- }
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_seek_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->seek,
- local->fd, local->cont.seek.offset,
- local->cont.seek.what, local->xdata_req);
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(seek, frame, local->op_ret, local->op_errno, 0, NULL);
return 0;
-}
+ }
+ STACK_WIND_COOKIE(
+ frame, afr_seek_cbk, (void *)(long)subvol, priv->children[subvol],
+ priv->children[subvol]->fops->seek, local->fd, local->cont.seek.offset,
+ local->cont.seek.what, local->xdata_req);
+ return 0;
+}
int
-afr_seek (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- gf_seek_what_t what, dict_t *xdata)
+afr_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t op_errno = 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_SEEK;
- local->fd = fd_ref (fd);
- local->cont.seek.offset = offset;
- local->cont.seek.what = what;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_SEEK;
+ local->fd = fd_ref(fd);
+ local->cont.seek.offset = offset;
+ local->cont.seek.what = what;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_read_txn (frame, this, fd->inode, afr_seek_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, fd->inode, afr_seek_wind, AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (seek, frame, -1, op_errno, 0, NULL);
+ AFR_STACK_UNWIND(seek, frame, -1, op_errno, 0, NULL);
- return 0;
+ return 0;
}
/* }}} */
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 755e928ef62..f9e2b302f8d 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include <libgen.h>
#include <unistd.h>
#include <fnmatch.h>
@@ -35,771 +34,736 @@
#include "afr-messages.h"
static void
-__afr_inode_write_finalize (call_frame_t *frame, xlator_t *this)
-{
- int i = 0;
- int ret = 0;
- int read_subvol = 0;
- struct iatt *stbuf = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_read_subvol_args_t args = {0,};
-
- local = frame->local;
- priv = this->private;
- GF_VALIDATE_OR_GOTO (this->name, local->inode, out);
-
- /*This code needs to stay till DHT sends fops on linked
- * inodes*/
- if (!inode_is_linked (local->inode)) {
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret == -1)
- continue;
- if (!gf_uuid_is_null
- (local->replies[i].poststat.ia_gfid)) {
- gf_uuid_copy (args.gfid,
- local->replies[i].poststat.ia_gfid);
- args.ia_type =
- local->replies[i].poststat.ia_type;
- break;
- } else {
- ret = dict_get_bin (local->replies[i].xdata,
- DHT_IATT_IN_XDATA_KEY,
- (void **) &stbuf);
- if (ret)
- continue;
- gf_uuid_copy (args.gfid, stbuf->ia_gfid);
- args.ia_type = stbuf->ia_type;
- break;
- }
- }
+__afr_inode_write_finalize(call_frame_t *frame, xlator_t *this)
+{
+ int i = 0;
+ int ret = 0;
+ int read_subvol = 0;
+ struct iatt *stbuf = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_read_subvol_args_t args = {
+ 0,
+ };
+
+ local = frame->local;
+ priv = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, local->inode, out);
+
+ /*This code needs to stay till DHT sends fops on linked
+ * inodes*/
+ if (!inode_is_linked(local->inode)) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret == -1)
+ continue;
+ if (!gf_uuid_is_null(local->replies[i].poststat.ia_gfid)) {
+ gf_uuid_copy(args.gfid, local->replies[i].poststat.ia_gfid);
+ args.ia_type = local->replies[i].poststat.ia_type;
+ break;
+ } else {
+ ret = dict_get_bin(local->replies[i].xdata,
+ DHT_IATT_IN_XDATA_KEY, (void **)&stbuf);
+ if (ret)
+ continue;
+ gf_uuid_copy(args.gfid, stbuf->ia_gfid);
+ args.ia_type = stbuf->ia_type;
+ break;
+ }
}
-
- if (local->transaction.type == AFR_METADATA_TRANSACTION) {
- read_subvol = afr_metadata_subvol_get (local->inode, this,
- NULL, local->readable,
- NULL, &args);
- } else {
- read_subvol = afr_data_subvol_get (local->inode, this,
- NULL, local->readable,
- NULL, &args);
+ }
+
+ if (local->transaction.type == AFR_METADATA_TRANSACTION) {
+ read_subvol = afr_metadata_subvol_get(local->inode, this, NULL,
+ local->readable, NULL, &args);
+ } else {
+ read_subvol = afr_data_subvol_get(local->inode, this, NULL,
+ local->readable, NULL, &args);
+ }
+
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ afr_pick_error_xdata(local, priv, local->inode, local->readable, NULL,
+ NULL);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret < 0)
+ continue;
+
+ /* Order of checks in the compound conditional
+ below is important.
+
+ - Highest precedence: largest op_ret
+ - Next precedence: if all op_rets are equal, read subvol
+ - Least precedence: any succeeded subvol
+ */
+ if ((local->op_ret < local->replies[i].op_ret) ||
+ ((local->op_ret == local->replies[i].op_ret) &&
+ (i == read_subvol))) {
+ local->op_ret = local->replies[i].op_ret;
+ local->op_errno = local->replies[i].op_errno;
+
+ local->cont.inode_wfop.prebuf = local->replies[i].prestat;
+ local->cont.inode_wfop.postbuf = local->replies[i].poststat;
+
+ if (local->replies[i].xdata) {
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ }
+ if (local->replies[i].xattr) {
+ if (local->xattr_rsp)
+ dict_unref(local->xattr_rsp);
+ local->xattr_rsp = dict_ref(local->replies[i].xattr);
+ }
}
+ }
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
- afr_pick_error_xdata (local, priv, local->inode, local->readable, NULL,
- NULL);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret < 0)
- continue;
-
- /* Order of checks in the compound conditional
- below is important.
-
- - Highest precedence: largest op_ret
- - Next precedence: if all op_rets are equal, read subvol
- - Least precedence: any succeeded subvol
- */
- if ((local->op_ret < local->replies[i].op_ret) ||
- ((local->op_ret == local->replies[i].op_ret) &&
- (i == read_subvol))) {
-
- local->op_ret = local->replies[i].op_ret;
- local->op_errno = local->replies[i].op_errno;
-
- local->cont.inode_wfop.prebuf =
- local->replies[i].prestat;
- local->cont.inode_wfop.postbuf =
- local->replies[i].poststat;
-
- if (local->replies[i].xdata) {
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- }
- if (local->replies[i].xattr) {
- if (local->xattr_rsp)
- dict_unref (local->xattr_rsp);
- local->xattr_rsp =
- dict_ref (local->replies[i].xattr);
- }
- }
- }
-
- afr_set_in_flight_sb_status (this, frame, local->inode);
+ afr_set_in_flight_sb_status(this, frame, local->inode);
out:
- return;
+ return;
}
-
static void
-__afr_inode_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
- int op_ret, int op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xattr, dict_t *xdata)
+__afr_inode_write_fill(call_frame_t *frame, xlator_t *this, int child_index,
+ int op_ret, int op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- local->replies[child_index].valid = 1;
+ local->replies[child_index].valid = 1;
- if (AFR_IS_ARBITER_BRICK(priv, child_index) && op_ret == 1)
- op_ret = iov_length (local->cont.writev.vector,
- local->cont.writev.count);
+ if (AFR_IS_ARBITER_BRICK(priv, child_index) && op_ret == 1)
+ op_ret = iov_length(local->cont.writev.vector,
+ local->cont.writev.count);
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
- if (op_ret >= 0) {
- if (prebuf)
- local->replies[child_index].prestat = *prebuf;
- if (postbuf)
- local->replies[child_index].poststat = *postbuf;
- if (xattr)
- local->replies[child_index].xattr = dict_ref (xattr);
- } else {
- afr_transaction_fop_failed (frame, this, child_index);
- }
+ if (op_ret >= 0) {
+ if (prebuf)
+ local->replies[child_index].prestat = *prebuf;
+ if (postbuf)
+ local->replies[child_index].poststat = *postbuf;
+ if (xattr)
+ local->replies[child_index].xattr = dict_ref(xattr);
+ } else {
+ afr_transaction_fop_failed(frame, this, child_index);
+ }
- return;
+ return;
}
-
static int
-__afr_inode_write_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xattr, dict_t *xdata)
+__afr_inode_write_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int child_index = (long) cookie;
- int call_count = -1;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int child_index = (long)cookie;
+ int call_count = -1;
+ afr_private_t *priv = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- __afr_inode_write_fill (frame, this, child_index, op_ret,
- op_errno, prebuf, postbuf, xattr,
- xdata);
- }
- UNLOCK (&frame->lock);
+ LOCK(&frame->lock);
+ {
+ __afr_inode_write_fill(frame, this, child_index, op_ret, op_errno,
+ prebuf, postbuf, xattr, xdata);
+ }
+ UNLOCK(&frame->lock);
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0) {
- __afr_inode_write_finalize (frame, this);
+ if (call_count == 0) {
+ __afr_inode_write_finalize(frame, this);
- if (afr_txn_nothing_failed (frame, this)) {
- /*if it did pre-op, it will do post-op changing ctime*/
- if (priv->consistent_metadata &&
- afr_needs_changelog_update (local))
- afr_zero_fill_stat (local);
- local->transaction.unwind (frame, this);
- }
-
- afr_transaction_resume (frame, this);
+ if (afr_txn_nothing_failed(frame, this)) {
+ /*if it did pre-op, it will do post-op changing ctime*/
+ if (priv->consistent_metadata && afr_needs_changelog_update(local))
+ afr_zero_fill_stat(local);
+ local->transaction.unwind(frame, this);
}
- return 0;
+ afr_transaction_resume(frame, this);
+ }
+
+ return 0;
}
/* {{{ writev */
void
-afr_writev_copy_outvars (call_frame_t *src_frame, call_frame_t *dst_frame)
+afr_writev_copy_outvars(call_frame_t *src_frame, call_frame_t *dst_frame)
{
- afr_local_t *src_local = NULL;
- afr_local_t *dst_local = NULL;
+ afr_local_t *src_local = NULL;
+ afr_local_t *dst_local = NULL;
- src_local = src_frame->local;
- dst_local = dst_frame->local;
+ src_local = src_frame->local;
+ dst_local = dst_frame->local;
- dst_local->op_ret = src_local->op_ret;
- dst_local->op_errno = src_local->op_errno;
- dst_local->cont.inode_wfop.prebuf = src_local->cont.inode_wfop.prebuf;
- dst_local->cont.inode_wfop.postbuf = src_local->cont.inode_wfop.postbuf;
- if (src_local->xdata_rsp)
- dst_local->xdata_rsp = dict_ref (src_local->xdata_rsp);
+ dst_local->op_ret = src_local->op_ret;
+ dst_local->op_errno = src_local->op_errno;
+ dst_local->cont.inode_wfop.prebuf = src_local->cont.inode_wfop.prebuf;
+ dst_local->cont.inode_wfop.postbuf = src_local->cont.inode_wfop.postbuf;
+ if (src_local->xdata_rsp)
+ dst_local->xdata_rsp = dict_ref(src_local->xdata_rsp);
}
void
-afr_writev_unwind (call_frame_t *frame, xlator_t *this)
+afr_writev_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- afr_private_t *priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = this->private;
- local = frame->local;
+ local = frame->local;
- if (priv->consistent_metadata)
- afr_zero_fill_stat (local);
+ if (priv->consistent_metadata)
+ afr_zero_fill_stat(local);
- AFR_STACK_UNWIND (writev, frame,
- local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf,
- local->xdata_rsp);
+ AFR_STACK_UNWIND(writev, frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
}
-
int
-afr_transaction_writev_unwind (call_frame_t *frame, xlator_t *this)
+afr_transaction_writev_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *fop_frame = NULL;
+ call_frame_t *fop_frame = NULL;
- fop_frame = afr_transaction_detach_fop_frame (frame);
+ fop_frame = afr_transaction_detach_fop_frame(frame);
- if (fop_frame) {
- afr_writev_copy_outvars (frame, fop_frame);
- afr_writev_unwind (fop_frame, this);
- }
- return 0;
+ if (fop_frame) {
+ afr_writev_copy_outvars(frame, fop_frame);
+ afr_writev_unwind(fop_frame, this);
+ }
+ return 0;
}
static void
-afr_writev_handle_short_writes (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
- /*
- * We already have the best case result of the writev calls staged
- * as the return value. Any writev that returns some value less
- * than the best case is now out of sync, so mark the fop as
- * failed. Note that fops that have returned with errors have
- * already been marked as failed.
- */
- for (i = 0; i < priv->child_count; i++) {
- if ((!local->replies[i].valid) ||
- (local->replies[i].op_ret == -1))
- continue;
-
- if (local->replies[i].op_ret < local->op_ret)
- afr_transaction_fop_failed (frame, this, i);
- }
+afr_writev_handle_short_writes(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+ /*
+ * We already have the best case result of the writev calls staged
+ * as the return value. Any writev that returns some value less
+ * than the best case is now out of sync, so mark the fop as
+ * failed. Note that fops that have returned with errors have
+ * already been marked as failed.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if ((!local->replies[i].valid) || (local->replies[i].op_ret == -1))
+ continue;
+
+ if (local->replies[i].op_ret < local->op_ret)
+ afr_transaction_fop_failed(frame, this, i);
+ }
}
void
-afr_inode_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
+afr_inode_write_fill(call_frame_t *frame, xlator_t *this, int child_index,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata)
{
- int ret = 0;
- afr_local_t *local = frame->local;
- uint32_t open_fd_count = 0;
- uint32_t write_is_append = 0;
-
- LOCK (&frame->lock);
- {
- __afr_inode_write_fill (frame, this, child_index, op_ret,
- op_errno, prebuf, postbuf, NULL, xdata);
- if (op_ret == -1 || !xdata)
- goto unlock;
-
- write_is_append = 0;
- ret = dict_get_uint32 (xdata, GLUSTERFS_WRITE_IS_APPEND,
- &write_is_append);
- if (ret || !write_is_append)
- local->append_write = _gf_false;
-
- ret = dict_get_uint32 (xdata, GLUSTERFS_ACTIVE_FD_COUNT,
- &open_fd_count);
- if (ret < 0)
- goto unlock;
- if (open_fd_count > local->open_fd_count) {
- local->open_fd_count = open_fd_count;
- local->update_open_fd_count = _gf_true;
- }
+ int ret = 0;
+ afr_local_t *local = frame->local;
+ uint32_t open_fd_count = 0;
+ uint32_t write_is_append = 0;
+
+ LOCK(&frame->lock);
+ {
+ __afr_inode_write_fill(frame, this, child_index, op_ret, op_errno,
+ prebuf, postbuf, NULL, xdata);
+ if (op_ret == -1 || !xdata)
+ goto unlock;
+
+ write_is_append = 0;
+ ret = dict_get_uint32(xdata, GLUSTERFS_WRITE_IS_APPEND,
+ &write_is_append);
+ if (ret || !write_is_append)
+ local->append_write = _gf_false;
+
+ ret = dict_get_uint32(xdata, GLUSTERFS_ACTIVE_FD_COUNT, &open_fd_count);
+ if (ret < 0)
+ goto unlock;
+ if (open_fd_count > local->open_fd_count) {
+ local->open_fd_count = open_fd_count;
+ local->update_open_fd_count = _gf_true;
}
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
}
void
-afr_process_post_writev (call_frame_t *frame, xlator_t *this)
+afr_process_post_writev(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local->stable_write && !local->append_write)
- /* An appended write removes the necessity to
- fsync() the file. This is because self-heal
- has the logic to check for larger file when
- the xattrs are not reliably pointing at
- a stale file.
- */
- afr_fd_report_unstable_write (this, local);
+ if (!local->stable_write && !local->append_write)
+ /* An appended write removes the necessity to
+ fsync() the file. This is because self-heal
+ has the logic to check for larger file when
+ the xattrs are not reliably pointing at
+ a stale file.
+ */
+ afr_fd_report_unstable_write(this, local);
- __afr_inode_write_finalize (frame, this);
+ __afr_inode_write_finalize(frame, this);
- afr_writev_handle_short_writes (frame, this);
-
- if (local->update_open_fd_count)
- local->inode_ctx->open_fd_count = local->open_fd_count;
+ afr_writev_handle_short_writes(frame, this);
+ if (local->update_open_fd_count)
+ local->inode_ctx->open_fd_count = local->open_fd_count;
}
int
-afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_writev_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- call_frame_t *fop_frame = NULL;
- int child_index = (long) cookie;
- int call_count = -1;
-
- afr_inode_write_fill (frame, this, child_index, op_ret, op_errno,
- prebuf, postbuf, xdata);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- afr_process_post_writev (frame, this);
-
- if (!afr_txn_nothing_failed (frame, this)) {
- //Don't unwind until post-op is complete
- afr_transaction_resume (frame, this);
- } else {
- /*
- * Generally inode-write fops do transaction.unwind then
- * transaction.resume, but writev needs to make sure that
- * delayed post-op frame is placed in fdctx before unwind
- * happens. This prevents the race of flush doing the
- * changelog wakeup first in fuse thread and then this
- * writev placing its delayed post-op frame in fdctx.
- * This helps flush make sure all the delayed post-ops are
- * completed.
- */
-
- fop_frame = afr_transaction_detach_fop_frame (frame);
- afr_writev_copy_outvars (frame, fop_frame);
- afr_transaction_resume (frame, this);
- afr_writev_unwind (fop_frame, this);
- }
+ call_frame_t *fop_frame = NULL;
+ int child_index = (long)cookie;
+ int call_count = -1;
+
+ afr_inode_write_fill(frame, this, child_index, op_ret, op_errno, prebuf,
+ postbuf, xdata);
+
+ call_count = afr_frame_return(frame);
+
+ if (call_count == 0) {
+ afr_process_post_writev(frame, this);
+
+ if (!afr_txn_nothing_failed(frame, this)) {
+ // Don't unwind until post-op is complete
+ afr_transaction_resume(frame, this);
+ } else {
+ /*
+ * Generally inode-write fops do transaction.unwind then
+ * transaction.resume, but writev needs to make sure that
+ * delayed post-op frame is placed in fdctx before unwind
+ * happens. This prevents the race of flush doing the
+ * changelog wakeup first in fuse thread and then this
+ * writev placing its delayed post-op frame in fdctx.
+ * This helps flush make sure all the delayed post-ops are
+ * completed.
+ */
+
+ fop_frame = afr_transaction_detach_fop_frame(frame);
+ afr_writev_copy_outvars(frame, fop_frame);
+ afr_transaction_resume(frame, this);
+ afr_writev_unwind(fop_frame, this);
}
- return 0;
+ }
+ return 0;
}
static int
-afr_arbiter_writev_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_arbiter_writev_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = frame->local;
- afr_private_t *priv = this->private;
- static char byte = 0xFF;
- static struct iovec vector = {&byte, 1};
- int32_t count = 1;
+ afr_local_t *local = frame->local;
+ afr_private_t *priv = this->private;
+ static char byte = 0xFF;
+ static struct iovec vector = {&byte, 1};
+ int32_t count = 1;
- STACK_WIND_COOKIE (frame, afr_writev_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->writev,
- local->fd, &vector, count, local->cont.writev.offset,
- local->cont.writev.flags, local->cont.writev.iobref,
- local->xdata_req);
+ STACK_WIND_COOKIE(
+ frame, afr_writev_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol], priv->children[subvol]->fops->writev, local->fd,
+ &vector, count, local->cont.writev.offset, local->cont.writev.flags,
+ local->cont.writev.iobref, local->xdata_req);
- return 0;
+ return 0;
}
int
-afr_writev_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_writev_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
-
- if (AFR_IS_ARBITER_BRICK(priv, subvol)) {
- afr_arbiter_writev_wind (frame, this, subvol);
- return 0;
- }
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_writev_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->writev,
- local->fd, local->cont.writev.vector,
- local->cont.writev.count, local->cont.writev.offset,
- local->cont.writev.flags, local->cont.writev.iobref,
- local->xdata_req);
+ if (AFR_IS_ARBITER_BRICK(priv, subvol)) {
+ afr_arbiter_writev_wind(frame, this, subvol);
return 0;
-}
+ }
+ STACK_WIND_COOKIE(frame, afr_writev_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->writev, local->fd,
+ local->cont.writev.vector, local->cont.writev.count,
+ local->cont.writev.offset, local->cont.writev.flags,
+ local->cont.writev.iobref, local->xdata_req);
+ return 0;
+}
int
-afr_do_writev (call_frame_t *frame, xlator_t *this)
+afr_do_writev(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *transaction_frame = NULL;
- afr_local_t *local = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = frame->local;
- transaction_frame->local = local;
- frame->local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- if (!AFR_FRAME_INIT (frame, op_errno))
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local->op = GF_FOP_WRITE;
+ local = frame->local;
+ transaction_frame->local = local;
+ frame->local = NULL;
- local->transaction.wind = afr_writev_wind;
- local->transaction.unwind = afr_transaction_writev_unwind;
+ if (!AFR_FRAME_INIT(frame, op_errno))
+ goto out;
- local->transaction.main_frame = frame;
+ local->op = GF_FOP_WRITE;
- if (local->fd->flags & O_APPEND) {
- /*
- * Backend vfs ignores the 'offset' for append mode fd so
- * locking just the region provided for the writev does not
- * give consistency guarantee. The actual write may happen at a
- * completely different range than the one provided by the
- * offset, len in the fop. So lock the entire file.
- */
- local->transaction.start = 0;
- local->transaction.len = 0;
- } else {
- local->transaction.start = local->cont.writev.offset;
- local->transaction.len = iov_length (local->cont.writev.vector,
- local->cont.writev.count);
- }
+ local->transaction.wind = afr_writev_wind;
+ local->transaction.unwind = afr_transaction_writev_unwind;
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ local->transaction.main_frame = frame;
- return 0;
+ if (local->fd->flags & O_APPEND) {
+ /*
+ * Backend vfs ignores the 'offset' for append mode fd so
+ * locking just the region provided for the writev does not
+ * give consistency guarantee. The actual write may happen at a
+ * completely different range than the one provided by the
+ * offset, len in the fop. So lock the entire file.
+ */
+ local->transaction.start = 0;
+ local->transaction.len = 0;
+ } else {
+ local->transaction.start = local->cont.writev.offset;
+ local->transaction.len = iov_length(local->cont.writev.vector,
+ local->cont.writev.count);
+ }
+
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
int
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
+afr_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int op_errno = ENOMEM;
- int ret = -1;
+ afr_local_t *local = NULL;
+ int op_errno = ENOMEM;
+ int ret = -1;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.writev.vector = iov_dup (vector, count);
- if (!local->cont.writev.vector)
- goto out;
- local->cont.writev.count = count;
- local->cont.writev.offset = offset;
- local->cont.writev.flags = flags;
- local->cont.writev.iobref = iobref_ref (iobref);
+ local->cont.writev.vector = iov_dup(vector, count);
+ if (!local->cont.writev.vector)
+ goto out;
+ local->cont.writev.count = count;
+ local->cont.writev.offset = offset;
+ local->cont.writev.flags = flags;
+ local->cont.writev.iobref = iobref_ref(iobref);
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- if (dict_set_uint32 (local->xdata_req, GLUSTERFS_ACTIVE_FD_COUNT, 4)) {
- op_errno = ENOMEM;
- goto out;
- }
+ if (dict_set_uint32(local->xdata_req, GLUSTERFS_ACTIVE_FD_COUNT, 4)) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- if (dict_set_uint32 (local->xdata_req, GLUSTERFS_WRITE_IS_APPEND, 4)) {
- op_errno = ENOMEM;
- goto out;
- }
+ if (dict_set_uint32(local->xdata_req, GLUSTERFS_WRITE_IS_APPEND, 4)) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- /* Set append_write to be true speculatively. If on any
- server it turns not be true, we unset it in the
- callback.
- */
- local->append_write = _gf_true;
+ /* Set append_write to be true speculatively. If on any
+ server it turns not be true, we unset it in the
+ callback.
+ */
+ local->append_write = _gf_true;
- /* detect here, but set it in writev_wind_cbk *after* the unstable
- write is performed
- */
- local->stable_write = !!((fd->flags|flags)&(O_SYNC|O_DSYNC));
+ /* detect here, but set it in writev_wind_cbk *after* the unstable
+ write is performed
+ */
+ local->stable_write = !!((fd->flags | flags) & (O_SYNC | O_DSYNC));
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_do_writev (frame, this);
+ afr_do_writev(frame, this);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
+ AFR_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
/* }}} */
/* {{{ truncate */
int
-afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
+afr_truncate_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (truncate, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(truncate, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_truncate_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
- local->stable_write = _gf_false;
+ if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
+ local->stable_write = _gf_false;
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_truncate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_truncate_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_truncate_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->truncate,
- &local->loc, local->cont.truncate.offset,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_truncate_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->truncate, &local->loc,
+ local->cont.truncate.offset, local->xdata_req);
+ return 0;
}
-
int
-afr_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset, dict_t *xdata)
+afr_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- afr_local_t * local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.truncate.offset = offset;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.truncate.offset = offset;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_truncate_wind;
- local->transaction.unwind = afr_truncate_unwind;
+ local->transaction.wind = afr_truncate_wind;
+ local->transaction.unwind = afr_truncate_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_TRUNCATE;
+ local->op = GF_FOP_TRUNCATE;
- local->transaction.main_frame = frame;
- local->transaction.start = offset;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = offset;
+ local->transaction.len = 0;
- /* Set it true speculatively, will get reset in afr_truncate_wind_cbk
- if truncate was not a NOP */
- local->stable_write = _gf_true;
+ /* Set it true speculatively, will get reset in afr_truncate_wind_cbk
+ if truncate was not a NOP */
+ local->stable_write = _gf_true;
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(truncate, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* }}} */
/* {{{ ftruncate */
-
int
-afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
+afr_ftruncate_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(ftruncate, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_ftruncate_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
- local->stable_write = _gf_false;
+ if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
+ local->stable_write = _gf_false;
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_ftruncate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_ftruncate_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_ftruncate_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->ftruncate,
- local->fd, local->cont.ftruncate.offset,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_ftruncate_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->ftruncate, local->fd,
+ local->cont.ftruncate.offset, local->xdata_req);
+ return 0;
}
-
int
-afr_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+afr_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.ftruncate.offset = offset;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.ftruncate.offset = offset;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FTRUNCATE;
+ local->op = GF_FOP_FTRUNCATE;
- local->transaction.wind = afr_ftruncate_wind;
- local->transaction.unwind = afr_ftruncate_unwind;
+ local->transaction.wind = afr_ftruncate_wind;
+ local->transaction.unwind = afr_ftruncate_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- local->transaction.start = local->cont.ftruncate.offset;
- local->transaction.len = 0;
+ local->transaction.start = local->cont.ftruncate.offset;
+ local->transaction.len = 0;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- /* Set it true speculatively, will get reset in afr_ftruncate_wind_cbk
- if truncate was not a NOP */
- local->stable_write = _gf_true;
+ /* Set it true speculatively, will get reset in afr_ftruncate_wind_cbk
+ if truncate was not a NOP */
+ local->stable_write = _gf_true;
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
+ AFR_STACK_UNWIND(ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
/* }}} */
@@ -807,1813 +771,1743 @@ out:
/* {{{ setattr */
int
-afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_setattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (setattr, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(setattr, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- struct iatt *preop, struct iatt *postop, dict_t *xdata)
+afr_setattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *preop,
+ struct iatt *postop, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- preop, postop, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, preop,
+ postop, NULL, xdata);
}
-
int
-afr_setattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_setattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_setattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->setattr,
- &local->loc, &local->cont.setattr.in_buf,
- local->cont.setattr.valid, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_setattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->setattr, &local->loc,
+ &local->cont.setattr.in_buf, local->cont.setattr.valid,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
- int32_t valid, dict_t *xdata)
+afr_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
+ int32_t valid, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.setattr.in_buf = *buf;
- local->cont.setattr.valid = valid;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.setattr.in_buf = *buf;
+ local->cont.setattr.valid = valid;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_setattr_wind;
- local->transaction.unwind = afr_setattr_unwind;
+ local->transaction.wind = afr_setattr_wind;
+ local->transaction.unwind = afr_setattr_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_SETATTR;
+ local->op = GF_FOP_SETATTR;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(setattr, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* {{{ fsetattr */
int
-afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsetattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (fsetattr, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(fsetattr, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop, dict_t *xdata)
+afr_fsetattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preop,
+ struct iatt *postop, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- preop, postop, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, preop,
+ postop, NULL, xdata);
}
-
int
-afr_fsetattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fsetattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fsetattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fsetattr,
- local->fd, &local->cont.fsetattr.in_buf,
- local->cont.fsetattr.valid, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fsetattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fsetattr, local->fd,
+ &local->cont.fsetattr.in_buf, local->cont.fsetattr.valid,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_fsetattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata)
+afr_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *buf,
+ int32_t valid, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.fsetattr.in_buf = *buf;
- local->cont.fsetattr.valid = valid;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.fsetattr.in_buf = *buf;
+ local->cont.fsetattr.valid = valid;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_fsetattr_wind;
- local->transaction.unwind = afr_fsetattr_unwind;
+ local->transaction.wind = afr_fsetattr_wind;
+ local->transaction.unwind = afr_fsetattr_unwind;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FSETATTR;
+ local->op = GF_FOP_FSETATTR;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* {{{ setxattr */
-
int
-afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_setxattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (setxattr, main_frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(setxattr, main_frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
+}
int
-afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_setxattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, NULL, xdata);
}
-
int
-afr_setxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_setxattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_setxattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->setxattr,
- &local->loc, local->cont.setxattr.dict,
- local->cont.setxattr.flags, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_setxattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->setxattr, &local->loc,
+ local->cont.setxattr.dict, local->cont.setxattr.flags,
+ local->xdata_req);
+ return 0;
}
int
-afr_emptyb_set_pending_changelog_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- dict_t *xattr, dict_t *xdata)
+afr_emptyb_set_pending_changelog_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i, ret = 0;
- char *op_type = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i, ret = 0;
+ char *op_type = NULL;
- local = frame->local;
- priv = this->private;
- i = (long) cookie;
+ local = frame->local;
+ priv = this->private;
+ i = (long)cookie;
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
- ret = dict_get_str (local->xdata_req, "replicate-brick-op", &op_type);
- if (ret)
- goto out;
+ ret = dict_get_str(local->xdata_req, "replicate-brick-op", &op_type);
+ if (ret)
+ goto out;
- gf_msg (this->name, op_ret ? GF_LOG_ERROR : GF_LOG_INFO,
- op_ret ? op_errno : 0,
- afr_get_msg_id (op_type),
- "Set of pending xattr %s on"
- " %s.", op_ret ? "failed" : "succeeded",
- priv->children[i]->name);
+ gf_msg(this->name, op_ret ? GF_LOG_ERROR : GF_LOG_INFO,
+ op_ret ? op_errno : 0, afr_get_msg_id(op_type),
+ "Set of pending xattr %s on"
+ " %s.",
+ op_ret ? "failed" : "succeeded", priv->children[i]->name);
out:
- syncbarrier_wake (&local->barrier);
- return 0;
+ syncbarrier_wake(&local->barrier);
+ return 0;
}
int
-afr_emptyb_set_pending_changelog (call_frame_t *frame, xlator_t *this,
- unsigned char *locked_nodes)
+afr_emptyb_set_pending_changelog(call_frame_t *frame, xlator_t *this,
+ unsigned char *locked_nodes)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int ret = 0, i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int ret = 0, i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- AFR_ONLIST (locked_nodes, frame, afr_emptyb_set_pending_changelog_cbk,
- xattrop, &local->loc, GF_XATTROP_ADD_ARRAY,
- local->xattr_req, NULL);
+ AFR_ONLIST(locked_nodes, frame, afr_emptyb_set_pending_changelog_cbk,
+ xattrop, &local->loc, GF_XATTROP_ADD_ARRAY, local->xattr_req,
+ NULL);
- /* It is sufficient if xattrop was successful on one child */
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
-
- if (local->replies[i].op_ret == 0) {
- ret = 0;
- goto out;
- } else {
- ret = afr_higher_errno (ret,
- local->replies[i].op_errno);
- }
+ /* It is sufficient if xattrop was successful on one child */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+
+ if (local->replies[i].op_ret == 0) {
+ ret = 0;
+ goto out;
+ } else {
+ ret = afr_higher_errno(ret, local->replies[i].op_errno);
}
+ }
out:
- return -ret;
+ return -ret;
}
int
-_afr_handle_empty_brick_type (xlator_t *this, call_frame_t *frame,
- loc_t *loc, int empty_index,
- afr_transaction_type type,
- char *op_type)
+_afr_handle_empty_brick_type(xlator_t *this, call_frame_t *frame, loc_t *loc,
+ int empty_index, afr_transaction_type type,
+ char *op_type)
{
- int count = 0;
- int ret = -ENOMEM;
- int idx = -1;
- int d_idx = -1;
- unsigned char *locked_nodes = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ int count = 0;
+ int ret = -ENOMEM;
+ int idx = -1;
+ int d_idx = -1;
+ unsigned char *locked_nodes = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- locked_nodes = alloca0 (priv->child_count);
+ locked_nodes = alloca0(priv->child_count);
- idx = afr_index_for_transaction_type (type);
- d_idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+ idx = afr_index_for_transaction_type(type);
+ d_idx = afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
- local->pending = afr_matrix_create (priv->child_count,
- AFR_NUM_CHANGE_LOGS);
- if (!local->pending)
- goto out;
+ local->pending = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!local->pending)
+ goto out;
- local->pending[empty_index][idx] = hton32 (1);
+ local->pending[empty_index][idx] = hton32(1);
- if ((priv->esh_granular) && (type == AFR_ENTRY_TRANSACTION))
- local->pending[empty_index][d_idx] = hton32 (1);
+ if ((priv->esh_granular) && (type == AFR_ENTRY_TRANSACTION))
+ local->pending[empty_index][d_idx] = hton32(1);
- local->xdata_req = dict_new ();
- if (!local->xdata_req)
- goto out;
+ local->xdata_req = dict_new();
+ if (!local->xdata_req)
+ goto out;
- ret = dict_set_str (local->xdata_req, "replicate-brick-op", op_type);
- if (ret)
- goto out;
+ ret = dict_set_str(local->xdata_req, "replicate-brick-op", op_type);
+ if (ret)
+ goto out;
- local->xattr_req = dict_new ();
- if (!local->xattr_req)
- goto out;
+ local->xattr_req = dict_new();
+ if (!local->xattr_req)
+ goto out;
- ret = afr_set_pending_dict (priv, local->xattr_req, local->pending);
- if (ret < 0)
- goto out;
+ ret = afr_set_pending_dict(priv, local->xattr_req, local->pending);
+ if (ret < 0)
+ goto out;
- if (AFR_ENTRY_TRANSACTION == type) {
- count = afr_selfheal_entrylk (frame, this, loc->inode,
- this->name, NULL, locked_nodes);
- } else {
- count = afr_selfheal_inodelk (frame, this, loc->inode,
- this->name, LLONG_MAX - 1, 0,
- locked_nodes);
- }
+ if (AFR_ENTRY_TRANSACTION == type) {
+ count = afr_selfheal_entrylk(frame, this, loc->inode, this->name, NULL,
+ locked_nodes);
+ } else {
+ count = afr_selfheal_inodelk(frame, this, loc->inode, this->name,
+ LLONG_MAX - 1, 0, locked_nodes);
+ }
- if (!count) {
- gf_msg (this->name, GF_LOG_ERROR, EAGAIN,
- AFR_MSG_REPLACE_BRICK_STATUS, "Couldn't acquire lock on"
- " any child.");
- ret = -EAGAIN;
- goto unlock;
- }
+ if (!count) {
+ gf_msg(this->name, GF_LOG_ERROR, EAGAIN, AFR_MSG_REPLACE_BRICK_STATUS,
+ "Couldn't acquire lock on"
+ " any child.");
+ ret = -EAGAIN;
+ goto unlock;
+ }
- ret = afr_emptyb_set_pending_changelog (frame, this, locked_nodes);
- if (ret)
- goto unlock;
- ret = 0;
+ ret = afr_emptyb_set_pending_changelog(frame, this, locked_nodes);
+ if (ret)
+ goto unlock;
+ ret = 0;
unlock:
- if (AFR_ENTRY_TRANSACTION == type) {
- afr_selfheal_unentrylk (frame, this, loc->inode, this->name,
- NULL, locked_nodes, NULL);
- } else {
- afr_selfheal_uninodelk (frame, this, loc->inode, this->name,
- LLONG_MAX - 1, 0, locked_nodes);
- }
+ if (AFR_ENTRY_TRANSACTION == type) {
+ afr_selfheal_unentrylk(frame, this, loc->inode, this->name, NULL,
+ locked_nodes, NULL);
+ } else {
+ afr_selfheal_uninodelk(frame, this, loc->inode, this->name,
+ LLONG_MAX - 1, 0, locked_nodes);
+ }
out:
- return ret;
+ return ret;
}
void
-afr_brick_args_cleanup (void *opaque)
+afr_brick_args_cleanup(void *opaque)
{
- afr_empty_brick_args_t *data = NULL;
+ afr_empty_brick_args_t *data = NULL;
- data = opaque;
- loc_wipe (&data->loc);
- GF_FREE (data);
+ data = opaque;
+ loc_wipe(&data->loc);
+ GF_FREE(data);
}
int
-_afr_handle_empty_brick_cbk (int ret, call_frame_t *frame, void *opaque)
+_afr_handle_empty_brick_cbk(int ret, call_frame_t *frame, void *opaque)
{
- afr_brick_args_cleanup (opaque);
- return 0;
+ afr_brick_args_cleanup(opaque);
+ return 0;
}
int
-_afr_handle_empty_brick (void *opaque)
+_afr_handle_empty_brick(void *opaque)
{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int empty_index = -1;
+ int ret = -1;
+ int op_errno = ENOMEM;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ char *op_type = NULL;
+ afr_empty_brick_args_t *data = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int empty_index = -1;
- int ret = -1;
- int op_errno = ENOMEM;
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- char *op_type = NULL;
- afr_empty_brick_args_t *data = NULL;
+ data = opaque;
+ frame = data->frame;
+ empty_index = data->empty_index;
+ op_type = data->op_type;
+ this = frame->this;
+ priv = this->private;
- data = opaque;
- frame = data->frame;
- empty_index = data->empty_index;
- op_type = data->op_type;
- this = frame->this;
- priv = this->private;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ loc_copy(&local->loc, &data->loc);
- loc_copy (&local->loc, &data->loc);
+ gf_msg(this->name, GF_LOG_INFO, 0, 0, "New brick is : %s",
+ priv->children[empty_index]->name);
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "New brick is : %s",
- priv->children[empty_index]->name);
+ ret = _afr_handle_empty_brick_type(this, frame, &local->loc, empty_index,
+ AFR_METADATA_TRANSACTION, op_type);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
- ret = _afr_handle_empty_brick_type (this, frame, &local->loc, empty_index,
- AFR_METADATA_TRANSACTION, op_type);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
-
- dict_unref (local->xdata_req);
- dict_unref (local->xattr_req);
- afr_matrix_cleanup (local->pending, priv->child_count);
- local->pending = NULL;
- local->xattr_req = NULL;
- local->xdata_req = NULL;
+ dict_unref(local->xdata_req);
+ dict_unref(local->xattr_req);
+ afr_matrix_cleanup(local->pending, priv->child_count);
+ local->pending = NULL;
+ local->xattr_req = NULL;
+ local->xdata_req = NULL;
- ret = _afr_handle_empty_brick_type (this, frame, &local->loc, empty_index,
- AFR_ENTRY_TRANSACTION, op_type);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
- ret = 0;
+ ret = _afr_handle_empty_brick_type(this, frame, &local->loc, empty_index,
+ AFR_ENTRY_TRANSACTION, op_type);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
out:
- AFR_STACK_UNWIND (setxattr, frame, ret, op_errno, NULL);
- return 0;
+ AFR_STACK_UNWIND(setxattr, frame, ret, op_errno, NULL);
+ return 0;
+}
+
+int
+afr_split_brain_resolve_do(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ char *data)
+{
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int op_errno = EINVAL;
+
+ local = frame->local;
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ ret = dict_set_int32(local->xdata_req, "heal-op",
+ GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ ret = dict_set_str(local->xdata_req, "child-name", data);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ /* set spb choice to -1 whether heal succeeds or not:
+ * If heal succeeds : spb-choice should be set to -1 as
+ * it is no longer valid; file is not
+ * in split-brain anymore.
+ * If heal doesn't succeed:
+ * spb-choice should be set to -1
+ * otherwise reads will be served
+ * from spb-choice which is misleading.
+ */
+ ret = afr_inode_split_brain_choice_set(loc->inode, this, -1);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Failed to set"
+ "split-brain choice to -1");
+ afr_heal_splitbrain_file(frame, this, loc);
+ ret = 0;
+out:
+ if (ret < 0)
+ AFR_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
+ return 0;
}
-
int
-afr_split_brain_resolve_do (call_frame_t *frame, xlator_t *this, loc_t *loc,
- char *data)
+afr_get_split_brain_child_index(xlator_t *this, void *value, size_t len)
{
- afr_local_t *local = NULL;
- int ret = -1;
- int op_errno = EINVAL;
+ int spb_child_index = -1;
+ char *spb_child_str = NULL;
- local = frame->local;
- local->xdata_req = dict_new ();
+ spb_child_str = alloca0(len + 1);
+ memcpy(spb_child_str, value, len);
- if (!local->xdata_req) {
- op_errno = ENOMEM;
- goto out;
- }
+ if (!strcmp(spb_child_str, "none"))
+ return -2;
- ret = dict_set_int32 (local->xdata_req, "heal-op",
- GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
- ret = dict_set_str (local->xdata_req, "child-name", data);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
- /* set spb choice to -1 whether heal succeeds or not:
- * If heal succeeds : spb-choice should be set to -1 as
- * it is no longer valid; file is not
- * in split-brain anymore.
- * If heal doesn't succeed:
- * spb-choice should be set to -1
- * otherwise reads will be served
- * from spb-choice which is misleading.
- */
- ret = afr_inode_split_brain_choice_set (loc->inode, this, -1);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR, "Failed to set"
- "split-brain choice to -1");
- afr_heal_splitbrain_file (frame, this, loc);
- ret = 0;
-out:
- if (ret < 0)
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
- return 0;
+ spb_child_index = afr_get_child_index_from_name(this, spb_child_str);
+ if (spb_child_index < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "Invalid subvol: %s", spb_child_str);
+ }
+ return spb_child_index;
}
int
-afr_get_split_brain_child_index (xlator_t *this, void *value, size_t len)
+afr_can_set_split_brain_choice(void *opaque)
{
- int spb_child_index = -1;
- char *spb_child_str = NULL;
+ afr_spbc_timeout_t *data = opaque;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ loc_t *loc = NULL;
+ int ret = -1;
- spb_child_str = alloca0 (len + 1);
- memcpy (spb_child_str, value, len);
+ frame = data->frame;
+ loc = data->loc;
+ this = frame->this;
- if (!strcmp (spb_child_str, "none"))
- return -2;
+ ret = afr_is_split_brain(frame, this, loc->inode, loc->gfid, &data->d_spb,
+ &data->m_spb);
- spb_child_index = afr_get_child_index_from_name (this,
- spb_child_str);
- if (spb_child_index < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL, "Invalid subvol: %s",
- spb_child_str);
- }
- return spb_child_index;
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Failed to determine if %s"
+ " is in split-brain. "
+ "Aborting split-brain-choice set.",
+ uuid_utoa(loc->gfid));
+ return ret;
}
int
-afr_can_set_split_brain_choice (void *opaque)
+afr_handle_split_brain_commands(xlator_t *this, call_frame_t *frame, loc_t *loc,
+ dict_t *dict)
{
- afr_spbc_timeout_t *data = opaque;
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- loc_t *loc = NULL;
- int ret = -1;
+ void *value = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_spbc_timeout_t *data = NULL;
+ int len = 0;
+ int spb_child_index = -1;
+ int ret = -1;
+ int op_errno = EINVAL;
- frame = data->frame;
- loc = data->loc;
- this = frame->this;
+ priv = this->private;
- ret = afr_is_split_brain (frame, this, loc->inode, loc->gfid,
- &data->d_spb, &data->m_spb);
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local) {
+ ret = 1;
+ goto out;
+ }
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- "Failed to determine if %s"
- " is in split-brain. "
- "Aborting split-brain-choice set.",
- uuid_utoa (loc->gfid));
- return ret;
-}
+ local->op = GF_FOP_SETXATTR;
-int
-afr_handle_split_brain_commands (xlator_t *this, call_frame_t *frame,
- loc_t *loc, dict_t *dict)
-{
- void *value = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_spbc_timeout_t *data = NULL;
- int len = 0;
- int spb_child_index = -1;
- int ret = -1;
- int op_errno = EINVAL;
-
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local) {
+ ret = dict_get_ptr_and_len(dict, GF_AFR_SBRAIN_CHOICE, &value, &len);
+ if (value) {
+ spb_child_index = afr_get_split_brain_child_index(this, value, len);
+ if (spb_child_index < 0) {
+ /* Case where value was "none" */
+ if (spb_child_index == -2)
+ spb_child_index = -1;
+ else {
ret = 1;
+ op_errno = EINVAL;
goto out;
+ }
}
- local->op = GF_FOP_SETXATTR;
-
- ret = dict_get_ptr_and_len (dict, GF_AFR_SBRAIN_CHOICE, &value,
- &len);
- if (value) {
- spb_child_index = afr_get_split_brain_child_index (this, value,
- len);
- if (spb_child_index < 0) {
- /* Case where value was "none" */
- if (spb_child_index == -2)
- spb_child_index = -1;
- else {
- ret = 1;
- op_errno = EINVAL;
- goto out;
- }
- }
-
- data = GF_CALLOC (1, sizeof (*data), gf_afr_mt_spbc_timeout_t);
- if (!data) {
- ret = 1;
- goto out;
- }
- data->spb_child_index = spb_child_index;
- data->frame = frame;
- loc_copy (&local->loc, loc);
- data->loc = &local->loc;
- ret = synctask_new (this->ctx->env,
- afr_can_set_split_brain_choice,
- afr_set_split_brain_choice, NULL, data);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- "Failed to create"
- " synctask. Aborting split-brain choice set"
- " for %s", loc->name);
- ret = 1;
- op_errno = ENOMEM;
- goto out;
- }
- ret = 0;
- goto out;
+ data = GF_CALLOC(1, sizeof(*data), gf_afr_mt_spbc_timeout_t);
+ if (!data) {
+ ret = 1;
+ goto out;
}
+ data->spb_child_index = spb_child_index;
+ data->frame = frame;
+ loc_copy(&local->loc, loc);
+ data->loc = &local->loc;
+ ret = synctask_new(this->ctx->env, afr_can_set_split_brain_choice,
+ afr_set_split_brain_choice, NULL, data);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Failed to create"
+ " synctask. Aborting split-brain choice set"
+ " for %s",
+ loc->name);
+ ret = 1;
+ op_errno = ENOMEM;
+ goto out;
+ }
+ ret = 0;
+ goto out;
+ }
- ret = dict_get_ptr_and_len (dict, GF_AFR_SBRAIN_RESOLVE, &value, &len);
- if (value) {
- spb_child_index = afr_get_split_brain_child_index (this, value,
- len);
- if (spb_child_index < 0) {
- ret = 1;
- goto out;
- }
-
- afr_split_brain_resolve_do (frame, this, loc,
- priv->children[spb_child_index]->name);
- ret = 0;
+ ret = dict_get_ptr_and_len(dict, GF_AFR_SBRAIN_RESOLVE, &value, &len);
+ if (value) {
+ spb_child_index = afr_get_split_brain_child_index(this, value, len);
+ if (spb_child_index < 0) {
+ ret = 1;
+ goto out;
}
+
+ afr_split_brain_resolve_do(frame, this, loc,
+ priv->children[spb_child_index]->name);
+ ret = 0;
+ }
out:
- /* key was correct but value was invalid when ret == 1 */
- if (ret == 1) {
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
- if (data)
- GF_FREE (data);
- ret = 0;
- }
- return ret;
+ /* key was correct but value was invalid when ret == 1 */
+ if (ret == 1) {
+ AFR_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
+ if (data)
+ GF_FREE(data);
+ ret = 0;
+ }
+ return ret;
}
int
-afr_handle_spb_choice_timeout (xlator_t *this, call_frame_t *frame,
- dict_t *dict)
+afr_handle_spb_choice_timeout(xlator_t *this, call_frame_t *frame, dict_t *dict)
{
- int ret = -1;
- int op_errno = 0;
- uint64_t timeout = 0;
- afr_private_t *priv = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ uint64_t timeout = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- ret = dict_get_uint64 (dict, GF_AFR_SPB_CHOICE_TIMEOUT, &timeout);
- if (!ret) {
- priv->spb_choice_timeout = timeout * 60;
- AFR_STACK_UNWIND (setxattr, frame, ret, op_errno, NULL);
- }
+ ret = dict_get_uint64(dict, GF_AFR_SPB_CHOICE_TIMEOUT, &timeout);
+ if (!ret) {
+ priv->spb_choice_timeout = timeout * 60;
+ AFR_STACK_UNWIND(setxattr, frame, ret, op_errno, NULL);
+ }
- return ret;
+ return ret;
}
int
-afr_handle_empty_brick (xlator_t *this, call_frame_t *frame, loc_t *loc,
- dict_t *dict)
+afr_handle_empty_brick(xlator_t *this, call_frame_t *frame, loc_t *loc,
+ dict_t *dict)
{
- int ret = -1;
- int ab_ret = -1;
- int empty_index = -1;
- int op_errno = EPERM;
- char *empty_brick = NULL;
- char *op_type = NULL;
- afr_empty_brick_args_t *data = NULL;
+ int ret = -1;
+ int ab_ret = -1;
+ int empty_index = -1;
+ int op_errno = EPERM;
+ char *empty_brick = NULL;
+ char *op_type = NULL;
+ afr_empty_brick_args_t *data = NULL;
- ret = dict_get_str (dict, GF_AFR_REPLACE_BRICK, &empty_brick);
- if (!ret)
- op_type = GF_AFR_REPLACE_BRICK;
+ ret = dict_get_str(dict, GF_AFR_REPLACE_BRICK, &empty_brick);
+ if (!ret)
+ op_type = GF_AFR_REPLACE_BRICK;
- ab_ret = dict_get_str (dict, GF_AFR_ADD_BRICK, &empty_brick);
- if (!ab_ret)
- op_type = GF_AFR_ADD_BRICK;
+ ab_ret = dict_get_str(dict, GF_AFR_ADD_BRICK, &empty_brick);
+ if (!ab_ret)
+ op_type = GF_AFR_ADD_BRICK;
- if (ret && ab_ret)
- goto out;
+ if (ret && ab_ret)
+ goto out;
- if (frame->root->pid != GF_CLIENT_PID_SELF_HEALD) {
- gf_msg (this->name, GF_LOG_ERROR, EPERM,
- afr_get_msg_id (op_type),
- "'%s' is an internal extended attribute.",
- op_type);
- ret = 1;
- goto out;
+ if (frame->root->pid != GF_CLIENT_PID_SELF_HEALD) {
+ gf_msg(this->name, GF_LOG_ERROR, EPERM, afr_get_msg_id(op_type),
+ "'%s' is an internal extended attribute.", op_type);
+ ret = 1;
+ goto out;
+ }
+ empty_index = afr_get_child_index_from_name(this, empty_brick);
+
+ if (empty_index < 0) {
+ /* Didn't belong to this replica pair
+ * Just do a no-op
+ */
+ AFR_STACK_UNWIND(setxattr, frame, 0, 0, NULL);
+ return 0;
+ } else {
+ data = GF_CALLOC(1, sizeof(*data), gf_afr_mt_empty_brick_t);
+ if (!data) {
+ ret = 1;
+ op_errno = ENOMEM;
+ goto out;
}
- empty_index = afr_get_child_index_from_name (this, empty_brick);
-
- if (empty_index < 0) {
- /* Didn't belong to this replica pair
- * Just do a no-op
- */
- AFR_STACK_UNWIND (setxattr, frame, 0, 0, NULL);
- return 0;
- } else {
- data = GF_CALLOC (1, sizeof (*data),
- gf_afr_mt_empty_brick_t);
- if (!data) {
- ret = 1;
- op_errno = ENOMEM;
- goto out;
- }
- data->frame = frame;
- loc_copy (&data->loc, loc);
- data->empty_index = empty_index;
- data->op_type = op_type;
- ret = synctask_new (this->ctx->env,
- _afr_handle_empty_brick,
- _afr_handle_empty_brick_cbk,
- NULL, data);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- afr_get_msg_id (op_type),
- "Failed to create synctask.");
- ret = 1;
- op_errno = ENOMEM;
- afr_brick_args_cleanup (data);
- goto out;
- }
+ data->frame = frame;
+ loc_copy(&data->loc, loc);
+ data->empty_index = empty_index;
+ data->op_type = op_type;
+ ret = synctask_new(this->ctx->env, _afr_handle_empty_brick,
+ _afr_handle_empty_brick_cbk, NULL, data);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, afr_get_msg_id(op_type),
+ "Failed to create synctask.");
+ ret = 1;
+ op_errno = ENOMEM;
+ afr_brick_args_cleanup(data);
+ goto out;
}
- ret = 0;
+ }
+ ret = 0;
out:
- if (ret == 1) {
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
- ret = 0;
- }
- return ret;
+ if (ret == 1) {
+ AFR_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
+ ret = 0;
+ }
+ return ret;
}
static int
-afr_handle_special_xattr (xlator_t *this, call_frame_t *frame, loc_t *loc,
- dict_t *dict)
+afr_handle_special_xattr(xlator_t *this, call_frame_t *frame, loc_t *loc,
+ dict_t *dict)
{
- int ret = -1;
+ int ret = -1;
- ret = afr_handle_split_brain_commands (this, frame, loc, dict);
- if (ret == 0)
- goto out;
+ ret = afr_handle_split_brain_commands(this, frame, loc, dict);
+ if (ret == 0)
+ goto out;
- ret = afr_handle_spb_choice_timeout (this, frame, dict);
- if (ret == 0)
- goto out;
+ ret = afr_handle_spb_choice_timeout(this, frame, dict);
+ if (ret == 0)
+ goto out;
- /* Applicable for replace-brick and add-brick commands */
- ret = afr_handle_empty_brick (this, frame, loc, dict);
+ /* Applicable for replace-brick and add-brick commands */
+ ret = afr_handle_empty_brick(this, frame, loc, dict);
out:
- return ret;
+ return ret;
}
int
-afr_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+afr_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = EINVAL;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = EINVAL;
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
- op_errno, out);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.afr.*", dict, op_errno, out);
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
- op_errno, out);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.glusterfs.afr.*", dict, op_errno, out);
- ret = afr_handle_special_xattr (this, frame, loc, dict);
- if (ret == 0)
- return 0;
+ ret = afr_handle_special_xattr(this, frame, loc, dict);
+ if (ret == 0)
+ return 0;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.setxattr.dict = dict_ref (dict);
- local->cont.setxattr.flags = flags;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.setxattr.dict = dict_ref(dict);
+ local->cont.setxattr.flags = flags;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_setxattr_wind;
- local->transaction.unwind = afr_setxattr_unwind;
+ local->transaction.wind = afr_setxattr_wind;
+ local->transaction.unwind = afr_setxattr_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- local->op = GF_FOP_SETXATTR;
+ local->op = GF_FOP_SETXATTR;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
/* {{{ fsetxattr */
-
int
-afr_fsetxattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsetxattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (fsetxattr, main_frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(fsetxattr, main_frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
+}
int
-afr_fsetxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_fsetxattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, NULL, xdata);
}
-
int
-afr_fsetxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fsetxattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fsetxattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fsetxattr,
- local->fd, local->cont.fsetxattr.dict,
- local->cont.fsetxattr.flags, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fsetxattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fsetxattr, local->fd,
+ local->cont.fsetxattr.dict, local->cont.fsetxattr.flags,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata)
+afr_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
- op_errno, out);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.afr.*", dict, op_errno, out);
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
- op_errno, out);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.glusterfs.afr.*", dict, op_errno, out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.fsetxattr.dict = dict_ref (dict);
- local->cont.fsetxattr.flags = flags;
+ local->cont.fsetxattr.dict = dict_ref(dict);
+ local->cont.fsetxattr.flags = flags;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_fsetxattr_wind;
- local->transaction.unwind = afr_fsetxattr_unwind;
+ local->transaction.wind = afr_fsetxattr_wind;
+ local->transaction.unwind = afr_fsetxattr_unwind;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FSETXATTR;
+ local->op = GF_FOP_FSETXATTR;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
- return 0;
+ AFR_STACK_UNWIND(fsetxattr, frame, -1, op_errno, NULL);
+ return 0;
}
/* }}} */
-
/* {{{ removexattr */
-
int
-afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_removexattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (removexattr, main_frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(removexattr, main_frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
+}
int
-afr_removexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_removexattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, NULL, xdata);
}
-
int
-afr_removexattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_removexattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_removexattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->removexattr,
- &local->loc, local->cont.removexattr.name,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_removexattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->removexattr, &local->loc,
+ local->cont.removexattr.name, local->xdata_req);
+ return 0;
}
-
int
-afr_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+afr_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
- name, op_errno, out);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.afr.*", name, op_errno, out);
- GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
- name, op_errno, out);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.glusterfs.afr.*", name, op_errno, out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.removexattr.name = gf_strdup (name);
+ local->cont.removexattr.name = gf_strdup(name);
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_removexattr_wind;
- local->transaction.unwind = afr_removexattr_unwind;
+ local->transaction.wind = afr_removexattr_wind;
+ local->transaction.unwind = afr_removexattr_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_REMOVEXATTR;
+ local->op = GF_FOP_REMOVEXATTR;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
- return 0;
+ AFR_STACK_UNWIND(removexattr, frame, -1, op_errno, NULL);
+ return 0;
}
/* ffremovexattr */
int
-afr_fremovexattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fremovexattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (fremovexattr, main_frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(fremovexattr, main_frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
+}
int
-afr_fremovexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+afr_fremovexattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, NULL, xdata);
}
-
int
-afr_fremovexattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fremovexattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fremovexattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fremovexattr,
- local->fd, local->cont.removexattr.name,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fremovexattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fremovexattr, local->fd,
+ local->cont.removexattr.name, local->xdata_req);
+ return 0;
}
-
int
-afr_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+afr_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
- name, op_errno, out);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.afr.*", name, op_errno, out);
- GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
- name, op_errno, out);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.glusterfs.afr.*", name, op_errno, out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.removexattr.name = gf_strdup (name);
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.removexattr.name = gf_strdup(name);
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_fremovexattr_wind;
- local->transaction.unwind = afr_fremovexattr_unwind;
+ local->transaction.wind = afr_fremovexattr_wind;
+ local->transaction.unwind = afr_fremovexattr_unwind;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FREMOVEXATTR;
+ local->op = GF_FOP_FREMOVEXATTR;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND(fremovexattr, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
int
-afr_fallocate_unwind (call_frame_t *frame, xlator_t *this)
+afr_fallocate_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (fallocate, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(fallocate, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_fallocate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_fallocate_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_fallocate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fallocate_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fallocate_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fallocate,
- local->fd, local->cont.fallocate.mode,
- local->cont.fallocate.offset,
- local->cont.fallocate.len, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fallocate_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fallocate, local->fd,
+ local->cont.fallocate.mode, local->cont.fallocate.offset,
+ local->cont.fallocate.len, local->xdata_req);
+ return 0;
}
-
int
-afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
+afr_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
{
- call_frame_t *transaction_frame = NULL;
- afr_local_t *local = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ call_frame_t *transaction_frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.fallocate.mode = mode;
- local->cont.fallocate.offset = offset;
- local->cont.fallocate.len = len;
+ local->cont.fallocate.mode = mode;
+ local->cont.fallocate.offset = offset;
+ local->cont.fallocate.len = len;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->op = GF_FOP_FALLOCATE;
+ local->op = GF_FOP_FALLOCATE;
- local->transaction.wind = afr_fallocate_wind;
- local->transaction.unwind = afr_fallocate_unwind;
+ local->transaction.wind = afr_fallocate_wind;
+ local->transaction.unwind = afr_fallocate_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- local->transaction.start = local->cont.fallocate.offset;
- local->transaction.len = 0;
+ local->transaction.start = local->cont.fallocate.offset;
+ local->transaction.len = 0;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* }}} */
/* {{{ discard */
int
-afr_discard_unwind (call_frame_t *frame, xlator_t *this)
+afr_discard_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (discard, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(discard, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_discard_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_discard_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_discard_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_discard_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_discard_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->discard,
- local->fd, local->cont.discard.offset,
- local->cont.discard.len, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_discard_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->discard, local->fd,
+ local->cont.discard.offset, local->cont.discard.len,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+afr_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.discard.offset = offset;
- local->cont.discard.len = len;
+ local->cont.discard.offset = offset;
+ local->cont.discard.len = len;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->op = GF_FOP_DISCARD;
+ local->op = GF_FOP_DISCARD;
- local->transaction.wind = afr_discard_wind;
- local->transaction.unwind = afr_discard_unwind;
+ local->transaction.wind = afr_discard_wind;
+ local->transaction.unwind = afr_discard_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- local->transaction.start = local->cont.discard.offset;
- local->transaction.len = 0;
+ local->transaction.start = local->cont.discard.offset;
+ local->transaction.len = 0;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(discard, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* {{{ zerofill */
int
-afr_zerofill_unwind (call_frame_t *frame, xlator_t *this)
+afr_zerofill_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (discard, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(discard, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_zerofill_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+afr_zerofill_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_zerofill_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_zerofill_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_zerofill_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->zerofill,
- local->fd, local->cont.zerofill.offset,
- local->cont.zerofill.len, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_zerofill_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->zerofill, local->fd,
+ local->cont.zerofill.offset, local->cont.zerofill.len,
+ local->xdata_req);
+ return 0;
}
int
-afr_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+afr_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
size_t len, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.zerofill.offset = offset;
- local->cont.zerofill.len = len;
+ local->cont.zerofill.offset = offset;
+ local->cont.zerofill.len = len;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->op = GF_FOP_ZEROFILL;
+ local->op = GF_FOP_ZEROFILL;
- local->transaction.wind = afr_zerofill_wind;
- local->transaction.unwind = afr_zerofill_unwind;
+ local->transaction.wind = afr_zerofill_wind;
+ local->transaction.unwind = afr_zerofill_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- local->transaction.start = local->cont.discard.offset;
- local->transaction.len = len;
+ local->transaction.start = local->cont.discard.offset;
+ local->transaction.len = len;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
int32_t
-afr_xattrop_wind_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr, dict_t *xdata)
+afr_xattrop_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, xattr, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, xattr, xdata);
}
int
-afr_xattrop_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_xattrop_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_xattrop_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->xattrop,
- &local->loc, local->cont.xattrop.optype,
- local->cont.xattrop.xattr, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_xattrop_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->xattrop, &local->loc,
+ local->cont.xattrop.optype, local->cont.xattrop.xattr,
+ local->xdata_req);
+ return 0;
}
int
-afr_xattrop_unwind (call_frame_t *frame, xlator_t *this)
+afr_xattrop_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (xattrop, main_frame, local->op_ret, local->op_errno,
- local->xattr_rsp, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
+
+ AFR_STACK_UNWIND(xattrop, main_frame, local->op_ret, local->op_errno,
+ local->xattr_rsp, local->xdata_rsp);
+ return 0;
}
int32_t
-afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+afr_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.xattrop.xattr = dict_ref (xattr);
- local->cont.xattrop.optype = optype;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->cont.xattrop.xattr = dict_ref(xattr);
+ local->cont.xattrop.optype = optype;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- local->transaction.wind = afr_xattrop_wind;
- local->transaction.unwind = afr_xattrop_unwind;
+ local->transaction.wind = afr_xattrop_wind;
+ local->transaction.unwind = afr_xattrop_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_XATTROP;
+ local->op = GF_FOP_XATTROP;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(xattrop, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
int32_t
-afr_fxattrop_wind_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr, dict_t *xdata)
+afr_fxattrop_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, xattr, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, xattr, xdata);
}
int
-afr_fxattrop_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fxattrop_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fxattrop_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fxattrop,
- local->fd, local->cont.xattrop.optype,
- local->cont.xattrop.xattr, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fxattrop_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fxattrop, local->fd,
+ local->cont.xattrop.optype, local->cont.xattrop.xattr,
+ local->xdata_req);
+ return 0;
}
int
-afr_fxattrop_unwind (call_frame_t *frame, xlator_t *this)
+afr_fxattrop_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (fxattrop, main_frame, local->op_ret, local->op_errno,
- local->xattr_rsp, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
+
+ AFR_STACK_UNWIND(fxattrop, main_frame, local->op_ret, local->op_errno,
+ local->xattr_rsp, local->xdata_rsp);
+ return 0;
}
int32_t
-afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+afr_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.xattrop.xattr = dict_ref (xattr);
- local->cont.xattrop.optype = optype;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->cont.xattrop.xattr = dict_ref(xattr);
+ local->cont.xattrop.optype = optype;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- local->transaction.wind = afr_fxattrop_wind;
- local->transaction.unwind = afr_fxattrop_unwind;
+ local->transaction.wind = afr_fxattrop_wind;
+ local->transaction.unwind = afr_fxattrop_unwind;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FXATTROP;
+ local->op = GF_FOP_FXATTROP;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this,
- AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(fxattrop, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
-
int
-afr_fsync_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsync_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
+ return 0;
- AFR_STACK_UNWIND (fsync, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ AFR_STACK_UNWIND(fsync, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
- return 0;
+ return 0;
}
-
int
-afr_fsync_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_fsync_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_fsync_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fsync_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fsync_wind_cbk, (void *)(long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fsync,
- local->fd, local->cont.fsync.datasync,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fsync_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fsync, local->fd,
+ local->cont.fsync.datasync, local->xdata_req);
+ return 0;
}
int
-afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
- dict_t *xdata)
+afr_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int32_t op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int32_t op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FSYNC;
- local->cont.fsync.datasync = datasync;
+ local->op = GF_FOP_FSYNC;
+ local->cont.fsync.datasync = datasync;
- if (afr_fd_has_witnessed_unstable_write (this, fd->inode)) {
- /* don't care. we only wanted to CLEAR the bit */
- }
+ if (afr_fd_has_witnessed_unstable_write(this, fd->inode)) {
+ /* don't care. we only wanted to CLEAR the bit */
+ }
- local->transaction.wind = afr_fsync_wind;
- local->transaction.unwind = afr_fsync_unwind;
+ local->transaction.wind = afr_fsync_wind;
+ local->transaction.unwind = afr_fsync_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
+ AFR_STACK_UNWIND(fsync, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-lk-common.c b/xlators/cluster/afr/src/afr-lk-common.c
index d7ff2b01a06..95e52ff4a09 100644
--- a/xlators/cluster/afr/src/afr-lk-common.c
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -18,1122 +18,1075 @@
#include <signal.h>
-
-#define LOCKED_NO 0x0 /* no lock held */
-#define LOCKED_YES 0x1 /* for DATA, METADATA, ENTRY and higher_path */
-#define LOCKED_LOWER 0x2 /* for lower path */
+#define LOCKED_NO 0x0 /* no lock held */
+#define LOCKED_YES 0x1 /* for DATA, METADATA, ENTRY and higher_path */
+#define LOCKED_LOWER 0x2 /* for lower path */
int
-afr_entry_lockee_cmp (const void *l1, const void *l2)
+afr_entry_lockee_cmp(const void *l1, const void *l2)
{
- const afr_entry_lockee_t *r1 = l1;
- const afr_entry_lockee_t *r2 = l2;
- int ret = 0;
- uuid_t gfid1 = {0};
- uuid_t gfid2 = {0};
-
- loc_gfid ((loc_t*)&r1->loc, gfid1);
- loc_gfid ((loc_t*)&r2->loc, gfid2);
- ret = gf_uuid_compare (gfid1, gfid2);
- /*Entrylks with NULL basename are the 'smallest'*/
- if (ret == 0) {
- if (!r1->basename)
- return -1;
- if (!r2->basename)
- return 1;
- ret = strcmp (r1->basename, r2->basename);
- }
-
- if (ret <= 0)
- return -1;
- else
- return 1;
+ const afr_entry_lockee_t *r1 = l1;
+ const afr_entry_lockee_t *r2 = l2;
+ int ret = 0;
+ uuid_t gfid1 = {0};
+ uuid_t gfid2 = {0};
+
+ loc_gfid((loc_t *)&r1->loc, gfid1);
+ loc_gfid((loc_t *)&r2->loc, gfid2);
+ ret = gf_uuid_compare(gfid1, gfid2);
+ /*Entrylks with NULL basename are the 'smallest'*/
+ if (ret == 0) {
+ if (!r1->basename)
+ return -1;
+ if (!r2->basename)
+ return 1;
+ ret = strcmp(r1->basename, r2->basename);
+ }
+
+ if (ret <= 0)
+ return -1;
+ else
+ return 1;
}
-int afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+int
+afr_lock_blocking(call_frame_t *frame, xlator_t *this, int child_index);
void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner)
+afr_set_lk_owner(call_frame_t *frame, xlator_t *this, void *lk_owner)
{
- gf_msg_trace (this->name, 0,
- "Setting lk-owner=%llu",
- (unsigned long long) (unsigned long)lk_owner);
+ gf_msg_trace(this->name, 0, "Setting lk-owner=%llu",
+ (unsigned long long)(unsigned long)lk_owner);
- set_lk_owner_from_ptr (&frame->root->lk_owner, lk_owner);
+ set_lk_owner_from_ptr(&frame->root->lk_owner, lk_owner);
}
int32_t
-internal_lock_count (call_frame_t *frame, xlator_t *this)
+internal_lock_count(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int32_t call_count = 0;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int32_t call_count = 0;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i])
- ++call_count;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i])
+ ++call_count;
+ }
- return call_count;
+ return call_count;
}
int
afr_is_inodelk_transaction(afr_transaction_type type)
{
- int ret = 0;
+ int ret = 0;
- switch (type) {
+ switch (type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
- ret = 1;
- break;
+ ret = 1;
+ break;
case AFR_ENTRY_RENAME_TRANSACTION:
case AFR_ENTRY_TRANSACTION:
- ret = 0;
- break;
+ ret = 0;
+ break;
+ }
- }
-
- return ret;
+ return ret;
}
int
-afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
- loc_t *loc, char *basename, int child_count)
+afr_init_entry_lockee(afr_entry_lockee_t *lockee, afr_local_t *local,
+ loc_t *loc, char *basename, int child_count)
{
- int ret = -1;
+ int ret = -1;
- loc_copy (&lockee->loc, loc);
- lockee->basename = (basename)? gf_strdup (basename): NULL;
- if (basename && !lockee->basename)
- goto out;
+ loc_copy(&lockee->loc, loc);
+ lockee->basename = (basename) ? gf_strdup(basename) : NULL;
+ if (basename && !lockee->basename)
+ goto out;
- lockee->locked_count = 0;
- lockee->locked_nodes = GF_CALLOC (child_count,
- sizeof (*lockee->locked_nodes),
- gf_afr_mt_afr_node_character);
+ lockee->locked_count = 0;
+ lockee->locked_nodes = GF_CALLOC(child_count, sizeof(*lockee->locked_nodes),
+ gf_afr_mt_afr_node_character);
- if (!lockee->locked_nodes)
- goto out;
+ if (!lockee->locked_nodes)
+ goto out;
- ret = 0;
+ ret = 0;
out:
- return ret;
-
+ return ret;
}
void
-afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock)
+afr_entry_lockee_cleanup(afr_internal_lock_t *int_lock)
{
- int i = 0;
-
- for (i = 0; i < int_lock->lockee_count; i++) {
- loc_wipe (&int_lock->lockee[i].loc);
- if (int_lock->lockee[i].basename)
- GF_FREE (int_lock->lockee[i].basename);
- if (int_lock->lockee[i].locked_nodes)
- GF_FREE (int_lock->lockee[i].locked_nodes);
- }
+ int i = 0;
+
+ for (i = 0; i < int_lock->lockee_count; i++) {
+ loc_wipe(&int_lock->lockee[i].loc);
+ if (int_lock->lockee[i].basename)
+ GF_FREE(int_lock->lockee[i].basename);
+ if (int_lock->lockee[i].locked_nodes)
+ GF_FREE(int_lock->lockee[i].locked_nodes);
+ }
- return;
+ return;
}
static int
-initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
+initialize_entrylk_variables(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
-
- int i = 0;
-
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- int_lock->entrylk_lock_count = 0;
- int_lock->lock_op_ret = -1;
- int_lock->lock_op_errno = 0;
-
- for (i = 0; i < AFR_LOCKEE_COUNT_MAX; i++) {
- if (!int_lock->lockee[i].locked_nodes)
- break;
- int_lock->lockee[i].locked_count = 0;
- memset (int_lock->lockee[i].locked_nodes, 0,
- sizeof (*int_lock->lockee[i].locked_nodes) *
- priv->child_count);
- }
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_private_t *priv = NULL;
- return 0;
+ int i = 0;
+
+ priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
+
+ int_lock->entrylk_lock_count = 0;
+ int_lock->lock_op_ret = -1;
+ int_lock->lock_op_errno = 0;
+
+ for (i = 0; i < AFR_LOCKEE_COUNT_MAX; i++) {
+ if (!int_lock->lockee[i].locked_nodes)
+ break;
+ int_lock->lockee[i].locked_count = 0;
+ memset(int_lock->lockee[i].locked_nodes, 0,
+ sizeof(*int_lock->lockee[i].locked_nodes) * priv->child_count);
+ }
+
+ return 0;
}
static int
-initialize_inodelk_variables (call_frame_t *frame, xlator_t *this)
+initialize_inodelk_variables(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_private_t *priv = NULL;
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
+ priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- int_lock->lock_count = 0;
- int_lock->lk_attempted_count = 0;
- int_lock->lock_op_ret = -1;
- int_lock->lock_op_errno = 0;
+ int_lock->lock_count = 0;
+ int_lock->lk_attempted_count = 0;
+ int_lock->lock_op_ret = -1;
+ int_lock->lock_op_errno = 0;
- memset (int_lock->locked_nodes, 0,
- sizeof (*int_lock->locked_nodes) * priv->child_count);
+ memset(int_lock->locked_nodes, 0,
+ sizeof(*int_lock->locked_nodes) * priv->child_count);
- return 0;
+ return 0;
}
int
-afr_lockee_locked_nodes_count (afr_internal_lock_t *int_lock)
+afr_lockee_locked_nodes_count(afr_internal_lock_t *int_lock)
{
- int call_count = 0;
- int i = 0;
+ int call_count = 0;
+ int i = 0;
- for (i = 0; i < int_lock->lockee_count; i++)
- call_count += int_lock->lockee[i].locked_count;
+ for (i = 0; i < int_lock->lockee_count; i++)
+ call_count += int_lock->lockee[i].locked_count;
- return call_count;
+ return call_count;
}
int
-afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
+afr_locked_nodes_count(unsigned char *locked_nodes, int child_count)
{
- int i = 0;
- int call_count = 0;
+ int i = 0;
+ int call_count = 0;
- for (i = 0; i < child_count; i++) {
- if (locked_nodes[i] & LOCKED_YES)
- call_count++;
- }
+ for (i = 0; i < child_count; i++) {
+ if (locked_nodes[i] & LOCKED_YES)
+ call_count++;
+ }
- return call_count;
+ return call_count;
}
/* FIXME: What if UNLOCK fails */
static int32_t
-afr_unlock_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_unlock_common_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- int call_count = 0;
- int ret = 0;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int call_count = 0;
+ int ret = 0;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- if (local->transaction.type == AFR_DATA_TRANSACTION && op_ret != 1)
- ret = afr_write_subvol_reset (frame, this);
+ if (local->transaction.type == AFR_DATA_TRANSACTION && op_ret != 1)
+ ret = afr_write_subvol_reset(frame, this);
- LOCK (&frame->lock);
- {
- call_count = --int_lock->lk_call_count;
- }
- UNLOCK (&frame->lock);
+ LOCK(&frame->lock);
+ {
+ call_count = --int_lock->lk_call_count;
+ }
+ UNLOCK(&frame->lock);
- if (call_count == 0) {
- gf_msg_trace (this->name, 0,
- "All internal locks unlocked");
- int_lock->lock_cbk (frame, this);
- }
+ if (call_count == 0) {
+ gf_msg_trace(this->name, 0, "All internal locks unlocked");
+ int_lock->lock_cbk(frame, this);
+ }
- return ret;
+ return ret;
}
void
-afr_update_uninodelk (afr_local_t *local, afr_internal_lock_t *int_lock,
- int32_t child_index)
+afr_update_uninodelk(afr_local_t *local, afr_internal_lock_t *int_lock,
+ int32_t child_index)
{
- int_lock->locked_nodes[child_index] &= LOCKED_NO;
-
+ int_lock->locked_nodes[child_index] &= LOCKED_NO;
}
static int32_t
-afr_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_unlock_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- int32_t child_index = (long)cookie;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- priv = this->private;
-
- if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- AFR_MSG_UNLOCK_FAIL,
- "path=%s gfid=%s: unlock failed on subvolume %s "
- "with lock owner %s", local->loc.path,
- loc_gfid_utoa (&(local->loc)),
- priv->children[child_index]->name,
- lkowner_utoa (&frame->root->lk_owner));
- }
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int32_t child_index = (long)cookie;
+ afr_private_t *priv = NULL;
- afr_update_uninodelk (local, int_lock, child_index);
+ local = frame->local;
+ int_lock = &local->internal_lock;
- afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, xdata);
+ priv = this->private;
- return 0;
+ if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_UNLOCK_FAIL,
+ "path=%s gfid=%s: unlock failed on subvolume %s "
+ "with lock owner %s",
+ local->loc.path, loc_gfid_utoa(&(local->loc)),
+ priv->children[child_index]->name,
+ lkowner_utoa(&frame->root->lk_owner));
+ }
+
+ afr_update_uninodelk(local, int_lock, child_index);
+
+ afr_unlock_common_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ return 0;
}
static int
-afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
+afr_unlock_inodelk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct gf_flock flock = {0,};
- int call_count = 0;
- int i = 0;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ int call_count = 0;
+ int i = 0;
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
- flock.l_start = int_lock->flock.l_start;
- flock.l_len = int_lock->flock.l_len;
- flock.l_type = F_UNLCK;
+ flock.l_start = int_lock->flock.l_start;
+ flock.l_len = int_lock->flock.l_len;
+ flock.l_type = F_UNLCK;
- call_count = afr_locked_nodes_count (int_lock->locked_nodes,
- priv->child_count);
+ call_count = afr_locked_nodes_count(int_lock->locked_nodes,
+ priv->child_count);
- int_lock->lk_call_count = call_count;
+ int_lock->lk_call_count = call_count;
- if (!call_count) {
- GF_ASSERT (!local->transaction.do_eager_unlock);
- gf_msg_trace (this->name, 0,
- "No internal locks unlocked");
+ if (!call_count) {
+ GF_ASSERT(!local->transaction.do_eager_unlock);
+ gf_msg_trace(this->name, 0, "No internal locks unlocked");
- int_lock->lock_cbk (frame, this);
- goto out;
- }
+ int_lock->lock_cbk(frame, this);
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if ((int_lock->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
- continue;
-
- if (local->fd) {
- STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
- (void *) (long)i,
- priv->children[i],
- priv->children[i]->fops->finodelk,
- int_lock->domain, local->fd,
- F_SETLK, &flock, NULL);
- } else {
- STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
- (void *) (long)i,
- priv->children[i],
- priv->children[i]->fops->inodelk,
- int_lock->domain, &local->loc,
- F_SETLK, &flock, NULL);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if ((int_lock->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
+ continue;
- if (!--call_count)
- break;
+ if (local->fd) {
+ STACK_WIND_COOKIE(
+ frame, afr_unlock_inodelk_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->finodelk,
+ int_lock->domain, local->fd, F_SETLK, &flock, NULL);
+ } else {
+ STACK_WIND_COOKIE(
+ frame, afr_unlock_inodelk_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->inodelk,
+ int_lock->domain, &local->loc, F_SETLK, &flock, NULL);
}
+
+ if (!--call_count)
+ break;
+ }
out:
- return 0;
+ return 0;
}
static int32_t
-afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_unlock_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_internal_lock_t *int_lock = NULL;
- int32_t child_index = 0;
- int lockee_no = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int32_t child_index = 0;
+ int lockee_no = 0;
- priv = this->private;
- lockee_no = (int)((long) cookie) / priv->child_count;
- child_index = (int) ((long) cookie) % priv->child_count;
+ priv = this->private;
+ lockee_no = (int)((long)cookie) / priv->child_count;
+ child_index = (int)((long)cookie) % priv->child_count;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- AFR_MSG_ENTRY_UNLOCK_FAIL,
- "%s: unlock failed on %s", local->loc.path,
- priv->children[child_index]->name);
- }
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_ENTRY_UNLOCK_FAIL,
+ "%s: unlock failed on %s", local->loc.path,
+ priv->children[child_index]->name);
+ }
- int_lock->lockee[lockee_no].locked_nodes[child_index] &= LOCKED_NO;
- afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, NULL);
+ int_lock->lockee[lockee_no].locked_nodes[child_index] &= LOCKED_NO;
+ afr_unlock_common_cbk(frame, cookie, this, op_ret, op_errno, NULL);
- return 0;
+ return 0;
}
static int
-afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
+afr_unlock_entrylk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int index = 0;
- int lockee_no = 0;
- int copies = 0;
- int i = -1;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
- copies = priv->child_count;
-
- call_count = afr_lockee_locked_nodes_count (int_lock);
-
- int_lock->lk_call_count = call_count;
-
- if (!call_count){
- gf_msg_trace (this->name, 0,
- "No internal locks unlocked");
- int_lock->lock_cbk (frame, this);
- goto out;
- }
-
- for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
- lockee_no = i / copies;
- index = i % copies;
- if (int_lock->lockee[lockee_no].locked_nodes[index] & LOCKED_YES) {
-
- STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
- (void *) (long) i,
- priv->children[index],
- priv->children[index]->fops->entrylk,
- int_lock->domain,
- &int_lock->lockee[lockee_no].loc,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
-
- if (!--call_count)
- break;
- }
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int index = 0;
+ int lockee_no = 0;
+ int copies = 0;
+ int i = -1;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+ copies = priv->child_count;
+
+ call_count = afr_lockee_locked_nodes_count(int_lock);
+
+ int_lock->lk_call_count = call_count;
+
+ if (!call_count) {
+ gf_msg_trace(this->name, 0, "No internal locks unlocked");
+ int_lock->lock_cbk(frame, this);
+ goto out;
+ }
+
+ for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+ lockee_no = i / copies;
+ index = i % copies;
+ if (int_lock->lockee[lockee_no].locked_nodes[index] & LOCKED_YES) {
+ STACK_WIND_COOKIE(
+ frame, afr_unlock_entrylk_cbk, (void *)(long)i,
+ priv->children[index], priv->children[index]->fops->entrylk,
+ int_lock->domain, &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename, ENTRYLK_UNLOCK,
+ ENTRYLK_WRLCK, NULL);
+
+ if (!--call_count)
+ break;
}
+ }
out:
- return 0;
-
+ return 0;
}
int32_t
-afr_unlock_now (call_frame_t *frame, xlator_t *this)
+afr_unlock_now(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = frame->local;
+ afr_local_t *local = frame->local;
- if (afr_is_inodelk_transaction(local->transaction.type))
- afr_unlock_inodelk (frame, this);
- else
- afr_unlock_entrylk (frame, this);
- return 0;
+ if (afr_is_inodelk_transaction(local->transaction.type))
+ afr_unlock_inodelk(frame, this);
+ else
+ afr_unlock_entrylk(frame, this);
+ return 0;
}
static int32_t
-afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int cky = (long) cookie;
- int child_index = 0;
- int lockee_no = 0;
-
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- child_index = ((int)cky) % priv->child_count;
- lockee_no = ((int)cky) / priv->child_count;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
- gf_msg (this->name, GF_LOG_ERROR, ENOSYS,
- AFR_MSG_LOCK_XLATOR_NOT_LOADED,
- "subvolume does not support locking. "
- "please load features/locks xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
- }
-
- local->op_errno = op_errno;
- int_lock->lock_op_errno = op_errno;
- }
-
- int_lock->lk_attempted_count++;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int cky = (long)cookie;
+ int child_index = 0;
+ int lockee_no = 0;
+
+ priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
+
+ child_index = ((int)cky) % priv->child_count;
+ lockee_no = ((int)cky) / priv->child_count;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ if (op_errno == ENOSYS) {
+ /* return ENOTSUP */
+ gf_msg(this->name, GF_LOG_ERROR, ENOSYS,
+ AFR_MSG_LOCK_XLATOR_NOT_LOADED,
+ "subvolume does not support locking. "
+ "please load features/locks xlator on server");
+ local->op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
+ }
+
+ local->op_errno = op_errno;
+ int_lock->lock_op_errno = op_errno;
}
- UNLOCK (&frame->lock);
- if ((op_ret == -1) &&
- (op_errno == ENOSYS)) {
- afr_unlock_now (frame, this);
- } else {
- if (op_ret == 0) {
- if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
- local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
- int_lock->lockee[lockee_no].locked_nodes[child_index] |= LOCKED_YES;
- int_lock->lockee[lockee_no].locked_count++;
- int_lock->entrylk_lock_count++;
- } else {
- int_lock->locked_nodes[child_index] |= LOCKED_YES;
- int_lock->lock_count++;
-
- if (local->transaction.type ==
- AFR_DATA_TRANSACTION) {
- LOCK(&local->inode->lock);
- {
- local->inode_ctx->lock_count++;
- }
- UNLOCK (&local->inode->lock);
- }
- }
+ int_lock->lk_attempted_count++;
+ }
+ UNLOCK(&frame->lock);
+
+ if ((op_ret == -1) && (op_errno == ENOSYS)) {
+ afr_unlock_now(frame, this);
+ } else {
+ if (op_ret == 0) {
+ if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+ local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+ int_lock->lockee[lockee_no]
+ .locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_count++;
+ int_lock->entrylk_lock_count++;
+ } else {
+ int_lock->locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lock_count++;
+
+ if (local->transaction.type == AFR_DATA_TRANSACTION) {
+ LOCK(&local->inode->lock);
+ {
+ local->inode_ctx->lock_count++;
+ }
+ UNLOCK(&local->inode->lock);
}
- afr_lock_blocking (frame, this, cky + 1);
+ }
}
+ afr_lock_blocking(frame, this, cky + 1);
+ }
- return 0;
+ return 0;
}
static int32_t
-afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_blocking_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- return 0;
-
+ afr_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ return 0;
}
static int32_t
-afr_blocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_blocking_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- return 0;
+ afr_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ return 0;
}
static gf_boolean_t
-afr_is_entrylk (afr_transaction_type trans_type)
+afr_is_entrylk(afr_transaction_type trans_type)
{
- if (afr_is_inodelk_transaction (trans_type))
- return _gf_false;
- return _gf_true;
+ if (afr_is_inodelk_transaction(trans_type))
+ return _gf_false;
+ return _gf_true;
}
static gf_boolean_t
-_is_lock_wind_needed (afr_local_t *local, int child_index)
+_is_lock_wind_needed(afr_local_t *local, int child_index)
{
- if (!local->child_up[child_index])
- return _gf_false;
+ if (!local->child_up[child_index])
+ return _gf_false;
- return _gf_true;
+ return _gf_true;
}
static void
afr_log_entry_locks_failure(xlator_t *this, afr_local_t *local,
afr_internal_lock_t *int_lock)
{
- const char *fop = NULL;
- char *pargfid = NULL;
- const char *name = NULL;
+ const char *fop = NULL;
+ char *pargfid = NULL;
+ const char *name = NULL;
- fop = gf_fop_list[local->op];
+ fop = gf_fop_list[local->op];
- switch (local->op) {
+ switch (local->op) {
case GF_FOP_LINK:
- pargfid = uuid_utoa(local->newloc.pargfid);
- name = local->newloc.name;
- break;
+ pargfid = uuid_utoa(local->newloc.pargfid);
+ name = local->newloc.name;
+ break;
default:
- pargfid = uuid_utoa(local->loc.pargfid);
- name = local->loc.name;
- break;
- }
-
- gf_msg (this->name, GF_LOG_WARNING, 0, AFR_MSG_BLOCKING_LKS_FAILED,
- "Unable to obtain sufficient blocking entry locks on at least "
- "one child while attempting %s on {pgfid:%s, name:%s}.", fop,
- pargfid, name);
+ pargfid = uuid_utoa(local->loc.pargfid);
+ name = local->loc.name;
+ break;
+ }
+
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_BLOCKING_LKS_FAILED,
+ "Unable to obtain sufficient blocking entry locks on at least "
+ "one child while attempting %s on {pgfid:%s, name:%s}.",
+ fop, pargfid, name);
}
static gf_boolean_t
-is_blocking_locks_count_sufficient (call_frame_t *frame, xlator_t *this)
+is_blocking_locks_count_sufficient(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_internal_lock_t *int_lock = NULL;
- gf_boolean_t is_entrylk = _gf_false;
- int child = 0;
- int nlockee = 0;
- int lockee_count = 0;
- gf_boolean_t ret = _gf_true;
-
- local = frame->local;
- priv = this->private;
- int_lock = &local->internal_lock;
- lockee_count = int_lock->lockee_count;
- is_entrylk = afr_is_entrylk (local->transaction.type);
-
- if (!is_entrylk) {
- if (int_lock->lock_count == 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_BLOCKING_LKS_FAILED, "Unable to obtain "
- "blocking inode lock on even one child for "
- "gfid:%s.", uuid_utoa (local->inode->gfid));
- return _gf_false;
- } else {
- /*inodelk succeeded on at least one child. */
- return _gf_true;
- }
-
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ gf_boolean_t is_entrylk = _gf_false;
+ int child = 0;
+ int nlockee = 0;
+ int lockee_count = 0;
+ gf_boolean_t ret = _gf_true;
+
+ local = frame->local;
+ priv = this->private;
+ int_lock = &local->internal_lock;
+ lockee_count = int_lock->lockee_count;
+ is_entrylk = afr_is_entrylk(local->transaction.type);
+
+ if (!is_entrylk) {
+ if (int_lock->lock_count == 0) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_BLOCKING_LKS_FAILED,
+ "Unable to obtain "
+ "blocking inode lock on even one child for "
+ "gfid:%s.",
+ uuid_utoa(local->inode->gfid));
+ return _gf_false;
} else {
- if (int_lock->entrylk_lock_count == 0) {
- afr_log_entry_locks_failure (this, local, int_lock);
- return _gf_false;
- }
- /* For FOPS that take multiple sets of locks (mkdir, rename),
- * there must be at least one brick on which the locks from
- * all lock sets were successful. */
- for (child = 0; child < priv->child_count; child++) {
- ret = _gf_true;
- for (nlockee = 0; nlockee < lockee_count; nlockee++) {
- if (!(int_lock->lockee[nlockee].locked_nodes[child] & LOCKED_YES))
- ret = _gf_false;
- }
- if (ret)
- return ret;
- }
- if (!ret)
- afr_log_entry_locks_failure (this, local, int_lock);
+ /*inodelk succeeded on at least one child. */
+ return _gf_true;
}
- return ret;
+ } else {
+ if (int_lock->entrylk_lock_count == 0) {
+ afr_log_entry_locks_failure(this, local, int_lock);
+ return _gf_false;
+ }
+ /* For FOPS that take multiple sets of locks (mkdir, rename),
+ * there must be at least one brick on which the locks from
+ * all lock sets were successful. */
+ for (child = 0; child < priv->child_count; child++) {
+ ret = _gf_true;
+ for (nlockee = 0; nlockee < lockee_count; nlockee++) {
+ if (!(int_lock->lockee[nlockee].locked_nodes[child] &
+ LOCKED_YES))
+ ret = _gf_false;
+ }
+ if (ret)
+ return ret;
+ }
+ if (!ret)
+ afr_log_entry_locks_failure(this, local, int_lock);
+ }
+ return ret;
}
int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int cookie)
+afr_lock_blocking(call_frame_t *frame, xlator_t *this, int cookie)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct gf_flock flock = {0,};
- uint64_t ctx = 0;
- int ret = 0;
- int child_index = 0;
- int lockee_no = 0;
- gf_boolean_t is_entrylk = _gf_false;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
- child_index = cookie % priv->child_count;
- lockee_no = cookie / priv->child_count;
- is_entrylk = afr_is_entrylk (local->transaction.type);
-
-
- if (!is_entrylk) {
- flock.l_start = int_lock->flock.l_start;
- flock.l_len = int_lock->flock.l_len;
- flock.l_type = int_lock->flock.l_type;
- }
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ uint64_t ctx = 0;
+ int ret = 0;
+ int child_index = 0;
+ int lockee_no = 0;
+ gf_boolean_t is_entrylk = _gf_false;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+ child_index = cookie % priv->child_count;
+ lockee_no = cookie / priv->child_count;
+ is_entrylk = afr_is_entrylk(local->transaction.type);
+
+ if (!is_entrylk) {
+ flock.l_start = int_lock->flock.l_start;
+ flock.l_len = int_lock->flock.l_len;
+ flock.l_type = int_lock->flock.l_type;
+ }
- if (local->fd) {
- ret = fd_ctx_get (local->fd, this, &ctx);
+ if (local->fd) {
+ ret = fd_ctx_get(local->fd, this, &ctx);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_FD_CTX_GET_FAILED,
- "unable to get fd ctx for fd=%p",
- local->fd);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_FD_CTX_GET_FAILED,
+ "unable to get fd ctx for fd=%p", local->fd);
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
- afr_unlock_now (frame, this);
+ afr_unlock_now(frame, this);
- return 0;
- }
+ return 0;
}
+ }
- if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
- if (!is_blocking_locks_count_sufficient (frame, this)) {
-
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
+ if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+ if (!is_blocking_locks_count_sufficient(frame, this)) {
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
- afr_unlock_now(frame, this);
+ afr_unlock_now(frame, this);
- return 0;
- }
+ return 0;
}
+ }
- if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
- /* we're done locking */
+ if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+ /* we're done locking */
- gf_msg_debug (this->name, 0,
- "we're done locking");
+ gf_msg_debug(this->name, 0, "we're done locking");
- int_lock->lock_op_ret = 0;
- int_lock->lock_cbk (frame, this);
- return 0;
- }
+ int_lock->lock_op_ret = 0;
+ int_lock->lock_cbk(frame, this);
+ return 0;
+ }
- if (!_is_lock_wind_needed (local, child_index)) {
- afr_lock_blocking (frame, this, cookie + 1);
- return 0;
- }
+ if (!_is_lock_wind_needed(local, child_index)) {
+ afr_lock_blocking(frame, this, cookie + 1);
+ return 0;
+ }
- switch (local->transaction.type) {
+ switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
- if (local->fd) {
- STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->finodelk,
- int_lock->domain, local->fd,
- F_SETLKW, &flock, NULL);
-
- } else {
- STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->inodelk,
- int_lock->domain, &local->loc,
- F_SETLKW, &flock, NULL);
- }
+ if (local->fd) {
+ STACK_WIND_COOKIE(
+ frame, afr_blocking_inodelk_cbk, (void *)(long)child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->finodelk,
+ int_lock->domain, local->fd, F_SETLKW, &flock, NULL);
- break;
+ } else {
+ STACK_WIND_COOKIE(
+ frame, afr_blocking_inodelk_cbk, (void *)(long)child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->inodelk,
+ int_lock->domain, &local->loc, F_SETLKW, &flock, NULL);
+ }
+
+ break;
case AFR_ENTRY_RENAME_TRANSACTION:
case AFR_ENTRY_TRANSACTION:
- /*Accounting for child_index increments on 'down'
- *and 'fd-less' children */
-
- if (local->fd) {
- STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) cookie,
- priv->children[child_index],
- priv->children[child_index]->fops->fentrylk,
- int_lock->domain, local->fd,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
- } else {
- STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) cookie,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- int_lock->domain,
- &int_lock->lockee[lockee_no].loc,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
- }
-
- break;
- }
-
- return 0;
+ /*Accounting for child_index increments on 'down'
+ *and 'fd-less' children */
+
+ if (local->fd) {
+ STACK_WIND_COOKIE(frame, afr_blocking_entrylk_cbk,
+ (void *)(long)cookie,
+ priv->children[child_index],
+ priv->children[child_index]->fops->fentrylk,
+ int_lock->domain, local->fd,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ } else {
+ STACK_WIND_COOKIE(
+ frame, afr_blocking_entrylk_cbk, (void *)(long)cookie,
+ priv->children[child_index],
+ priv->children[child_index]->fops->entrylk,
+ int_lock->domain, &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename, ENTRYLK_LOCK,
+ ENTRYLK_WRLCK, NULL);
+ }
+
+ break;
+ }
+
+ return 0;
}
int32_t
-afr_blocking_lock (call_frame_t *frame, xlator_t *this)
+afr_blocking_lock(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int up_count = 0;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int up_count = 0;
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
+ priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- switch (local->transaction.type) {
+ switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
- initialize_inodelk_variables (frame, this);
- break;
+ initialize_inodelk_variables(frame, this);
+ break;
case AFR_ENTRY_RENAME_TRANSACTION:
case AFR_ENTRY_TRANSACTION:
- up_count = AFR_COUNT (local->child_up, priv->child_count);
- int_lock->lk_call_count = int_lock->lk_expected_count
- = (int_lock->lockee_count *
- up_count);
- initialize_entrylk_variables (frame, this);
- break;
- }
+ up_count = AFR_COUNT(local->child_up, priv->child_count);
+ int_lock->lk_call_count = int_lock->lk_expected_count =
+ (int_lock->lockee_count * up_count);
+ initialize_entrylk_variables(frame, this);
+ break;
+ }
- afr_lock_blocking (frame, this, 0);
+ afr_lock_blocking(frame, this, 0);
- return 0;
+ return 0;
}
static int32_t
-afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_nonblocking_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
- int copies = 0;
- int index = 0;
- int lockee_no = 0;
- afr_private_t *priv = NULL;
-
- priv = this->private;
-
- copies = priv->child_count;
- index = child_index % copies;
- lockee_no = child_index / copies;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- LOCK (&frame->lock);
- {
- if (op_ret < 0 ) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
- gf_msg (this->name, GF_LOG_ERROR,
- ENOSYS, AFR_MSG_LOCK_XLATOR_NOT_LOADED,
- "subvolume does not support "
- "locking. please load features/locks"
- " xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
-
- int_lock->lock_op_errno = op_errno;
- local->op_errno = op_errno;
- }
- } else if (op_ret == 0) {
- int_lock->lockee[lockee_no].locked_nodes[index] |= \
- LOCKED_YES;
- int_lock->lockee[lockee_no].locked_count++;
- int_lock->entrylk_lock_count++;
- }
-
- call_count = --int_lock->lk_call_count;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ int call_count = 0;
+ int child_index = (long)cookie;
+ int copies = 0;
+ int index = 0;
+ int lockee_no = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ copies = priv->child_count;
+ index = child_index % copies;
+ lockee_no = child_index / copies;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret < 0) {
+ if (op_errno == ENOSYS) {
+ /* return ENOTSUP */
+ gf_msg(this->name, GF_LOG_ERROR, ENOSYS,
+ AFR_MSG_LOCK_XLATOR_NOT_LOADED,
+ "subvolume does not support "
+ "locking. please load features/locks"
+ " xlator on server");
+ local->op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
+
+ int_lock->lock_op_errno = op_errno;
+ local->op_errno = op_errno;
+ }
+ } else if (op_ret == 0) {
+ int_lock->lockee[lockee_no].locked_nodes[index] |= LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_count++;
+ int_lock->entrylk_lock_count++;
}
- UNLOCK (&frame->lock);
-
- if (call_count == 0) {
- gf_msg_trace (this->name, 0,
- "Last locking reply received");
- /* all locks successful. Proceed to call FOP */
- if (int_lock->entrylk_lock_count ==
- int_lock->lk_expected_count) {
- gf_msg_trace (this->name, 0,
- "All servers locked. Calling the cbk");
- int_lock->lock_op_ret = 0;
- int_lock->lock_cbk (frame, this);
- }
- /* Not all locks were successful. Unlock and try locking
- again, this time with serially blocking locks */
- else {
- gf_msg_trace (this->name, 0,
- "%d servers locked. Trying again "
- "with blocking calls",
- int_lock->lock_count);
-
- afr_unlock_now(frame, this);
- }
+
+ call_count = --int_lock->lk_call_count;
+ }
+ UNLOCK(&frame->lock);
+
+ if (call_count == 0) {
+ gf_msg_trace(this->name, 0, "Last locking reply received");
+ /* all locks successful. Proceed to call FOP */
+ if (int_lock->entrylk_lock_count == int_lock->lk_expected_count) {
+ gf_msg_trace(this->name, 0, "All servers locked. Calling the cbk");
+ int_lock->lock_op_ret = 0;
+ int_lock->lock_cbk(frame, this);
+ }
+ /* Not all locks were successful. Unlock and try locking
+ again, this time with serially blocking locks */
+ else {
+ gf_msg_trace(this->name, 0,
+ "%d servers locked. Trying again "
+ "with blocking calls",
+ int_lock->lock_count);
+
+ afr_unlock_now(frame, this);
}
+ }
- return 0;
+ return 0;
}
int
-afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
+afr_nonblocking_entrylk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- int copies = 0;
- int index = 0;
- int lockee_no = 0;
- int32_t call_count = 0;
- int i = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- copies = priv->child_count;
- initialize_entrylk_variables (frame, this);
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int copies = 0;
+ int index = 0;
+ int lockee_no = 0;
+ int32_t call_count = 0;
+ int i = 0;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+
+ copies = priv->child_count;
+ initialize_entrylk_variables(frame, this);
+
+ if (local->fd) {
+ fd_ctx = afr_fd_ctx_get(local->fd, this);
+ if (!fd_ctx) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_FD_CTX_GET_FAILED,
+ "unable to get fd ctx for fd=%p", local->fd);
+
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
+ local->op_errno = EINVAL;
+ int_lock->lock_op_errno = EINVAL;
+
+ afr_unlock_now(frame, this);
+ return -1;
+ }
- if (local->fd) {
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- if (!fd_ctx) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_FD_CTX_GET_FAILED,
- "unable to get fd ctx for fd=%p",
- local->fd);
-
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
- local->op_errno = EINVAL;
- int_lock->lock_op_errno = EINVAL;
-
- afr_unlock_now (frame, this);
- return -1;
- }
+ call_count = int_lock->lockee_count * internal_lock_count(frame, this);
+ int_lock->lk_call_count = call_count;
+ int_lock->lk_expected_count = call_count;
- call_count = int_lock->lockee_count * internal_lock_count (frame, this);
- int_lock->lk_call_count = call_count;
- int_lock->lk_expected_count = call_count;
+ if (!call_count) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_INFO_COMMON,
+ "fd not open on any subvolumes. aborting.");
+ afr_unlock_now(frame, this);
+ goto out;
+ }
- if (!call_count) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_INFO_COMMON,
- "fd not open on any subvolumes. aborting.");
- afr_unlock_now (frame, this);
- goto out;
- }
+ /* Send non-blocking entrylk calls only on up children
+ and where the fd has been opened */
+ for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+ index = i % copies;
+ lockee_no = i / copies;
+ if (local->child_up[index]) {
+ STACK_WIND_COOKIE(frame, afr_nonblocking_entrylk_cbk,
+ (void *)(long)i, priv->children[index],
+ priv->children[index]->fops->fentrylk,
+ this->name, local->fd,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
+ if (!--call_count)
+ break;
+ }
+ }
+ } else {
+ call_count = int_lock->lockee_count * internal_lock_count(frame, this);
+ int_lock->lk_call_count = call_count;
+ int_lock->lk_expected_count = call_count;
- /* Send non-blocking entrylk calls only on up children
- and where the fd has been opened */
- for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
- index = i%copies;
- lockee_no = i/copies;
- if (local->child_up[index]) {
- STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
- (void *) (long) i,
- priv->children[index],
- priv->children[index]->fops->fentrylk,
- this->name, local->fd,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
- NULL);
- if (!--call_count)
- break;
- }
- }
- } else {
- call_count = int_lock->lockee_count * internal_lock_count (frame, this);
- int_lock->lk_call_count = call_count;
- int_lock->lk_expected_count = call_count;
-
- for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
- index = i%copies;
- lockee_no = i/copies;
- if (local->child_up[index]) {
- STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
- (void *) (long) i,
- priv->children[index],
- priv->children[index]->fops->entrylk,
- this->name, &int_lock->lockee[lockee_no].loc,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
- NULL);
-
- if (!--call_count)
- break;
- }
- }
+ for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+ index = i % copies;
+ lockee_no = i / copies;
+ if (local->child_up[index]) {
+ STACK_WIND_COOKIE(frame, afr_nonblocking_entrylk_cbk,
+ (void *)(long)i, priv->children[index],
+ priv->children[index]->fops->entrylk,
+ this->name, &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
+
+ if (!--call_count)
+ break;
+ }
}
+ }
out:
- return 0;
+ return 0;
}
int32_t
-afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_nonblocking_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ int call_count = 0;
+ int child_index = (long)cookie;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- if (op_ret == 0 && local->transaction.type == AFR_DATA_TRANSACTION) {
- LOCK (&local->inode->lock);
- {
- local->inode_ctx->lock_count++;
- }
- UNLOCK (&local->inode->lock);
+ if (op_ret == 0 && local->transaction.type == AFR_DATA_TRANSACTION) {
+ LOCK(&local->inode->lock);
+ {
+ local->inode_ctx->lock_count++;
}
+ UNLOCK(&local->inode->lock);
+ }
- LOCK (&frame->lock);
- {
- if (op_ret < 0) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
- gf_msg (this->name, GF_LOG_ERROR, ENOSYS,
- AFR_MSG_LOCK_XLATOR_NOT_LOADED,
- "subvolume does not support "
- "locking. please load features/locks"
- " xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
- int_lock->lock_op_errno = op_errno;
- local->op_errno = op_errno;
- }
- } else {
- int_lock->locked_nodes[child_index] |= LOCKED_YES;
- int_lock->lock_count++;
- }
-
- call_count = --int_lock->lk_call_count;
+ LOCK(&frame->lock);
+ {
+ if (op_ret < 0) {
+ if (op_errno == ENOSYS) {
+ /* return ENOTSUP */
+ gf_msg(this->name, GF_LOG_ERROR, ENOSYS,
+ AFR_MSG_LOCK_XLATOR_NOT_LOADED,
+ "subvolume does not support "
+ "locking. please load features/locks"
+ " xlator on server");
+ local->op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
+ int_lock->lock_op_errno = op_errno;
+ local->op_errno = op_errno;
+ }
+ } else {
+ int_lock->locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lock_count++;
}
- UNLOCK (&frame->lock);
-
- if (call_count == 0) {
- gf_msg_trace (this->name, 0,
- "Last inode locking reply received");
- /* all locks successful. Proceed to call FOP */
- if (int_lock->lock_count == int_lock->lk_expected_count) {
- gf_msg_trace (this->name, 0,
- "All servers locked. Calling the cbk");
- int_lock->lock_op_ret = 0;
- int_lock->lock_cbk (frame, this);
- }
- /* Not all locks were successful. Unlock and try locking
- again, this time with serially blocking locks */
- else {
- gf_msg_trace (this->name, 0,
- "%d servers locked. "
- "Trying again with blocking calls",
- int_lock->lock_count);
-
- afr_unlock_now(frame, this);
- }
+
+ call_count = --int_lock->lk_call_count;
+ }
+ UNLOCK(&frame->lock);
+
+ if (call_count == 0) {
+ gf_msg_trace(this->name, 0, "Last inode locking reply received");
+ /* all locks successful. Proceed to call FOP */
+ if (int_lock->lock_count == int_lock->lk_expected_count) {
+ gf_msg_trace(this->name, 0, "All servers locked. Calling the cbk");
+ int_lock->lock_op_ret = 0;
+ int_lock->lock_cbk(frame, this);
+ }
+ /* Not all locks were successful. Unlock and try locking
+ again, this time with serially blocking locks */
+ else {
+ gf_msg_trace(this->name, 0,
+ "%d servers locked. "
+ "Trying again with blocking calls",
+ int_lock->lock_count);
+
+ afr_unlock_now(frame, this);
}
+ }
- return 0;
+ return 0;
}
int
-afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
+afr_nonblocking_inodelk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- int32_t call_count = 0;
- int i = 0;
- int ret = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- initialize_inodelk_variables (frame, this);
-
- if (local->fd) {
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- if (!fd_ctx) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_FD_CTX_GET_FAILED,
- "unable to get fd ctx for fd=%p",
- local->fd);
-
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
- local->op_errno = EINVAL;
- int_lock->lock_op_errno = EINVAL;
-
- afr_unlock_now (frame, this);
- ret = -1;
- goto out;
- }
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int32_t call_count = 0;
+ int i = 0;
+ int ret = 0;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+
+ initialize_inodelk_variables(frame, this);
+
+ if (local->fd) {
+ fd_ctx = afr_fd_ctx_get(local->fd, this);
+ if (!fd_ctx) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_FD_CTX_GET_FAILED,
+ "unable to get fd ctx for fd=%p", local->fd);
+
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
+ local->op_errno = EINVAL;
+ int_lock->lock_op_errno = EINVAL;
+
+ afr_unlock_now(frame, this);
+ ret = -1;
+ goto out;
}
+ }
- call_count = internal_lock_count (frame, this);
- int_lock->lk_call_count = call_count;
- int_lock->lk_expected_count = call_count;
+ call_count = internal_lock_count(frame, this);
+ int_lock->lk_call_count = call_count;
+ int_lock->lk_expected_count = call_count;
- if (!call_count) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_SUBVOLS_DOWN,
- "All bricks are down, aborting.");
- afr_unlock_now (frame, this);
- goto out;
- }
+ if (!call_count) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOLS_DOWN,
+ "All bricks are down, aborting.");
+ afr_unlock_now(frame, this);
+ goto out;
+ }
- /* Send non-blocking inodelk calls only on up children
- and where the fd has been opened */
- for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
- continue;
-
- if (local->fd) {
- STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->finodelk,
- int_lock->domain, local->fd,
- F_SETLK, &int_lock->flock, NULL);
- } else {
-
- STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->inodelk,
- int_lock->domain, &local->loc,
- F_SETLK, &int_lock->flock, NULL);
- }
- if (!--call_count)
- break;
+ /* Send non-blocking inodelk calls only on up children
+ and where the fd has been opened */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->child_up[i])
+ continue;
+
+ if (local->fd) {
+ STACK_WIND_COOKIE(
+ frame, afr_nonblocking_inodelk_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->finodelk,
+ int_lock->domain, local->fd, F_SETLK, &int_lock->flock, NULL);
+ } else {
+ STACK_WIND_COOKIE(
+ frame, afr_nonblocking_inodelk_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->inodelk,
+ int_lock->domain, &local->loc, F_SETLK, &int_lock->flock, NULL);
}
+ if (!--call_count)
+ break;
+ }
out:
- return ret;
+ return ret;
}
int32_t
-afr_unlock (call_frame_t *frame, xlator_t *this)
+afr_unlock(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_lock_t *lock = NULL;
-
- local = frame->local;
-
- if (!local->transaction.eager_lock_on)
- goto out;
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- list_del_init (&local->transaction.owner_list);
- if (list_empty (&lock->owners) && list_empty (&lock->post_op)) {
- local->transaction.do_eager_unlock = _gf_true;
- /*TODO: Need to get metadata use on_disk and inherit/uninherit
- *GF_ASSERT (!local->inode_ctx->on_disk[local->transaction.type]);
- *GF_ASSERT (!local->inode_ctx->inherited[local->transaction.type]);
- */
- GF_ASSERT (lock->release);
- }
- }
- UNLOCK (&local->inode->lock);
- if (!local->transaction.do_eager_unlock) {
- local->internal_lock.lock_cbk (frame, this);
- return 0;
+ afr_local_t *local = NULL;
+ afr_lock_t *lock = NULL;
+
+ local = frame->local;
+
+ if (!local->transaction.eager_lock_on)
+ goto out;
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ list_del_init(&local->transaction.owner_list);
+ if (list_empty(&lock->owners) && list_empty(&lock->post_op)) {
+ local->transaction.do_eager_unlock = _gf_true;
+ /*TODO: Need to get metadata use on_disk and inherit/uninherit
+ *GF_ASSERT (!local->inode_ctx->on_disk[local->transaction.type]);
+ *GF_ASSERT (!local->inode_ctx->inherited[local->transaction.type]);
+ */
+ GF_ASSERT(lock->release);
}
+ }
+ UNLOCK(&local->inode->lock);
+ if (!local->transaction.do_eager_unlock) {
+ local->internal_lock.lock_cbk(frame, this);
+ return 0;
+ }
out:
- afr_unlock_now (frame, this);
- return 0;
+ afr_unlock_now(frame, this);
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-open.c b/xlators/cluster/afr/src/afr-open.c
index 3057deed604..a202d528ec8 100644
--- a/xlators/cluster/afr/src/afr-open.c
+++ b/xlators/cluster/afr/src/afr-open.c
@@ -36,336 +36,318 @@
#include "afr-dir-write.h"
#include "afr-transaction.h"
-
gf_boolean_t
-afr_is_fd_fixable (fd_t *fd)
+afr_is_fd_fixable(fd_t *fd)
{
- if (!fd || !fd->inode)
- return _gf_false;
- else if (fd_is_anonymous (fd))
- return _gf_false;
- else if (gf_uuid_is_null (fd->inode->gfid))
- return _gf_false;
-
- return _gf_true;
+ if (!fd || !fd->inode)
+ return _gf_false;
+ else if (fd_is_anonymous(fd))
+ return _gf_false;
+ else if (gf_uuid_is_null(fd->inode->gfid))
+ return _gf_false;
+
+ return _gf_true;
}
-
int
-afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_open_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_local_t * local = frame->local;
+ afr_local_t *local = frame->local;
- AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
- local->fd, xdata);
- return 0;
+ AFR_STACK_UNWIND(open, frame, local->op_ret, local->op_errno, local->fd,
+ xdata);
+ return 0;
}
int
-afr_open_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd, dict_t *xdata)
+afr_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
- int child_index = (long) cookie;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = frame->local;
- fd_ctx = local->fd_ctx;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
- } else {
- local->op_ret = op_ret;
- fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int child_index = (long)cookie;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ local = frame->local;
+ fd_ctx = local->fd_ctx;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+ } else {
+ local->op_ret = op_ret;
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
}
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- if ((fd_ctx->flags & O_TRUNC) && (local->op_ret >= 0)) {
- STACK_WIND (frame, afr_open_ftruncate_cbk,
- this, this->fops->ftruncate,
- fd, 0, NULL);
- } else {
- AFR_STACK_UNWIND (open, frame, local->op_ret,
- local->op_errno, local->cont.open.fd,
- local->xdata_rsp);
- }
+ }
+ UNLOCK(&frame->lock);
+
+ call_count = afr_frame_return(frame);
+
+ if (call_count == 0) {
+ if ((fd_ctx->flags & O_TRUNC) && (local->op_ret >= 0)) {
+ STACK_WIND(frame, afr_open_ftruncate_cbk, this,
+ this->fops->ftruncate, fd, 0, NULL);
+ } else {
+ AFR_STACK_UNWIND(open, frame, local->op_ret, local->op_errno,
+ local->cont.open.fd, local->xdata_rsp);
}
+ }
- return 0;
+ return 0;
}
-
int
-afr_open_continue (call_frame_t *frame, xlator_t *this, int err)
+afr_open_continue(call_frame_t *frame, xlator_t *this, int err)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (err) {
- AFR_STACK_UNWIND (open, frame, -1, err, NULL, NULL);
- } else {
- local->call_count = AFR_COUNT (local->child_up,
- priv->child_count);
- call_count = local->call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_open_cbk,
- (void *)(long)i,
- priv->children[i],
- priv->children[i]->fops->open,
- &local->loc,
- (local->cont.open.flags & ~O_TRUNC),
- local->cont.open.fd,
- local->xdata_req);
- if (!--call_count)
- break;
- }
- }
+ if (err) {
+ AFR_STACK_UNWIND(open, frame, -1, err, NULL, NULL);
+ } else {
+ local->call_count = AFR_COUNT(local->child_up, priv->child_count);
+ call_count = local->call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, afr_open_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->open, &local->loc,
+ (local->cont.open.flags & ~O_TRUNC),
+ local->cont.open.fd, local->xdata_req);
+ if (!--call_count)
+ break;
+ }
}
- return 0;
+ }
+ return 0;
}
int
-afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+afr_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int spb_choice = 0;
- int event_generation = 0;
- int ret = 0;
- int32_t op_errno = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- //We can't let truncation to happen outside transaction.
-
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_OPEN;
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx) {
- op_errno = ENOMEM;
- goto out;
- }
-
- if (!afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
-
- local->inode = inode_ref (loc->inode);
- loc_copy (&local->loc, loc);
- local->fd_ctx = fd_ctx;
- fd_ctx->flags = flags;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
-
- local->cont.open.flags = flags;
- local->cont.open.fd = fd_ref (fd);
-
- ret = afr_inode_get_readable (frame, local->inode, this,
- NULL, &event_generation,
- AFR_DATA_TRANSACTION);
- if ((ret < 0) &&
- (afr_inode_split_brain_choice_get (local->inode,
- this, &spb_choice) == 0) &&
- spb_choice < 0) {
- afr_inode_refresh (frame, this, local->inode,
- local->inode->gfid, afr_open_continue);
- } else {
- afr_open_continue (frame, this, 0);
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int spb_choice = 0;
+ int event_generation = 0;
+ int ret = 0;
+ int32_t op_errno = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ // We can't let truncation to happen outside transaction.
+
+ priv = this->private;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_OPEN;
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
+
+ local->inode = inode_ref(loc->inode);
+ loc_copy(&local->loc, loc);
+ local->fd_ctx = fd_ctx;
+ fd_ctx->flags = flags;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+
+ local->cont.open.flags = flags;
+ local->cont.open.fd = fd_ref(fd);
+
+ ret = afr_inode_get_readable(frame, local->inode, this, NULL,
+ &event_generation, AFR_DATA_TRANSACTION);
+ if ((ret < 0) &&
+ (afr_inode_split_brain_choice_get(local->inode, this, &spb_choice) ==
+ 0) &&
+ spb_choice < 0) {
+ afr_inode_refresh(frame, this, local->inode, local->inode->gfid,
+ afr_open_continue);
+ } else {
+ afr_open_continue(frame, this, 0);
+ }
+
+ return 0;
out:
- AFR_STACK_UNWIND (open, frame, -1, op_errno, fd, NULL);
+ AFR_STACK_UNWIND(open, frame, -1, op_errno, fd, NULL);
- return 0;
+ return 0;
}
int
-afr_openfd_fix_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- dict_t *xdata)
+afr_openfd_fix_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
-
- priv = this->private;
- local = frame->local;
-
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int call_count = 0;
+ int child_index = (long)cookie;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (op_ret >= 0) {
+ gf_msg_debug(this->name, 0,
+ "fd for %s opened "
+ "successfully on subvolume %s",
+ local->loc.path, priv->children[child_index]->name);
+ } else {
+ gf_msg(this->name, fop_log_level(GF_FOP_OPEN, op_errno), op_errno,
+ AFR_MSG_OPEN_FAIL,
+ "Failed to open %s on "
+ "subvolume %s",
+ local->loc.path, priv->children[child_index]->name);
+ }
+
+ fd_ctx = local->fd_ctx;
+
+ LOCK(&local->fd->lock);
+ {
if (op_ret >= 0) {
- gf_msg_debug (this->name, 0, "fd for %s opened "
- "successfully on subvolume %s", local->loc.path,
- priv->children[child_index]->name);
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
} else {
- gf_msg (this->name, fop_log_level (GF_FOP_OPEN, op_errno),
- op_errno, AFR_MSG_OPEN_FAIL, "Failed to open %s on "
- "subvolume %s", local->loc.path,
- priv->children[child_index]->name);
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
}
+ }
+ UNLOCK(&local->fd->lock);
- fd_ctx = local->fd_ctx;
+ call_count = afr_frame_return(frame);
+ if (call_count == 0)
+ AFR_STACK_DESTROY(frame);
- LOCK (&local->fd->lock);
- {
- if (op_ret >= 0) {
- fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- } else {
- fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
- }
- }
- UNLOCK (&local->fd->lock);
+ return 0;
+}
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- AFR_STACK_DESTROY (frame);
+static int
+afr_fd_ctx_need_open(fd_t *fd, xlator_t *this, unsigned char *need_open)
+{
+ afr_fd_ctx_t *fd_ctx = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int count = 0;
+ priv = this->private;
+
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx)
return 0;
-}
+ LOCK(&fd->lock);
+ {
+ for (i = 0; i < priv->child_count; i++) {
+ if (fd_ctx->opened_on[i] == AFR_FD_NOT_OPENED &&
+ priv->child_up[i]) {
+ fd_ctx->opened_on[i] = AFR_FD_OPENING;
+ need_open[i] = 1;
+ count++;
+ } else {
+ need_open[i] = 0;
+ }
+ }
+ }
+ UNLOCK(&fd->lock);
-static int
-afr_fd_ctx_need_open (fd_t *fd, xlator_t *this, unsigned char *need_open)
-{
- afr_fd_ctx_t *fd_ctx = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int count = 0;
-
- priv = this->private;
-
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx)
- return 0;
-
- LOCK (&fd->lock);
- {
- for (i = 0; i < priv->child_count; i++) {
- if (fd_ctx->opened_on[i] == AFR_FD_NOT_OPENED &&
- priv->child_up[i]) {
- fd_ctx->opened_on[i] = AFR_FD_OPENING;
- need_open[i] = 1;
- count++;
- } else {
- need_open[i] = 0;
- }
- }
- }
- UNLOCK (&fd->lock);
-
- return count;
+ return count;
}
-
void
-afr_fix_open (fd_t *fd, xlator_t *this)
+afr_fix_open(fd_t *fd, xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = 0;
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
- int ret = -1;
- int32_t op_errno = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- unsigned char *need_open = NULL;
- int call_count = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ unsigned char *need_open = NULL;
+ int call_count = 0;
- priv = this->private;
+ priv = this->private;
- if (!afr_is_fd_fixable (fd))
- goto out;
+ if (!afr_is_fd_fixable(fd))
+ goto out;
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx)
- goto out;
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx)
+ goto out;
- need_open = alloca0 (priv->child_count);
+ need_open = alloca0(priv->child_count);
- call_count = afr_fd_ctx_need_open (fd, this, need_open);
- if (!call_count)
- goto out;
+ call_count = afr_fd_ctx_need_open(fd, this, need_open);
+ if (!call_count)
+ goto out;
- frame = create_frame (this, this->ctx->pool);
- if (!frame)
- goto out;
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame)
+ goto out;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->loc.inode = inode_ref (fd->inode);
- ret = loc_path (&local->loc, NULL);
- if (ret < 0)
- goto out;
+ local->loc.inode = inode_ref(fd->inode);
+ ret = loc_path(&local->loc, NULL);
+ if (ret < 0)
+ goto out;
- local->fd = fd_ref (fd);
- local->fd_ctx = fd_ctx;
+ local->fd = fd_ref(fd);
+ local->fd_ctx = fd_ctx;
- local->call_count = call_count;
+ local->call_count = call_count;
- gf_msg_debug (this->name, 0, "need open count: %d",
- call_count);
+ gf_msg_debug(this->name, 0, "need open count: %d", call_count);
- for (i = 0; i < priv->child_count; i++) {
- if (!need_open[i])
- continue;
-
- if (IA_IFDIR == fd->inode->ia_type) {
- gf_msg_debug (this->name, 0,
- "opening fd for dir %s on subvolume %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
- (void*) (long) i,
- priv->children[i],
- priv->children[i]->fops->opendir,
- &local->loc, local->fd,
- NULL);
- } else {
- gf_msg_debug (this->name, 0,
- "opening fd for file %s on subvolume %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
- (void *)(long) i,
- priv->children[i],
- priv->children[i]->fops->open,
- &local->loc,
- fd_ctx->flags & (~O_TRUNC),
- local->fd, NULL);
- }
-
- if (!--call_count)
- break;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!need_open[i])
+ continue;
+
+ if (IA_IFDIR == fd->inode->ia_type) {
+ gf_msg_debug(this->name, 0, "opening fd for dir %s on subvolume %s",
+ local->loc.path, priv->children[i]->name);
+
+ STACK_WIND_COOKIE(frame, afr_openfd_fix_open_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->opendir, &local->loc,
+ local->fd, NULL);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "opening fd for file %s on subvolume %s",
+ local->loc.path, priv->children[i]->name);
+
+ STACK_WIND_COOKIE(frame, afr_openfd_fix_open_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->open,
+ &local->loc, fd_ctx->flags & (~O_TRUNC),
+ local->fd, NULL);
}
- return;
+ if (!--call_count)
+ break;
+ }
+
+ return;
out:
- if (frame)
- AFR_STACK_DESTROY (frame);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
}
diff --git a/xlators/cluster/afr/src/afr-read-txn.c b/xlators/cluster/afr/src/afr-read-txn.c
index 945c050da03..1df39c35fce 100644
--- a/xlators/cluster/afr/src/afr-read-txn.c
+++ b/xlators/cluster/afr/src/afr-read-txn.c
@@ -13,347 +13,343 @@
#include "afr-messages.h"
void
-afr_pending_read_increment (afr_private_t *priv, int child_index)
+afr_pending_read_increment(afr_private_t *priv, int child_index)
{
- if (child_index < 0 || child_index > priv->child_count)
- return;
+ if (child_index < 0 || child_index > priv->child_count)
+ return;
- GF_ATOMIC_INC(priv->pending_reads[child_index]);
+ GF_ATOMIC_INC(priv->pending_reads[child_index]);
}
void
-afr_pending_read_decrement (afr_private_t *priv, int child_index)
+afr_pending_read_decrement(afr_private_t *priv, int child_index)
{
- if (child_index < 0 || child_index > priv->child_count)
- return;
+ if (child_index < 0 || child_index > priv->child_count)
+ return;
- GF_ATOMIC_DEC(priv->pending_reads[child_index]);
+ GF_ATOMIC_DEC(priv->pending_reads[child_index]);
}
static gf_boolean_t
-afr_ta_dict_contains_pending_xattr (dict_t *dict, afr_private_t *priv,
- int child)
+afr_ta_dict_contains_pending_xattr(dict_t *dict, afr_private_t *priv, int child)
{
- int *pending = NULL;
- int ret = 0;
- int i = 0;
-
- ret = dict_get_ptr (dict, priv->pending_key[child], (void *)&pending);
- if (ret == 0) {
- for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
- /* Not doing a ntoh32(pending) as we just want to check
- * if it is non-zero or not. */
- if (pending[i]) {
- return _gf_true;
- }
- }
+ int *pending = NULL;
+ int ret = 0;
+ int i = 0;
+
+ ret = dict_get_ptr(dict, priv->pending_key[child], (void *)&pending);
+ if (ret == 0) {
+ for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
+ /* Not doing a ntoh32(pending) as we just want to check
+ * if it is non-zero or not. */
+ if (pending[i]) {
+ return _gf_true;
+ }
}
+ }
- return _gf_false;
+ return _gf_false;
}
void
-afr_read_txn_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_read_txn_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- afr_pending_read_decrement (priv, local->read_subvol);
- local->read_subvol = subvol;
- afr_pending_read_increment (priv, subvol);
- local->readfn (frame, this, subvol);
+ afr_pending_read_decrement(priv, local->read_subvol);
+ local->read_subvol = subvol;
+ afr_pending_read_increment(priv, subvol);
+ local->readfn(frame, this, subvol);
}
int
-afr_read_txn_next_subvol (call_frame_t *frame, xlator_t *this)
+afr_read_txn_next_subvol(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int subvol = -1;
-
- local = frame->local;
- priv = this->private;
-
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->readable[i]) {
- /* don't even bother trying here.
- just mark as attempted and move on. */
- local->read_attempted[i] = 1;
- continue;
- }
-
- if (!local->read_attempted[i]) {
- subvol = i;
- break;
- }
- }
-
- /* If no more subvols were available for reading, we leave
- @subvol as -1, which is an indication we have run out of
- readable subvols. */
- if (subvol != -1)
- local->read_attempted[subvol] = 1;
- afr_read_txn_wind (frame, this, subvol);
-
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int subvol = -1;
+
+ local = frame->local;
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->readable[i]) {
+ /* don't even bother trying here.
+ just mark as attempted and move on. */
+ local->read_attempted[i] = 1;
+ continue;
+ }
+
+ if (!local->read_attempted[i]) {
+ subvol = i;
+ break;
+ }
+ }
+
+ /* If no more subvols were available for reading, we leave
+ @subvol as -1, which is an indication we have run out of
+ readable subvols. */
+ if (subvol != -1)
+ local->read_attempted[subvol] = 1;
+ afr_read_txn_wind(frame, this, subvol);
+
+ return 0;
}
static int
-afr_ta_read_txn_done (int ret, call_frame_t *ta_frame, void *opaque)
+afr_ta_read_txn_done(int ret, call_frame_t *ta_frame, void *opaque)
{
- STACK_DESTROY(ta_frame->root);
- return 0;
+ STACK_DESTROY(ta_frame->root);
+ return 0;
}
static int
-afr_ta_read_txn (void *opaque)
+afr_ta_read_txn(void *opaque)
{
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- int read_subvol = -1;
- int up_child = AFR_CHILD_UNKNOWN;
- int possible_bad_child = AFR_CHILD_UNKNOWN;
- int ret = 0;
- int op_errno = ENOMEM;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct gf_flock flock = {0, };
- dict_t *xdata_req = NULL;
- dict_t *xdata_rsp = NULL;
- int **pending = NULL;
- loc_t loc = {0,};
-
- frame = (call_frame_t *)opaque;
- this = frame->this;
- local = frame->local;
- priv = this->private;
-
- if (local->child_up[AFR_CHILD_ZERO]) {
- up_child = AFR_CHILD_ZERO;
- possible_bad_child = AFR_CHILD_ONE;
- } else if (local->child_up[AFR_CHILD_ONE]) {
- up_child = AFR_CHILD_ONE;
- possible_bad_child = AFR_CHILD_ZERO;
- }
-
- GF_ASSERT (up_child != AFR_CHILD_UNKNOWN);
-
- /* Query the up_child to see if it blames the down one. */
- xdata_req = dict_new();
- if (!xdata_req)
- goto out;
-
- pending = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
- if (!pending)
- goto out;
-
- ret = afr_set_pending_dict (priv, xdata_req, pending);
- if (ret < 0)
- goto out;
-
- if (local->fd) {
- ret = syncop_fxattrop (priv->children[up_child], local->fd,
- GF_XATTROP_ADD_ARRAY, xdata_req, NULL,
- &xdata_rsp, NULL);
- } else {
- ret = syncop_xattrop (priv->children[up_child], &local->loc,
- GF_XATTROP_ADD_ARRAY, xdata_req, NULL,
- &xdata_rsp, NULL);
- }
- if (ret || !xdata_rsp) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed xattrop for gfid %s on %s",
- uuid_utoa (local->inode->gfid),
- priv->children[up_child]->name);
- op_errno = -ret;
- goto out;
- }
-
- if (afr_ta_dict_contains_pending_xattr (xdata_rsp, priv,
- possible_bad_child)) {
- read_subvol = up_child;
- goto out;
- }
- dict_unref (xdata_rsp);
- /* Query thin-arbiter to see if it blames any data brick. */
- ret = afr_fill_ta_loc (this, &loc);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed to populate thin-arbiter loc for: %s.",
- loc.name);
- goto out;
- }
- flock.l_type = F_WRLCK;/*start and length are already zero. */
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_MODIFY, &loc, F_SETLKW, &flock,
- NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "gfid:%s: Failed to get AFR_TA_DOM_MODIFY lock on %s.",
- uuid_utoa (local->inode->gfid),
- priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
- op_errno = -ret;
- goto out;
- }
-
- ret = syncop_xattrop (priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ int read_subvol = -1;
+ int up_child = AFR_CHILD_UNKNOWN;
+ int possible_bad_child = AFR_CHILD_UNKNOWN;
+ int ret = 0;
+ int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ dict_t *xdata_req = NULL;
+ dict_t *xdata_rsp = NULL;
+ int **pending = NULL;
+ loc_t loc = {
+ 0,
+ };
+
+ frame = (call_frame_t *)opaque;
+ this = frame->this;
+ local = frame->local;
+ priv = this->private;
+
+ if (local->child_up[AFR_CHILD_ZERO]) {
+ up_child = AFR_CHILD_ZERO;
+ possible_bad_child = AFR_CHILD_ONE;
+ } else if (local->child_up[AFR_CHILD_ONE]) {
+ up_child = AFR_CHILD_ONE;
+ possible_bad_child = AFR_CHILD_ZERO;
+ }
+
+ GF_ASSERT(up_child != AFR_CHILD_UNKNOWN);
+
+ /* Query the up_child to see if it blames the down one. */
+ xdata_req = dict_new();
+ if (!xdata_req)
+ goto out;
+
+ pending = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!pending)
+ goto out;
+
+ ret = afr_set_pending_dict(priv, xdata_req, pending);
+ if (ret < 0)
+ goto out;
+
+ if (local->fd) {
+ ret = syncop_fxattrop(priv->children[up_child], local->fd,
GF_XATTROP_ADD_ARRAY, xdata_req, NULL, &xdata_rsp,
NULL);
- if (ret || !xdata_rsp) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "gfid:%s: Failed xattrop on %s.",
- uuid_utoa (local->inode->gfid),
- priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
- op_errno = -ret;
- goto unlock;
- }
-
- if (!afr_ta_dict_contains_pending_xattr(xdata_rsp, priv, up_child)) {
- read_subvol = up_child;
- } else {
- gf_msg (this->name, GF_LOG_ERROR, EIO, AFR_MSG_THIN_ARB,
- "Failing read for gfid %s since good brick %s is down",
- uuid_utoa (local->inode->gfid),
- priv->children[possible_bad_child]->name);
- op_errno = EIO;
- }
+ } else {
+ ret = syncop_xattrop(priv->children[up_child], &local->loc,
+ GF_XATTROP_ADD_ARRAY, xdata_req, NULL, &xdata_rsp,
+ NULL);
+ }
+ if (ret || !xdata_rsp) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed xattrop for gfid %s on %s",
+ uuid_utoa(local->inode->gfid), priv->children[up_child]->name);
+ op_errno = -ret;
+ goto out;
+ }
+
+ if (afr_ta_dict_contains_pending_xattr(xdata_rsp, priv,
+ possible_bad_child)) {
+ read_subvol = up_child;
+ goto out;
+ }
+ dict_unref(xdata_rsp);
+ /* Query thin-arbiter to see if it blames any data brick. */
+ ret = afr_fill_ta_loc(this, &loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate thin-arbiter loc for: %s.", loc.name);
+ goto out;
+ }
+ flock.l_type = F_WRLCK; /*start and length are already zero. */
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_MODIFY, &loc, F_SETLKW, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "gfid:%s: Failed to get AFR_TA_DOM_MODIFY lock on %s.",
+ uuid_utoa(local->inode->gfid),
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
+ op_errno = -ret;
+ goto out;
+ }
+
+ ret = syncop_xattrop(priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
+ GF_XATTROP_ADD_ARRAY, xdata_req, NULL, &xdata_rsp,
+ NULL);
+ if (ret || !xdata_rsp) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "gfid:%s: Failed xattrop on %s.", uuid_utoa(local->inode->gfid),
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
+ op_errno = -ret;
+ goto unlock;
+ }
+
+ if (!afr_ta_dict_contains_pending_xattr(xdata_rsp, priv, up_child)) {
+ read_subvol = up_child;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, EIO, AFR_MSG_THIN_ARB,
+ "Failing read for gfid %s since good brick %s is down",
+ uuid_utoa(local->inode->gfid),
+ priv->children[possible_bad_child]->name);
+ op_errno = EIO;
+ }
unlock:
- flock.l_type = F_UNLCK;
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_MODIFY, &loc, F_SETLK, &flock,
- NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "gfid:%s: Failed to unlock AFR_TA_DOM_MODIFY lock on "
- "%s.", uuid_utoa (local->inode->gfid),
- priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
- }
+ flock.l_type = F_UNLCK;
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_MODIFY, &loc, F_SETLK, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "gfid:%s: Failed to unlock AFR_TA_DOM_MODIFY lock on "
+ "%s.",
+ uuid_utoa(local->inode->gfid),
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
+ }
out:
- if (xdata_req)
- dict_unref(xdata_req);
- if (xdata_rsp)
- dict_unref(xdata_rsp);
- if (pending)
- afr_matrix_cleanup (pending, priv->child_count);
- loc_wipe (&loc);
-
- if (read_subvol == -1) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- }
- afr_read_txn_wind (frame, this, read_subvol);
- return ret;
+ if (xdata_req)
+ dict_unref(xdata_req);
+ if (xdata_rsp)
+ dict_unref(xdata_rsp);
+ if (pending)
+ afr_matrix_cleanup(pending, priv->child_count);
+ loc_wipe(&loc);
+
+ if (read_subvol == -1) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ }
+ afr_read_txn_wind(frame, this, read_subvol);
+ return ret;
}
void
-afr_ta_read_txn_synctask (call_frame_t *frame, xlator_t *this)
+afr_ta_read_txn_synctask(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *ta_frame = NULL;
- afr_local_t *local = NULL;
- int ret = 0;
-
- local = frame->local;
- ta_frame = afr_ta_frame_create(this);
- if (!ta_frame) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto out;
- }
- ret = synctask_new (this->ctx->env, afr_ta_read_txn,
- afr_ta_read_txn_done, ta_frame, frame);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- AFR_MSG_THIN_ARB, "Failed to launch "
- "afr_ta_read_txn synctask for gfid %s.",
- uuid_utoa(local->inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- STACK_DESTROY(ta_frame->root);
- goto out;
- }
- return;
+ call_frame_t *ta_frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = 0;
+
+ local = frame->local;
+ ta_frame = afr_ta_frame_create(this);
+ if (!ta_frame) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto out;
+ }
+ ret = synctask_new(this->ctx->env, afr_ta_read_txn, afr_ta_read_txn_done,
+ ta_frame, frame);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_THIN_ARB,
+ "Failed to launch "
+ "afr_ta_read_txn synctask for gfid %s.",
+ uuid_utoa(local->inode->gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ STACK_DESTROY(ta_frame->root);
+ goto out;
+ }
+ return;
out:
- afr_read_txn_wind (frame, this, -1);
+ afr_read_txn_wind(frame, this, -1);
}
int
-afr_read_txn_refresh_done (call_frame_t *frame, xlator_t *this, int err)
+afr_read_txn_refresh_done(call_frame_t *frame, xlator_t *this, int err)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int read_subvol = -1;
- inode_t *inode = NULL;
- int ret = -1;
- int spb_choice = -1;
-
- local = frame->local;
- inode = local->inode;
- priv = this->private;
-
- if (err) {
- if (!priv->thin_arbiter_count)
- goto readfn;
- if (err != EINVAL)
- goto readfn;
- /* We need to query the good bricks and/or thin-arbiter.*/
- afr_ta_read_txn_synctask (frame, this);
- return 0;
- }
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int read_subvol = -1;
+ inode_t *inode = NULL;
+ int ret = -1;
+ int spb_choice = -1;
+
+ local = frame->local;
+ inode = local->inode;
+ priv = this->private;
+
+ if (err) {
+ if (!priv->thin_arbiter_count)
+ goto readfn;
+ if (err != EINVAL)
+ goto readfn;
+ /* We need to query the good bricks and/or thin-arbiter.*/
+ afr_ta_read_txn_synctask(frame, this);
+ return 0;
+ }
- read_subvol = afr_read_subvol_select_by_policy (inode, this,
- local->readable, NULL);
- if (read_subvol == -1) {
- err = EIO;
- goto readfn;
- }
+ read_subvol = afr_read_subvol_select_by_policy(inode, this, local->readable,
+ NULL);
+ if (read_subvol == -1) {
+ err = EIO;
+ goto readfn;
+ }
- if (local->read_attempted[read_subvol]) {
- afr_read_txn_next_subvol (frame, this);
- return 0;
- }
+ if (local->read_attempted[read_subvol]) {
+ afr_read_txn_next_subvol(frame, this);
+ return 0;
+ }
- local->read_attempted[read_subvol] = 1;
+ local->read_attempted[read_subvol] = 1;
readfn:
- if (read_subvol == -1) {
- ret = afr_inode_split_brain_choice_get (inode, this,
- &spb_choice);
- if ((ret == 0) && spb_choice >= 0)
- read_subvol = spb_choice;
- }
-
- if (read_subvol == -1) {
- AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN (-1, err);
- }
- afr_read_txn_wind (frame, this, read_subvol);
-
- return 0;
+ if (read_subvol == -1) {
+ ret = afr_inode_split_brain_choice_get(inode, this, &spb_choice);
+ if ((ret == 0) && spb_choice >= 0)
+ read_subvol = spb_choice;
+ }
+
+ if (read_subvol == -1) {
+ AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN(-1, err);
+ }
+ afr_read_txn_wind(frame, this, read_subvol);
+
+ return 0;
}
int
-afr_read_txn_continue (call_frame_t *frame, xlator_t *this, int subvol)
+afr_read_txn_continue(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local->refreshed) {
- local->refreshed = _gf_true;
- afr_inode_refresh (frame, this, local->inode, NULL,
- afr_read_txn_refresh_done);
- } else {
- afr_read_txn_next_subvol (frame, this);
- }
+ if (!local->refreshed) {
+ local->refreshed = _gf_true;
+ afr_inode_refresh(frame, this, local->inode, NULL,
+ afr_read_txn_refresh_done);
+ } else {
+ afr_read_txn_next_subvol(frame, this);
+ }
- return 0;
+ return 0;
}
-
/* afr_read_txn_wipe:
clean internal variables in @local in order to make
@@ -362,24 +358,24 @@ afr_read_txn_continue (call_frame_t *frame, xlator_t *this, int subvol)
*/
void
-afr_read_txn_wipe (call_frame_t *frame, xlator_t *this)
+afr_read_txn_wipe(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- local->readfn = NULL;
+ local->readfn = NULL;
- if (local->inode)
- inode_unref (local->inode);
+ if (local->inode)
+ inode_unref(local->inode);
- for (i = 0; i < priv->child_count; i++) {
- local->read_attempted[i] = 0;
- local->readable[i] = 0;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ local->read_attempted[i] = 0;
+ local->readable[i] = 0;
+ }
}
/*
@@ -410,92 +406,94 @@ afr_read_txn_wipe (call_frame_t *frame, xlator_t *this)
*/
int
-afr_read_txn (call_frame_t *frame, xlator_t *this, inode_t *inode,
- afr_read_txn_wind_t readfn, afr_transaction_type type)
+afr_read_txn(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ afr_read_txn_wind_t readfn, afr_transaction_type type)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- unsigned char *data = NULL;
- unsigned char *metadata = NULL;
- int read_subvol = -1;
- int event_generation = 0;
- int ret = -1;
-
- priv = this->private;
- local = frame->local;
- data = alloca0 (priv->child_count);
- metadata = alloca0 (priv->child_count);
-
- afr_read_txn_wipe (frame, this);
-
- local->readfn = readfn;
- local->inode = inode_ref (inode);
- local->is_read_txn = _gf_true;
- local->transaction.type = type;
-
- if (priv->quorum_count && !afr_has_quorum (local->child_up, this)) {
- local->op_ret = -1;
- local->op_errno = afr_quorum_errno(priv);
- goto read;
- }
-
- if (!afr_is_consistent_io_possible (local, priv, &local->op_errno)) {
- local->op_ret = -1;
- goto read;
- }
-
- if (priv->thin_arbiter_count &&
- AFR_COUNT (local->child_up, priv->child_count) !=
- priv->child_count) {
- afr_ta_read_txn_synctask (frame, this);
- return 0;
- }
-
- ret = afr_inode_read_subvol_get (inode, this, data, metadata,
- &event_generation);
- if (ret == -1)
- /* very first transaction on this inode */
- goto refresh;
- AFR_INTERSECT (local->readable, data, metadata, priv->child_count);
-
- gf_msg_debug (this->name, 0, "%s: generation now vs cached: %d, "
- "%d", uuid_utoa (inode->gfid), local->event_generation,
- event_generation);
- if (afr_is_inode_refresh_reqd (inode, this, local->event_generation,
- event_generation))
- /* servers have disconnected / reconnected, and possibly
- rebooted, very likely changing the state of freshness
- of copies */
- goto refresh;
-
- read_subvol = afr_read_subvol_select_by_policy (inode, this,
- local->readable, NULL);
-
- if (read_subvol < 0 || read_subvol > priv->child_count) {
- gf_msg_debug (this->name, 0, "Unreadable subvolume %d found "
- "with event generation %d for gfid %s.",
- read_subvol, event_generation, uuid_utoa(inode->gfid));
- goto refresh;
- }
-
- if (!local->child_up[read_subvol]) {
- /* should never happen, just in case */
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_READ_SUBVOL_ERROR, "subvolume %d is the "
- "read subvolume in this generation, but is not up",
- read_subvol);
- goto refresh;
- }
-
- local->read_attempted[read_subvol] = 1;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ unsigned char *data = NULL;
+ unsigned char *metadata = NULL;
+ int read_subvol = -1;
+ int event_generation = 0;
+ int ret = -1;
+
+ priv = this->private;
+ local = frame->local;
+ data = alloca0(priv->child_count);
+ metadata = alloca0(priv->child_count);
+
+ afr_read_txn_wipe(frame, this);
+
+ local->readfn = readfn;
+ local->inode = inode_ref(inode);
+ local->is_read_txn = _gf_true;
+ local->transaction.type = type;
+
+ if (priv->quorum_count && !afr_has_quorum(local->child_up, this)) {
+ local->op_ret = -1;
+ local->op_errno = afr_quorum_errno(priv);
+ goto read;
+ }
+
+ if (!afr_is_consistent_io_possible(local, priv, &local->op_errno)) {
+ local->op_ret = -1;
+ goto read;
+ }
+
+ if (priv->thin_arbiter_count &&
+ AFR_COUNT(local->child_up, priv->child_count) != priv->child_count) {
+ afr_ta_read_txn_synctask(frame, this);
+ return 0;
+ }
+
+ ret = afr_inode_read_subvol_get(inode, this, data, metadata,
+ &event_generation);
+ if (ret == -1)
+ /* very first transaction on this inode */
+ goto refresh;
+ AFR_INTERSECT(local->readable, data, metadata, priv->child_count);
+
+ gf_msg_debug(this->name, 0,
+ "%s: generation now vs cached: %d, "
+ "%d",
+ uuid_utoa(inode->gfid), local->event_generation,
+ event_generation);
+ if (afr_is_inode_refresh_reqd(inode, this, local->event_generation,
+ event_generation))
+ /* servers have disconnected / reconnected, and possibly
+ rebooted, very likely changing the state of freshness
+ of copies */
+ goto refresh;
+
+ read_subvol = afr_read_subvol_select_by_policy(inode, this, local->readable,
+ NULL);
+
+ if (read_subvol < 0 || read_subvol > priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "Unreadable subvolume %d found "
+ "with event generation %d for gfid %s.",
+ read_subvol, event_generation, uuid_utoa(inode->gfid));
+ goto refresh;
+ }
+
+ if (!local->child_up[read_subvol]) {
+ /* should never happen, just in case */
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_READ_SUBVOL_ERROR,
+ "subvolume %d is the "
+ "read subvolume in this generation, but is not up",
+ read_subvol);
+ goto refresh;
+ }
+
+ local->read_attempted[read_subvol] = 1;
read:
- afr_read_txn_wind (frame, this, read_subvol);
+ afr_read_txn_wind(frame, this, read_subvol);
- return 0;
+ return 0;
refresh:
- afr_inode_refresh (frame, this, inode, NULL, afr_read_txn_refresh_done);
+ afr_inode_refresh(frame, this, inode, NULL, afr_read_txn_refresh_done);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index be5dd327c9d..c48c47683c3 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
#include "byte-order.h"
@@ -17,757 +16,767 @@
#include "events.h"
void
-afr_heal_synctask (xlator_t *this, afr_local_t *local);
+afr_heal_synctask(xlator_t *this, afr_local_t *local);
int
-afr_lookup_and_heal_gfid (xlator_t *this, inode_t *parent, const char *name,
- inode_t *inode, struct afr_reply *replies,
- int source, unsigned char *sources, void *gfid)
-{
- afr_private_t *priv = NULL;
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
- unsigned char *wind_on = NULL;
- ia_type_t ia_type = IA_INVAL;
- dict_t *xdata = NULL;
- loc_t loc = {0, };
- int ret = 0;
- int i = 0;
-
- priv = this->private;
- wind_on = alloca0 (priv->child_count);
- ia_type = replies[source].poststat.ia_type;
- if ((ia_type == IA_INVAL) &&
- (AFR_COUNT(sources, priv->child_count) == priv->child_count)) {
- /* If a file is present on some bricks of the replica but parent
- * dir does not have pending xattrs, all bricks are sources and
- * the 'source' we selected earlier might be one where the file
- * is not actually present. Hence check if file is present in
- * any of the sources.*/
- for (i = 0; i < priv->child_count; i++) {
- if (i == source)
- continue;
- if (sources[i] && replies[i].valid &&
- replies[i].op_ret == 0) {
- ia_type = replies[i].poststat.ia_type;
- break;
- }
- }
- }
-
- /* gfid heal on those subvolumes that do not have gfid associated
- * with the inode and update those replies.
- */
+afr_lookup_and_heal_gfid(xlator_t *this, inode_t *parent, const char *name,
+ inode_t *inode, struct afr_reply *replies, int source,
+ unsigned char *sources, void *gfid)
+{
+ afr_private_t *priv = NULL;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ unsigned char *wind_on = NULL;
+ ia_type_t ia_type = IA_INVAL;
+ dict_t *xdata = NULL;
+ loc_t loc = {
+ 0,
+ };
+ int ret = 0;
+ int i = 0;
+
+ priv = this->private;
+ wind_on = alloca0(priv->child_count);
+ ia_type = replies[source].poststat.ia_type;
+ if ((ia_type == IA_INVAL) &&
+ (AFR_COUNT(sources, priv->child_count) == priv->child_count)) {
+ /* If a file is present on some bricks of the replica but parent
+ * dir does not have pending xattrs, all bricks are sources and
+ * the 'source' we selected earlier might be one where the file
+ * is not actually present. Hence check if file is present in
+ * any of the sources.*/
for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
- if (!gf_uuid_is_null (replies[i].poststat.ia_gfid) ||
- replies[i].poststat.ia_type != ia_type)
- continue;
-
- wind_on[i] = 1;
+ if (i == source)
+ continue;
+ if (sources[i] && replies[i].valid && replies[i].op_ret == 0) {
+ ia_type = replies[i].poststat.ia_type;
+ break;
+ }
}
+ }
- if (AFR_COUNT(wind_on, priv->child_count) == 0)
- return 0;
+ /* gfid heal on those subvolumes that do not have gfid associated
+ * with the inode and update those replies.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
+ if (!gf_uuid_is_null(replies[i].poststat.ia_gfid) ||
+ replies[i].poststat.ia_type != ia_type)
+ continue;
- xdata = dict_new ();
- if (!xdata) {
- ret = -ENOMEM;
- goto out;
- }
-
- ret = dict_set_gfuuid (xdata, "gfid-req", gfid, true);
- if (ret) {
- ret = -ENOMEM;
- goto out;
- }
+ wind_on[i] = 1;
+ }
- frame = afr_frame_create (this, &ret);
- if (!frame) {
- ret = -ret;
- goto out;
- }
+ if (AFR_COUNT(wind_on, priv->child_count) == 0)
+ return 0;
- local = frame->local;
- loc.parent = inode_ref (parent);
- gf_uuid_copy (loc.pargfid, parent->gfid);
- loc.name = name;
- loc.inode = inode_ref (inode);
+ xdata = dict_new();
+ if (!xdata) {
+ ret = -ENOMEM;
+ goto out;
+ }
- AFR_ONLIST (wind_on, frame, afr_selfheal_discover_cbk, lookup,
- &loc, xdata);
+ ret = dict_set_gfuuid(xdata, "gfid-req", gfid, true);
+ if (ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!wind_on[i])
- continue;
- afr_reply_wipe (&replies[i]);
- afr_reply_copy (&replies[i], &local->replies[i]);
- }
+ frame = afr_frame_create(this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+
+ local = frame->local;
+ loc.parent = inode_ref(parent);
+ gf_uuid_copy(loc.pargfid, parent->gfid);
+ loc.name = name;
+ loc.inode = inode_ref(inode);
+
+ AFR_ONLIST(wind_on, frame, afr_selfheal_discover_cbk, lookup, &loc, xdata);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!wind_on[i])
+ continue;
+ afr_reply_wipe(&replies[i]);
+ afr_reply_copy(&replies[i], &local->replies[i]);
+ }
out:
- loc_wipe (&loc);
- if (frame)
- AFR_STACK_DESTROY (frame);
- if (xdata)
- dict_unref (xdata);
+ loc_wipe(&loc);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ if (xdata)
+ dict_unref(xdata);
- return ret;
+ return ret;
}
int
-afr_gfid_sbrain_source_from_src_brick (xlator_t *this,
- struct afr_reply *replies,
- char *src_brick)
+afr_gfid_sbrain_source_from_src_brick(xlator_t *this, struct afr_reply *replies,
+ char *src_brick)
{
- int i = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (strcmp (priv->children[i]->name, src_brick) == 0)
- return i;
- }
- return -1;
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (strcmp(priv->children[i]->name, src_brick) == 0)
+ return i;
+ }
+ return -1;
}
int
-afr_selfheal_gfid_mismatch_by_majority (struct afr_reply *replies,
- int child_count)
-{
- int j = 0;
- int i = 0;
- int src = -1;
- int votes[child_count];
-
- for (i = 0; i < child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
-
- votes[i] = 1;
- for (j = i+1; j < child_count; j++) {
- if ((!gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[j].poststat.ia_gfid)))
- votes[i]++;
- if (votes[i] > child_count / 2) {
- src = i;
- goto out;
- }
- }
+afr_selfheal_gfid_mismatch_by_majority(struct afr_reply *replies,
+ int child_count)
+{
+ int j = 0;
+ int i = 0;
+ int src = -1;
+ int votes[child_count];
+
+ for (i = 0; i < child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+
+ votes[i] = 1;
+ for (j = i + 1; j < child_count; j++) {
+ if ((!gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[j].poststat.ia_gfid)))
+ votes[i]++;
+ if (votes[i] > child_count / 2) {
+ src = i;
+ goto out;
+ }
}
+ }
out:
- return src;
+ return src;
}
-int afr_gfid_sbrain_source_from_bigger_file (struct afr_reply *replies,
- int child_count)
+int
+afr_gfid_sbrain_source_from_bigger_file(struct afr_reply *replies,
+ int child_count)
{
- int i = 0;
- int src = -1;
- uint64_t size = 0;
+ int i = 0;
+ int src = -1;
+ uint64_t size = 0;
- for (i = 0; i < child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (size < replies[i].poststat.ia_size) {
- src = i;
- size = replies[i].poststat.ia_size;
- } else if (replies[i].poststat.ia_size == size) {
- src = -1;
- }
- }
- return src;
-}
-
-int afr_gfid_sbrain_source_from_latest_mtime (struct afr_reply *replies,
- int child_count)
-{
- int i = 0;
- int src = -1;
- uint32_t mtime = 0;
- uint32_t mtime_nsec = 0;
-
- for (i = 0; i < child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
- if ((mtime < replies[i].poststat.ia_mtime) ||
- ((mtime == replies[i].poststat.ia_mtime) &&
- (mtime_nsec < replies[i].poststat.ia_mtime_nsec))) {
- src = i;
- mtime = replies[i].poststat.ia_mtime;
- mtime_nsec = replies[i].poststat.ia_mtime_nsec;
- } else if ((mtime == replies[i].poststat.ia_mtime) &&
- (mtime_nsec == replies[i].poststat.ia_mtime_nsec)) {
- src = -1;
- }
+ for (i = 0; i < child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (size < replies[i].poststat.ia_size) {
+ src = i;
+ size = replies[i].poststat.ia_size;
+ } else if (replies[i].poststat.ia_size == size) {
+ src = -1;
}
- return src;
+ }
+ return src;
}
int
-afr_gfid_split_brain_source (xlator_t *this, struct afr_reply *replies,
- inode_t *inode, uuid_t pargfid, const char *bname,
- int src_idx, int child_idx,
- unsigned char *locked_on, int *src, dict_t *xdata)
-{
- afr_private_t *priv = NULL;
- char g1[64] = {0,};
- char g2[64] = {0,};
- int up_count = 0;
- int heal_op = -1;
- int ret = -1;
- char *src_brick = NULL;
-
- *src = -1;
- priv = this->private;
- up_count = AFR_COUNT (locked_on, priv->child_count);
- if (up_count != priv->child_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
- "All the bricks should be up to resolve the gfid split "
- "barin");
- if (xdata) {
- ret = dict_set_str (xdata, "gfid-heal-msg", "All the "
- "bricks should be up to resolve the"
- " gfid split barin");
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED, "Error setting"
- " gfid-heal-msg dict");
- }
- goto out;
+afr_gfid_sbrain_source_from_latest_mtime(struct afr_reply *replies,
+ int child_count)
+{
+ int i = 0;
+ int src = -1;
+ uint32_t mtime = 0;
+ uint32_t mtime_nsec = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
+ if ((mtime < replies[i].poststat.ia_mtime) ||
+ ((mtime == replies[i].poststat.ia_mtime) &&
+ (mtime_nsec < replies[i].poststat.ia_mtime_nsec))) {
+ src = i;
+ mtime = replies[i].poststat.ia_mtime;
+ mtime_nsec = replies[i].poststat.ia_mtime_nsec;
+ } else if ((mtime == replies[i].poststat.ia_mtime) &&
+ (mtime_nsec == replies[i].poststat.ia_mtime_nsec)) {
+ src = -1;
}
+ }
+ return src;
+}
+int
+afr_gfid_split_brain_source(xlator_t *this, struct afr_reply *replies,
+ inode_t *inode, uuid_t pargfid, const char *bname,
+ int src_idx, int child_idx,
+ unsigned char *locked_on, int *src, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ char g1[64] = {
+ 0,
+ };
+ char g2[64] = {
+ 0,
+ };
+ int up_count = 0;
+ int heal_op = -1;
+ int ret = -1;
+ char *src_brick = NULL;
+
+ *src = -1;
+ priv = this->private;
+ up_count = AFR_COUNT(locked_on, priv->child_count);
+ if (up_count != priv->child_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "All the bricks should be up to resolve the gfid split "
+ "barin");
if (xdata) {
- ret = dict_get_int32 (xdata, "heal-op", &heal_op);
- if (ret)
- goto fav_child;
- } else {
- goto fav_child;
+ ret = dict_set_str(xdata, "gfid-heal-msg",
+ "All the "
+ "bricks should be up to resolve the"
+ " gfid split barin");
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_SET_FAILED,
+ "Error setting"
+ " gfid-heal-msg dict");
}
+ goto out;
+ }
- switch (heal_op) {
+ if (xdata) {
+ ret = dict_get_int32(xdata, "heal-op", &heal_op);
+ if (ret)
+ goto fav_child;
+ } else {
+ goto fav_child;
+ }
+
+ switch (heal_op) {
case GF_SHD_OP_SBRAIN_HEAL_FROM_BIGGER_FILE:
- *src = afr_gfid_sbrain_source_from_bigger_file (replies,
- priv->child_count);
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "No bigger file");
- if (xdata) {
- ret = dict_set_str (xdata, "gfid-heal-msg",
- "No bigger file");
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED, "Error"
- " setting gfid-heal-msg dict");
- }
+ *src = afr_gfid_sbrain_source_from_bigger_file(replies,
+ priv->child_count);
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "No bigger file");
+ if (xdata) {
+ ret = dict_set_str(xdata, "gfid-heal-msg",
+ "No bigger file");
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_DICT_SET_FAILED,
+ "Error"
+ " setting gfid-heal-msg dict");
}
- break;
+ }
+ break;
case GF_SHD_OP_SBRAIN_HEAL_FROM_LATEST_MTIME:
- *src = afr_gfid_sbrain_source_from_latest_mtime (replies,
- priv->child_count);
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "No difference in mtime");
- if (xdata) {
- ret = dict_set_str (xdata, "gfid-heal-msg",
- "No difference in mtime");
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED, "Error"
- "setting gfid-heal-msg dict");
- }
+ *src = afr_gfid_sbrain_source_from_latest_mtime(replies,
+ priv->child_count);
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "No difference in mtime");
+ if (xdata) {
+ ret = dict_set_str(xdata, "gfid-heal-msg",
+ "No difference in mtime");
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_DICT_SET_FAILED,
+ "Error"
+ "setting gfid-heal-msg dict");
}
- break;
+ }
+ break;
case GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK:
- ret = dict_get_str (xdata, "child-name", &src_brick);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "Error getting the source "
- "brick");
- break;
- }
- *src = afr_gfid_sbrain_source_from_src_brick (this, replies,
- src_brick);
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "Error getting the source "
- "brick");
- if (xdata) {
- ret = dict_set_str (xdata, "gfid-heal-msg",
- "Error getting the source "
- "brick");
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED, "Error"
- " setting gfid-heal-msg dict");
- }
- }
+ ret = dict_get_str(xdata, "child-name", &src_brick);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Error getting the source "
+ "brick");
break;
+ }
+ *src = afr_gfid_sbrain_source_from_src_brick(this, replies,
+ src_brick);
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Error getting the source "
+ "brick");
+ if (xdata) {
+ ret = dict_set_str(xdata, "gfid-heal-msg",
+ "Error getting the source "
+ "brick");
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_DICT_SET_FAILED,
+ "Error"
+ " setting gfid-heal-msg dict");
+ }
+ }
+ break;
default:
- break;
- }
- goto out;
+ break;
+ }
+ goto out;
fav_child:
- switch (priv->fav_child_policy) {
+ switch (priv->fav_child_policy) {
case AFR_FAV_CHILD_BY_SIZE:
- *src = afr_sh_fav_by_size (this, replies, inode);
- break;
+ *src = afr_sh_fav_by_size(this, replies, inode);
+ break;
case AFR_FAV_CHILD_BY_MTIME:
- *src = afr_sh_fav_by_mtime (this, replies, inode);
- break;
+ *src = afr_sh_fav_by_mtime(this, replies, inode);
+ break;
case AFR_FAV_CHILD_BY_CTIME:
- *src = afr_sh_fav_by_ctime(this, replies, inode);
- break;
+ *src = afr_sh_fav_by_ctime(this, replies, inode);
+ break;
case AFR_FAV_CHILD_BY_MAJORITY:
- if (priv->child_count != 2)
- *src = afr_selfheal_gfid_mismatch_by_majority (replies,
- priv->child_count);
- else
- *src = -1;
-
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "No majority to resolve "
- "gfid split brain");
- }
- break;
+ if (priv->child_count != 2)
+ *src = afr_selfheal_gfid_mismatch_by_majority(
+ replies, priv->child_count);
+ else
+ *src = -1;
+
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "No majority to resolve "
+ "gfid split brain");
+ }
+ break;
default:
- break;
- }
+ break;
+ }
out:
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
- "Gfid mismatch detected for <gfid:%s>/%s>, %s on %s and"
- " %s on %s.", uuid_utoa (pargfid), bname,
- uuid_utoa_r (replies[child_idx].poststat.ia_gfid, g1),
- priv->children[child_idx]->name,
- uuid_utoa_r (replies[src_idx].poststat.ia_gfid, g2),
- priv->children[src_idx]->name);
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;type=gfid;file="
- "<gfid:%s>/%s>;count=2;child-%d=%s;gfid-%d=%s;"
- "child-%d=%s;gfid-%d=%s", this->name,
- uuid_utoa (pargfid), bname, child_idx,
- priv->children[child_idx]->name, child_idx,
- uuid_utoa_r (replies[child_idx].poststat.ia_gfid, g1),
- src_idx, priv->children[src_idx]->name, src_idx,
- uuid_utoa_r (replies[src_idx].poststat.ia_gfid, g2));
- return -1;
- }
- return 0;
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Gfid mismatch detected for <gfid:%s>/%s>, %s on %s and"
+ " %s on %s.",
+ uuid_utoa(pargfid), bname,
+ uuid_utoa_r(replies[child_idx].poststat.ia_gfid, g1),
+ priv->children[child_idx]->name,
+ uuid_utoa_r(replies[src_idx].poststat.ia_gfid, g2),
+ priv->children[src_idx]->name);
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "subvol=%s;type=gfid;file="
+ "<gfid:%s>/%s>;count=2;child-%d=%s;gfid-%d=%s;"
+ "child-%d=%s;gfid-%d=%s",
+ this->name, uuid_utoa(pargfid), bname, child_idx,
+ priv->children[child_idx]->name, child_idx,
+ uuid_utoa_r(replies[child_idx].poststat.ia_gfid, g1), src_idx,
+ priv->children[src_idx]->name, src_idx,
+ uuid_utoa_r(replies[src_idx].poststat.ia_gfid, g2));
+ return -1;
+ }
+ return 0;
}
-
int
-afr_selfheal_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+afr_selfheal_post_op_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- syncbarrier_wake (&local->barrier);
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ syncbarrier_wake(&local->barrier);
- return 0;
+ return 0;
}
-
int
-afr_selfheal_post_op (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int subvol, dict_t *xattr, dict_t *xdata)
+afr_selfheal_post_op(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int subvol, dict_t *xattr, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- loc_t loc = {0, };
- int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ loc_t loc = {
+ 0,
+ };
+ int ret = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- local->op_ret = 0;
+ local->op_ret = 0;
- STACK_WIND (frame, afr_selfheal_post_op_cbk, priv->children[subvol],
- priv->children[subvol]->fops->xattrop, &loc,
- GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ STACK_WIND(frame, afr_selfheal_post_op_cbk, priv->children[subvol],
+ priv->children[subvol]->fops->xattrop, &loc,
+ GF_XATTROP_ADD_ARRAY, xattr, xdata);
- syncbarrier_wait (&local->barrier, 1);
- if (local->op_ret < 0)
- ret = -local->op_errno;
+ syncbarrier_wait(&local->barrier, 1);
+ if (local->op_ret < 0)
+ ret = -local->op_errno;
- loc_wipe (&loc);
- local->op_ret = 0;
+ loc_wipe(&loc);
+ local->op_ret = 0;
- return ret;
+ return ret;
}
int
-afr_check_stale_error (struct afr_reply *replies, afr_private_t *priv)
+afr_check_stale_error(struct afr_reply *replies, afr_private_t *priv)
{
- int i = 0;
- int op_errno = 0;
- int tmp_errno = 0;
- int stale_count = 0;
+ int i = 0;
+ int op_errno = 0;
+ int tmp_errno = 0;
+ int stale_count = 0;
- for (i = 0; i < priv->child_count; i++) {
- tmp_errno = replies[i].op_errno;
- if (tmp_errno == ENOENT || tmp_errno == ESTALE) {
- op_errno = afr_higher_errno (op_errno, tmp_errno);
- stale_count++;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ tmp_errno = replies[i].op_errno;
+ if (tmp_errno == ENOENT || tmp_errno == ESTALE) {
+ op_errno = afr_higher_errno(op_errno, tmp_errno);
+ stale_count++;
}
- if (stale_count != priv->child_count)
- return -ENOTCONN;
- else
- return -op_errno;
+ }
+ if (stale_count != priv->child_count)
+ return -ENOTCONN;
+ else
+ return -op_errno;
}
int
-afr_sh_generic_fop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *pre, struct iatt *post,
- dict_t *xdata)
+afr_sh_generic_fop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- int i = (long) cookie;
- afr_local_t *local = NULL;
+ int i = (long)cookie;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
- if (pre)
- local->replies[i].prestat = *pre;
- if (post)
- local->replies[i].poststat = *post;
- if (xdata)
- local->replies[i].xdata = dict_ref (xdata);
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
+ if (pre)
+ local->replies[i].prestat = *pre;
+ if (post)
+ local->replies[i].poststat = *post;
+ if (xdata)
+ local->replies[i].xdata = dict_ref(xdata);
- syncbarrier_wake (&local->barrier);
+ syncbarrier_wake(&local->barrier);
- return 0;
+ return 0;
}
int
-afr_selfheal_restore_time (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int source, unsigned char *healed_sinks,
- struct afr_reply *replies)
+afr_selfheal_restore_time(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int source, unsigned char *healed_sinks,
+ struct afr_reply *replies)
{
- loc_t loc = {0, };
+ loc_t loc = {
+ 0,
+ };
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONLIST (healed_sinks, frame, afr_sh_generic_fop_cbk, setattr, &loc,
- &replies[source].poststat,
- (GF_SET_ATTR_ATIME|GF_SET_ATTR_MTIME), NULL);
+ AFR_ONLIST(healed_sinks, frame, afr_sh_generic_fop_cbk, setattr, &loc,
+ &replies[source].poststat,
+ (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME), NULL);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return 0;
+ return 0;
}
dict_t *
-afr_selfheal_output_xattr (xlator_t *this, gf_boolean_t is_full_crawl,
- afr_transaction_type type, int *output_dirty,
- int **output_matrix, int subvol,
- int **full_heal_mtx_out)
-{
- int j = 0;
- int idx = 0;
- int d_idx = 0;
- int ret = 0;
- int *raw = 0;
- dict_t *xattr = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- idx = afr_index_for_transaction_type (type);
- d_idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
-
- xattr = dict_new ();
- if (!xattr)
- return NULL;
-
- /* clear dirty */
- raw = GF_CALLOC (sizeof(int), AFR_NUM_CHANGE_LOGS, gf_afr_mt_int32_t);
- if (!raw)
- goto err;
-
- raw[idx] = hton32 (output_dirty[subvol]);
- ret = dict_set_bin (xattr, AFR_DIRTY, raw,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret) {
- GF_FREE (raw);
- goto err;
- }
+afr_selfheal_output_xattr(xlator_t *this, gf_boolean_t is_full_crawl,
+ afr_transaction_type type, int *output_dirty,
+ int **output_matrix, int subvol,
+ int **full_heal_mtx_out)
+{
+ int j = 0;
+ int idx = 0;
+ int d_idx = 0;
+ int ret = 0;
+ int *raw = 0;
+ dict_t *xattr = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ idx = afr_index_for_transaction_type(type);
+ d_idx = afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
+
+ xattr = dict_new();
+ if (!xattr)
+ return NULL;
- /* clear/set pending */
- for (j = 0; j < priv->child_count; j++) {
- raw = GF_CALLOC (sizeof(int), AFR_NUM_CHANGE_LOGS,
- gf_afr_mt_int32_t);
- if (!raw)
- goto err;
-
- raw[idx] = hton32 (output_matrix[subvol][j]);
- if (is_full_crawl)
- raw[d_idx] = hton32 (full_heal_mtx_out[subvol][j]);
-
- ret = dict_set_bin (xattr, priv->pending_key[j],
- raw, sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret) {
- GF_FREE (raw);
- goto err;
- }
- }
+ /* clear dirty */
+ raw = GF_CALLOC(sizeof(int), AFR_NUM_CHANGE_LOGS, gf_afr_mt_int32_t);
+ if (!raw)
+ goto err;
+
+ raw[idx] = hton32(output_dirty[subvol]);
+ ret = dict_set_bin(xattr, AFR_DIRTY, raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ GF_FREE(raw);
+ goto err;
+ }
+
+ /* clear/set pending */
+ for (j = 0; j < priv->child_count; j++) {
+ raw = GF_CALLOC(sizeof(int), AFR_NUM_CHANGE_LOGS, gf_afr_mt_int32_t);
+ if (!raw)
+ goto err;
+
+ raw[idx] = hton32(output_matrix[subvol][j]);
+ if (is_full_crawl)
+ raw[d_idx] = hton32(full_heal_mtx_out[subvol][j]);
+
+ ret = dict_set_bin(xattr, priv->pending_key[j], raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ GF_FREE(raw);
+ goto err;
+ }
+ }
- return xattr;
+ return xattr;
err:
- if (xattr)
- dict_unref (xattr);
- return NULL;
+ if (xattr)
+ dict_unref(xattr);
+ return NULL;
}
-
int
-afr_selfheal_undo_pending (call_frame_t *frame, xlator_t *this, inode_t *inode,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- afr_transaction_type type, struct afr_reply *replies,
- unsigned char *locked_on)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int j = 0;
- unsigned char *pending = NULL;
- int *input_dirty = NULL;
- int **input_matrix = NULL;
- int **full_heal_mtx_in = NULL;
- int **full_heal_mtx_out = NULL;
- int *output_dirty = NULL;
- int **output_matrix = NULL;
- dict_t *xattr = NULL;
- dict_t *xdata = NULL;
-
- priv = this->private;
- local = frame->local;
-
- pending = alloca0 (priv->child_count);
-
- input_dirty = alloca0 (priv->child_count * sizeof (int));
- input_matrix = ALLOC_MATRIX (priv->child_count, int);
- full_heal_mtx_in = ALLOC_MATRIX (priv->child_count, int);
- full_heal_mtx_out = ALLOC_MATRIX (priv->child_count, int);
- output_dirty = alloca0 (priv->child_count * sizeof (int));
- output_matrix = ALLOC_MATRIX (priv->child_count, int);
-
- xdata = dict_new ();
- if (!xdata)
- return -1;
+afr_selfheal_undo_pending(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ afr_transaction_type type, struct afr_reply *replies,
+ unsigned char *locked_on)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int j = 0;
+ unsigned char *pending = NULL;
+ int *input_dirty = NULL;
+ int **input_matrix = NULL;
+ int **full_heal_mtx_in = NULL;
+ int **full_heal_mtx_out = NULL;
+ int *output_dirty = NULL;
+ int **output_matrix = NULL;
+ dict_t *xattr = NULL;
+ dict_t *xdata = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ pending = alloca0(priv->child_count);
+
+ input_dirty = alloca0(priv->child_count * sizeof(int));
+ input_matrix = ALLOC_MATRIX(priv->child_count, int);
+ full_heal_mtx_in = ALLOC_MATRIX(priv->child_count, int);
+ full_heal_mtx_out = ALLOC_MATRIX(priv->child_count, int);
+ output_dirty = alloca0(priv->child_count * sizeof(int));
+ output_matrix = ALLOC_MATRIX(priv->child_count, int);
+
+ xdata = dict_new();
+ if (!xdata)
+ return -1;
- afr_selfheal_extract_xattr (this, replies, type, input_dirty,
- input_matrix);
-
- if (local->need_full_crawl)
- afr_selfheal_extract_xattr (this, replies, AFR_DATA_TRANSACTION,
- NULL, full_heal_mtx_in);
-
- for (i = 0; i < priv->child_count; i++)
- if (sinks[i] && !healed_sinks[i])
- pending[i] = 1;
-
- for (i = 0; i < priv->child_count; i++) {
- for (j = 0; j < priv->child_count; j++) {
- if (pending[j]) {
- output_matrix[i][j] = 1;
- if (type == AFR_ENTRY_TRANSACTION)
- full_heal_mtx_out[i][j] = 1;
- } else if (locked_on[j]) {
- output_matrix[i][j] = -input_matrix[i][j];
- if (type == AFR_ENTRY_TRANSACTION)
- full_heal_mtx_out[i][j] = -full_heal_mtx_in[i][j];
- }
- }
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (!pending[i])
- output_dirty[i] = -input_dirty[i];
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (!locked_on[i])
- /* perform post-op only on subvols we had locked
- and inspected on.
- */
- continue;
- if (undid_pending[i])
- /* We already unset the pending xattrs in
- * _afr_fav_child_reset_sink_xattrs(). */
- continue;
-
- xattr = afr_selfheal_output_xattr (this, local->need_full_crawl,
- type, output_dirty,
- output_matrix, i,
- full_heal_mtx_out);
- if (!xattr) {
- continue;
- }
-
- if ((type == AFR_ENTRY_TRANSACTION) && (priv->esh_granular)) {
- if (xdata &&
- dict_set_int8 (xdata, GF_XATTROP_PURGE_INDEX, 1))
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_DICT_SET_FAILED, "Failed to set"
- " dict value for %s",
- GF_XATTROP_PURGE_INDEX);
- }
+ afr_selfheal_extract_xattr(this, replies, type, input_dirty, input_matrix);
+
+ if (local->need_full_crawl)
+ afr_selfheal_extract_xattr(this, replies, AFR_DATA_TRANSACTION, NULL,
+ full_heal_mtx_in);
- afr_selfheal_post_op (frame, this, inode, i, xattr, xdata);
- dict_unref (xattr);
- }
+ for (i = 0; i < priv->child_count; i++)
+ if (sinks[i] && !healed_sinks[i])
+ pending[i] = 1;
+
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++) {
+ if (pending[j]) {
+ output_matrix[i][j] = 1;
+ if (type == AFR_ENTRY_TRANSACTION)
+ full_heal_mtx_out[i][j] = 1;
+ } else if (locked_on[j]) {
+ output_matrix[i][j] = -input_matrix[i][j];
+ if (type == AFR_ENTRY_TRANSACTION)
+ full_heal_mtx_out[i][j] = -full_heal_mtx_in[i][j];
+ }
+ }
+ }
- if (xdata)
- dict_unref (xdata);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!pending[i])
+ output_dirty[i] = -input_dirty[i];
+ }
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!locked_on[i])
+ /* perform post-op only on subvols we had locked
+ and inspected on.
+ */
+ continue;
+ if (undid_pending[i])
+ /* We already unset the pending xattrs in
+ * _afr_fav_child_reset_sink_xattrs(). */
+ continue;
+
+ xattr = afr_selfheal_output_xattr(this, local->need_full_crawl, type,
+ output_dirty, output_matrix, i,
+ full_heal_mtx_out);
+ if (!xattr) {
+ continue;
+ }
+
+ if ((type == AFR_ENTRY_TRANSACTION) && (priv->esh_granular)) {
+ if (xdata && dict_set_int8(xdata, GF_XATTROP_PURGE_INDEX, 1))
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_DICT_SET_FAILED,
+ "Failed to set"
+ " dict value for %s",
+ GF_XATTROP_PURGE_INDEX);
+ }
+
+ afr_selfheal_post_op(frame, this, inode, i, xattr, xdata);
+ dict_unref(xattr);
+ }
+
+ if (xdata)
+ dict_unref(xdata);
+
+ return 0;
}
void
-afr_reply_copy (struct afr_reply *dst, struct afr_reply *src)
-{
- dict_t *xdata = NULL;
-
- dst->valid = src->valid;
- dst->op_ret = src->op_ret;
- dst->op_errno = src->op_errno;
- dst->prestat = src->prestat;
- dst->poststat = src->poststat;
- dst->preparent = src->preparent;
- dst->postparent = src->postparent;
- dst->preparent2 = src->preparent2;
- dst->postparent2 = src->postparent2;
- if (src->xdata)
- xdata = dict_ref (src->xdata);
- else
- xdata = NULL;
- if (dst->xdata)
- dict_unref (dst->xdata);
- dst->xdata = xdata;
- if (xdata && dict_get_str_boolean (xdata, "fips-mode-rchecksum",
- _gf_false) == _gf_true) {
- memcpy (dst->checksum, src->checksum, SHA256_DIGEST_LENGTH);
- } else {
- memcpy (dst->checksum, src->checksum, MD5_DIGEST_LENGTH);
- }
- dst->fips_mode_rchecksum = src->fips_mode_rchecksum;
+afr_reply_copy(struct afr_reply *dst, struct afr_reply *src)
+{
+ dict_t *xdata = NULL;
+
+ dst->valid = src->valid;
+ dst->op_ret = src->op_ret;
+ dst->op_errno = src->op_errno;
+ dst->prestat = src->prestat;
+ dst->poststat = src->poststat;
+ dst->preparent = src->preparent;
+ dst->postparent = src->postparent;
+ dst->preparent2 = src->preparent2;
+ dst->postparent2 = src->postparent2;
+ if (src->xdata)
+ xdata = dict_ref(src->xdata);
+ else
+ xdata = NULL;
+ if (dst->xdata)
+ dict_unref(dst->xdata);
+ dst->xdata = xdata;
+ if (xdata && dict_get_str_boolean(xdata, "fips-mode-rchecksum",
+ _gf_false) == _gf_true) {
+ memcpy(dst->checksum, src->checksum, SHA256_DIGEST_LENGTH);
+ } else {
+ memcpy(dst->checksum, src->checksum, MD5_DIGEST_LENGTH);
+ }
+ dst->fips_mode_rchecksum = src->fips_mode_rchecksum;
}
void
-afr_replies_copy (struct afr_reply *dst, struct afr_reply *src, int count)
+afr_replies_copy(struct afr_reply *dst, struct afr_reply *src, int count)
{
- int i = 0;
+ int i = 0;
- if (dst == src)
- return;
+ if (dst == src)
+ return;
- for (i = 0; i < count; i++) {
- afr_reply_copy (&dst[i], &src[i]);
- }
+ for (i = 0; i < count; i++) {
+ afr_reply_copy(&dst[i], &src[i]);
+ }
}
int
-afr_selfheal_fill_dirty (xlator_t *this, int *dirty, int subvol,
- int idx, dict_t *xdata)
+afr_selfheal_fill_dirty(xlator_t *this, int *dirty, int subvol, int idx,
+ dict_t *xdata)
{
- void *pending_raw = NULL;
- int pending[3] = {0, };
+ void *pending_raw = NULL;
+ int pending[3] = {
+ 0,
+ };
- if (!dirty)
- return 0;
+ if (!dirty)
+ return 0;
- if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw))
- return -1;
+ if (dict_get_ptr(xdata, AFR_DIRTY, &pending_raw))
+ return -1;
- if (!pending_raw)
- return -1;
+ if (!pending_raw)
+ return -1;
- memcpy (pending, pending_raw, sizeof(pending));
+ memcpy(pending, pending_raw, sizeof(pending));
- dirty[subvol] = ntoh32 (pending[idx]);
+ dirty[subvol] = ntoh32(pending[idx]);
- return 0;
+ return 0;
}
-
int
-afr_selfheal_fill_matrix (xlator_t *this, int **matrix, int subvol,
- int idx, dict_t *xdata)
+afr_selfheal_fill_matrix(xlator_t *this, int **matrix, int subvol, int idx,
+ dict_t *xdata)
{
- int i = 0;
- void *pending_raw = NULL;
- int pending[3] = {0, };
- afr_private_t *priv = NULL;
+ int i = 0;
+ void *pending_raw = NULL;
+ int pending[3] = {
+ 0,
+ };
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!matrix)
- return 0;
+ if (!matrix)
+ return 0;
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get_ptr (xdata, priv->pending_key[i], &pending_raw))
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_get_ptr(xdata, priv->pending_key[i], &pending_raw))
+ continue;
- if (!pending_raw)
- continue;
+ if (!pending_raw)
+ continue;
- memcpy (pending, pending_raw, sizeof(pending));
+ memcpy(pending, pending_raw, sizeof(pending));
- matrix[subvol][i] = ntoh32 (pending[idx]);
- }
+ matrix[subvol][i] = ntoh32(pending[idx]);
+ }
- return 0;
+ return 0;
}
-
int
-afr_selfheal_extract_xattr (xlator_t *this, struct afr_reply *replies,
- afr_transaction_type type, int *dirty, int **matrix)
+afr_selfheal_extract_xattr(xlator_t *this, struct afr_reply *replies,
+ afr_transaction_type type, int *dirty, int **matrix)
{
- afr_private_t *priv = NULL;
- int i = 0;
- dict_t *xdata = NULL;
- int idx = -1;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ dict_t *xdata = NULL;
+ int idx = -1;
- idx = afr_index_for_transaction_type (type);
+ idx = afr_index_for_transaction_type(type);
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
- if (!replies[i].xdata)
- continue;
+ if (!replies[i].xdata)
+ continue;
- xdata = replies[i].xdata;
+ xdata = replies[i].xdata;
- afr_selfheal_fill_dirty (this, dirty, i, idx, xdata);
- afr_selfheal_fill_matrix (this, matrix, i, idx, xdata);
- }
+ afr_selfheal_fill_dirty(this, dirty, i, idx, xdata);
+ afr_selfheal_fill_matrix(this, matrix, i, idx, xdata);
+ }
- return 0;
+ return 0;
}
/*
@@ -777,573 +786,547 @@ afr_selfheal_extract_xattr (xlator_t *this, struct afr_reply *replies,
* This can happen if data was directly modified in the backend or for snapshots
*/
void
-afr_mark_largest_file_as_source (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies)
+afr_mark_largest_file_as_source(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies)
{
- int i = 0;
- afr_private_t *priv = NULL;
- uint64_t size = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ uint64_t size = 0;
- /* Find source with biggest file size */
- priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (!replies[i].valid || replies[i].op_ret != 0) {
- sources[i] = 0;
- continue;
- }
- if (size <= replies[i].poststat.ia_size) {
- size = replies[i].poststat.ia_size;
- }
+ /* Find source with biggest file size */
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (!replies[i].valid || replies[i].op_ret != 0) {
+ sources[i] = 0;
+ continue;
}
-
- /* Mark sources with less size as not source */
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (size > replies[i].poststat.ia_size)
- sources[i] = 0;
+ if (size <= replies[i].poststat.ia_size) {
+ size = replies[i].poststat.ia_size;
}
+ }
+
+ /* Mark sources with less size as not source */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (size > replies[i].poststat.ia_size)
+ sources[i] = 0;
+ }
}
void
-afr_mark_latest_mtime_file_as_source (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies)
-{
- int i = 0;
- afr_private_t *priv = NULL;
- uint32_t mtime = 0;
- uint32_t mtime_nsec = 0;
-
- priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (!replies[i].valid || replies[i].op_ret != 0) {
- sources[i] = 0;
- continue;
- }
- if ((mtime < replies[i].poststat.ia_mtime) ||
- ((mtime == replies[i].poststat.ia_mtime) &&
- (mtime_nsec < replies[i].poststat.ia_mtime_nsec))) {
- mtime = replies[i].poststat.ia_mtime;
- mtime_nsec = replies[i].poststat.ia_mtime_nsec;
- }
- }
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if ((mtime > replies[i].poststat.ia_mtime) ||
- ((mtime == replies[i].poststat.ia_mtime) &&
- (mtime_nsec > replies[i].poststat.ia_mtime_nsec))) {
- sources[i] = 0;
- }
- }
+afr_mark_latest_mtime_file_as_source(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies)
+{
+ int i = 0;
+ afr_private_t *priv = NULL;
+ uint32_t mtime = 0;
+ uint32_t mtime_nsec = 0;
+
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (!replies[i].valid || replies[i].op_ret != 0) {
+ sources[i] = 0;
+ continue;
+ }
+ if ((mtime < replies[i].poststat.ia_mtime) ||
+ ((mtime == replies[i].poststat.ia_mtime) &&
+ (mtime_nsec < replies[i].poststat.ia_mtime_nsec))) {
+ mtime = replies[i].poststat.ia_mtime;
+ mtime_nsec = replies[i].poststat.ia_mtime_nsec;
+ }
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if ((mtime > replies[i].poststat.ia_mtime) ||
+ ((mtime == replies[i].poststat.ia_mtime) &&
+ (mtime_nsec > replies[i].poststat.ia_mtime_nsec))) {
+ sources[i] = 0;
+ }
+ }
}
void
-afr_mark_active_sinks (xlator_t *this, unsigned char *sources,
- unsigned char *locked_on, unsigned char *sinks)
+afr_mark_active_sinks(xlator_t *this, unsigned char *sources,
+ unsigned char *locked_on, unsigned char *sinks)
{
- int i = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i] && locked_on[i])
- sinks[i] = 1;
- else
- sinks[i] = 0;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i] && locked_on[i])
+ sinks[i] = 1;
+ else
+ sinks[i] = 0;
+ }
}
gf_boolean_t
-afr_dict_contains_heal_op (call_frame_t *frame)
+afr_dict_contains_heal_op(call_frame_t *frame)
{
- afr_local_t *local = NULL;
- dict_t *xdata_req = NULL;
- int ret = 0;
- int heal_op = -1;
+ afr_local_t *local = NULL;
+ dict_t *xdata_req = NULL;
+ int ret = 0;
+ int heal_op = -1;
- local = frame->local;
- xdata_req = local->xdata_req;
- ret = dict_get_int32 (xdata_req, "heal-op", &heal_op);
- if (ret)
- return _gf_false;
- if (local->xdata_rsp == NULL) {
- local->xdata_rsp = dict_new();
- if (!local->xdata_rsp)
- return _gf_true;
- }
- ret = dict_set_str (local->xdata_rsp, "sh-fail-msg",
- "File not in split-brain");
+ local = frame->local;
+ xdata_req = local->xdata_req;
+ ret = dict_get_int32(xdata_req, "heal-op", &heal_op);
+ if (ret)
+ return _gf_false;
+ if (local->xdata_rsp == NULL) {
+ local->xdata_rsp = dict_new();
+ if (!local->xdata_rsp)
+ return _gf_true;
+ }
+ ret = dict_set_str(local->xdata_rsp, "sh-fail-msg",
+ "File not in split-brain");
- return _gf_true;
+ return _gf_true;
}
gf_boolean_t
-afr_can_decide_split_brain_source_sinks (struct afr_reply *replies,
- int child_count)
+afr_can_decide_split_brain_source_sinks(struct afr_reply *replies,
+ int child_count)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < child_count; i++)
- if (replies[i].valid != 1 || replies[i].op_ret != 0)
- return _gf_false;
+ for (i = 0; i < child_count; i++)
+ if (replies[i].valid != 1 || replies[i].op_ret != 0)
+ return _gf_false;
- return _gf_true;
+ return _gf_true;
}
int
-afr_mark_split_brain_source_sinks_by_heal_op (call_frame_t *frame,
- xlator_t *this, unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type, int heal_op)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xdata_req = NULL;
- dict_t *xdata_rsp = NULL;
- int ret = 0;
- int i = 0;
- char *name = NULL;
- int source = -1;
-
- local = frame->local;
- priv = this->private;
- xdata_req = local->xdata_req;
-
- for (i = 0; i < priv->child_count; i++) {
- if (locked_on[i])
- if (sources[i] || !sinks[i] || !healed_sinks[i]) {
- ret = -1;
- goto out;
- }
- }
- if (local->xdata_rsp == NULL) {
- local->xdata_rsp = dict_new();
- if (!local->xdata_rsp) {
- ret = -1;
- goto out;
- }
- }
- xdata_rsp = local->xdata_rsp;
-
- if (!afr_can_decide_split_brain_source_sinks (replies,
- priv->child_count)) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- SBRAIN_HEAL_NO_GO_MSG);
+afr_mark_split_brain_source_sinks_by_heal_op(
+ call_frame_t *frame, xlator_t *this, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ struct afr_reply *replies, afr_transaction_type type, int heal_op)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xdata_req = NULL;
+ dict_t *xdata_rsp = NULL;
+ int ret = 0;
+ int i = 0;
+ char *name = NULL;
+ int source = -1;
+
+ local = frame->local;
+ priv = this->private;
+ xdata_req = local->xdata_req;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (locked_on[i])
+ if (sources[i] || !sinks[i] || !healed_sinks[i]) {
ret = -1;
goto out;
- }
+ }
+ }
+ if (local->xdata_rsp == NULL) {
+ local->xdata_rsp = dict_new();
+ if (!local->xdata_rsp) {
+ ret = -1;
+ goto out;
+ }
+ }
+ xdata_rsp = local->xdata_rsp;
+
+ if (!afr_can_decide_split_brain_source_sinks(replies, priv->child_count)) {
+ ret = dict_set_str(xdata_rsp, "sh-fail-msg", SBRAIN_HEAL_NO_GO_MSG);
+ ret = -1;
+ goto out;
+ }
- for (i = 0 ; i < priv->child_count; i++)
- if (locked_on[i])
- sources[i] = 1;
- switch (heal_op) {
+ for (i = 0; i < priv->child_count; i++)
+ if (locked_on[i])
+ sources[i] = 1;
+ switch (heal_op) {
case GF_SHD_OP_SBRAIN_HEAL_FROM_BIGGER_FILE:
- if (type == AFR_METADATA_TRANSACTION) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "Use source-brick option to"
- " heal metadata split-brain");
- if (!ret)
- ret = -1;
- goto out;
- }
- afr_mark_largest_file_as_source (this, sources, replies);
- if (AFR_COUNT (sources, priv->child_count) != 1) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "No bigger file");
- if (!ret)
- ret = -1;
- goto out;
- }
- break;
+ if (type == AFR_METADATA_TRANSACTION) {
+ ret = dict_set_str(xdata_rsp, "sh-fail-msg",
+ "Use source-brick option to"
+ " heal metadata split-brain");
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ afr_mark_largest_file_as_source(this, sources, replies);
+ if (AFR_COUNT(sources, priv->child_count) != 1) {
+ ret = dict_set_str(xdata_rsp, "sh-fail-msg", "No bigger file");
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ break;
case GF_SHD_OP_SBRAIN_HEAL_FROM_LATEST_MTIME:
- if (type == AFR_METADATA_TRANSACTION) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "Use source-brick option to"
- " heal metadata split-brain");
- if (!ret)
- ret = -1;
- goto out;
- }
- afr_mark_latest_mtime_file_as_source (this, sources, replies);
- if (AFR_COUNT (sources, priv->child_count) != 1) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "No difference in mtime");
- if (!ret)
- ret = -1;
- goto out;
- }
- break;
+ if (type == AFR_METADATA_TRANSACTION) {
+ ret = dict_set_str(xdata_rsp, "sh-fail-msg",
+ "Use source-brick option to"
+ " heal metadata split-brain");
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ afr_mark_latest_mtime_file_as_source(this, sources, replies);
+ if (AFR_COUNT(sources, priv->child_count) != 1) {
+ ret = dict_set_str(xdata_rsp, "sh-fail-msg",
+ "No difference in mtime");
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ break;
case GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK:
- ret = dict_get_str (xdata_req, "child-name", &name);
- if (ret)
- goto out;
- source = afr_get_child_index_from_name (this, name);
- if (source < 0) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "Invalid brick name");
- if (!ret)
- ret = -1;
- goto out;
- }
- if (locked_on[source] != 1) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "Brick is not up");
- if (!ret)
- ret = -1;
- goto out;
- }
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- sources[source] = 1;
- break;
- default:
- ret = -1;
+ ret = dict_get_str(xdata_req, "child-name", &name);
+ if (ret)
goto out;
- }
- for (i = 0 ; i < priv->child_count; i++) {
- if (sources[i]) {
- source = i;
- break;
- }
- }
- sinks[source] = 0;
- healed_sinks[source] = 0;
- ret = source;
+ source = afr_get_child_index_from_name(this, name);
+ if (source < 0) {
+ ret = dict_set_str(xdata_rsp, "sh-fail-msg",
+ "Invalid brick name");
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ if (locked_on[source] != 1) {
+ ret = dict_set_str(xdata_rsp, "sh-fail-msg", "Brick is not up");
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ sources[source] = 1;
+ break;
+ default:
+ ret = -1;
+ goto out;
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i]) {
+ source = i;
+ break;
+ }
+ }
+ sinks[source] = 0;
+ healed_sinks[source] = 0;
+ ret = source;
out:
- if (ret < 0)
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- return ret;
-
+ if (ret < 0)
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ return ret;
}
int
-afr_sh_fav_by_majority (xlator_t *this, struct afr_reply *replies,
- inode_t *inode)
-{
- afr_private_t *priv;
- int vote_count = -1;
- int fav_child = -1;
- int i = 0;
- int k = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid == 1) {
- gf_msg_debug (this->name, 0, "Child:%s "
- "mtime_sec = %ld, size = %lu for gfid %s",
- priv->children[i]->name,
- replies[i].poststat.ia_mtime,
- replies[i].poststat.ia_size,
- uuid_utoa (inode->gfid));
- vote_count = 0;
- for (k = 0; k < priv->child_count; k++) {
- if ((replies[k].poststat.ia_mtime ==
- replies[i].poststat.ia_mtime) &&
- (replies[k].poststat.ia_size ==
- replies[i].poststat.ia_size)
- ) {
- vote_count++;
- }
- }
- if (vote_count > priv->child_count/2) {
- fav_child = i;
- break;
- }
+afr_sh_fav_by_majority(xlator_t *this, struct afr_reply *replies,
+ inode_t *inode)
+{
+ afr_private_t *priv;
+ int vote_count = -1;
+ int fav_child = -1;
+ int i = 0;
+ int k = 0;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid == 1) {
+ gf_msg_debug(this->name, 0,
+ "Child:%s "
+ "mtime_sec = %ld, size = %lu for gfid %s",
+ priv->children[i]->name, replies[i].poststat.ia_mtime,
+ replies[i].poststat.ia_size, uuid_utoa(inode->gfid));
+ vote_count = 0;
+ for (k = 0; k < priv->child_count; k++) {
+ if ((replies[k].poststat.ia_mtime ==
+ replies[i].poststat.ia_mtime) &&
+ (replies[k].poststat.ia_size ==
+ replies[i].poststat.ia_size)) {
+ vote_count++;
}
+ }
+ if (vote_count > priv->child_count / 2) {
+ fav_child = i;
+ break;
+ }
}
- return fav_child;
+ }
+ return fav_child;
}
/*
* afr_sh_fav_by_mtime: Choose favorite child by mtime.
*/
int
-afr_sh_fav_by_mtime (xlator_t *this, struct afr_reply *replies, inode_t *inode)
-{
- afr_private_t *priv;
- int fav_child = -1;
- int i = 0;
- uint32_t cmp_mtime = 0;
- uint32_t cmp_mtime_nsec = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid == 1) {
- gf_msg_debug (this->name, 0, "Child:%s "
- "mtime = %ld, mtime_nsec = %d for gfid %s",
- priv->children[i]->name,
- replies[i].poststat.ia_mtime,
- replies[i].poststat.ia_mtime_nsec,
- uuid_utoa (inode->gfid));
- if (replies[i].poststat.ia_mtime > cmp_mtime) {
- cmp_mtime = replies[i].poststat.ia_mtime;
- cmp_mtime_nsec =
- replies[i].poststat.ia_mtime_nsec;
- fav_child = i;
- } else if ((replies[i].poststat.ia_mtime == cmp_mtime)
- && (replies[i].poststat.ia_mtime_nsec >
- cmp_mtime_nsec)) {
- cmp_mtime = replies[i].poststat.ia_mtime;
- cmp_mtime_nsec =
- replies[i].poststat.ia_mtime_nsec;
- fav_child = i;
- }
- }
- }
- return fav_child;
+afr_sh_fav_by_mtime(xlator_t *this, struct afr_reply *replies, inode_t *inode)
+{
+ afr_private_t *priv;
+ int fav_child = -1;
+ int i = 0;
+ uint32_t cmp_mtime = 0;
+ uint32_t cmp_mtime_nsec = 0;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid == 1) {
+ gf_msg_debug(this->name, 0,
+ "Child:%s "
+ "mtime = %ld, mtime_nsec = %d for gfid %s",
+ priv->children[i]->name, replies[i].poststat.ia_mtime,
+ replies[i].poststat.ia_mtime_nsec,
+ uuid_utoa(inode->gfid));
+ if (replies[i].poststat.ia_mtime > cmp_mtime) {
+ cmp_mtime = replies[i].poststat.ia_mtime;
+ cmp_mtime_nsec = replies[i].poststat.ia_mtime_nsec;
+ fav_child = i;
+ } else if ((replies[i].poststat.ia_mtime == cmp_mtime) &&
+ (replies[i].poststat.ia_mtime_nsec > cmp_mtime_nsec)) {
+ cmp_mtime = replies[i].poststat.ia_mtime;
+ cmp_mtime_nsec = replies[i].poststat.ia_mtime_nsec;
+ fav_child = i;
+ }
+ }
+ }
+ return fav_child;
}
/*
* afr_sh_fav_by_ctime: Choose favorite child by ctime.
*/
int
-afr_sh_fav_by_ctime (xlator_t *this, struct afr_reply *replies, inode_t *inode)
-{
- afr_private_t *priv;
- int fav_child = -1;
- int i = 0;
- uint32_t cmp_ctime = 0;
- uint32_t cmp_ctime_nsec = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid == 1) {
- gf_msg_debug (this->name, 0, "Child:%s "
- "ctime = %ld, ctime_nsec = %d for gfid %s",
- priv->children[i]->name,
- replies[i].poststat.ia_ctime,
- replies[i].poststat.ia_ctime_nsec,
- uuid_utoa (inode->gfid));
- if (replies[i].poststat.ia_ctime > cmp_ctime) {
- cmp_ctime = replies[i].poststat.ia_ctime;
- cmp_ctime_nsec =
- replies[i].poststat.ia_ctime_nsec;
- fav_child = i;
- } else if ((replies[i].poststat.ia_ctime == cmp_ctime)
- && (replies[i].poststat.ia_ctime_nsec >
- cmp_ctime_nsec)) {
- cmp_ctime = replies[i].poststat.ia_ctime;
- cmp_ctime_nsec =
- replies[i].poststat.ia_ctime_nsec;
- fav_child = i;
- }
- }
- }
- return fav_child;
+afr_sh_fav_by_ctime(xlator_t *this, struct afr_reply *replies, inode_t *inode)
+{
+ afr_private_t *priv;
+ int fav_child = -1;
+ int i = 0;
+ uint32_t cmp_ctime = 0;
+ uint32_t cmp_ctime_nsec = 0;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid == 1) {
+ gf_msg_debug(this->name, 0,
+ "Child:%s "
+ "ctime = %ld, ctime_nsec = %d for gfid %s",
+ priv->children[i]->name, replies[i].poststat.ia_ctime,
+ replies[i].poststat.ia_ctime_nsec,
+ uuid_utoa(inode->gfid));
+ if (replies[i].poststat.ia_ctime > cmp_ctime) {
+ cmp_ctime = replies[i].poststat.ia_ctime;
+ cmp_ctime_nsec = replies[i].poststat.ia_ctime_nsec;
+ fav_child = i;
+ } else if ((replies[i].poststat.ia_ctime == cmp_ctime) &&
+ (replies[i].poststat.ia_ctime_nsec > cmp_ctime_nsec)) {
+ cmp_ctime = replies[i].poststat.ia_ctime;
+ cmp_ctime_nsec = replies[i].poststat.ia_ctime_nsec;
+ fav_child = i;
+ }
+ }
+ }
+ return fav_child;
}
/*
* afr_sh_fav_by_size: Choose favorite child by size.
*/
int
-afr_sh_fav_by_size (xlator_t *this, struct afr_reply *replies, inode_t *inode)
+afr_sh_fav_by_size(xlator_t *this, struct afr_reply *replies, inode_t *inode)
{
- afr_private_t *priv;
- int fav_child = -1;
- int i = 0;
- uint64_t cmp_sz = 0;
+ afr_private_t *priv;
+ int fav_child = -1;
+ int i = 0;
+ uint64_t cmp_sz = 0;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid == 1) {
- gf_msg_debug (this->name, 0, "Child:%s "
- "file size = %lu for gfid %s",
- priv->children[i]->name,
- replies[i].poststat.ia_size,
- uuid_utoa (inode->gfid));
- if (replies[i].poststat.ia_size > cmp_sz) {
- cmp_sz = replies[i].poststat.ia_size;
- fav_child = i;
- }
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid == 1) {
+ gf_msg_debug(this->name, 0,
+ "Child:%s "
+ "file size = %lu for gfid %s",
+ priv->children[i]->name, replies[i].poststat.ia_size,
+ uuid_utoa(inode->gfid));
+ if (replies[i].poststat.ia_size > cmp_sz) {
+ cmp_sz = replies[i].poststat.ia_size;
+ fav_child = i;
+ }
}
- return fav_child;
+ }
+ return fav_child;
}
int
-afr_sh_get_fav_by_policy (xlator_t *this, struct afr_reply *replies,
- inode_t *inode, char **policy_str)
+afr_sh_get_fav_by_policy(xlator_t *this, struct afr_reply *replies,
+ inode_t *inode, char **policy_str)
{
- afr_private_t *priv = NULL;
- int fav_child = -1;
+ afr_private_t *priv = NULL;
+ int fav_child = -1;
- priv = this->private;
- if (!afr_can_decide_split_brain_source_sinks (replies,
- priv->child_count)) {
- return -1;
- }
+ priv = this->private;
+ if (!afr_can_decide_split_brain_source_sinks(replies, priv->child_count)) {
+ return -1;
+ }
- switch (priv->fav_child_policy) {
+ switch (priv->fav_child_policy) {
case AFR_FAV_CHILD_BY_SIZE:
- fav_child = afr_sh_fav_by_size (this, replies, inode);
- if (policy_str && fav_child >= 0) {
- *policy_str = "SIZE";
- }
- break;
+ fav_child = afr_sh_fav_by_size(this, replies, inode);
+ if (policy_str && fav_child >= 0) {
+ *policy_str = "SIZE";
+ }
+ break;
case AFR_FAV_CHILD_BY_CTIME:
- fav_child = afr_sh_fav_by_ctime (this, replies, inode);
- if (policy_str && fav_child >= 0) {
- *policy_str = "CTIME";
- }
- break;
+ fav_child = afr_sh_fav_by_ctime(this, replies, inode);
+ if (policy_str && fav_child >= 0) {
+ *policy_str = "CTIME";
+ }
+ break;
case AFR_FAV_CHILD_BY_MTIME:
- fav_child = afr_sh_fav_by_mtime (this, replies, inode);
- if (policy_str && fav_child >= 0) {
- *policy_str = "MTIME";
- }
- break;
+ fav_child = afr_sh_fav_by_mtime(this, replies, inode);
+ if (policy_str && fav_child >= 0) {
+ *policy_str = "MTIME";
+ }
+ break;
case AFR_FAV_CHILD_BY_MAJORITY:
- fav_child = afr_sh_fav_by_majority (this, replies, inode);
- if (policy_str && fav_child >= 0) {
- *policy_str = "MAJORITY";
- }
- break;
+ fav_child = afr_sh_fav_by_majority(this, replies, inode);
+ if (policy_str && fav_child >= 0) {
+ *policy_str = "MAJORITY";
+ }
+ break;
case AFR_FAV_CHILD_NONE:
default:
- break;
- }
+ break;
+ }
- return fav_child;
+ return fav_child;
}
int
-afr_mark_split_brain_source_sinks_by_policy (call_frame_t *frame,
- xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type)
-{
- afr_private_t *priv = NULL;
- int fav_child = -1;
- char mtime_str[256];
- char ctime_str[256];
- char *policy_str = NULL;
- struct tm *tm_ptr;
- time_t time;
-
- priv = this->private;
-
- fav_child = afr_sh_get_fav_by_policy (this, replies, inode,
- &policy_str);
- if (fav_child > priv->child_count - 1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SBRAIN_FAV_CHILD_POLICY, "Invalid child (%d) "
- "selected by policy %s.", fav_child, policy_str);
- } else if (fav_child >= 0) {
- time = replies[fav_child].poststat.ia_mtime;
- tm_ptr = localtime (&time);
- strftime (mtime_str, sizeof (mtime_str), "%Y-%m-%d %H:%M:%S",
- tm_ptr);
- time = replies[fav_child].poststat.ia_ctime;
- tm_ptr = localtime (&time);
- strftime (ctime_str, sizeof (ctime_str), "%Y-%m-%d %H:%M:%S",
- tm_ptr);
-
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SBRAIN_FAV_CHILD_POLICY, "Source %s "
- "selected as authentic to resolve conflicting "
- "data in file (gfid:%s) by %s (%lu bytes @ %s mtime, "
- "%s ctime).",
- priv->children[fav_child]->name,
- uuid_utoa (inode->gfid),
- policy_str,
- replies[fav_child].poststat.ia_size,
- mtime_str,
- ctime_str);
-
- sources[fav_child] = 1;
- sinks[fav_child] = 0;
- healed_sinks[fav_child] = 0;
- }
- return fav_child;
+afr_mark_split_brain_source_sinks_by_policy(
+ call_frame_t *frame, xlator_t *this, inode_t *inode, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ struct afr_reply *replies, afr_transaction_type type)
+{
+ afr_private_t *priv = NULL;
+ int fav_child = -1;
+ char mtime_str[256];
+ char ctime_str[256];
+ char *policy_str = NULL;
+ struct tm *tm_ptr;
+ time_t time;
+
+ priv = this->private;
+
+ fav_child = afr_sh_get_fav_by_policy(this, replies, inode, &policy_str);
+ if (fav_child > priv->child_count - 1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SBRAIN_FAV_CHILD_POLICY,
+ "Invalid child (%d) "
+ "selected by policy %s.",
+ fav_child, policy_str);
+ } else if (fav_child >= 0) {
+ time = replies[fav_child].poststat.ia_mtime;
+ tm_ptr = localtime(&time);
+ strftime(mtime_str, sizeof(mtime_str), "%Y-%m-%d %H:%M:%S", tm_ptr);
+ time = replies[fav_child].poststat.ia_ctime;
+ tm_ptr = localtime(&time);
+ strftime(ctime_str, sizeof(ctime_str), "%Y-%m-%d %H:%M:%S", tm_ptr);
+
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SBRAIN_FAV_CHILD_POLICY,
+ "Source %s "
+ "selected as authentic to resolve conflicting "
+ "data in file (gfid:%s) by %s (%lu bytes @ %s mtime, "
+ "%s ctime).",
+ priv->children[fav_child]->name, uuid_utoa(inode->gfid),
+ policy_str, replies[fav_child].poststat.ia_size, mtime_str,
+ ctime_str);
+
+ sources[fav_child] = 1;
+ sinks[fav_child] = 0;
+ healed_sinks[fav_child] = 0;
+ }
+ return fav_child;
}
gf_boolean_t
-afr_is_file_empty_on_all_children (afr_private_t *priv,
- struct afr_reply *replies)
+afr_is_file_empty_on_all_children(afr_private_t *priv,
+ struct afr_reply *replies)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < priv->child_count; i++) {
- if ((!replies[i].valid) || (replies[i].op_ret != 0) ||
- (replies[i].poststat.ia_size != 0))
- return _gf_false;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if ((!replies[i].valid) || (replies[i].op_ret != 0) ||
+ (replies[i].poststat.ia_size != 0))
+ return _gf_false;
+ }
- return _gf_true;
+ return _gf_true;
}
int
-afr_mark_source_sinks_if_file_empty (xlator_t *this, unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type)
-{
- int source = -1;
- int i = 0;
- afr_private_t *priv = this->private;
- struct iatt stbuf = {0, };
-
- if ((AFR_COUNT (locked_on, priv->child_count) < priv->child_count) ||
- (afr_success_count(replies, priv->child_count) < priv->child_count))
- return -1;
+afr_mark_source_sinks_if_file_empty(xlator_t *this, unsigned char *sources,
+ unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on,
+ struct afr_reply *replies,
+ afr_transaction_type type)
+{
+ int source = -1;
+ int i = 0;
+ afr_private_t *priv = this->private;
+ struct iatt stbuf = {
+ 0,
+ };
+
+ if ((AFR_COUNT(locked_on, priv->child_count) < priv->child_count) ||
+ (afr_success_count(replies, priv->child_count) < priv->child_count))
+ return -1;
- if (type == AFR_DATA_TRANSACTION) {
- if (!afr_is_file_empty_on_all_children(priv, replies))
- return -1;
- goto mark;
- }
-
- /*For AFR_METADATA_TRANSACTION, metadata must be same on all bricks.*/
- stbuf = replies[0].poststat;
- for (i = 1; i < priv->child_count; i++) {
- if ((!IA_EQUAL (stbuf, replies[i].poststat, type)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, uid)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, gid)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, prot)))
- return -1;
- }
- for (i = 1; i < priv->child_count; i++) {
- if (!afr_xattrs_are_equal (replies[0].xdata,
- replies[i].xdata))
- return -1;
- }
+ if (type == AFR_DATA_TRANSACTION) {
+ if (!afr_is_file_empty_on_all_children(priv, replies))
+ return -1;
+ goto mark;
+ }
+
+ /*For AFR_METADATA_TRANSACTION, metadata must be same on all bricks.*/
+ stbuf = replies[0].poststat;
+ for (i = 1; i < priv->child_count; i++) {
+ if ((!IA_EQUAL(stbuf, replies[i].poststat, type)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, uid)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, gid)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, prot)))
+ return -1;
+ }
+ for (i = 1; i < priv->child_count; i++) {
+ if (!afr_xattrs_are_equal(replies[0].xdata, replies[i].xdata))
+ return -1;
+ }
mark:
- /* data/metadata is same on all bricks. Pick one of them as source. Rest
- * are sinks.*/
- for (i = 0 ; i < priv->child_count; i++) {
- if (source == -1) {
- source = i;
- sources[i] = 1;
- sinks[i] = 0;
- healed_sinks[i] = 0;
- continue;
- }
- sources[i] = 0;
- sinks[i] = 1;
- healed_sinks[i] = 1;
+ /* data/metadata is same on all bricks. Pick one of them as source. Rest
+ * are sinks.*/
+ for (i = 0; i < priv->child_count; i++) {
+ if (source == -1) {
+ source = i;
+ sources[i] = 1;
+ sinks[i] = 0;
+ healed_sinks[i] = 0;
+ continue;
}
+ sources[i] = 0;
+ sinks[i] = 1;
+ healed_sinks[i] = 1;
+ }
- return source;
+ return source;
}
/* Return a source depending on the type of heal_op, and set sources[source],
@@ -1354,171 +1337,156 @@ mark:
* sinks[node] are 1. This should be the case if the file is in split-brain.
*/
int
-afr_mark_split_brain_source_sinks (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xdata_req = NULL;
- int heal_op = -1;
- int ret = -1;
- int source = -1;
-
- local = frame->local;
- priv = this->private;
- xdata_req = local->xdata_req;
-
- source = afr_mark_source_sinks_if_file_empty (this, sources, sinks,
- healed_sinks, locked_on,
- replies, type);
- if (source >= 0)
- return source;
-
- ret = dict_get_int32 (xdata_req, "heal-op", &heal_op);
- if (ret)
- goto autoheal;
-
- source = afr_mark_split_brain_source_sinks_by_heal_op (frame, this,
- sources, sinks,
- healed_sinks,
- locked_on, replies,
- type, heal_op);
+afr_mark_split_brain_source_sinks(
+ call_frame_t *frame, xlator_t *this, inode_t *inode, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ struct afr_reply *replies, afr_transaction_type type)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xdata_req = NULL;
+ int heal_op = -1;
+ int ret = -1;
+ int source = -1;
+
+ local = frame->local;
+ priv = this->private;
+ xdata_req = local->xdata_req;
+
+ source = afr_mark_source_sinks_if_file_empty(
+ this, sources, sinks, healed_sinks, locked_on, replies, type);
+ if (source >= 0)
return source;
+ ret = dict_get_int32(xdata_req, "heal-op", &heal_op);
+ if (ret)
+ goto autoheal;
+
+ source = afr_mark_split_brain_source_sinks_by_heal_op(
+ frame, this, sources, sinks, healed_sinks, locked_on, replies, type,
+ heal_op);
+ return source;
+
autoheal:
- /* Automatically heal if fav_child_policy is set. */
- if (priv->fav_child_policy != AFR_FAV_CHILD_NONE) {
- source = afr_mark_split_brain_source_sinks_by_policy (frame,
- this,
- inode,
- sources,
- sinks,
- healed_sinks,
- locked_on,
- replies,
- type);
- if (source != -1) {
- ret = dict_set_int32 (xdata_req, "fav-child-policy", 1);
- if (ret)
- return -1;
- }
+ /* Automatically heal if fav_child_policy is set. */
+ if (priv->fav_child_policy != AFR_FAV_CHILD_NONE) {
+ source = afr_mark_split_brain_source_sinks_by_policy(
+ frame, this, inode, sources, sinks, healed_sinks, locked_on,
+ replies, type);
+ if (source != -1) {
+ ret = dict_set_int32(xdata_req, "fav-child-policy", 1);
+ if (ret)
+ return -1;
}
+ }
- return source;
+ return source;
}
int
-_afr_fav_child_reset_sink_xattrs (call_frame_t *frame, xlator_t *this,
- inode_t *inode, int source,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- afr_transaction_type type,
- unsigned char *locked_on,
- struct afr_reply *replies)
+_afr_fav_child_reset_sink_xattrs(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, int source,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ afr_transaction_type type,
+ unsigned char *locked_on,
+ struct afr_reply *replies)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int *input_dirty = NULL;
- int **input_matrix = NULL;
- int *output_dirty = NULL;
- int **output_matrix = NULL;
- dict_t *xattr = NULL;
- dict_t *xdata = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int *input_dirty = NULL;
+ int **input_matrix = NULL;
+ int *output_dirty = NULL;
+ int **output_matrix = NULL;
+ dict_t *xattr = NULL;
+ dict_t *xdata = NULL;
+ int i = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- if (!dict_get (local->xdata_req, "fav-child-policy"))
- return 0;
+ if (!dict_get(local->xdata_req, "fav-child-policy"))
+ return 0;
- xdata = dict_new();
- if (!xdata)
- return -1;
+ xdata = dict_new();
+ if (!xdata)
+ return -1;
- input_dirty = alloca0 (priv->child_count * sizeof (int));
- input_matrix = ALLOC_MATRIX (priv->child_count, int);
- output_dirty = alloca0 (priv->child_count * sizeof (int));
- output_matrix = ALLOC_MATRIX (priv->child_count, int);
+ input_dirty = alloca0(priv->child_count * sizeof(int));
+ input_matrix = ALLOC_MATRIX(priv->child_count, int);
+ output_dirty = alloca0(priv->child_count * sizeof(int));
+ output_matrix = ALLOC_MATRIX(priv->child_count, int);
- afr_selfheal_extract_xattr (this, replies, type, input_dirty,
- input_matrix);
+ afr_selfheal_extract_xattr(this, replies, type, input_dirty, input_matrix);
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || !healed_sinks[i])
- continue;
- output_dirty[i] = -input_dirty[i];
- output_matrix[i][source] = -input_matrix[i][source];
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source || !healed_sinks[i])
+ continue;
+ output_dirty[i] = -input_dirty[i];
+ output_matrix[i][source] = -input_matrix[i][source];
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i] || !locked_on[i])
- continue;
- xattr = afr_selfheal_output_xattr (this, _gf_false, type,
- output_dirty, output_matrix,
- i, NULL);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i] || !locked_on[i])
+ continue;
+ xattr = afr_selfheal_output_xattr(this, _gf_false, type, output_dirty,
+ output_matrix, i, NULL);
- afr_selfheal_post_op (frame, this, inode, i, xattr, xdata);
+ afr_selfheal_post_op(frame, this, inode, i, xattr, xdata);
- undid_pending[i] = 1;
- dict_unref (xattr);
- }
+ undid_pending[i] = 1;
+ dict_unref(xattr);
+ }
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- return 0;
+ return 0;
}
gf_boolean_t
-afr_does_witness_exist (xlator_t *this, uint64_t *witness)
+afr_does_witness_exist(xlator_t *this, uint64_t *witness)
{
- int i = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (witness[i])
- return _gf_true;
- }
- return _gf_false;
+ for (i = 0; i < priv->child_count; i++) {
+ if (witness[i])
+ return _gf_true;
+ }
+ return _gf_false;
}
unsigned int
-afr_get_quorum_count (afr_private_t *priv)
+afr_get_quorum_count(afr_private_t *priv)
{
- if (priv->quorum_count == AFR_QUORUM_AUTO) {
- return priv->child_count/2 + 1;
- } else {
- return priv->quorum_count;
- }
+ if (priv->quorum_count == AFR_QUORUM_AUTO) {
+ return priv->child_count / 2 + 1;
+ } else {
+ return priv->quorum_count;
+ }
}
void
-afr_selfheal_post_op_failure_accounting (afr_private_t *priv, char *accused,
- unsigned char *sources,
- unsigned char *locked_on)
+afr_selfheal_post_op_failure_accounting(afr_private_t *priv, char *accused,
+ unsigned char *sources,
+ unsigned char *locked_on)
{
- int i = 0;
- unsigned int quorum_count = 0;
+ int i = 0;
+ unsigned int quorum_count = 0;
- if (AFR_COUNT (sources, priv->child_count) != 0)
- return;
+ if (AFR_COUNT(sources, priv->child_count) != 0)
+ return;
- quorum_count = afr_get_quorum_count (priv);
- for (i = 0; i < priv->child_count; i++) {
- if ((accused[i] < quorum_count) && locked_on[i]) {
- sources[i] = 1;
- }
+ quorum_count = afr_get_quorum_count(priv);
+ for (i = 0; i < priv->child_count; i++) {
+ if ((accused[i] < quorum_count) && locked_on[i]) {
+ sources[i] = 1;
}
- return;
+ }
+ return;
}
/*
@@ -1541,663 +1509,675 @@ afr_selfheal_post_op_failure_accounting (afr_private_t *priv, char *accused,
*/
int
-afr_selfheal_find_direction (call_frame_t *frame, xlator_t *this,
- struct afr_reply *replies,
- afr_transaction_type type,
- unsigned char *locked_on, unsigned char *sources,
- unsigned char *sinks, uint64_t *witness,
- gf_boolean_t *pflag)
-{
- afr_private_t *priv = NULL;
- int i = 0;
- int j = 0;
- int *dirty = NULL; /* Denotes if dirty xattr is set */
- int **matrix = NULL;/* Changelog matrix */
- char *accused = NULL;/* Accused others without any self-accusal */
- char *pending = NULL;/* Have pending operations on others */
- char *self_accused = NULL; /* Accused itself */
- int min_participants = -1;
-
- priv = this->private;
-
- dirty = alloca0 (priv->child_count * sizeof (int));
- accused = alloca0 (priv->child_count);
- pending = alloca0 (priv->child_count);
- self_accused = alloca0 (priv->child_count);
- matrix = ALLOC_MATRIX(priv->child_count, int);
- memset (witness, 0, sizeof (*witness) * priv->child_count);
-
- /* First construct the pending matrix for further analysis */
- afr_selfheal_extract_xattr (this, replies, type, dirty, matrix);
-
- if (pflag) {
- for (i = 0; i < priv->child_count; i++) {
- for (j = 0; j < priv->child_count; j++)
- if (matrix[i][j])
- *pflag = _gf_true;
- if (*pflag)
- break;
- }
- }
-
- if (type == AFR_DATA_TRANSACTION) {
- min_participants = priv->child_count;
- } else {
- min_participants = AFR_SH_MIN_PARTICIPANTS;
+afr_selfheal_find_direction(call_frame_t *frame, xlator_t *this,
+ struct afr_reply *replies,
+ afr_transaction_type type, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ uint64_t *witness, gf_boolean_t *pflag)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int j = 0;
+ int *dirty = NULL; /* Denotes if dirty xattr is set */
+ int **matrix = NULL; /* Changelog matrix */
+ char *accused = NULL; /* Accused others without any self-accusal */
+ char *pending = NULL; /* Have pending operations on others */
+ char *self_accused = NULL; /* Accused itself */
+ int min_participants = -1;
+
+ priv = this->private;
+
+ dirty = alloca0(priv->child_count * sizeof(int));
+ accused = alloca0(priv->child_count);
+ pending = alloca0(priv->child_count);
+ self_accused = alloca0(priv->child_count);
+ matrix = ALLOC_MATRIX(priv->child_count, int);
+ memset(witness, 0, sizeof(*witness) * priv->child_count);
+
+ /* First construct the pending matrix for further analysis */
+ afr_selfheal_extract_xattr(this, replies, type, dirty, matrix);
+
+ if (pflag) {
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++)
+ if (matrix[i][j])
+ *pflag = _gf_true;
+ if (*pflag)
+ break;
}
- if (afr_success_count (replies,
- priv->child_count) < min_participants) {
- /* Treat this just like locks not being acquired */
- return -ENOTCONN;
+ }
+
+ if (type == AFR_DATA_TRANSACTION) {
+ min_participants = priv->child_count;
+ } else {
+ min_participants = AFR_SH_MIN_PARTICIPANTS;
+ }
+ if (afr_success_count(replies, priv->child_count) < min_participants) {
+ /* Treat this just like locks not being acquired */
+ return -ENOTCONN;
+ }
+
+ /* short list all self-accused */
+ for (i = 0; i < priv->child_count; i++) {
+ if (matrix[i][i])
+ self_accused[i] = 1;
+ }
+
+ /* Next short list all accused to exclude them from being sources */
+ /* Self-accused can't accuse others as they are FOOLs */
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++) {
+ if (matrix[i][j]) {
+ if (!self_accused[i])
+ accused[j] += 1;
+ if (i != j)
+ pending[i] += 1;
+ }
}
+ }
- /* short list all self-accused */
+ /* Short list all non-accused as sources */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!accused[i] && locked_on[i])
+ sources[i] = 1;
+ else
+ sources[i] = 0;
+ }
+
+ /* Everyone accused by non-self-accused sources are sinks */
+ memset(sinks, 0, priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (self_accused[i])
+ continue;
+ for (j = 0; j < priv->child_count; j++) {
+ if (matrix[i][j])
+ sinks[j] = 1;
+ }
+ }
+
+ /* For breaking ties provide with number of fops they witnessed */
+
+ /*
+ * count the pending fops witnessed from itself to others when it is
+ * self-accused
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!self_accused[i])
+ continue;
+ for (j = 0; j < priv->child_count; j++) {
+ if (i == j)
+ continue;
+ witness[i] += matrix[i][j];
+ }
+ }
+
+ if (type == AFR_DATA_TRANSACTION)
+ afr_selfheal_post_op_failure_accounting(priv, accused, sources,
+ locked_on);
+
+ /* If no sources, all locked nodes are sinks - split brain */
+ if (AFR_COUNT(sources, priv->child_count) == 0) {
for (i = 0; i < priv->child_count; i++) {
- if (matrix[i][i])
- self_accused[i] = 1;
+ if (locked_on[i])
+ sinks[i] = 1;
}
+ }
- /* Next short list all accused to exclude them from being sources */
- /* Self-accused can't accuse others as they are FOOLs */
- for (i = 0; i < priv->child_count; i++) {
- for (j = 0; j < priv->child_count; j++) {
- if (matrix[i][j]) {
- if (!self_accused[i])
- accused[j] += 1;
- if (i != j)
- pending[i] += 1;
- }
- }
- }
-
- /* Short list all non-accused as sources */
- for (i = 0; i < priv->child_count; i++) {
- if (!accused[i] && locked_on[i])
- sources[i] = 1;
- else
- sources[i] = 0;
- }
-
- /* Everyone accused by non-self-accused sources are sinks */
- memset (sinks, 0, priv->child_count);
+ /* One more class of witness similar to dirty in v2 is where no pending
+ * exists but we have self-accusing markers. This can happen in afr-v1
+ * if the brick crashes just after doing xattrop on self but
+ * before xattrop on the other xattrs on the brick in pre-op. */
+ if (AFR_COUNT(pending, priv->child_count) == 0) {
for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (self_accused[i])
- continue;
- for (j = 0; j < priv->child_count; j++) {
- if (matrix[i][j])
- sinks[j] = 1;
- }
+ if (self_accused[i])
+ witness[i] += matrix[i][i];
}
-
- /* For breaking ties provide with number of fops they witnessed */
-
- /*
- * count the pending fops witnessed from itself to others when it is
- * self-accused
+ } else {
+ /* In afr-v1 if a file is self-accused and has pending
+ * operations on others then it is similar to 'dirty' in afr-v2.
+ * Consider such cases as witness.
*/
for (i = 0; i < priv->child_count; i++) {
- if (!self_accused[i])
- continue;
- for (j = 0; j < priv->child_count; j++) {
- if (i == j)
- continue;
- witness[i] += matrix[i][j];
- }
+ if (self_accused[i] && pending[i])
+ witness[i] += matrix[i][i];
}
+ }
- if (type == AFR_DATA_TRANSACTION)
- afr_selfheal_post_op_failure_accounting (priv, accused,
- sources, locked_on);
+ /* count the number of dirty fops witnessed */
+ for (i = 0; i < priv->child_count; i++)
+ witness[i] += dirty[i];
- /* If no sources, all locked nodes are sinks - split brain */
- if (AFR_COUNT (sources, priv->child_count) == 0) {
- for (i = 0; i < priv->child_count; i++) {
- if (locked_on[i])
- sinks[i] = 1;
- }
- }
-
- /* One more class of witness similar to dirty in v2 is where no pending
- * exists but we have self-accusing markers. This can happen in afr-v1
- * if the brick crashes just after doing xattrop on self but
- * before xattrop on the other xattrs on the brick in pre-op. */
- if (AFR_COUNT (pending, priv->child_count) == 0) {
- for (i = 0; i < priv->child_count; i++) {
- if (self_accused[i])
- witness[i] += matrix[i][i];
- }
- } else {
- /* In afr-v1 if a file is self-accused and has pending
- * operations on others then it is similar to 'dirty' in afr-v2.
- * Consider such cases as witness.
- */
- for (i = 0; i < priv->child_count; i++) {
- if (self_accused[i] && pending[i])
- witness[i] += matrix[i][i];
- }
- }
-
-
- /* count the number of dirty fops witnessed */
- for (i = 0; i < priv->child_count; i++)
- witness[i] += dirty[i];
-
- return 0;
+ return 0;
}
void
-afr_log_selfheal (uuid_t gfid, xlator_t *this, int ret, char *type,
- int source, unsigned char *sources,
- unsigned char *healed_sinks)
-{
- char *status = NULL;
- char *sinks_str = NULL;
- char *p = NULL;
- char *sources_str = NULL;
- char *q = NULL;
- afr_private_t *priv = NULL;
- gf_loglevel_t loglevel = GF_LOG_NONE;
- int i = 0;
-
- priv = this->private;
- sinks_str = alloca0 (priv->child_count * 8);
- p = sinks_str;
- sources_str = alloca0 (priv->child_count * 8);
- q = sources_str;
- for (i = 0; i < priv->child_count; i++) {
- if (healed_sinks[i])
- p += sprintf (p, "%d ", i);
- if (sources[i]) {
- if (source == i) {
- q += sprintf (q, "[%d] ", i);
- } else {
- q += sprintf (q, "%d ", i);
- }
- }
- }
-
- if (ret < 0) {
- status = "Failed";
- loglevel = GF_LOG_DEBUG;
- } else {
- status = "Completed";
- loglevel = GF_LOG_INFO;
- }
-
- gf_msg (this->name, loglevel, 0,
- AFR_MSG_SELF_HEAL_INFO, "%s %s selfheal on %s. "
- "sources=%s sinks=%s", status, type, uuid_utoa (gfid),
- sources_str, sinks_str);
+afr_log_selfheal(uuid_t gfid, xlator_t *this, int ret, char *type, int source,
+ unsigned char *sources, unsigned char *healed_sinks)
+{
+ char *status = NULL;
+ char *sinks_str = NULL;
+ char *p = NULL;
+ char *sources_str = NULL;
+ char *q = NULL;
+ afr_private_t *priv = NULL;
+ gf_loglevel_t loglevel = GF_LOG_NONE;
+ int i = 0;
+
+ priv = this->private;
+ sinks_str = alloca0(priv->child_count * 8);
+ p = sinks_str;
+ sources_str = alloca0(priv->child_count * 8);
+ q = sources_str;
+ for (i = 0; i < priv->child_count; i++) {
+ if (healed_sinks[i])
+ p += sprintf(p, "%d ", i);
+ if (sources[i]) {
+ if (source == i) {
+ q += sprintf(q, "[%d] ", i);
+ } else {
+ q += sprintf(q, "%d ", i);
+ }
+ }
+ }
+
+ if (ret < 0) {
+ status = "Failed";
+ loglevel = GF_LOG_DEBUG;
+ } else {
+ status = "Completed";
+ loglevel = GF_LOG_INFO;
+ }
+
+ gf_msg(this->name, loglevel, 0, AFR_MSG_SELF_HEAL_INFO,
+ "%s %s selfheal on %s. "
+ "sources=%s sinks=%s",
+ status, type, uuid_utoa(gfid), sources_str, sinks_str);
}
int
-afr_selfheal_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *parbuf)
-{
- afr_local_t *local = NULL;
- int i = -1;
- GF_UNUSED int ret = -1;
- int8_t need_heal = 1;
-
- local = frame->local;
- i = (long) cookie;
-
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
- if (buf)
- local->replies[i].poststat = *buf;
- if (parbuf)
- local->replies[i].postparent = *parbuf;
- if (xdata) {
- local->replies[i].xdata = dict_ref (xdata);
- ret = dict_get_int8 (xdata, "link-count", &need_heal);
- local->replies[i].need_heal = need_heal;
- } else {
- local->replies[i].need_heal = need_heal;
- }
-
- syncbarrier_wake (&local->barrier);
-
- return 0;
+afr_selfheal_discover_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *parbuf)
+{
+ afr_local_t *local = NULL;
+ int i = -1;
+ GF_UNUSED int ret = -1;
+ int8_t need_heal = 1;
+
+ local = frame->local;
+ i = (long)cookie;
+
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
+ if (buf)
+ local->replies[i].poststat = *buf;
+ if (parbuf)
+ local->replies[i].postparent = *parbuf;
+ if (xdata) {
+ local->replies[i].xdata = dict_ref(xdata);
+ ret = dict_get_int8(xdata, "link-count", &need_heal);
+ local->replies[i].need_heal = need_heal;
+ } else {
+ local->replies[i].need_heal = need_heal;
+ }
+
+ syncbarrier_wake(&local->barrier);
+
+ return 0;
}
-
inode_t *
-afr_selfheal_unlocked_lookup_on (call_frame_t *frame, inode_t *parent,
- const char *name, struct afr_reply *replies,
- unsigned char *lookup_on, dict_t *xattr)
-{
- loc_t loc = {0, };
- dict_t *xattr_req = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- inode_t *inode = NULL;
-
- local = frame->local;
- priv = frame->this->private;
-
- xattr_req = dict_new ();
- if (!xattr_req)
- return NULL;
+afr_selfheal_unlocked_lookup_on(call_frame_t *frame, inode_t *parent,
+ const char *name, struct afr_reply *replies,
+ unsigned char *lookup_on, dict_t *xattr)
+{
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr_req = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ inode_t *inode = NULL;
+
+ local = frame->local;
+ priv = frame->this->private;
+
+ xattr_req = dict_new();
+ if (!xattr_req)
+ return NULL;
- if (xattr)
- dict_copy (xattr, xattr_req);
+ if (xattr)
+ dict_copy(xattr, xattr_req);
- if (afr_xattr_req_prepare (frame->this, xattr_req) != 0) {
- dict_unref (xattr_req);
- return NULL;
- }
+ if (afr_xattr_req_prepare(frame->this, xattr_req) != 0) {
+ dict_unref(xattr_req);
+ return NULL;
+ }
- inode = inode_new (parent->table);
- if (!inode) {
- dict_unref (xattr_req);
- return NULL;
- }
+ inode = inode_new(parent->table);
+ if (!inode) {
+ dict_unref(xattr_req);
+ return NULL;
+ }
- loc.parent = inode_ref (parent);
- gf_uuid_copy (loc.pargfid, parent->gfid);
- loc.name = name;
- loc.inode = inode_ref (inode);
+ loc.parent = inode_ref(parent);
+ gf_uuid_copy(loc.pargfid, parent->gfid);
+ loc.name = name;
+ loc.inode = inode_ref(inode);
- AFR_ONLIST (lookup_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
- xattr_req);
+ AFR_ONLIST(lookup_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+ xattr_req);
- afr_replies_copy (replies, local->replies, priv->child_count);
+ afr_replies_copy(replies, local->replies, priv->child_count);
- loc_wipe (&loc);
- dict_unref (xattr_req);
+ loc_wipe(&loc);
+ dict_unref(xattr_req);
- return inode;
+ return inode;
}
int
-afr_selfheal_unlocked_discover_on (call_frame_t *frame, inode_t *inode,
- uuid_t gfid, struct afr_reply *replies,
- unsigned char *discover_on)
+afr_selfheal_unlocked_discover_on(call_frame_t *frame, inode_t *inode,
+ uuid_t gfid, struct afr_reply *replies,
+ unsigned char *discover_on)
{
- loc_t loc = {0, };
- dict_t *xattr_req = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr_req = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = frame->this->private;
+ local = frame->local;
+ priv = frame->this->private;
- xattr_req = dict_new ();
- if (!xattr_req)
- return -ENOMEM;
+ xattr_req = dict_new();
+ if (!xattr_req)
+ return -ENOMEM;
- if (afr_xattr_req_prepare (frame->this, xattr_req) != 0) {
- dict_unref (xattr_req);
- return -ENOMEM;
- }
+ if (afr_xattr_req_prepare(frame->this, xattr_req) != 0) {
+ dict_unref(xattr_req);
+ return -ENOMEM;
+ }
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, gfid);
- AFR_ONLIST (discover_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
- xattr_req);
+ AFR_ONLIST(discover_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+ xattr_req);
- afr_replies_copy (replies, local->replies, priv->child_count);
+ afr_replies_copy(replies, local->replies, priv->child_count);
- loc_wipe (&loc);
- dict_unref (xattr_req);
+ loc_wipe(&loc);
+ dict_unref(xattr_req);
- return 0;
+ return 0;
}
int
-afr_selfheal_unlocked_discover (call_frame_t *frame, inode_t *inode,
- uuid_t gfid, struct afr_reply *replies)
+afr_selfheal_unlocked_discover(call_frame_t *frame, inode_t *inode, uuid_t gfid,
+ struct afr_reply *replies)
{
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
- priv = frame->this->private;
+ priv = frame->this->private;
- return afr_selfheal_unlocked_discover_on (frame, inode, gfid, replies,
- priv->child_up);
+ return afr_selfheal_unlocked_discover_on(frame, inode, gfid, replies,
+ priv->child_up);
}
unsigned int
-afr_success_count (struct afr_reply *replies, unsigned int count)
+afr_success_count(struct afr_reply *replies, unsigned int count)
{
- int i = 0;
- unsigned int success = 0;
+ int i = 0;
+ unsigned int success = 0;
- for (i = 0; i < count; i++)
- if (replies[i].valid && replies[i].op_ret == 0)
- success++;
- return success;
+ for (i = 0; i < count; i++)
+ if (replies[i].valid && replies[i].op_ret == 0)
+ success++;
+ return success;
}
int
-afr_selfheal_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+afr_selfheal_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int i = 0;
+ afr_local_t *local = NULL;
+ int i = 0;
- local = frame->local;
- i = (long) cookie;
+ local = frame->local;
+ i = (long)cookie;
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
- syncbarrier_wake (&local->barrier);
+ syncbarrier_wake(&local->barrier);
- return 0;
+ return 0;
}
-
int
-afr_locked_fill (call_frame_t *frame, xlator_t *this,
- unsigned char *locked_on)
+afr_locked_fill(call_frame_t *frame, xlator_t *this, unsigned char *locked_on)
{
- int i = 0;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int count = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int count = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].valid && local->replies[i].op_ret == 0) {
- locked_on[i] = 1;
- count++;
- } else {
- locked_on[i] = 0;
- }
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].valid && local->replies[i].op_ret == 0) {
+ locked_on[i] = 1;
+ count++;
+ } else {
+ locked_on[i] = 0;
+ }
+ }
- return count;
+ return count;
}
-
int
-afr_selfheal_tryinodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- unsigned char *locked_on)
+afr_selfheal_tryinodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ unsigned char *locked_on)
{
- loc_t loc = {0,};
- struct gf_flock flock = {0, };
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- flock.l_type = F_WRLCK;
- flock.l_start = off;
- flock.l_len = size;
+ flock.l_type = F_WRLCK;
+ flock.l_start = off;
+ flock.l_len = size;
- AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLK, &flock, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLK, &flock,
+ NULL);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
-
int
-afr_selfheal_inodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- unsigned char *locked_on)
+afr_selfheal_inodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ unsigned char *locked_on)
{
- loc_t loc = {0,};
- struct gf_flock flock = {0, };
- afr_local_t *local = NULL;
- int i = 0;
- afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
+ afr_local_t *local = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- flock.l_type = F_WRLCK;
- flock.l_start = off;
- flock.l_len = size;
+ flock.l_type = F_WRLCK;
+ flock.l_start = off;
+ flock.l_len = size;
- AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLK, &flock, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLK, &flock,
+ NULL);
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].op_ret == -1 &&
- local->replies[i].op_errno == EAGAIN) {
- afr_locked_fill (frame, this, locked_on);
- afr_selfheal_uninodelk (frame, this, inode, dom, off,
- size, locked_on);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].op_ret == -1 &&
+ local->replies[i].op_errno == EAGAIN) {
+ afr_locked_fill(frame, this, locked_on);
+ afr_selfheal_uninodelk(frame, this, inode, dom, off, size,
+ locked_on);
- AFR_SEQ (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLKW, &flock, NULL);
- break;
- }
- }
+ AFR_SEQ(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLKW,
+ &flock, NULL);
+ break;
+ }
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
static void
-afr_get_lock_and_eagain_counts (afr_private_t *priv, struct afr_reply *replies,
- int *lock_count, int *eagain_count)
-{
- int i = 0;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
- if (replies[i].op_ret == 0) {
- (*lock_count)++;
- } else if (replies[i].op_ret == -1 &&
- replies[i].op_errno == EAGAIN) {
- (*eagain_count)++;
- }
- }
+afr_get_lock_and_eagain_counts(afr_private_t *priv, struct afr_reply *replies,
+ int *lock_count, int *eagain_count)
+{
+ int i = 0;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (replies[i].op_ret == 0) {
+ (*lock_count)++;
+ } else if (replies[i].op_ret == -1 && replies[i].op_errno == EAGAIN) {
+ (*eagain_count)++;
+ }
+ }
}
/*Do blocking locks if number of locks acquired is majority and there were some
* EAGAINs. Useful for odd-way replication*/
int
-afr_selfheal_tie_breaker_inodelk (call_frame_t *frame, xlator_t *this,
- inode_t *inode, char *dom, off_t off,
- size_t size, unsigned char *locked_on)
+afr_selfheal_tie_breaker_inodelk(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, char *dom, off_t off,
+ size_t size, unsigned char *locked_on)
{
- loc_t loc = {0,};
- struct gf_flock flock = {0, };
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int lock_count = 0;
- int eagain_count = 0;
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int lock_count = 0;
+ int eagain_count = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- flock.l_type = F_WRLCK;
- flock.l_start = off;
- flock.l_len = size;
+ flock.l_type = F_WRLCK;
+ flock.l_start = off;
+ flock.l_len = size;
- AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLK, &flock, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLK, &flock,
+ NULL);
- afr_get_lock_and_eagain_counts (priv, local->replies, &lock_count,
- &eagain_count);
+ afr_get_lock_and_eagain_counts(priv, local->replies, &lock_count,
+ &eagain_count);
- if (lock_count > priv->child_count/2 && eagain_count) {
- afr_locked_fill (frame, this, locked_on);
- afr_selfheal_uninodelk (frame, this, inode, dom, off,
- size, locked_on);
+ if (lock_count > priv->child_count / 2 && eagain_count) {
+ afr_locked_fill(frame, this, locked_on);
+ afr_selfheal_uninodelk(frame, this, inode, dom, off, size, locked_on);
- AFR_SEQ (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLKW, &flock, NULL);
- }
+ AFR_SEQ(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLKW,
+ &flock, NULL);
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
int
-afr_selfheal_uninodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- const unsigned char *locked_on)
+afr_selfheal_uninodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ const unsigned char *locked_on)
{
- loc_t loc = {0,};
- struct gf_flock flock = {0, };
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ flock.l_type = F_UNLCK;
+ flock.l_start = off;
+ flock.l_len = size;
- flock.l_type = F_UNLCK;
- flock.l_start = off;
- flock.l_len = size;
+ AFR_ONLIST(locked_on, frame, afr_selfheal_lock_cbk, inodelk, dom, &loc,
+ F_SETLK, &flock, NULL);
- AFR_ONLIST (locked_on, frame, afr_selfheal_lock_cbk, inodelk,
- dom, &loc, F_SETLK, &flock, NULL);
+ loc_wipe(&loc);
- loc_wipe (&loc);
-
- return 0;
+ return 0;
}
-
int
-afr_selfheal_tryentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on)
+afr_selfheal_tryentrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on)
{
- loc_t loc = {0,};
+ loc_t loc = {
+ 0,
+ };
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONALL (frame, afr_selfheal_lock_cbk, entrylk, dom,
- &loc, name, ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
-
int
-afr_selfheal_entrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on)
+afr_selfheal_entrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on)
{
- loc_t loc = {0,};
- afr_local_t *local = NULL;
- int i = 0;
- afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ afr_local_t *local = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONALL (frame, afr_selfheal_lock_cbk, entrylk, dom, &loc,
- name, ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].op_ret == -1 &&
- local->replies[i].op_errno == EAGAIN) {
- afr_locked_fill (frame, this, locked_on);
- afr_selfheal_unentrylk (frame, this, inode, dom, name,
- locked_on, NULL);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].op_ret == -1 &&
+ local->replies[i].op_errno == EAGAIN) {
+ afr_locked_fill(frame, this, locked_on);
+ afr_selfheal_unentrylk(frame, this, inode, dom, name, locked_on,
+ NULL);
- AFR_SEQ (frame, afr_selfheal_lock_cbk, entrylk, dom,
- &loc, name, ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
- break;
- }
- }
+ AFR_SEQ(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ break;
+ }
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
int
-afr_selfheal_tie_breaker_entrylk (call_frame_t *frame, xlator_t *this,
- inode_t *inode, char *dom, const char *name,
- unsigned char *locked_on)
+afr_selfheal_tie_breaker_entrylk(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, char *dom, const char *name,
+ unsigned char *locked_on)
{
- loc_t loc = {0,};
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int lock_count = 0;
- int eagain_count = 0;
+ loc_t loc = {
+ 0,
+ };
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int lock_count = 0;
+ int eagain_count = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONALL (frame, afr_selfheal_lock_cbk, entrylk, dom, &loc,
- name, ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
- afr_get_lock_and_eagain_counts (priv, local->replies, &lock_count,
- &eagain_count);
+ afr_get_lock_and_eagain_counts(priv, local->replies, &lock_count,
+ &eagain_count);
- if (lock_count > priv->child_count/2 && eagain_count) {
- afr_locked_fill (frame, this, locked_on);
- afr_selfheal_unentrylk (frame, this, inode, dom, name,
- locked_on, NULL);
+ if (lock_count > priv->child_count / 2 && eagain_count) {
+ afr_locked_fill(frame, this, locked_on);
+ afr_selfheal_unentrylk(frame, this, inode, dom, name, locked_on, NULL);
- AFR_SEQ (frame, afr_selfheal_lock_cbk, entrylk, dom,
- &loc, name, ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
- }
+ AFR_SEQ(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
-
int
-afr_selfheal_unentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on,
- dict_t *xdata)
+afr_selfheal_unentrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on,
+ dict_t *xdata)
{
- loc_t loc = {0,};
+ loc_t loc = {
+ 0,
+ };
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONLIST (locked_on, frame, afr_selfheal_lock_cbk, entrylk,
- dom, &loc, name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ AFR_ONLIST(locked_on, frame, afr_selfheal_lock_cbk, entrylk, dom, &loc,
+ name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return 0;
+ return 0;
}
gf_boolean_t
-afr_is_data_set (xlator_t *this, dict_t *xdata)
+afr_is_data_set(xlator_t *this, dict_t *xdata)
{
- return afr_is_pending_set (this, xdata, AFR_DATA_TRANSACTION);
+ return afr_is_pending_set(this, xdata, AFR_DATA_TRANSACTION);
}
gf_boolean_t
-afr_is_metadata_set (xlator_t *this, dict_t *xdata)
+afr_is_metadata_set(xlator_t *this, dict_t *xdata)
{
- return afr_is_pending_set (this, xdata, AFR_METADATA_TRANSACTION);
+ return afr_is_pending_set(this, xdata, AFR_METADATA_TRANSACTION);
}
gf_boolean_t
-afr_is_entry_set (xlator_t *this, dict_t *xdata)
+afr_is_entry_set(xlator_t *this, dict_t *xdata)
{
- return afr_is_pending_set (this, xdata, AFR_ENTRY_TRANSACTION);
+ return afr_is_pending_set(this, xdata, AFR_ENTRY_TRANSACTION);
}
/*
@@ -2210,317 +2190,307 @@ afr_is_entry_set (xlator_t *this, dict_t *xdata)
*/
int
-afr_selfheal_unlocked_inspect (call_frame_t *frame, xlator_t *this,
- uuid_t gfid, inode_t **link_inode,
- gf_boolean_t *data_selfheal,
- gf_boolean_t *metadata_selfheal,
- gf_boolean_t *entry_selfheal)
-{
- afr_private_t *priv = NULL;
- inode_t *inode = NULL;
- int i = 0;
- int valid_cnt = 0;
- struct iatt first = {0, };
- int first_idx = 0;
- struct afr_reply *replies = NULL;
- int ret = -1;
-
- priv = this->private;
-
- inode = afr_inode_find (this, gfid);
- if (!inode)
- goto out;
+afr_selfheal_unlocked_inspect(call_frame_t *frame, xlator_t *this, uuid_t gfid,
+ inode_t **link_inode, gf_boolean_t *data_selfheal,
+ gf_boolean_t *metadata_selfheal,
+ gf_boolean_t *entry_selfheal)
+{
+ afr_private_t *priv = NULL;
+ inode_t *inode = NULL;
+ int i = 0;
+ int valid_cnt = 0;
+ struct iatt first = {
+ 0,
+ };
+ int first_idx = 0;
+ struct afr_reply *replies = NULL;
+ int ret = -1;
+
+ priv = this->private;
+
+ inode = afr_inode_find(this, gfid);
+ if (!inode)
+ goto out;
- replies = alloca0 (sizeof (*replies) * priv->child_count);
+ replies = alloca0(sizeof(*replies) * priv->child_count);
- ret = afr_selfheal_unlocked_discover (frame, inode, gfid, replies);
- if (ret)
- goto out;
+ ret = afr_selfheal_unlocked_discover(frame, inode, gfid, replies);
+ if (ret)
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
- if (replies[i].op_ret == -1)
- continue;
-
- /* The data segment of the changelog can be non-zero to indicate
- * the directory needs a full heal. So the check below ensures
- * it's not a directory before setting the data_selfheal boolean.
- */
- if (data_selfheal && !IA_ISDIR (replies[i].poststat.ia_type) &&
- afr_is_data_set (this, replies[i].xdata))
- *data_selfheal = _gf_true;
-
- if (metadata_selfheal &&
- afr_is_metadata_set (this, replies[i].xdata))
- *metadata_selfheal = _gf_true;
-
- if (entry_selfheal && afr_is_entry_set (this, replies[i].xdata))
- *entry_selfheal = _gf_true;
-
- valid_cnt++;
- if (valid_cnt == 1) {
- first = replies[i].poststat;
- first_idx = i;
- continue;
- }
-
- if (!IA_EQUAL (first, replies[i].poststat, type)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN,
- "TYPE mismatch %d vs %d on %s for gfid:%s",
- (int) first.ia_type,
- (int) replies[i].poststat.ia_type,
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;"
- "type=file;gfid=%s;"
- "ia_type-%d=%s;ia_type-%d=%s",
- this->name,
- uuid_utoa (replies[i].poststat.ia_gfid),
- first_idx,
- gf_inode_type_to_str (first.ia_type), i,
- gf_inode_type_to_str (replies[i].poststat.ia_type));
- ret = -EIO;
- goto out;
- }
-
- if (!IA_EQUAL (first, replies[i].poststat, uid)) {
- gf_msg_debug (this->name, 0,
- "UID mismatch "
- "%d vs %d on %s for gfid:%s",
- (int) first.ia_uid,
- (int) replies[i].poststat.ia_uid,
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
-
- if (metadata_selfheal)
- *metadata_selfheal = _gf_true;
- }
-
- if (!IA_EQUAL (first, replies[i].poststat, gid)) {
- gf_msg_debug (this->name, 0,
- "GID mismatch "
- "%d vs %d on %s for gfid:%s",
- (int) first.ia_uid,
- (int) replies[i].poststat.ia_uid,
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
-
- if (metadata_selfheal)
- *metadata_selfheal = _gf_true;
- }
-
- if (!IA_EQUAL (first, replies[i].poststat, prot)) {
- gf_msg_debug (this->name, 0,
- "MODE mismatch "
- "%d vs %d on %s for gfid:%s",
- (int) st_mode_from_ia (first.ia_prot, 0),
- (int) st_mode_from_ia
- (replies[i].poststat.ia_prot, 0),
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
-
- if (metadata_selfheal)
- *metadata_selfheal = _gf_true;
- }
-
- if (IA_ISREG(first.ia_type) &&
- !IA_EQUAL (first, replies[i].poststat, size)) {
- gf_msg_debug (this->name, 0,
- "SIZE mismatch "
- "%lld vs %lld on %s for gfid:%s",
- (long long) first.ia_size,
- (long long) replies[i].poststat.ia_size,
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
-
- if (data_selfheal)
- *data_selfheal = _gf_true;
- }
- }
-
- if (valid_cnt > 0 && link_inode) {
- *link_inode = inode_link (inode, NULL, NULL, &first);
- if (!*link_inode) {
- ret = -EINVAL;
- goto out;
- }
- } else if (valid_cnt < 2) {
- ret = afr_check_stale_error (replies, priv);
- goto out;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (replies[i].op_ret == -1)
+ continue;
- ret = 0;
+ /* The data segment of the changelog can be non-zero to indicate
+ * the directory needs a full heal. So the check below ensures
+ * it's not a directory before setting the data_selfheal boolean.
+ */
+ if (data_selfheal && !IA_ISDIR(replies[i].poststat.ia_type) &&
+ afr_is_data_set(this, replies[i].xdata))
+ *data_selfheal = _gf_true;
+
+ if (metadata_selfheal && afr_is_metadata_set(this, replies[i].xdata))
+ *metadata_selfheal = _gf_true;
+
+ if (entry_selfheal && afr_is_entry_set(this, replies[i].xdata))
+ *entry_selfheal = _gf_true;
+
+ valid_cnt++;
+ if (valid_cnt == 1) {
+ first = replies[i].poststat;
+ first_idx = i;
+ continue;
+ }
+
+ if (!IA_EQUAL(first, replies[i].poststat, type)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "TYPE mismatch %d vs %d on %s for gfid:%s",
+ (int)first.ia_type, (int)replies[i].poststat.ia_type,
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "subvol=%s;"
+ "type=file;gfid=%s;"
+ "ia_type-%d=%s;ia_type-%d=%s",
+ this->name, uuid_utoa(replies[i].poststat.ia_gfid),
+ first_idx, gf_inode_type_to_str(first.ia_type), i,
+ gf_inode_type_to_str(replies[i].poststat.ia_type));
+ ret = -EIO;
+ goto out;
+ }
+
+ if (!IA_EQUAL(first, replies[i].poststat, uid)) {
+ gf_msg_debug(this->name, 0,
+ "UID mismatch "
+ "%d vs %d on %s for gfid:%s",
+ (int)first.ia_uid, (int)replies[i].poststat.ia_uid,
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+
+ if (metadata_selfheal)
+ *metadata_selfheal = _gf_true;
+ }
+
+ if (!IA_EQUAL(first, replies[i].poststat, gid)) {
+ gf_msg_debug(this->name, 0,
+ "GID mismatch "
+ "%d vs %d on %s for gfid:%s",
+ (int)first.ia_uid, (int)replies[i].poststat.ia_uid,
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+
+ if (metadata_selfheal)
+ *metadata_selfheal = _gf_true;
+ }
+
+ if (!IA_EQUAL(first, replies[i].poststat, prot)) {
+ gf_msg_debug(this->name, 0,
+ "MODE mismatch "
+ "%d vs %d on %s for gfid:%s",
+ (int)st_mode_from_ia(first.ia_prot, 0),
+ (int)st_mode_from_ia(replies[i].poststat.ia_prot, 0),
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+
+ if (metadata_selfheal)
+ *metadata_selfheal = _gf_true;
+ }
+
+ if (IA_ISREG(first.ia_type) &&
+ !IA_EQUAL(first, replies[i].poststat, size)) {
+ gf_msg_debug(this->name, 0,
+ "SIZE mismatch "
+ "%lld vs %lld on %s for gfid:%s",
+ (long long)first.ia_size,
+ (long long)replies[i].poststat.ia_size,
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+
+ if (data_selfheal)
+ *data_selfheal = _gf_true;
+ }
+ }
+
+ if (valid_cnt > 0 && link_inode) {
+ *link_inode = inode_link(inode, NULL, NULL, &first);
+ if (!*link_inode) {
+ ret = -EINVAL;
+ goto out;
+ }
+ } else if (valid_cnt < 2) {
+ ret = afr_check_stale_error(replies, priv);
+ goto out;
+ }
+
+ ret = 0;
out:
- if (inode)
- inode_unref (inode);
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
+ if (inode)
+ inode_unref(inode);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
- return ret;
+ return ret;
}
-
inode_t *
-afr_inode_find (xlator_t *this, uuid_t gfid)
+afr_inode_find(xlator_t *this, uuid_t gfid)
{
- inode_table_t *table = NULL;
- inode_t *inode = NULL;
+ inode_table_t *table = NULL;
+ inode_t *inode = NULL;
- table = this->itable;
- if (!table)
- return NULL;
+ table = this->itable;
+ if (!table)
+ return NULL;
- inode = inode_find (table, gfid);
- if (inode)
- return inode;
+ inode = inode_find(table, gfid);
+ if (inode)
+ return inode;
- inode = inode_new (table);
- if (!inode)
- return NULL;
+ inode = inode_new(table);
+ if (!inode)
+ return NULL;
- gf_uuid_copy (inode->gfid, gfid);
+ gf_uuid_copy(inode->gfid, gfid);
- return inode;
+ return inode;
}
-
call_frame_t *
-afr_frame_create (xlator_t *this, int32_t *op_errno)
+afr_frame_create(xlator_t *this, int32_t *op_errno)
{
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
- pid_t pid = GF_CLIENT_PID_SELF_HEALD;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ pid_t pid = GF_CLIENT_PID_SELF_HEALD;
- frame = create_frame (this, this->ctx->pool);
- if (!frame)
- return NULL;
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame)
+ return NULL;
- local = AFR_FRAME_INIT (frame, (*op_errno));
- if (!local) {
- STACK_DESTROY (frame->root);
- return NULL;
- }
+ local = AFR_FRAME_INIT(frame, (*op_errno));
+ if (!local) {
+ STACK_DESTROY(frame->root);
+ return NULL;
+ }
- syncopctx_setfspid (&pid);
+ syncopctx_setfspid(&pid);
- frame->root->pid = pid;
+ frame->root->pid = pid;
- afr_set_lk_owner (frame, this, frame->root);
+ afr_set_lk_owner(frame, this, frame->root);
- return frame;
+ return frame;
}
int
-afr_selfheal_newentry_mark (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int source, struct afr_reply *replies,
- unsigned char *sources, unsigned char *newentry)
+afr_selfheal_newentry_mark(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int source, struct afr_reply *replies,
+ unsigned char *sources, unsigned char *newentry)
{
- int ret = 0;
- int i = 0;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- int **changelog = NULL;
+ int ret = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int **changelog = NULL;
- priv = this->private;
+ priv = this->private;
- gf_uuid_copy (inode->gfid, replies[source].poststat.ia_gfid);
+ gf_uuid_copy(inode->gfid, replies[source].poststat.ia_gfid);
- xattr = dict_new();
- if (!xattr)
- return -ENOMEM;
+ xattr = dict_new();
+ if (!xattr)
+ return -ENOMEM;
- changelog = afr_mark_pending_changelog (priv, newentry, xattr,
- replies[source].poststat.ia_type);
+ changelog = afr_mark_pending_changelog(priv, newentry, xattr,
+ replies[source].poststat.ia_type);
- if (!changelog) {
- ret = -ENOMEM;
- goto out;
- }
+ if (!changelog) {
+ ret = -ENOMEM;
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- ret |= afr_selfheal_post_op (frame, this, inode, i, xattr,
- NULL);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ ret |= afr_selfheal_post_op(frame, this, inode, i, xattr, NULL);
+ }
out:
- if (changelog)
- afr_matrix_cleanup (changelog, priv->child_count);
- if (xattr)
- dict_unref (xattr);
- return ret;
+ if (changelog)
+ afr_matrix_cleanup(changelog, priv->child_count);
+ if (xattr)
+ dict_unref(xattr);
+ return ret;
}
int
-afr_selfheal_do (call_frame_t *frame, xlator_t *this, uuid_t gfid)
-{
- int ret = -1;
- int entry_ret = 1;
- int metadata_ret = 1;
- int data_ret = 1;
- int or_ret = 0;
- inode_t *inode = NULL;
- fd_t *fd = NULL;
- gf_boolean_t data_selfheal = _gf_false;
- gf_boolean_t metadata_selfheal = _gf_false;
- gf_boolean_t entry_selfheal = _gf_false;
- afr_private_t *priv = NULL;
- gf_boolean_t dataheal_enabled = _gf_false;
-
- priv = this->private;
-
- ret = gf_string2boolean (priv->data_self_heal, &dataheal_enabled);
- if (ret)
- goto out;
+afr_selfheal_do(call_frame_t *frame, xlator_t *this, uuid_t gfid)
+{
+ int ret = -1;
+ int entry_ret = 1;
+ int metadata_ret = 1;
+ int data_ret = 1;
+ int or_ret = 0;
+ inode_t *inode = NULL;
+ fd_t *fd = NULL;
+ gf_boolean_t data_selfheal = _gf_false;
+ gf_boolean_t metadata_selfheal = _gf_false;
+ gf_boolean_t entry_selfheal = _gf_false;
+ afr_private_t *priv = NULL;
+ gf_boolean_t dataheal_enabled = _gf_false;
+
+ priv = this->private;
+
+ ret = gf_string2boolean(priv->data_self_heal, &dataheal_enabled);
+ if (ret)
+ goto out;
- ret = afr_selfheal_unlocked_inspect (frame, this, gfid, &inode,
- &data_selfheal,
- &metadata_selfheal,
- &entry_selfheal);
- if (ret)
- goto out;
+ ret = afr_selfheal_unlocked_inspect(frame, this, gfid, &inode,
+ &data_selfheal, &metadata_selfheal,
+ &entry_selfheal);
+ if (ret)
+ goto out;
- if (!(data_selfheal || metadata_selfheal || entry_selfheal)) {
- ret = 2;
- goto out;
- }
+ if (!(data_selfheal || metadata_selfheal || entry_selfheal)) {
+ ret = 2;
+ goto out;
+ }
- if (inode->ia_type == IA_IFREG) {
- ret = afr_selfheal_data_open (this, inode, &fd);
- if (!fd) {
- ret = -EIO;
- goto out;
- }
+ if (inode->ia_type == IA_IFREG) {
+ ret = afr_selfheal_data_open(this, inode, &fd);
+ if (!fd) {
+ ret = -EIO;
+ goto out;
}
+ }
- if (data_selfheal && dataheal_enabled)
- data_ret = afr_selfheal_data (frame, this, fd);
+ if (data_selfheal && dataheal_enabled)
+ data_ret = afr_selfheal_data(frame, this, fd);
- if (metadata_selfheal && priv->metadata_self_heal)
- metadata_ret = afr_selfheal_metadata (frame, this, inode);
+ if (metadata_selfheal && priv->metadata_self_heal)
+ metadata_ret = afr_selfheal_metadata(frame, this, inode);
- if (entry_selfheal && priv->entry_self_heal)
- entry_ret = afr_selfheal_entry (frame, this, inode);
+ if (entry_selfheal && priv->entry_self_heal)
+ entry_ret = afr_selfheal_entry(frame, this, inode);
- or_ret = (data_ret | metadata_ret | entry_ret);
+ or_ret = (data_ret | metadata_ret | entry_ret);
- if (data_ret == -EIO || metadata_ret == -EIO || entry_ret == -EIO)
- ret = -EIO;
- else if (data_ret == 1 && metadata_ret == 1 && entry_ret == 1)
- ret = 1;
- else if (or_ret < 0)
- ret = or_ret;
- else
- ret = 0;
+ if (data_ret == -EIO || metadata_ret == -EIO || entry_ret == -EIO)
+ ret = -EIO;
+ else if (data_ret == 1 && metadata_ret == 1 && entry_ret == 1)
+ ret = 1;
+ else if (or_ret < 0)
+ ret = or_ret;
+ else
+ ret = 0;
out:
- if (inode)
- inode_unref (inode);
- if (fd)
- fd_unref (fd);
- return ret;
+ if (inode)
+ inode_unref(inode);
+ if (fd)
+ fd_unref(fd);
+ return ret;
}
/*
* This is the entry point for healing a given GFID. The return values for this
@@ -2532,160 +2502,160 @@ out:
*/
int
-afr_selfheal (xlator_t *this, uuid_t gfid)
+afr_selfheal(xlator_t *this, uuid_t gfid)
{
- int ret = -1;
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
+ int ret = -1;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
- frame = afr_frame_create (this, NULL);
- if (!frame)
- return ret;
+ frame = afr_frame_create(this, NULL);
+ if (!frame)
+ return ret;
- local = frame->local;
- local->xdata_req = dict_new();
+ local = frame->local;
+ local->xdata_req = dict_new();
- ret = afr_selfheal_do (frame, this, gfid);
+ ret = afr_selfheal_do(frame, this, gfid);
- if (frame)
- AFR_STACK_DESTROY (frame);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
- return ret;
+ return ret;
}
-afr_local_t*
-__afr_dequeue_heals (afr_private_t *priv)
+afr_local_t *
+__afr_dequeue_heals(afr_private_t *priv)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- if (list_empty (&priv->heal_waiting))
- goto none;
- if ((priv->background_self_heal_count > 0) &&
- (priv->healers >= priv->background_self_heal_count))
- goto none;
+ if (list_empty(&priv->heal_waiting))
+ goto none;
+ if ((priv->background_self_heal_count > 0) &&
+ (priv->healers >= priv->background_self_heal_count))
+ goto none;
- local = list_entry (priv->heal_waiting.next, afr_local_t, healer);
- priv->heal_waiters--;
- GF_ASSERT (priv->heal_waiters >= 0);
- list_del_init(&local->healer);
- list_add(&local->healer, &priv->healing);
- priv->healers++;
- return local;
+ local = list_entry(priv->heal_waiting.next, afr_local_t, healer);
+ priv->heal_waiters--;
+ GF_ASSERT(priv->heal_waiters >= 0);
+ list_del_init(&local->healer);
+ list_add(&local->healer, &priv->healing);
+ priv->healers++;
+ return local;
none:
- gf_msg_debug (THIS->name, 0, "Nothing dequeued. "
- "Num healers: %d, Num Waiters: %d",
- priv->healers, priv->heal_waiters);
- return NULL;
+ gf_msg_debug(THIS->name, 0,
+ "Nothing dequeued. "
+ "Num healers: %d, Num Waiters: %d",
+ priv->healers, priv->heal_waiters);
+ return NULL;
}
int
-afr_refresh_selfheal_wrap (void *opaque)
+afr_refresh_selfheal_wrap(void *opaque)
{
- call_frame_t *heal_frame = opaque;
- afr_local_t *local = heal_frame->local;
- int ret = 0;
+ call_frame_t *heal_frame = opaque;
+ afr_local_t *local = heal_frame->local;
+ int ret = 0;
- ret = afr_selfheal (heal_frame->this, local->refreshinode->gfid);
- return ret;
+ ret = afr_selfheal(heal_frame->this, local->refreshinode->gfid);
+ return ret;
}
int
-afr_refresh_heal_done (int ret, call_frame_t *frame, void *opaque)
-{
- call_frame_t *heal_frame = opaque;
- xlator_t *this = heal_frame->this;
- afr_private_t *priv = this->private;
- afr_local_t *local = heal_frame->local;
-
- LOCK (&priv->lock);
- {
- list_del_init(&local->healer);
- priv->healers--;
- GF_ASSERT (priv->healers >= 0);
- local = __afr_dequeue_heals (priv);
- }
- UNLOCK (&priv->lock);
+afr_refresh_heal_done(int ret, call_frame_t *frame, void *opaque)
+{
+ call_frame_t *heal_frame = opaque;
+ xlator_t *this = heal_frame->this;
+ afr_private_t *priv = this->private;
+ afr_local_t *local = heal_frame->local;
- AFR_STACK_DESTROY (heal_frame);
+ LOCK(&priv->lock);
+ {
+ list_del_init(&local->healer);
+ priv->healers--;
+ GF_ASSERT(priv->healers >= 0);
+ local = __afr_dequeue_heals(priv);
+ }
+ UNLOCK(&priv->lock);
- if (local)
- afr_heal_synctask (this, local);
- return 0;
+ AFR_STACK_DESTROY(heal_frame);
+ if (local)
+ afr_heal_synctask(this, local);
+ return 0;
}
void
-afr_heal_synctask (xlator_t *this, afr_local_t *local)
+afr_heal_synctask(xlator_t *this, afr_local_t *local)
{
- int ret = 0;
- call_frame_t *heal_frame = NULL;
+ int ret = 0;
+ call_frame_t *heal_frame = NULL;
- heal_frame = local->heal_frame;
- ret = synctask_new (this->ctx->env, afr_refresh_selfheal_wrap,
- afr_refresh_heal_done, heal_frame, heal_frame);
- if (ret < 0)
- /* Heal not launched. Will be queued when the next inode
- * refresh happens and shd hasn't healed it yet. */
- afr_refresh_heal_done (ret, heal_frame, heal_frame);
+ heal_frame = local->heal_frame;
+ ret = synctask_new(this->ctx->env, afr_refresh_selfheal_wrap,
+ afr_refresh_heal_done, heal_frame, heal_frame);
+ if (ret < 0)
+ /* Heal not launched. Will be queued when the next inode
+ * refresh happens and shd hasn't healed it yet. */
+ afr_refresh_heal_done(ret, heal_frame, heal_frame);
}
gf_boolean_t
-afr_throttled_selfheal (call_frame_t *frame, xlator_t *this)
-{
- gf_boolean_t can_heal = _gf_true;
- afr_private_t *priv = this->private;
- afr_local_t *local = frame->local;
-
- LOCK (&priv->lock);
- {
- if ((priv->background_self_heal_count > 0) &&
- (priv->heal_wait_qlen + priv->background_self_heal_count) >
- (priv->heal_waiters + priv->healers)) {
- list_add_tail(&local->healer, &priv->heal_waiting);
- priv->heal_waiters++;
- local = __afr_dequeue_heals (priv);
- } else {
- can_heal = _gf_false;
- }
- }
- UNLOCK (&priv->lock);
-
- if (can_heal) {
- if (local)
- afr_heal_synctask (this, local);
- else
- gf_msg_debug (this->name, 0, "Max number of heals are "
- "pending, background self-heal rejected.");
+afr_throttled_selfheal(call_frame_t *frame, xlator_t *this)
+{
+ gf_boolean_t can_heal = _gf_true;
+ afr_private_t *priv = this->private;
+ afr_local_t *local = frame->local;
+
+ LOCK(&priv->lock);
+ {
+ if ((priv->background_self_heal_count > 0) &&
+ (priv->heal_wait_qlen + priv->background_self_heal_count) >
+ (priv->heal_waiters + priv->healers)) {
+ list_add_tail(&local->healer, &priv->heal_waiting);
+ priv->heal_waiters++;
+ local = __afr_dequeue_heals(priv);
+ } else {
+ can_heal = _gf_false;
}
+ }
+ UNLOCK(&priv->lock);
+
+ if (can_heal) {
+ if (local)
+ afr_heal_synctask(this, local);
+ else
+ gf_msg_debug(this->name, 0,
+ "Max number of heals are "
+ "pending, background self-heal rejected.");
+ }
- return can_heal;
+ return can_heal;
}
int
-afr_choose_source_by_policy (afr_private_t *priv, unsigned char *sources,
- afr_transaction_type type)
+afr_choose_source_by_policy(afr_private_t *priv, unsigned char *sources,
+ afr_transaction_type type)
{
- int source = -1;
- int i = 0;
+ int source = -1;
+ int i = 0;
- /* Give preference to local child to save on bandwidth */
- for (i = 0; i < priv->child_count; i++) {
- if (priv->local[i] && sources[i]) {
- if ((type == AFR_DATA_TRANSACTION) &&
- AFR_IS_ARBITER_BRICK (priv, i))
- continue;
+ /* Give preference to local child to save on bandwidth */
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->local[i] && sources[i]) {
+ if ((type == AFR_DATA_TRANSACTION) && AFR_IS_ARBITER_BRICK(priv, i))
+ continue;
- source = i;
- goto out;
- }
+ source = i;
+ goto out;
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i]) {
- source = i;
- goto out;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i]) {
+ source = i;
+ goto out;
}
+ }
out:
- return source;
+ return source;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index c83ef0b7e30..a477fae8039 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
#include "byte-order.h"
@@ -17,596 +16,571 @@
#include "events.h"
enum {
- AFR_SELFHEAL_DATA_FULL = 0,
- AFR_SELFHEAL_DATA_DIFF,
+ AFR_SELFHEAL_DATA_FULL = 0,
+ AFR_SELFHEAL_DATA_DIFF,
};
-
#define HAS_HOLES(i) ((i->ia_blocks * 512) < (i->ia_size))
static int
-__checksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, uint32_t weak, uint8_t *strong,
- dict_t *xdata)
+__checksum_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, uint32_t weak, uint8_t *strong, dict_t *xdata)
{
- afr_local_t *local = NULL;
- struct afr_reply *replies = NULL;
- int i = (long) cookie;
-
- local = frame->local;
- replies = local->replies;
-
- replies[i].valid = 1;
- replies[i].op_ret = op_ret;
- replies[i].op_errno = op_errno;
- if (xdata) {
- replies[i].buf_has_zeroes = dict_get_str_boolean (xdata,
- "buf-has-zeroes", _gf_false);
- replies[i].fips_mode_rchecksum = dict_get_str_boolean (xdata,
- "fips-mode-rchecksum", _gf_false);
- }
- if (strong) {
- if (replies[i].fips_mode_rchecksum) {
- memcpy (local->replies[i].checksum, strong,
- SHA256_DIGEST_LENGTH);
- } else {
- memcpy (local->replies[i].checksum, strong,
- MD5_DIGEST_LENGTH);
- }
+ afr_local_t *local = NULL;
+ struct afr_reply *replies = NULL;
+ int i = (long)cookie;
+
+ local = frame->local;
+ replies = local->replies;
+
+ replies[i].valid = 1;
+ replies[i].op_ret = op_ret;
+ replies[i].op_errno = op_errno;
+ if (xdata) {
+ replies[i].buf_has_zeroes = dict_get_str_boolean(
+ xdata, "buf-has-zeroes", _gf_false);
+ replies[i].fips_mode_rchecksum = dict_get_str_boolean(
+ xdata, "fips-mode-rchecksum", _gf_false);
+ }
+ if (strong) {
+ if (replies[i].fips_mode_rchecksum) {
+ memcpy(local->replies[i].checksum, strong, SHA256_DIGEST_LENGTH);
+ } else {
+ memcpy(local->replies[i].checksum, strong, MD5_DIGEST_LENGTH);
}
+ }
- syncbarrier_wake (&local->barrier);
- return 0;
+ syncbarrier_wake(&local->barrier);
+ return 0;
}
static gf_boolean_t
-__afr_can_skip_data_block_heal (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *healed_sinks,
- off_t offset, size_t size,
- struct iatt *poststat)
+__afr_can_skip_data_block_heal(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int source, unsigned char *healed_sinks,
+ off_t offset, size_t size, struct iatt *poststat)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- unsigned char *wind_subvols = NULL;
- gf_boolean_t checksum_match = _gf_true;
- struct afr_reply *replies = NULL;
- dict_t *xdata = NULL;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
- replies = local->replies;
-
- xdata = dict_new();
- if (!xdata)
- goto out;
- if (dict_set_int32 (xdata, "check-zero-filled", 1)) {
- dict_unref (xdata);
- goto out;
- }
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ unsigned char *wind_subvols = NULL;
+ gf_boolean_t checksum_match = _gf_true;
+ struct afr_reply *replies = NULL;
+ dict_t *xdata = NULL;
+ int i = 0;
+
+ priv = this->private;
+ local = frame->local;
+ replies = local->replies;
+
+ xdata = dict_new();
+ if (!xdata)
+ goto out;
+ if (dict_set_int32(xdata, "check-zero-filled", 1)) {
+ dict_unref(xdata);
+ goto out;
+ }
+
+ wind_subvols = alloca0(priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source || healed_sinks[i])
+ wind_subvols[i] = 1;
+ }
+
+ AFR_ONLIST(wind_subvols, frame, __checksum_cbk, rchecksum, fd, offset, size,
+ xdata);
+ if (xdata)
+ dict_unref(xdata);
+
+ if (!replies[source].valid || replies[source].op_ret != 0)
+ return _gf_false;
- wind_subvols = alloca0 (priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || healed_sinks[i])
- wind_subvols[i] = 1;
- }
-
- AFR_ONLIST (wind_subvols, frame, __checksum_cbk, rchecksum, fd,
- offset, size, xdata);
- if (xdata)
- dict_unref (xdata);
-
- if (!replies[source].valid || replies[source].op_ret != 0)
- return _gf_false;
-
- for (i = 0; i < priv->child_count; i++) {
- if (i == source)
- continue;
- if (replies[i].valid) {
- if (memcmp (replies[source].checksum,
- replies[i].checksum,
- replies[source].fips_mode_rchecksum ?
- SHA256_DIGEST_LENGTH : MD5_DIGEST_LENGTH)) {
- checksum_match = _gf_false;
- break;
- }
- }
- }
-
- if (checksum_match) {
- if (HAS_HOLES (poststat))
- return _gf_true;
-
- /* For non-sparse files, we might be better off writing the
- * zeroes to sinks to avoid mismatch of disk-usage in bricks. */
- if (local->replies[source].buf_has_zeroes)
- return _gf_false;
- else
- return _gf_true;
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source)
+ continue;
+ if (replies[i].valid) {
+ if (memcmp(replies[source].checksum, replies[i].checksum,
+ replies[source].fips_mode_rchecksum
+ ? SHA256_DIGEST_LENGTH
+ : MD5_DIGEST_LENGTH)) {
+ checksum_match = _gf_false;
+ break;
+ }
}
+ }
+
+ if (checksum_match) {
+ if (HAS_HOLES(poststat))
+ return _gf_true;
+
+ /* For non-sparse files, we might be better off writing the
+ * zeroes to sinks to avoid mismatch of disk-usage in bricks. */
+ if (local->replies[source].buf_has_zeroes)
+ return _gf_false;
+ else
+ return _gf_true;
+ }
out:
- return _gf_false;
+ return _gf_false;
}
-
static gf_boolean_t
-__afr_is_sink_zero_filled (xlator_t *this, fd_t *fd, size_t size,
- off_t offset, int sink)
+__afr_is_sink_zero_filled(xlator_t *this, fd_t *fd, size_t size, off_t offset,
+ int sink)
{
- afr_private_t *priv = NULL;
- struct iobref *iobref = NULL;
- struct iovec *iovec = NULL;
- int count = 0;
- int ret = 0;
- gf_boolean_t zero_filled = _gf_false;
-
- priv = this->private;
- ret = syncop_readv (priv->children[sink], fd, size, offset, 0, &iovec,
- &count, &iobref, NULL, NULL, NULL);
- if (ret < 0)
- goto out;
- ret = iov_0filled (iovec, count);
- if (!ret)
- zero_filled = _gf_true;
+ afr_private_t *priv = NULL;
+ struct iobref *iobref = NULL;
+ struct iovec *iovec = NULL;
+ int count = 0;
+ int ret = 0;
+ gf_boolean_t zero_filled = _gf_false;
+
+ priv = this->private;
+ ret = syncop_readv(priv->children[sink], fd, size, offset, 0, &iovec,
+ &count, &iobref, NULL, NULL, NULL);
+ if (ret < 0)
+ goto out;
+ ret = iov_0filled(iovec, count);
+ if (!ret)
+ zero_filled = _gf_true;
out:
- if (iovec)
- GF_FREE (iovec);
- if (iobref)
- iobref_unref (iobref);
- return zero_filled;
+ if (iovec)
+ GF_FREE(iovec);
+ if (iobref)
+ iobref_unref(iobref);
+ return zero_filled;
}
static int
-__afr_selfheal_data_read_write (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *healed_sinks,
- off_t offset, size_t size,
- struct afr_reply *replies, int type)
+__afr_selfheal_data_read_write(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int source, unsigned char *healed_sinks,
+ off_t offset, size_t size,
+ struct afr_reply *replies, int type)
{
- struct iovec *iovec = NULL;
- int count = 0;
- struct iobref *iobref = NULL;
- int ret = 0;
- int i = 0;
- afr_private_t *priv = NULL;
-
- priv = this->private;
-
- ret = syncop_readv (priv->children[source], fd, size, offset, 0,
- &iovec, &count, &iobref, NULL, NULL, NULL);
- if (ret <= 0)
- return ret;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i])
- continue;
-
- /*
- * TODO: Use fiemap() and discard() to heal holes
- * in the future.
- *
- * For now,
- *
- * - if the source had any holes at all,
- * AND
- * - if we are writing past the original file size
- * of the sink
- * AND
- * - is NOT the last block of the source file. if
- * the block contains EOF, it has to be written
- * in order to set the file size even if the
- * last block is 0-filled.
- * AND
- * - if the read buffer is filled with only 0's
- *
- * then, skip writing to this source. We don't depend
- * on the write to happen to update the size as we
- * have performed an ftruncate() upfront anyways.
- */
-#define is_last_block(o,b,s) ((s >= o) && (s <= (o + b)))
- if (HAS_HOLES ((&replies[source].poststat)) &&
- offset >= replies[i].poststat.ia_size &&
- !is_last_block (offset, size,
- replies[source].poststat.ia_size) &&
- (iov_0filled (iovec, count) == 0))
- continue;
-
- /* Avoid filling up sparse regions of the sink with 0-filled
- * writes.*/
- if (type == AFR_SELFHEAL_DATA_FULL &&
- HAS_HOLES ((&replies[source].poststat)) &&
- ((offset + size) <= replies[i].poststat.ia_size) &&
- (iov_0filled (iovec, count) == 0) &&
- __afr_is_sink_zero_filled (this, fd, size, offset, i)) {
- continue;
- }
-
- ret = syncop_writev (priv->children[i], fd, iovec, count,
- offset, iobref, 0, NULL, NULL, NULL, NULL);
- if (ret != iov_length (iovec, count)) {
- /* write() failed on this sink. unset the corresponding
- member in sinks[] (which is healed_sinks[] in the
- caller) so that this server does NOT get considered
- as successfully healed.
- */
- healed_sinks[i] = 0;
- }
- }
- if (iovec)
- GF_FREE (iovec);
- if (iobref)
- iobref_unref (iobref);
-
- return ret;
+ struct iovec *iovec = NULL;
+ int count = 0;
+ struct iobref *iobref = NULL;
+ int ret = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ ret = syncop_readv(priv->children[source], fd, size, offset, 0, &iovec,
+ &count, &iobref, NULL, NULL, NULL);
+ if (ret <= 0)
+ return ret;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i])
+ continue;
+
+ /*
+ * TODO: Use fiemap() and discard() to heal holes
+ * in the future.
+ *
+ * For now,
+ *
+ * - if the source had any holes at all,
+ * AND
+ * - if we are writing past the original file size
+ * of the sink
+ * AND
+ * - is NOT the last block of the source file. if
+ * the block contains EOF, it has to be written
+ * in order to set the file size even if the
+ * last block is 0-filled.
+ * AND
+ * - if the read buffer is filled with only 0's
+ *
+ * then, skip writing to this source. We don't depend
+ * on the write to happen to update the size as we
+ * have performed an ftruncate() upfront anyways.
+ */
+#define is_last_block(o, b, s) ((s >= o) && (s <= (o + b)))
+ if (HAS_HOLES((&replies[source].poststat)) &&
+ offset >= replies[i].poststat.ia_size &&
+ !is_last_block(offset, size, replies[source].poststat.ia_size) &&
+ (iov_0filled(iovec, count) == 0))
+ continue;
+
+ /* Avoid filling up sparse regions of the sink with 0-filled
+ * writes.*/
+ if (type == AFR_SELFHEAL_DATA_FULL &&
+ HAS_HOLES((&replies[source].poststat)) &&
+ ((offset + size) <= replies[i].poststat.ia_size) &&
+ (iov_0filled(iovec, count) == 0) &&
+ __afr_is_sink_zero_filled(this, fd, size, offset, i)) {
+ continue;
+ }
+
+ ret = syncop_writev(priv->children[i], fd, iovec, count, offset, iobref,
+ 0, NULL, NULL, NULL, NULL);
+ if (ret != iov_length(iovec, count)) {
+ /* write() failed on this sink. unset the corresponding
+ member in sinks[] (which is healed_sinks[] in the
+ caller) so that this server does NOT get considered
+ as successfully healed.
+ */
+ healed_sinks[i] = 0;
+ }
+ }
+ if (iovec)
+ GF_FREE(iovec);
+ if (iobref)
+ iobref_unref(iobref);
+
+ return ret;
}
static int
-afr_selfheal_data_block (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *healed_sinks, off_t offset,
- size_t size, int type, struct afr_reply *replies)
+afr_selfheal_data_block(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int source, unsigned char *healed_sinks, off_t offset,
+ size_t size, int type, struct afr_reply *replies)
{
- int ret = -1;
- int sink_count = 0;
- afr_private_t *priv = NULL;
- unsigned char *data_lock = NULL;
-
- priv = this->private;
- sink_count = AFR_COUNT (healed_sinks, priv->child_count);
- data_lock = alloca0 (priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name,
- offset, size, data_lock);
- {
- if (ret < sink_count) {
- ret = -ENOTCONN;
- goto unlock;
- }
-
- if (type == AFR_SELFHEAL_DATA_DIFF &&
- __afr_can_skip_data_block_heal (frame, this, fd, source,
- healed_sinks, offset, size,
- &replies[source].poststat)) {
- ret = 0;
- goto unlock;
- }
-
- ret = __afr_selfheal_data_read_write (frame, this, fd, source,
- healed_sinks, offset, size,
- replies, type);
- }
-unlock:
- afr_selfheal_uninodelk (frame, this, fd->inode, this->name,
- offset, size, data_lock);
- return ret;
-}
+ int ret = -1;
+ int sink_count = 0;
+ afr_private_t *priv = NULL;
+ unsigned char *data_lock = NULL;
+
+ priv = this->private;
+ sink_count = AFR_COUNT(healed_sinks, priv->child_count);
+ data_lock = alloca0(priv->child_count);
+
+ ret = afr_selfheal_inodelk(frame, this, fd->inode, this->name, offset, size,
+ data_lock);
+ {
+ if (ret < sink_count) {
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+ if (type == AFR_SELFHEAL_DATA_DIFF &&
+ __afr_can_skip_data_block_heal(frame, this, fd, source,
+ healed_sinks, offset, size,
+ &replies[source].poststat)) {
+ ret = 0;
+ goto unlock;
+ }
+ ret = __afr_selfheal_data_read_write(
+ frame, this, fd, source, healed_sinks, offset, size, replies, type);
+ }
+unlock:
+ afr_selfheal_uninodelk(frame, this, fd->inode, this->name, offset, size,
+ data_lock);
+ return ret;
+}
static int
-afr_selfheal_data_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
- unsigned char *healed_sinks)
+afr_selfheal_data_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ unsigned char *healed_sinks)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (!priv->ensure_durability)
- return 0;
-
- AFR_ONLIST (healed_sinks, frame, afr_sh_generic_fop_cbk, fsync, fd, 0,
- NULL);
-
- for (i = 0; i < priv->child_count; i++)
- if (healed_sinks[i] && local->replies[i].op_ret != 0)
- /* fsync() failed. Do NOT consider this server
- as successfully healed. Mark it so.
- */
- healed_sinks[i] = 0;
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (!priv->ensure_durability)
+ return 0;
+
+ AFR_ONLIST(healed_sinks, frame, afr_sh_generic_fop_cbk, fsync, fd, 0, NULL);
+
+ for (i = 0; i < priv->child_count; i++)
+ if (healed_sinks[i] && local->replies[i].op_ret != 0)
+ /* fsync() failed. Do NOT consider this server
+ as successfully healed. Mark it so.
+ */
+ healed_sinks[i] = 0;
+ return 0;
}
static int
-afr_data_self_heal_type_get (afr_private_t *priv, unsigned char *healed_sinks,
- int source, struct afr_reply *replies)
+afr_data_self_heal_type_get(afr_private_t *priv, unsigned char *healed_sinks,
+ int source, struct afr_reply *replies)
{
- int type = AFR_SELFHEAL_DATA_FULL;
- int i = 0;
-
- if (priv->data_self_heal_algorithm == NULL) {
- type = AFR_SELFHEAL_DATA_FULL;
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i] && i != source)
- continue;
- if (replies[i].poststat.ia_size) {
- type = AFR_SELFHEAL_DATA_DIFF;
- break;
- }
- }
- } else if (strcmp (priv->data_self_heal_algorithm, "full") == 0) {
- type = AFR_SELFHEAL_DATA_FULL;
- } else if (strcmp (priv->data_self_heal_algorithm, "diff") == 0) {
+ int type = AFR_SELFHEAL_DATA_FULL;
+ int i = 0;
+
+ if (priv->data_self_heal_algorithm == NULL) {
+ type = AFR_SELFHEAL_DATA_FULL;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i] && i != source)
+ continue;
+ if (replies[i].poststat.ia_size) {
type = AFR_SELFHEAL_DATA_DIFF;
+ break;
+ }
}
- return type;
+ } else if (strcmp(priv->data_self_heal_algorithm, "full") == 0) {
+ type = AFR_SELFHEAL_DATA_FULL;
+ } else if (strcmp(priv->data_self_heal_algorithm, "diff") == 0) {
+ type = AFR_SELFHEAL_DATA_DIFF;
+ }
+ return type;
}
static int
-afr_selfheal_data_do (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *healed_sinks,
- struct afr_reply *replies)
+afr_selfheal_data_do(call_frame_t *frame, xlator_t *this, fd_t *fd, int source,
+ unsigned char *healed_sinks, struct afr_reply *replies)
{
- afr_private_t *priv = NULL;
- off_t off = 0;
- size_t block = 0;
- int type = AFR_SELFHEAL_DATA_FULL;
- int ret = -1;
- call_frame_t *iter_frame = NULL;
- unsigned char arbiter_sink_status = 0;
-
- priv = this->private;
- if (priv->arbiter_count) {
- arbiter_sink_status = healed_sinks[ARBITER_BRICK_INDEX];
- healed_sinks[ARBITER_BRICK_INDEX] = 0;
+ afr_private_t *priv = NULL;
+ off_t off = 0;
+ size_t block = 0;
+ int type = AFR_SELFHEAL_DATA_FULL;
+ int ret = -1;
+ call_frame_t *iter_frame = NULL;
+ unsigned char arbiter_sink_status = 0;
+
+ priv = this->private;
+ if (priv->arbiter_count) {
+ arbiter_sink_status = healed_sinks[ARBITER_BRICK_INDEX];
+ healed_sinks[ARBITER_BRICK_INDEX] = 0;
+ }
+
+ block = 128 * 1024 * priv->data_self_heal_window_size;
+
+ type = afr_data_self_heal_type_get(priv, healed_sinks, source, replies);
+
+ iter_frame = afr_copy_frame(frame);
+ if (!iter_frame) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ for (off = 0; off < replies[source].poststat.ia_size; off += block) {
+ if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
+ ret = -ENOTCONN;
+ goto out;
}
- block = 128 * 1024 * priv->data_self_heal_window_size;
-
- type = afr_data_self_heal_type_get (priv, healed_sinks, source,
- replies);
+ ret = afr_selfheal_data_block(iter_frame, this, fd, source,
+ healed_sinks, off, block, type, replies);
+ if (ret < 0)
+ goto out;
- iter_frame = afr_copy_frame (frame);
- if (!iter_frame) {
- ret = -ENOMEM;
- goto out;
+ AFR_STACK_RESET(iter_frame);
+ if (iter_frame->local == NULL) {
+ ret = -ENOTCONN;
+ goto out;
}
+ }
- for (off = 0; off < replies[source].poststat.ia_size; off += block) {
- if (AFR_COUNT (healed_sinks, priv->child_count) == 0) {
- ret = -ENOTCONN;
- goto out;
- }
-
- ret = afr_selfheal_data_block (iter_frame, this, fd, source,
- healed_sinks, off, block, type,
- replies);
- if (ret < 0)
- goto out;
-
- AFR_STACK_RESET (iter_frame);
- if (iter_frame->local == NULL) {
- ret = -ENOTCONN;
- goto out;
- }
- }
-
- ret = afr_selfheal_data_fsync (frame, this, fd, healed_sinks);
+ ret = afr_selfheal_data_fsync(frame, this, fd, healed_sinks);
out:
- if (arbiter_sink_status)
- healed_sinks[ARBITER_BRICK_INDEX] = arbiter_sink_status;
+ if (arbiter_sink_status)
+ healed_sinks[ARBITER_BRICK_INDEX] = arbiter_sink_status;
- if (iter_frame)
- AFR_STACK_DESTROY (iter_frame);
- return ret;
+ if (iter_frame)
+ AFR_STACK_DESTROY(iter_frame);
+ return ret;
}
-
static int
-__afr_selfheal_truncate_sinks (call_frame_t *frame, xlator_t *this,
- fd_t *fd, unsigned char *healed_sinks,
- uint64_t size)
+__afr_selfheal_truncate_sinks(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ unsigned char *healed_sinks, uint64_t size)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- unsigned char arbiter_sink_status = 0;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (priv->arbiter_count) {
- arbiter_sink_status = healed_sinks[ARBITER_BRICK_INDEX];
- healed_sinks[ARBITER_BRICK_INDEX] = 0;
- }
-
- AFR_ONLIST (healed_sinks, frame, afr_sh_generic_fop_cbk, ftruncate, fd,
- size, NULL);
-
- for (i = 0; i < priv->child_count; i++)
- if (healed_sinks[i] && local->replies[i].op_ret == -1)
- /* truncate() failed. Do NOT consider this server
- as successfully healed. Mark it so.
- */
- healed_sinks[i] = 0;
-
- if (arbiter_sink_status)
- healed_sinks[ARBITER_BRICK_INDEX] = arbiter_sink_status;
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ unsigned char arbiter_sink_status = 0;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (priv->arbiter_count) {
+ arbiter_sink_status = healed_sinks[ARBITER_BRICK_INDEX];
+ healed_sinks[ARBITER_BRICK_INDEX] = 0;
+ }
+
+ AFR_ONLIST(healed_sinks, frame, afr_sh_generic_fop_cbk, ftruncate, fd, size,
+ NULL);
+
+ for (i = 0; i < priv->child_count; i++)
+ if (healed_sinks[i] && local->replies[i].op_ret == -1)
+ /* truncate() failed. Do NOT consider this server
+ as successfully healed. Mark it so.
+ */
+ healed_sinks[i] = 0;
+
+ if (arbiter_sink_status)
+ healed_sinks[ARBITER_BRICK_INDEX] = arbiter_sink_status;
+ return 0;
}
gf_boolean_t
-afr_has_source_witnesses (xlator_t *this, unsigned char *sources,
- uint64_t *witness)
+afr_has_source_witnesses(xlator_t *this, unsigned char *sources,
+ uint64_t *witness)
{
- int i = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i] && witness[i])
- return _gf_true;
- }
- return _gf_false;
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i] && witness[i])
+ return _gf_true;
+ }
+ return _gf_false;
}
static gf_boolean_t
-afr_does_size_mismatch (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies)
+afr_does_size_mismatch(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies)
{
- int i = 0;
- afr_private_t *priv = NULL;
- struct iatt *min = NULL;
- struct iatt *max = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ struct iatt *min = NULL;
+ struct iatt *max = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- if (replies[i].op_ret < 0)
- continue;
+ if (replies[i].op_ret < 0)
+ continue;
- if (!sources[i])
- continue;
+ if (!sources[i])
+ continue;
- if (AFR_IS_ARBITER_BRICK (priv, i) &&
- (replies[i].poststat.ia_size == 0))
- continue;
+ if (AFR_IS_ARBITER_BRICK(priv, i) && (replies[i].poststat.ia_size == 0))
+ continue;
- if (!min)
- min = &replies[i].poststat;
+ if (!min)
+ min = &replies[i].poststat;
- if (!max)
- max = &replies[i].poststat;
+ if (!max)
+ max = &replies[i].poststat;
- if (min->ia_size > replies[i].poststat.ia_size)
- min = &replies[i].poststat;
+ if (min->ia_size > replies[i].poststat.ia_size)
+ min = &replies[i].poststat;
- if (max->ia_size < replies[i].poststat.ia_size)
- max = &replies[i].poststat;
- }
+ if (max->ia_size < replies[i].poststat.ia_size)
+ max = &replies[i].poststat;
+ }
- if (min && max) {
- if (min->ia_size != max->ia_size)
- return _gf_true;
- }
+ if (min && max) {
+ if (min->ia_size != max->ia_size)
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static void
-afr_mark_biggest_witness_as_source (xlator_t *this, unsigned char *sources,
- uint64_t *witness)
+afr_mark_biggest_witness_as_source(xlator_t *this, unsigned char *sources,
+ uint64_t *witness)
{
- int i = 0;
- afr_private_t *priv = NULL;
- uint64_t biggest_witness = 0;
-
- priv = this->private;
- /* Find source with biggest witness count */
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (biggest_witness < witness[i])
- biggest_witness = witness[i];
- }
-
- /* Mark files with less witness count as not source */
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (witness[i] < biggest_witness)
- sources[i] = 0;
- }
-
- return;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ uint64_t biggest_witness = 0;
+
+ priv = this->private;
+ /* Find source with biggest witness count */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (biggest_witness < witness[i])
+ biggest_witness = witness[i];
+ }
+
+ /* Mark files with less witness count as not source */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (witness[i] < biggest_witness)
+ sources[i] = 0;
+ }
+
+ return;
}
/* This is a tie breaker function. Only one source be assigned here */
static void
-afr_mark_newest_file_as_source (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies)
+afr_mark_newest_file_as_source(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies)
{
- int i = 0;
- afr_private_t *priv = NULL;
- int source = -1;
- uint32_t max_ctime = 0;
-
- priv = this->private;
- /* Find source with latest ctime */
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
-
- if (max_ctime <= replies[i].poststat.ia_ctime) {
- source = i;
- max_ctime = replies[i].poststat.ia_ctime;
- }
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int source = -1;
+ uint32_t max_ctime = 0;
+
+ priv = this->private;
+ /* Find source with latest ctime */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+
+ if (max_ctime <= replies[i].poststat.ia_ctime) {
+ source = i;
+ max_ctime = replies[i].poststat.ia_ctime;
}
+ }
- /* Only mark one of the files as source to break ties */
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- sources[source] = 1;
+ /* Only mark one of the files as source to break ties */
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ sources[source] = 1;
}
static int
-__afr_selfheal_data_finalize_source (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- unsigned char *undid_pending,
- struct afr_reply *replies,
- uint64_t *witness)
+__afr_selfheal_data_finalize_source(
+ call_frame_t *frame, xlator_t *this, inode_t *inode, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ unsigned char *undid_pending, struct afr_reply *replies, uint64_t *witness)
{
- afr_private_t *priv = NULL;
- int source = -1;
- int sources_count = 0;
- priv = this->private;
-
- sources_count = AFR_COUNT (sources, priv->child_count);
-
- if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count) {
- /* split brain */
- source = afr_mark_split_brain_source_sinks (frame, this, inode,
- sources, sinks,
- healed_sinks,
- locked_on, replies,
- AFR_DATA_TRANSACTION);
- if (source < 0) {
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;type=data;"
- "file=%s", this->name, uuid_utoa(inode->gfid));
- return -EIO;
- }
-
- _afr_fav_child_reset_sink_xattrs (frame, this, inode, source,
- healed_sinks, undid_pending,
- AFR_DATA_TRANSACTION,
- locked_on, replies);
- goto out;
- }
-
- /* No split brain at this point. If we were called from
- * afr_heal_splitbrain_file(), abort.*/
- if (afr_dict_contains_heal_op(frame))
- return -EIO;
-
- /* If there are no witnesses/size-mismatches on sources we are done*/
- if (!afr_does_size_mismatch (this, sources, replies) &&
- !afr_has_source_witnesses (this, sources, witness))
- goto out;
-
- afr_mark_largest_file_as_source (this, sources, replies);
- afr_mark_biggest_witness_as_source (this, sources, witness);
- afr_mark_newest_file_as_source (this, sources, replies);
- if (priv->arbiter_count)
- /* Choose non-arbiter brick as source for empty files. */
- afr_mark_source_sinks_if_file_empty (this, sources, sinks,
- healed_sinks, locked_on,
- replies,
- AFR_DATA_TRANSACTION);
+ afr_private_t *priv = NULL;
+ int source = -1;
+ int sources_count = 0;
+ priv = this->private;
+
+ sources_count = AFR_COUNT(sources, priv->child_count);
+
+ if ((AFR_CMP(locked_on, healed_sinks, priv->child_count) == 0) ||
+ !sources_count) {
+ /* split brain */
+ source = afr_mark_split_brain_source_sinks(
+ frame, this, inode, sources, sinks, healed_sinks, locked_on,
+ replies, AFR_DATA_TRANSACTION);
+ if (source < 0) {
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "subvol=%s;type=data;"
+ "file=%s",
+ this->name, uuid_utoa(inode->gfid));
+ return -EIO;
+ }
+
+ _afr_fav_child_reset_sink_xattrs(
+ frame, this, inode, source, healed_sinks, undid_pending,
+ AFR_DATA_TRANSACTION, locked_on, replies);
+ goto out;
+ }
+
+ /* No split brain at this point. If we were called from
+ * afr_heal_splitbrain_file(), abort.*/
+ if (afr_dict_contains_heal_op(frame))
+ return -EIO;
+
+ /* If there are no witnesses/size-mismatches on sources we are done*/
+ if (!afr_does_size_mismatch(this, sources, replies) &&
+ !afr_has_source_witnesses(this, sources, witness))
+ goto out;
+
+ afr_mark_largest_file_as_source(this, sources, replies);
+ afr_mark_biggest_witness_as_source(this, sources, witness);
+ afr_mark_newest_file_as_source(this, sources, replies);
+ if (priv->arbiter_count)
+ /* Choose non-arbiter brick as source for empty files. */
+ afr_mark_source_sinks_if_file_empty(this, sources, sinks, healed_sinks,
+ locked_on, replies,
+ AFR_DATA_TRANSACTION);
out:
- afr_mark_active_sinks (this, sources, locked_on, healed_sinks);
- source = afr_choose_source_by_policy (priv, sources,
- AFR_DATA_TRANSACTION);
+ afr_mark_active_sinks(this, sources, locked_on, healed_sinks);
+ source = afr_choose_source_by_policy(priv, sources, AFR_DATA_TRANSACTION);
- return source;
+ return source;
}
/*
@@ -619,296 +593,287 @@ out:
* for self-healing, or -1 if no healing is necessary/split brain.
*/
int
-__afr_selfheal_data_prepare (call_frame_t *frame, xlator_t *this,
- inode_t *inode, unsigned char *locked_on,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- struct afr_reply *replies, gf_boolean_t *pflag)
+__afr_selfheal_data_prepare(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ unsigned char *locked_on, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ struct afr_reply *replies, gf_boolean_t *pflag)
{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- uint64_t *witness = NULL;
-
- priv = this->private;
-
- ret = afr_selfheal_unlocked_discover (frame, inode, inode->gfid,
- replies);
-
- if (ret)
- return ret;
-
- witness = alloca0(priv->child_count * sizeof (*witness));
- ret = afr_selfheal_find_direction (frame, this, replies,
- AFR_DATA_TRANSACTION,
- locked_on, sources, sinks, witness,
- pflag);
- if (ret)
- return ret;
-
- /* Initialize the healed_sinks[] array optimistically to
- the intersection of to-be-healed (i.e sinks[]) and
- the list of servers which are up (i.e locked_on[]).
- As we encounter failures in the healing process, we
- will unmark the respective servers in the healed_sinks[]
- array.
- */
- AFR_INTERSECT (healed_sinks, sinks, locked_on, priv->child_count);
-
- source = __afr_selfheal_data_finalize_source (frame, this, inode,
- sources, sinks,
- healed_sinks,
- locked_on, undid_pending,
- replies, witness);
- if (source < 0)
- return -EIO;
-
- return source;
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ uint64_t *witness = NULL;
+
+ priv = this->private;
+
+ ret = afr_selfheal_unlocked_discover(frame, inode, inode->gfid, replies);
+
+ if (ret)
+ return ret;
+
+ witness = alloca0(priv->child_count * sizeof(*witness));
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_DATA_TRANSACTION, locked_on, sources,
+ sinks, witness, pflag);
+ if (ret)
+ return ret;
+
+ /* Initialize the healed_sinks[] array optimistically to
+ the intersection of to-be-healed (i.e sinks[]) and
+ the list of servers which are up (i.e locked_on[]).
+ As we encounter failures in the healing process, we
+ will unmark the respective servers in the healed_sinks[]
+ array.
+ */
+ AFR_INTERSECT(healed_sinks, sinks, locked_on, priv->child_count);
+
+ source = __afr_selfheal_data_finalize_source(
+ frame, this, inode, sources, sinks, healed_sinks, locked_on,
+ undid_pending, replies, witness);
+ if (source < 0)
+ return -EIO;
+
+ return source;
}
-
static int
-__afr_selfheal_data (call_frame_t *frame, xlator_t *this, fd_t *fd,
- unsigned char *locked_on)
+__afr_selfheal_data(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ unsigned char *locked_on)
{
- afr_private_t *priv = NULL;
- int ret = -1;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *data_lock = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
- int source = -1;
- gf_boolean_t did_sh = _gf_true;
- gf_boolean_t is_arbiter_the_only_sink = _gf_false;
- gf_boolean_t empty_file = _gf_false;
-
- priv = this->private;
-
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- data_lock = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name, 0, 0,
- data_lock);
- {
- if (ret < priv->child_count) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "self-heal as only %d number "
- "of subvolumes "
- "could be locked",
- uuid_utoa (fd->inode->gfid),
- ret);
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_data_prepare (frame, this, fd->inode,
- data_lock, sources, sinks,
- healed_sinks, undid_pending,
- locked_replies, NULL);
- if (ret < 0)
- goto unlock;
-
- if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
- did_sh = _gf_false;
- goto unlock;
- }
-
- source = ret;
-
- if (AFR_IS_ARBITER_BRICK(priv, source)) {
- empty_file = afr_is_file_empty_on_all_children (priv,
- locked_replies);
- if (empty_file)
- goto restore_time;
-
- did_sh = _gf_false;
- goto unlock;
- }
-
- if (priv->arbiter_count &&
- AFR_COUNT (healed_sinks, priv->child_count) == 1 &&
- healed_sinks[ARBITER_BRICK_INDEX]) {
- is_arbiter_the_only_sink = _gf_true;
- goto restore_time;
- }
-
- ret = __afr_selfheal_truncate_sinks (frame, this, fd, healed_sinks,
- locked_replies[source].poststat.ia_size);
- if (ret < 0)
- goto unlock;
-
- ret = 0;
-
- }
-unlock:
- afr_selfheal_uninodelk (frame, this, fd->inode, this->name, 0, 0,
- data_lock);
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *data_lock = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+ int source = -1;
+ gf_boolean_t did_sh = _gf_true;
+ gf_boolean_t is_arbiter_the_only_sink = _gf_false;
+ gf_boolean_t empty_file = _gf_false;
+
+ priv = this->private;
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ data_lock = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = afr_selfheal_inodelk(frame, this, fd->inode, this->name, 0, 0,
+ data_lock);
+ {
+ if (ret < priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "self-heal as only %d number "
+ "of subvolumes "
+ "could be locked",
+ uuid_utoa(fd->inode->gfid), ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_data_prepare(frame, this, fd->inode, data_lock,
+ sources, sinks, healed_sinks,
+ undid_pending, locked_replies, NULL);
if (ret < 0)
- goto out;
+ goto unlock;
- if (!did_sh)
- goto out;
+ if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
+ did_sh = _gf_false;
+ goto unlock;
+ }
- ret = afr_selfheal_data_do (frame, this, fd, source, healed_sinks,
- locked_replies);
- if (ret)
- goto out;
+ source = ret;
+
+ if (AFR_IS_ARBITER_BRICK(priv, source)) {
+ empty_file = afr_is_file_empty_on_all_children(priv,
+ locked_replies);
+ if (empty_file)
+ goto restore_time;
+
+ did_sh = _gf_false;
+ goto unlock;
+ }
+
+ if (priv->arbiter_count &&
+ AFR_COUNT(healed_sinks, priv->child_count) == 1 &&
+ healed_sinks[ARBITER_BRICK_INDEX]) {
+ is_arbiter_the_only_sink = _gf_true;
+ goto restore_time;
+ }
+
+ ret = __afr_selfheal_truncate_sinks(
+ frame, this, fd, healed_sinks,
+ locked_replies[source].poststat.ia_size);
+ if (ret < 0)
+ goto unlock;
+
+ ret = 0;
+ }
+unlock:
+ afr_selfheal_uninodelk(frame, this, fd->inode, this->name, 0, 0, data_lock);
+ if (ret < 0)
+ goto out;
+
+ if (!did_sh)
+ goto out;
+
+ ret = afr_selfheal_data_do(frame, this, fd, source, healed_sinks,
+ locked_replies);
+ if (ret)
+ goto out;
restore_time:
- afr_selfheal_restore_time (frame, this, fd->inode, source,
- healed_sinks, locked_replies);
-
- if (!is_arbiter_the_only_sink || !empty_file) {
- ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name,
- 0, 0, data_lock);
- if (ret < priv->child_count) {
- ret = -ENOTCONN;
- did_sh = _gf_false;
- goto skip_undo_pending;
- }
+ afr_selfheal_restore_time(frame, this, fd->inode, source, healed_sinks,
+ locked_replies);
+
+ if (!is_arbiter_the_only_sink || !empty_file) {
+ ret = afr_selfheal_inodelk(frame, this, fd->inode, this->name, 0, 0,
+ data_lock);
+ if (ret < priv->child_count) {
+ ret = -ENOTCONN;
+ did_sh = _gf_false;
+ goto skip_undo_pending;
}
- ret = afr_selfheal_undo_pending (frame, this, fd->inode,
- sources, sinks, healed_sinks,
- undid_pending, AFR_DATA_TRANSACTION,
- locked_replies, data_lock);
+ }
+ ret = afr_selfheal_undo_pending(
+ frame, this, fd->inode, sources, sinks, healed_sinks, undid_pending,
+ AFR_DATA_TRANSACTION, locked_replies, data_lock);
skip_undo_pending:
- afr_selfheal_uninodelk (frame, this, fd->inode, this->name, 0, 0,
- data_lock);
+ afr_selfheal_uninodelk(frame, this, fd->inode, this->name, 0, 0, data_lock);
out:
- if (did_sh)
- afr_log_selfheal (fd->inode->gfid, this, ret, "data", source,
- sources, healed_sinks);
- else
- ret = 1;
+ if (did_sh)
+ afr_log_selfheal(fd->inode->gfid, this, ret, "data", source, sources,
+ healed_sinks);
+ else
+ ret = 1;
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
- return ret;
+ return ret;
}
int
-afr_selfheal_data_open_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd, dict_t *xdata)
+afr_selfheal_data_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int i = (long) cookie;
+ afr_local_t *local = NULL;
+ int i = (long)cookie;
- local = frame->local;
+ local = frame->local;
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
- syncbarrier_wake (&local->barrier);
+ syncbarrier_wake(&local->barrier);
- return 0;
+ return 0;
}
int
-afr_selfheal_data_open (xlator_t *this, inode_t *inode, fd_t **fd)
+afr_selfheal_data_open(xlator_t *this, inode_t *inode, fd_t **fd)
{
- int ret = 0;
- fd_t *fd_tmp = NULL;
- loc_t loc = {0,};
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- priv = this->private;
-
- fd_tmp = fd_create (inode, 0);
- if (!fd_tmp)
- return -ENOMEM;
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- frame = afr_frame_create (this, &ret);
- if (!frame) {
- ret = -ret;
- fd_unref (fd_tmp);
- goto out;
+ int ret = 0;
+ fd_t *fd_tmp = NULL;
+ loc_t loc = {
+ 0,
+ };
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ priv = this->private;
+
+ fd_tmp = fd_create(inode, 0);
+ if (!fd_tmp)
+ return -ENOMEM;
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ frame = afr_frame_create(this, &ret);
+ if (!frame) {
+ ret = -ret;
+ fd_unref(fd_tmp);
+ goto out;
+ }
+ local = frame->local;
+
+ AFR_ONLIST(local->child_up, frame, afr_selfheal_data_open_cbk, open, &loc,
+ O_RDWR | O_LARGEFILE, fd_tmp, NULL);
+
+ ret = -ENOTCONN;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+
+ if (local->replies[i].op_ret < 0) {
+ ret = -local->replies[i].op_errno;
+ continue;
}
- local = frame->local;
-
- AFR_ONLIST (local->child_up, frame, afr_selfheal_data_open_cbk, open,
- &loc, O_RDWR|O_LARGEFILE, fd_tmp, NULL);
-
- ret = -ENOTCONN;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret < 0) {
- ret = -local->replies[i].op_errno;
- continue;
- }
+ ret = 0;
+ break;
+ }
- ret = 0;
- break;
- }
-
- if (ret < 0) {
- fd_unref (fd_tmp);
- goto out;
- } else {
- fd_bind (fd_tmp);
- }
+ if (ret < 0) {
+ fd_unref(fd_tmp);
+ goto out;
+ } else {
+ fd_bind(fd_tmp);
+ }
- *fd = fd_tmp;
+ *fd = fd_tmp;
out:
- loc_wipe (&loc);
- if (frame)
- AFR_STACK_DESTROY (frame);
- return ret;
+ loc_wipe(&loc);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ return ret;
}
int
-afr_selfheal_data (call_frame_t *frame, xlator_t *this, fd_t *fd)
+afr_selfheal_data(call_frame_t *frame, xlator_t *this, fd_t *fd)
{
- afr_private_t *priv = NULL;
- unsigned char *locked_on = NULL;
- int ret = 0;
- inode_t *inode = fd->inode;
-
- priv = this->private;
-
- locked_on = alloca0 (priv->child_count);
-
- ret = afr_selfheal_tie_breaker_inodelk (frame, this, inode,
- priv->sh_domain, 0, 0,
- locked_on);
- {
- if (ret < priv->child_count) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "self-heal as only %d number of "
- "subvolumes could be locked",
- uuid_utoa (fd->inode->gfid),
- ret);
- /* Either less than two subvols available, or another
- selfheal (from another server) is in progress. Skip
- for now in any case there isn't anything to do.
- */
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_data (frame, this, fd, locked_on);
- }
+ afr_private_t *priv = NULL;
+ unsigned char *locked_on = NULL;
+ int ret = 0;
+ inode_t *inode = fd->inode;
+
+ priv = this->private;
+
+ locked_on = alloca0(priv->child_count);
+
+ ret = afr_selfheal_tie_breaker_inodelk(frame, this, inode, priv->sh_domain,
+ 0, 0, locked_on);
+ {
+ if (ret < priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "self-heal as only %d number of "
+ "subvolumes could be locked",
+ uuid_utoa(fd->inode->gfid), ret);
+ /* Either less than two subvols available, or another
+ selfheal (from another server) is in progress. Skip
+ for now in any case there isn't anything to do.
+ */
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_data(frame, this, fd, locked_on);
+ }
unlock:
- afr_selfheal_uninodelk (frame, this, inode, priv->sh_domain, 0, 0,
- locked_on);
+ afr_selfheal_uninodelk(frame, this, inode, priv->sh_domain, 0, 0,
+ locked_on);
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index e6022cc939b..bf7a6b9d1e8 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
#include "byte-order.h"
@@ -18,1105 +17,1090 @@
#include "events.h"
static int
-afr_selfheal_entry_delete (xlator_t *this, inode_t *dir, const char *name,
- inode_t *inode, int child, struct afr_reply *replies)
+afr_selfheal_entry_delete(xlator_t *this, inode_t *dir, const char *name,
+ inode_t *inode, int child, struct afr_reply *replies)
{
- afr_private_t *priv = NULL;
- xlator_t *subvol = NULL;
- int ret = 0;
- loc_t loc = {0, };
- char g[64];
-
- priv = this->private;
-
- subvol = priv->children[child];
-
- loc.parent = inode_ref (dir);
- gf_uuid_copy (loc.pargfid, dir->gfid);
- loc.name = name;
- loc.inode = inode_ref (inode);
-
- if (replies[child].valid && replies[child].op_ret == 0) {
- switch (replies[child].poststat.ia_type) {
- case IA_IFDIR:
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- "expunging dir %s/%s (%s) on %s",
- uuid_utoa (dir->gfid), name,
- uuid_utoa_r (replies[child].poststat.ia_gfid, g),
- subvol->name);
- ret = syncop_rmdir (subvol, &loc, 1, NULL, NULL);
- break;
- default:
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- "expunging file %s/%s (%s) on %s",
- uuid_utoa (dir->gfid), name,
- uuid_utoa_r (replies[child].poststat.ia_gfid, g),
- subvol->name);
- ret = syncop_unlink (subvol, &loc, NULL, NULL);
- break;
- }
- }
-
- loc_wipe (&loc);
-
- return ret;
+ afr_private_t *priv = NULL;
+ xlator_t *subvol = NULL;
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ char g[64];
+
+ priv = this->private;
+
+ subvol = priv->children[child];
+
+ loc.parent = inode_ref(dir);
+ gf_uuid_copy(loc.pargfid, dir->gfid);
+ loc.name = name;
+ loc.inode = inode_ref(inode);
+
+ if (replies[child].valid && replies[child].op_ret == 0) {
+ switch (replies[child].poststat.ia_type) {
+ case IA_IFDIR:
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "expunging dir %s/%s (%s) on %s", uuid_utoa(dir->gfid),
+ name, uuid_utoa_r(replies[child].poststat.ia_gfid, g),
+ subvol->name);
+ ret = syncop_rmdir(subvol, &loc, 1, NULL, NULL);
+ break;
+ default:
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "expunging file %s/%s (%s) on %s", uuid_utoa(dir->gfid),
+ name, uuid_utoa_r(replies[child].poststat.ia_gfid, g),
+ subvol->name);
+ ret = syncop_unlink(subvol, &loc, NULL, NULL);
+ break;
+ }
+ }
+
+ loc_wipe(&loc);
+
+ return ret;
}
int
-afr_selfheal_recreate_entry (call_frame_t *frame, int dst, int source,
- unsigned char *sources, inode_t *dir,
- const char *name, inode_t *inode,
- struct afr_reply *replies)
+afr_selfheal_recreate_entry(call_frame_t *frame, int dst, int source,
+ unsigned char *sources, inode_t *dir,
+ const char *name, inode_t *inode,
+ struct afr_reply *replies)
{
- int ret = 0;
- loc_t loc = {0,};
- loc_t srcloc = {0,};
- xlator_t *this = frame->this;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- struct iatt *iatt = NULL;
- char *linkname = NULL;
- mode_t mode = 0;
- struct iatt newent = {0,};
- unsigned char *newentry = NULL;
-
- priv = this->private;
- iatt = &replies[source].poststat;
- if (iatt->ia_type == IA_INVAL || gf_uuid_is_null (iatt->ia_gfid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_SELF_HEAL_FAILED,
- "Invalid ia_type (%d) or gfid(%s). source brick=%d, "
- "pargfid=%s, name=%s", iatt->ia_type,
- uuid_utoa(iatt->ia_gfid), source,
- uuid_utoa(dir->gfid), name);
- ret = -EINVAL;
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ loc_t srcloc = {
+ 0,
+ };
+ xlator_t *this = frame->this;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ struct iatt *iatt = NULL;
+ char *linkname = NULL;
+ mode_t mode = 0;
+ struct iatt newent = {
+ 0,
+ };
+ unsigned char *newentry = NULL;
+
+ priv = this->private;
+ iatt = &replies[source].poststat;
+ if (iatt->ia_type == IA_INVAL || gf_uuid_is_null(iatt->ia_gfid)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SELF_HEAL_FAILED,
+ "Invalid ia_type (%d) or gfid(%s). source brick=%d, "
+ "pargfid=%s, name=%s",
+ iatt->ia_type, uuid_utoa(iatt->ia_gfid), source,
+ uuid_utoa(dir->gfid), name);
+ ret = -EINVAL;
+ goto out;
+ }
+
+ xdata = dict_new();
+ if (!xdata)
+ return -ENOMEM;
+ newentry = alloca0(priv->child_count);
+ loc.parent = inode_ref(dir);
+ gf_uuid_copy(loc.pargfid, dir->gfid);
+ loc.name = name;
+ loc.inode = inode_ref(inode);
+
+ ret = afr_selfheal_entry_delete(this, dir, name, inode, dst, replies);
+ if (ret)
+ goto out;
+
+ ret = dict_set_gfuuid(xdata, "gfid-req", replies[source].poststat.ia_gfid,
+ true);
+ if (ret)
+ goto out;
+
+ srcloc.inode = inode_ref(inode);
+ gf_uuid_copy(srcloc.gfid, iatt->ia_gfid);
+ if (iatt->ia_type != IA_IFDIR)
+ ret = syncop_lookup(priv->children[dst], &srcloc, 0, 0, 0, 0);
+ if (iatt->ia_type == IA_IFDIR || ret == -ENOENT || ret == -ESTALE) {
+ newentry[dst] = 1;
+ ret = afr_selfheal_newentry_mark(frame, this, inode, source, replies,
+ sources, newentry);
+ if (ret)
+ goto out;
+ }
+
+ mode = st_mode_from_ia(iatt->ia_prot, iatt->ia_type);
+
+ switch (iatt->ia_type) {
+ case IA_IFDIR:
+ ret = syncop_mkdir(priv->children[dst], &loc, mode, 0, xdata, NULL);
+ break;
+ case IA_IFLNK:
+ if (!newentry[dst]) {
+ ret = syncop_link(priv->children[dst], &srcloc, &loc, &newent,
+ NULL, NULL);
+ } else {
+ ret = syncop_readlink(priv->children[source], &srcloc,
+ &linkname, 4096, NULL, NULL);
+ if (ret <= 0)
+ goto out;
+ ret = syncop_symlink(priv->children[dst], &loc, linkname, NULL,
+ xdata, NULL);
+ }
+ break;
+ default:
+ ret = dict_set_int32(xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
+ if (ret)
goto out;
- }
-
- xdata = dict_new();
- if (!xdata)
- return -ENOMEM;
- newentry = alloca0 (priv->child_count);
- loc.parent = inode_ref (dir);
- gf_uuid_copy (loc.pargfid, dir->gfid);
- loc.name = name;
- loc.inode = inode_ref (inode);
-
- ret = afr_selfheal_entry_delete (this, dir, name, inode, dst, replies);
- if (ret)
- goto out;
-
- ret = dict_set_gfuuid (xdata, "gfid-req",
- replies[source].poststat.ia_gfid, true);
- if (ret)
- goto out;
-
- srcloc.inode = inode_ref (inode);
- gf_uuid_copy (srcloc.gfid, iatt->ia_gfid);
- if (iatt->ia_type != IA_IFDIR)
- ret = syncop_lookup (priv->children[dst], &srcloc, 0, 0, 0, 0);
- if (iatt->ia_type == IA_IFDIR || ret == -ENOENT || ret == -ESTALE) {
- newentry[dst] = 1;
- ret = afr_selfheal_newentry_mark (frame, this, inode, source,
- replies, sources, newentry);
- if (ret)
- goto out;
- }
-
- mode = st_mode_from_ia (iatt->ia_prot, iatt->ia_type);
-
- switch (iatt->ia_type) {
- case IA_IFDIR:
- ret = syncop_mkdir (priv->children[dst], &loc, mode, 0,
- xdata, NULL);
- break;
- case IA_IFLNK:
- if (!newentry[dst]) {
- ret = syncop_link (priv->children[dst], &srcloc, &loc,
- &newent, NULL, NULL);
- } else {
- ret = syncop_readlink (priv->children[source], &srcloc,
- &linkname, 4096, NULL, NULL);
- if (ret <= 0)
- goto out;
- ret = syncop_symlink (priv->children[dst], &loc,
- linkname, NULL, xdata, NULL);
- }
- break;
- default:
- ret = dict_set_int32 (xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
- if (ret)
- goto out;
- ret = syncop_mknod (priv->children[dst], &loc, mode,
- makedev (ia_major(iatt->ia_rdev), ia_minor (iatt->ia_rdev)),
- &newent, xdata, NULL);
- break;
- }
+ ret = syncop_mknod(
+ priv->children[dst], &loc, mode,
+ makedev(ia_major(iatt->ia_rdev), ia_minor(iatt->ia_rdev)),
+ &newent, xdata, NULL);
+ break;
+ }
out:
- if (xdata)
- dict_unref (xdata);
- GF_FREE (linkname);
- loc_wipe (&loc);
- loc_wipe (&srcloc);
- return ret;
+ if (xdata)
+ dict_unref(xdata);
+ GF_FREE(linkname);
+ loc_wipe(&loc);
+ loc_wipe(&srcloc);
+ return ret;
}
static int
-__afr_selfheal_heal_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
- char *name, inode_t *inode, int source,
- unsigned char *sources, unsigned char *healed_sinks,
- unsigned char *locked_on, struct afr_reply *replies)
+__afr_selfheal_heal_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *name, inode_t *inode, int source,
+ unsigned char *sources, unsigned char *healed_sinks,
+ unsigned char *locked_on, struct afr_reply *replies)
{
- int ret = 0;
- afr_private_t *priv = NULL;
- int i = 0;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
- priv = this->private;
+ priv = this->private;
- if (!replies[source].valid)
- return -EIO;
+ if (!replies[source].valid)
+ return -EIO;
- /* Skip healing this entry if the last lookup on it failed for reasons
- * other than ENOENT.
- */
- if ((replies[source].op_ret < 0) &&
- (replies[source].op_errno != ENOENT))
- return -replies[source].op_errno;
-
- if (replies[source].op_ret == 0) {
- ret = afr_lookup_and_heal_gfid (this, fd->inode, name,
- inode, replies, source, sources,
- &replies[source].poststat.ia_gfid);
- if (ret)
- return ret;
+ /* Skip healing this entry if the last lookup on it failed for reasons
+ * other than ENOENT.
+ */
+ if ((replies[source].op_ret < 0) && (replies[source].op_errno != ENOENT))
+ return -replies[source].op_errno;
+
+ if (replies[source].op_ret == 0) {
+ ret = afr_lookup_and_heal_gfid(this, fd->inode, name, inode, replies,
+ source, sources,
+ &replies[source].poststat.ia_gfid);
+ if (ret)
+ return ret;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i])
+ continue;
+ if (replies[source].op_ret == -1 &&
+ replies[source].op_errno == ENOENT) {
+ ret = afr_selfheal_entry_delete(this, fd->inode, name, inode, i,
+ replies);
+ } else {
+ if (!gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[source].poststat.ia_gfid))
+ continue;
+
+ ret = afr_selfheal_recreate_entry(frame, i, source, sources,
+ fd->inode, name, inode, replies);
}
+ if (ret < 0)
+ break;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i])
- continue;
- if (replies[source].op_ret == -1 &&
- replies[source].op_errno == ENOENT) {
- ret = afr_selfheal_entry_delete (this, fd->inode, name,
- inode, i, replies);
- } else {
- if (!gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[source].poststat.ia_gfid))
- continue;
-
- ret = afr_selfheal_recreate_entry (frame, i, source,
- sources, fd->inode,
- name, inode,
- replies);
- }
- if (ret < 0)
- break;
- }
-
- return ret;
+ return ret;
}
static int
-afr_selfheal_detect_gfid_and_type_mismatch (xlator_t *this,
- struct afr_reply *replies,
- inode_t *inode,
- uuid_t pargfid,
- char *bname, int src_idx,
- unsigned char *locked_on,
- int *src)
+afr_selfheal_detect_gfid_and_type_mismatch(xlator_t *this,
+ struct afr_reply *replies,
+ inode_t *inode, uuid_t pargfid,
+ char *bname, int src_idx,
+ unsigned char *locked_on, int *src)
{
- int i = 0;
- int ret = -1;
- afr_private_t *priv = NULL;
- void *gfid = NULL;
- ia_type_t ia_type = IA_INVAL;
-
- priv = this->private;
- gfid = &replies[src_idx].poststat.ia_gfid;
- ia_type = replies[src_idx].poststat.ia_type;
+ int i = 0;
+ int ret = -1;
+ afr_private_t *priv = NULL;
+ void *gfid = NULL;
+ ia_type_t ia_type = IA_INVAL;
+
+ priv = this->private;
+ gfid = &replies[src_idx].poststat.ia_gfid;
+ ia_type = replies[src_idx].poststat.ia_type;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == src_idx)
+ continue;
+
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret != 0)
+ continue;
+
+ if (gf_uuid_compare(gfid, replies[i].poststat.ia_gfid) &&
+ (ia_type == replies[i].poststat.ia_type)) {
+ ret = afr_gfid_split_brain_source(this, replies, inode, pargfid,
+ bname, src_idx, i, locked_on, src,
+ NULL);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Skipping conservative merge on the "
+ "file.");
+ return ret;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (i == src_idx)
- continue;
-
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret != 0)
- continue;
-
- if (gf_uuid_compare (gfid, replies[i].poststat.ia_gfid) &&
- (ia_type == replies[i].poststat.ia_type)) {
- ret = afr_gfid_split_brain_source (this, replies, inode,
- pargfid, bname,
- src_idx, i,
- locked_on, src,
- NULL);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN,
- "Skipping conservative merge on the "
- "file.");
- return ret;
- }
-
- if (ia_type != replies[i].poststat.ia_type) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "Type mismatch detected "
- "for <gfid:%s>/%s>, %s on %s and %s on %s. "
- "Skipping conservative merge on the file.",
- uuid_utoa (pargfid), bname,
- gf_inode_type_to_str (replies[i].poststat.ia_type),
- priv->children[i]->name,
- gf_inode_type_to_str (replies[src_idx].poststat.ia_type),
- priv->children[src_idx]->name);
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;type=file;"
- "file=<gfid:%s>/%s>;count=2;child-%d=%s;type-"
- "%d=%s;child-%d=%s;type-%d=%s",
- this->name, uuid_utoa (pargfid), bname, i,
- priv->children[i]->name, i,
- gf_inode_type_to_str(replies[i].poststat.ia_type),
- src_idx, priv->children[src_idx]->name, src_idx,
- gf_inode_type_to_str(replies[src_idx].poststat.ia_type));
- return -1;
- }
+ if (ia_type != replies[i].poststat.ia_type) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Type mismatch detected "
+ "for <gfid:%s>/%s>, %s on %s and %s on %s. "
+ "Skipping conservative merge on the file.",
+ uuid_utoa(pargfid), bname,
+ gf_inode_type_to_str(replies[i].poststat.ia_type),
+ priv->children[i]->name,
+ gf_inode_type_to_str(replies[src_idx].poststat.ia_type),
+ priv->children[src_idx]->name);
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "subvol=%s;type=file;"
+ "file=<gfid:%s>/%s>;count=2;child-%d=%s;type-"
+ "%d=%s;child-%d=%s;type-%d=%s",
+ this->name, uuid_utoa(pargfid), bname, i,
+ priv->children[i]->name, i,
+ gf_inode_type_to_str(replies[i].poststat.ia_type), src_idx,
+ priv->children[src_idx]->name, src_idx,
+ gf_inode_type_to_str(replies[src_idx].poststat.ia_type));
+ return -1;
}
+ }
- return 0;
+ return 0;
}
static int
-__afr_selfheal_merge_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
- char *name, inode_t *inode, unsigned char *sources,
- unsigned char *healed_sinks, unsigned char *locked_on,
- struct afr_reply *replies)
+__afr_selfheal_merge_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *name, inode_t *inode, unsigned char *sources,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on, struct afr_reply *replies)
{
- int ret = 0;
- int i = 0;
- int source = -1;
- int src = -1;
- afr_private_t *priv = NULL;
+ int ret = 0;
+ int i = 0;
+ int source = -1;
+ int src = -1;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid && replies[i].op_ret == 0) {
+ source = i;
+ break;
+ }
+ }
- priv = this->private;
+ if (source == -1) {
+ /* entry got deleted in the mean time? */
+ return 0;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid && replies[i].op_ret == 0) {
- source = i;
- break;
- }
- }
-
- if (source == -1) {
- /* entry got deleted in the mean time? */
- return 0;
- }
-
- /* Set all the sources as 1, otheriwse newentry_mark won't be set */
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid && replies[i].op_ret == 0) {
- sources[i] = 1;
- }
- }
-
- ret = afr_lookup_and_heal_gfid (this, fd->inode, name, inode, replies,
- source, sources,
- &replies[source].poststat.ia_gfid);
- if (ret)
- return ret;
+ /* Set all the sources as 1, otheriwse newentry_mark won't be set */
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid && replies[i].op_ret == 0) {
+ sources[i] = 1;
+ }
+ }
- /* In case of type mismatch / unable to resolve gfid mismatch on the
- * entry, return -1.*/
- ret = afr_selfheal_detect_gfid_and_type_mismatch (this, replies, inode,
- fd->inode->gfid,
- name, source,
- locked_on, &src);
+ ret = afr_lookup_and_heal_gfid(this, fd->inode, name, inode, replies,
+ source, sources,
+ &replies[source].poststat.ia_gfid);
+ if (ret)
+ return ret;
- if (ret < 0)
- return ret;
- if (src != -1) {
- source = src;
- for (i = 0; i < priv->child_count; i++) {
- if (i != src && replies[i].valid &&
- gf_uuid_compare (replies[src].poststat.ia_gfid,
- replies[i].poststat.ia_gfid)) {
- sources[i] = 0;
- }
- }
+ /* In case of type mismatch / unable to resolve gfid mismatch on the
+ * entry, return -1.*/
+ ret = afr_selfheal_detect_gfid_and_type_mismatch(
+ this, replies, inode, fd->inode->gfid, name, source, locked_on, &src);
+
+ if (ret < 0)
+ return ret;
+ if (src != -1) {
+ source = src;
+ for (i = 0; i < priv->child_count; i++) {
+ if (i != src && replies[i].valid &&
+ gf_uuid_compare(replies[src].poststat.ia_gfid,
+ replies[i].poststat.ia_gfid)) {
+ sources[i] = 0;
+ }
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || !healed_sinks[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source || !healed_sinks[i])
+ continue;
- if (src != -1) {
- if (!gf_uuid_compare (replies[src].poststat.ia_gfid,
- replies[i].poststat.ia_gfid))
- continue;
- } else if (replies[i].op_errno != ENOENT) {
- continue;
- }
+ if (src != -1) {
+ if (!gf_uuid_compare(replies[src].poststat.ia_gfid,
+ replies[i].poststat.ia_gfid))
+ continue;
+ } else if (replies[i].op_errno != ENOENT) {
+ continue;
+ }
- ret |= afr_selfheal_recreate_entry (frame, i, source, sources,
- fd->inode, name, inode,
- replies);
- }
+ ret |= afr_selfheal_recreate_entry(frame, i, source, sources, fd->inode,
+ name, inode, replies);
+ }
- return ret;
+ return ret;
}
-
static int
-__afr_selfheal_entry_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
- char *name, inode_t *inode, int source,
- unsigned char *sources, unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies)
+__afr_selfheal_entry_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *name, inode_t *inode, int source,
+ unsigned char *sources, unsigned char *healed_sinks,
+ unsigned char *locked_on, struct afr_reply *replies)
{
- int ret = -1;
-
- if (source < 0)
- ret = __afr_selfheal_merge_dirent (frame, this, fd, name, inode,
- sources, healed_sinks,
- locked_on, replies);
- else
- ret = __afr_selfheal_heal_dirent (frame, this, fd, name, inode,
- source, sources, healed_sinks,
- locked_on, replies);
- return ret;
+ int ret = -1;
+
+ if (source < 0)
+ ret = __afr_selfheal_merge_dirent(frame, this, fd, name, inode, sources,
+ healed_sinks, locked_on, replies);
+ else
+ ret = __afr_selfheal_heal_dirent(frame, this, fd, name, inode, source,
+ sources, healed_sinks, locked_on,
+ replies);
+ return ret;
}
static gf_boolean_t
-is_full_heal_marker_present (xlator_t *this, dict_t *xdata, int idx)
+is_full_heal_marker_present(xlator_t *this, dict_t *xdata, int idx)
{
- int i = 0;
- int pending[3] = {0,};
- void *pending_raw = NULL;
- afr_private_t *priv = NULL;
+ int i = 0;
+ int pending[3] = {
+ 0,
+ };
+ void *pending_raw = NULL;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!xdata)
- return _gf_false;
+ if (!xdata)
+ return _gf_false;
- /* Iterate over each of the priv->pending_keys[] elements and then
- * see if any of them have data segment non-zero. If they do, return
- * true. Else return false.
- */
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get_ptr (xdata, priv->pending_key[i], &pending_raw))
- continue;
+ /* Iterate over each of the priv->pending_keys[] elements and then
+ * see if any of them have data segment non-zero. If they do, return
+ * true. Else return false.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_get_ptr(xdata, priv->pending_key[i], &pending_raw))
+ continue;
- if (!pending_raw)
- continue;
+ if (!pending_raw)
+ continue;
- memcpy (pending, pending_raw, sizeof (pending));
- if (ntoh32 (pending[idx]))
- return _gf_true;
- }
+ memcpy(pending, pending_raw, sizeof(pending));
+ if (ntoh32(pending[idx]))
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static gf_boolean_t
-afr_need_full_heal (xlator_t *this, struct afr_reply *replies, int source,
- unsigned char *healed_sinks, afr_transaction_type type)
+afr_need_full_heal(xlator_t *this, struct afr_reply *replies, int source,
+ unsigned char *healed_sinks, afr_transaction_type type)
{
- int i = 0;
- int idx = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ int idx = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!priv->esh_granular)
- return _gf_true;
+ if (!priv->esh_granular)
+ return _gf_true;
- if (type != AFR_ENTRY_TRANSACTION)
- return _gf_true;
+ if (type != AFR_ENTRY_TRANSACTION)
+ return _gf_true;
- priv = this->private;
- idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+ priv = this->private;
+ idx = afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
- /* If there is a clear source, check whether the full-heal-indicator
- * is present in its xdata. Otherwise, we need to examine all the
- * participating bricks and then figure if *even* one of them has a
- * full-heal-indicator.
- */
+ /* If there is a clear source, check whether the full-heal-indicator
+ * is present in its xdata. Otherwise, we need to examine all the
+ * participating bricks and then figure if *even* one of them has a
+ * full-heal-indicator.
+ */
- if (source != -1) {
- if (is_full_heal_marker_present (this, replies[source].xdata,
- idx))
- return _gf_true;
- }
+ if (source != -1) {
+ if (is_full_heal_marker_present(this, replies[source].xdata, idx))
+ return _gf_true;
+ }
- /* else ..*/
+ /* else ..*/
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i])
+ continue;
- if (is_full_heal_marker_present (this, replies[i].xdata, idx))
- return _gf_true;
- }
+ if (is_full_heal_marker_present(this, replies[i].xdata, idx))
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static int
-__afr_selfheal_entry_finalize_source (xlator_t *this, unsigned char *sources,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- uint64_t *witness)
+__afr_selfheal_entry_finalize_source(xlator_t *this, unsigned char *sources,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on,
+ struct afr_reply *replies,
+ uint64_t *witness)
{
- afr_private_t *priv = NULL;
- int source = -1;
- int sources_count = 0;
+ afr_private_t *priv = NULL;
+ int source = -1;
+ int sources_count = 0;
- priv = this->private;
+ priv = this->private;
- sources_count = AFR_COUNT (sources, priv->child_count);
+ sources_count = AFR_COUNT(sources, priv->child_count);
- if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count || afr_does_witness_exist (this, witness)) {
+ if ((AFR_CMP(locked_on, healed_sinks, priv->child_count) == 0) ||
+ !sources_count || afr_does_witness_exist(this, witness)) {
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ afr_mark_active_sinks(this, sources, locked_on, healed_sinks);
+ return -1;
+ }
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- afr_mark_active_sinks (this, sources, locked_on, healed_sinks);
- return -1;
- }
-
- source = afr_choose_source_by_policy (priv, sources,
- AFR_ENTRY_TRANSACTION);
- return source;
+ source = afr_choose_source_by_policy(priv, sources, AFR_ENTRY_TRANSACTION);
+ return source;
}
int
-__afr_selfheal_entry_prepare (call_frame_t *frame, xlator_t *this,
- inode_t *inode, unsigned char *locked_on,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks,
- struct afr_reply *replies, int *source_p,
- gf_boolean_t *pflag)
+__afr_selfheal_entry_prepare(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ struct afr_reply *replies, int *source_p,
+ gf_boolean_t *pflag)
{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- uint64_t *witness = NULL;
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ uint64_t *witness = NULL;
- priv = this->private;
+ priv = this->private;
- ret = afr_selfheal_unlocked_discover (frame, inode, inode->gfid,
- replies);
- if (ret)
- return ret;
-
- witness = alloca0 (sizeof (*witness) * priv->child_count);
- ret = afr_selfheal_find_direction (frame, this, replies,
- AFR_ENTRY_TRANSACTION,
- locked_on, sources, sinks, witness,
- pflag);
- if (ret)
- return ret;
-
- /* Initialize the healed_sinks[] array optimistically to
- the intersection of to-be-healed (i.e sinks[]) and
- the list of servers which are up (i.e locked_on[]).
-
- As we encounter failures in the healing process, we
- will unmark the respective servers in the healed_sinks[]
- array.
- */
- AFR_INTERSECT (healed_sinks, sinks, locked_on, priv->child_count);
-
- source = __afr_selfheal_entry_finalize_source (this, sources,
- healed_sinks,
- locked_on, replies,
- witness);
-
- if (source < 0) {
- /* If source is < 0 (typically split-brain), we perform a
- conservative merge of entries rather than erroring out */
- }
- *source_p = source;
-
- return ret;
-}
+ ret = afr_selfheal_unlocked_discover(frame, inode, inode->gfid, replies);
+ if (ret)
+ return ret;
-static int
-afr_selfheal_entry_dirent (call_frame_t *frame, xlator_t *this,
- fd_t *fd, char *name, inode_t *parent_idx_inode,
- xlator_t *subvol, gf_boolean_t full_crawl)
-{
- int ret = 0;
- int source = -1;
- unsigned char *locked_on = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- inode_t *inode = NULL;
- struct afr_reply *replies = NULL;
- struct afr_reply *par_replies = NULL;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
-
- priv = this->private;
-
- xattr = dict_new ();
- if (!xattr)
- return -ENOMEM;
- ret = dict_set_int32 (xattr, GF_GFIDLESS_LOOKUP, 1);
- if (ret) {
- dict_unref (xattr);
- return -1;
- }
+ witness = alloca0(sizeof(*witness) * priv->child_count);
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_ENTRY_TRANSACTION, locked_on, sources,
+ sinks, witness, pflag);
+ if (ret)
+ return ret;
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- locked_on = alloca0 (priv->child_count);
-
- replies = alloca0 (priv->child_count * sizeof(*replies));
- par_replies = alloca0 (priv->child_count * sizeof(*par_replies));
-
- ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
- locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "entry self-heal as only %d sub-volumes "
- " could be locked in %s domain",
- uuid_utoa (fd->inode->gfid),
- ret, this->name);
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_entry_prepare (frame, this, fd->inode,
- locked_on,
- sources, sinks,
- healed_sinks, par_replies,
- &source, NULL);
- if (ret < 0)
- goto unlock;
-
- inode = afr_selfheal_unlocked_lookup_on (frame, fd->inode, name,
- replies, locked_on,
- xattr);
- if (!inode) {
- ret = -ENOMEM;
- goto unlock;
- }
-
- ret = __afr_selfheal_entry_dirent (frame, this, fd, name, inode,
- source, sources, healed_sinks,
- locked_on, replies);
-
- if ((ret == 0) && (priv->esh_granular) && parent_idx_inode) {
- ret = afr_shd_index_purge (subvol, parent_idx_inode,
- name, inode->ia_type);
- /* Why is ret force-set to 0? We do not care about
- * index purge failing for full heal as it is quite
- * possible during replace-brick that not all files
- * and directories have their name indices present in
- * entry-changes/.
- */
- ret = 0;
- }
- }
+ /* Initialize the healed_sinks[] array optimistically to
+ the intersection of to-be-healed (i.e sinks[]) and
+ the list of servers which are up (i.e locked_on[]).
-unlock:
- afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
- locked_on, NULL);
- if (inode)
- inode_unref (inode);
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
- if (par_replies)
- afr_replies_wipe (par_replies, priv->child_count);
- if (xattr)
- dict_unref (xattr);
-
- return ret;
-}
+ As we encounter failures in the healing process, we
+ will unmark the respective servers in the healed_sinks[]
+ array.
+ */
+ AFR_INTERSECT(healed_sinks, sinks, locked_on, priv->child_count);
+ source = __afr_selfheal_entry_finalize_source(this, sources, healed_sinks,
+ locked_on, replies, witness);
-static inode_t *
-afr_shd_entry_changes_index_inode (xlator_t *this, xlator_t *subvol,
- uuid_t pargfid)
+ if (source < 0) {
+ /* If source is < 0 (typically split-brain), we perform a
+ conservative merge of entries rather than erroring out */
+ }
+ *source_p = source;
+
+ return ret;
+}
+
+static int
+afr_selfheal_entry_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *name, inode_t *parent_idx_inode,
+ xlator_t *subvol, gf_boolean_t full_crawl)
{
- int ret = -1;
- void *index_gfid = NULL;
- loc_t rootloc = {0,};
- loc_t loc = {0,};
- dict_t *xattr = NULL;
- inode_t *inode = NULL;
- struct iatt iatt = {0,};
-
- rootloc.inode = inode_ref (this->itable->root);
- gf_uuid_copy (rootloc.gfid, rootloc.inode->gfid);
-
- ret = syncop_getxattr (subvol, &rootloc, &xattr,
- GF_XATTROP_ENTRY_CHANGES_GFID, NULL, NULL);
- if (ret || !xattr) {
- errno = -ret;
- goto out;
+ int ret = 0;
+ int source = -1;
+ unsigned char *locked_on = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ inode_t *inode = NULL;
+ struct afr_reply *replies = NULL;
+ struct afr_reply *par_replies = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+
+ priv = this->private;
+
+ xattr = dict_new();
+ if (!xattr)
+ return -ENOMEM;
+ ret = dict_set_int32(xattr, GF_GFIDLESS_LOOKUP, 1);
+ if (ret) {
+ dict_unref(xattr);
+ return -1;
+ }
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ locked_on = alloca0(priv->child_count);
+
+ replies = alloca0(priv->child_count * sizeof(*replies));
+ par_replies = alloca0(priv->child_count * sizeof(*par_replies));
+
+ ret = afr_selfheal_entrylk(frame, this, fd->inode, this->name, NULL,
+ locked_on);
+ {
+ if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "entry self-heal as only %d sub-volumes "
+ " could be locked in %s domain",
+ uuid_utoa(fd->inode->gfid), ret, this->name);
+ ret = -ENOTCONN;
+ goto unlock;
}
- ret = dict_get_ptr (xattr, GF_XATTROP_ENTRY_CHANGES_GFID, &index_gfid);
- if (ret) {
- errno = EINVAL;
- goto out;
- }
+ ret = __afr_selfheal_entry_prepare(frame, this, fd->inode, locked_on,
+ sources, sinks, healed_sinks,
+ par_replies, &source, NULL);
+ if (ret < 0)
+ goto unlock;
- loc.inode = inode_new (this->itable);
- if (!loc.inode) {
- errno = ENOMEM;
- goto out;
+ inode = afr_selfheal_unlocked_lookup_on(frame, fd->inode, name, replies,
+ locked_on, xattr);
+ if (!inode) {
+ ret = -ENOMEM;
+ goto unlock;
}
- gf_uuid_copy (loc.pargfid, index_gfid);
- loc.name = gf_strdup (uuid_utoa (pargfid));
-
- ret = syncop_lookup (subvol, &loc, &iatt, NULL, NULL, NULL);
- if (ret < 0) {
- errno = -ret;
- goto out;
+ ret = __afr_selfheal_entry_dirent(frame, this, fd, name, inode, source,
+ sources, healed_sinks, locked_on,
+ replies);
+
+ if ((ret == 0) && (priv->esh_granular) && parent_idx_inode) {
+ ret = afr_shd_index_purge(subvol, parent_idx_inode, name,
+ inode->ia_type);
+ /* Why is ret force-set to 0? We do not care about
+ * index purge failing for full heal as it is quite
+ * possible during replace-brick that not all files
+ * and directories have their name indices present in
+ * entry-changes/.
+ */
+ ret = 0;
}
+ }
- inode = inode_link (loc.inode, NULL, NULL, &iatt);
+unlock:
+ afr_selfheal_unentrylk(frame, this, fd->inode, this->name, NULL, locked_on,
+ NULL);
+ if (inode)
+ inode_unref(inode);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
+ if (par_replies)
+ afr_replies_wipe(par_replies, priv->child_count);
+ if (xattr)
+ dict_unref(xattr);
+
+ return ret;
+}
+
+static inode_t *
+afr_shd_entry_changes_index_inode(xlator_t *this, xlator_t *subvol,
+ uuid_t pargfid)
+{
+ int ret = -1;
+ void *index_gfid = NULL;
+ loc_t rootloc = {
+ 0,
+ };
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ inode_t *inode = NULL;
+ struct iatt iatt = {
+ 0,
+ };
+
+ rootloc.inode = inode_ref(this->itable->root);
+ gf_uuid_copy(rootloc.gfid, rootloc.inode->gfid);
+
+ ret = syncop_getxattr(subvol, &rootloc, &xattr,
+ GF_XATTROP_ENTRY_CHANGES_GFID, NULL, NULL);
+ if (ret || !xattr) {
+ errno = -ret;
+ goto out;
+ }
+
+ ret = dict_get_ptr(xattr, GF_XATTROP_ENTRY_CHANGES_GFID, &index_gfid);
+ if (ret) {
+ errno = EINVAL;
+ goto out;
+ }
+
+ loc.inode = inode_new(this->itable);
+ if (!loc.inode) {
+ errno = ENOMEM;
+ goto out;
+ }
+
+ gf_uuid_copy(loc.pargfid, index_gfid);
+ loc.name = gf_strdup(uuid_utoa(pargfid));
+
+ ret = syncop_lookup(subvol, &loc, &iatt, NULL, NULL, NULL);
+ if (ret < 0) {
+ errno = -ret;
+ goto out;
+ }
+
+ inode = inode_link(loc.inode, NULL, NULL, &iatt);
out:
- if (xattr)
- dict_unref (xattr);
- loc_wipe (&rootloc);
- GF_FREE ((char *)loc.name);
- loc_wipe (&loc);
+ if (xattr)
+ dict_unref(xattr);
+ loc_wipe(&rootloc);
+ GF_FREE((char *)loc.name);
+ loc_wipe(&loc);
- return inode;
+ return inode;
}
static int
-afr_selfheal_entry_do_subvol (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int child)
+afr_selfheal_entry_do_subvol(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int child)
{
- int ret = 0;
- gf_dirent_t entries;
- gf_dirent_t *entry = NULL;
- off_t offset = 0;
- call_frame_t *iter_frame = NULL;
- xlator_t *subvol = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t mismatch = _gf_false;
- afr_local_t *local = NULL;
- loc_t loc = {0,};
-
- priv = this->private;
- subvol = priv->children[child];
-
- INIT_LIST_HEAD (&entries.list);
-
- local = frame->local;
-
- iter_frame = afr_copy_frame (frame);
- if (!iter_frame)
- return -ENOMEM;
-
- loc.inode = afr_shd_entry_changes_index_inode (this, subvol,
- fd->inode->gfid);
-
- while ((ret = syncop_readdir (subvol, fd, 131072, offset, &entries,
- NULL, NULL))) {
- if (ret > 0)
- ret = 0;
- list_for_each_entry (entry, &entries.list, list) {
- offset = entry->d_off;
-
- if (!strcmp (entry->d_name, ".") ||
- !strcmp (entry->d_name, ".."))
- continue;
-
- if (__is_root_gfid (fd->inode->gfid) &&
- !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR))
- continue;
-
- ret = afr_selfheal_entry_dirent (iter_frame, this, fd,
- entry->d_name,
- loc.inode, subvol,
- local->need_full_crawl);
- AFR_STACK_RESET (iter_frame);
- if (iter_frame->local == NULL) {
- ret = -ENOTCONN;
- break;
- }
-
- if (ret == -1) {
- /* gfid or type mismatch. */
- mismatch = _gf_true;
- ret = 0;
- }
- if (ret)
- break;
- }
-
- gf_dirent_free (&entries);
- if (ret)
- break;
- }
-
- loc_wipe (&loc);
-
- AFR_STACK_DESTROY (iter_frame);
- if (mismatch == _gf_true)
- /* undo pending will be skipped */
- ret = -1;
- return ret;
-}
+ int ret = 0;
+ gf_dirent_t entries;
+ gf_dirent_t *entry = NULL;
+ off_t offset = 0;
+ call_frame_t *iter_frame = NULL;
+ xlator_t *subvol = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t mismatch = _gf_false;
+ afr_local_t *local = NULL;
+ loc_t loc = {
+ 0,
+ };
+
+ priv = this->private;
+ subvol = priv->children[child];
+
+ INIT_LIST_HEAD(&entries.list);
+
+ local = frame->local;
+
+ iter_frame = afr_copy_frame(frame);
+ if (!iter_frame)
+ return -ENOMEM;
+
+ loc.inode = afr_shd_entry_changes_index_inode(this, subvol,
+ fd->inode->gfid);
+
+ while ((ret = syncop_readdir(subvol, fd, 131072, offset, &entries, NULL,
+ NULL))) {
+ if (ret > 0)
+ ret = 0;
+ list_for_each_entry(entry, &entries.list, list)
+ {
+ offset = entry->d_off;
-static int
-afr_selfheal_entry_granular_dirent (xlator_t *subvol, gf_dirent_t *entry,
- loc_t *parent, void *data)
-{
- int ret = 0;
- loc_t loc = {0,};
- struct iatt iatt = {0,};
- afr_granular_esh_args_t *args = data;
-
- /* Look up the actual inode associated with entry. If the lookup returns
- * ESTALE or ENOENT, then it means we have a stale index. Remove it.
- * This is analogous to the check in afr_shd_index_heal() except that
- * here it is achieved through LOOKUP and in afr_shd_index_heal() through
- * a GETXATTR.
- */
+ if (!strcmp(entry->d_name, ".") || !strcmp(entry->d_name, ".."))
+ continue;
+
+ if (__is_root_gfid(fd->inode->gfid) &&
+ !strcmp(entry->d_name, GF_REPLICATE_TRASH_DIR))
+ continue;
- loc.inode = inode_new (args->xl->itable);
- loc.parent = inode_ref (args->heal_fd->inode);
- gf_uuid_copy (loc.pargfid, loc.parent->gfid);
- loc.name = entry->d_name;
-
- ret = syncop_lookup (args->xl, &loc, &iatt, NULL, NULL, NULL);
- if ((ret == -ENOENT) || (ret == -ESTALE)) {
- /* The name indices under the pgfid index dir are guaranteed
- * to be regular files. Hence the hardcoding.
- */
- afr_shd_index_purge (subvol, parent->inode, entry->d_name,
- IA_IFREG);
+ ret = afr_selfheal_entry_dirent(iter_frame, this, fd, entry->d_name,
+ loc.inode, subvol,
+ local->need_full_crawl);
+ AFR_STACK_RESET(iter_frame);
+ if (iter_frame->local == NULL) {
+ ret = -ENOTCONN;
+ break;
+ }
+
+ if (ret == -1) {
+ /* gfid or type mismatch. */
+ mismatch = _gf_true;
ret = 0;
- goto out;
+ }
+ if (ret)
+ break;
}
- /* TBD: afr_shd_zero_xattrop? */
- ret = afr_selfheal_entry_dirent (args->frame, args->xl, args->heal_fd,
- entry->d_name, parent->inode, subvol,
- _gf_false);
- AFR_STACK_RESET (args->frame);
- if (args->frame->local == NULL)
- ret = -ENOTCONN;
+ gf_dirent_free(&entries);
+ if (ret)
+ break;
+ }
- if (ret == -1)
- args->mismatch = _gf_true;
+ loc_wipe(&loc);
-out:
- loc_wipe (&loc);
- return 0;
+ AFR_STACK_DESTROY(iter_frame);
+ if (mismatch == _gf_true)
+ /* undo pending will be skipped */
+ ret = -1;
+ return ret;
}
static int
-afr_selfheal_entry_granular (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int subvol_idx, gf_boolean_t is_src)
+afr_selfheal_entry_granular_dirent(xlator_t *subvol, gf_dirent_t *entry,
+ loc_t *parent, void *data)
{
- int ret = 0;
- loc_t loc = {0,};
- xlator_t *subvol = NULL;
- afr_private_t *priv = NULL;
- afr_granular_esh_args_t args = {0,};
-
- priv = this->private;
- subvol = priv->children[subvol_idx];
-
- args.frame = afr_copy_frame (frame);
- args.xl = this;
- /* args.heal_fd represents the fd associated with the original directory
- * on which entry heal is being attempted.
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ struct iatt iatt = {
+ 0,
+ };
+ afr_granular_esh_args_t *args = data;
+
+ /* Look up the actual inode associated with entry. If the lookup returns
+ * ESTALE or ENOENT, then it means we have a stale index. Remove it.
+ * This is analogous to the check in afr_shd_index_heal() except that
+ * here it is achieved through LOOKUP and in afr_shd_index_heal() through
+ * a GETXATTR.
+ */
+
+ loc.inode = inode_new(args->xl->itable);
+ loc.parent = inode_ref(args->heal_fd->inode);
+ gf_uuid_copy(loc.pargfid, loc.parent->gfid);
+ loc.name = entry->d_name;
+
+ ret = syncop_lookup(args->xl, &loc, &iatt, NULL, NULL, NULL);
+ if ((ret == -ENOENT) || (ret == -ESTALE)) {
+ /* The name indices under the pgfid index dir are guaranteed
+ * to be regular files. Hence the hardcoding.
*/
- args.heal_fd = fd;
+ afr_shd_index_purge(subvol, parent->inode, entry->d_name, IA_IFREG);
+ ret = 0;
+ goto out;
+ }
+ /* TBD: afr_shd_zero_xattrop? */
+
+ ret = afr_selfheal_entry_dirent(args->frame, args->xl, args->heal_fd,
+ entry->d_name, parent->inode, subvol,
+ _gf_false);
+ AFR_STACK_RESET(args->frame);
+ if (args->frame->local == NULL)
+ ret = -ENOTCONN;
+
+ if (ret == -1)
+ args->mismatch = _gf_true;
- /* @subvol here represents the subvolume of AFR where
- * indices/entry-changes/<pargfid> will be processed
+out:
+ loc_wipe(&loc);
+ return 0;
+}
+
+static int
+afr_selfheal_entry_granular(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int subvol_idx, gf_boolean_t is_src)
+{
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ xlator_t *subvol = NULL;
+ afr_private_t *priv = NULL;
+ afr_granular_esh_args_t args = {
+ 0,
+ };
+
+ priv = this->private;
+ subvol = priv->children[subvol_idx];
+
+ args.frame = afr_copy_frame(frame);
+ args.xl = this;
+ /* args.heal_fd represents the fd associated with the original directory
+ * on which entry heal is being attempted.
+ */
+ args.heal_fd = fd;
+
+ /* @subvol here represents the subvolume of AFR where
+ * indices/entry-changes/<pargfid> will be processed
+ */
+ loc.inode = afr_shd_entry_changes_index_inode(this, subvol,
+ fd->inode->gfid);
+ if (!loc.inode) {
+ /* If granular heal failed on the sink (as it might sometimes
+ * because it is the src that would mostly contain the granular
+ * changelogs and the sink's entry-changes would be empty),
+ * do not treat heal as failure.
*/
- loc.inode = afr_shd_entry_changes_index_inode (this, subvol,
- fd->inode->gfid);
- if (!loc.inode) {
- /* If granular heal failed on the sink (as it might sometimes
- * because it is the src that would mostly contain the granular
- * changelogs and the sink's entry-changes would be empty),
- * do not treat heal as failure.
- */
- if (is_src)
- return -errno;
- else
- return 0;
- }
+ if (is_src)
+ return -errno;
+ else
+ return 0;
+ }
- ret = syncop_dir_scan (subvol, &loc, GF_CLIENT_PID_SELF_HEALD,
- &args, afr_selfheal_entry_granular_dirent);
+ ret = syncop_dir_scan(subvol, &loc, GF_CLIENT_PID_SELF_HEALD, &args,
+ afr_selfheal_entry_granular_dirent);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- if (args.mismatch == _gf_true)
- ret = -1;
+ if (args.mismatch == _gf_true)
+ ret = -1;
- return ret;
+ return ret;
}
static int
-afr_selfheal_entry_do (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *sources,
- unsigned char *healed_sinks)
+afr_selfheal_entry_do(call_frame_t *frame, xlator_t *this, fd_t *fd, int source,
+ unsigned char *sources, unsigned char *healed_sinks)
{
- int i = 0;
- int ret = 0;
- gf_boolean_t mismatch = _gf_false;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- local = frame->local;
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_SELF_HEAL_INFO, "performing entry selfheal on %s",
- uuid_utoa (fd->inode->gfid));
-
- for (i = 0; i < priv->child_count; i++) {
- /* Expunge */
- if (!healed_sinks[i])
- continue;
-
- if (!local->need_full_crawl)
- /* Why call afr_selfheal_entry_granular() on a "healed sink",
- * given that it is the source that contains the granular
- * indices?
- * If the index for this directory is non-existent or empty on
- * this subvol (=> clear sink), then it will return early
- * without failure status.
- * If the index is non-empty and it is yet a 'healed sink', then
- * it is due to a split-brain in which case we anyway need to
- * crawl the indices/entry-changes/pargfid directory.
- */
- ret = afr_selfheal_entry_granular (frame, this, fd, i,
- _gf_false);
- else
- ret = afr_selfheal_entry_do_subvol (frame, this, fd, i);
-
- if (ret == -1) {
- /* gfid or type mismatch. */
- mismatch = _gf_true;
- ret = 0;
- }
- if (ret)
- break;
- }
-
- if (!ret && source != -1) {
- /* Impunge */
- if (local->need_full_crawl)
- ret = afr_selfheal_entry_do_subvol (frame, this, fd,
- source);
- else
- ret = afr_selfheal_entry_granular (frame, this, fd,
- source, _gf_true);
+ int i = 0;
+ int ret = 0;
+ gf_boolean_t mismatch = _gf_false;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "performing entry selfheal on %s", uuid_utoa(fd->inode->gfid));
+
+ for (i = 0; i < priv->child_count; i++) {
+ /* Expunge */
+ if (!healed_sinks[i])
+ continue;
+
+ if (!local->need_full_crawl)
+ /* Why call afr_selfheal_entry_granular() on a "healed sink",
+ * given that it is the source that contains the granular
+ * indices?
+ * If the index for this directory is non-existent or empty on
+ * this subvol (=> clear sink), then it will return early
+ * without failure status.
+ * If the index is non-empty and it is yet a 'healed sink', then
+ * it is due to a split-brain in which case we anyway need to
+ * crawl the indices/entry-changes/pargfid directory.
+ */
+ ret = afr_selfheal_entry_granular(frame, this, fd, i, _gf_false);
+ else
+ ret = afr_selfheal_entry_do_subvol(frame, this, fd, i);
+
+ if (ret == -1) {
+ /* gfid or type mismatch. */
+ mismatch = _gf_true;
+ ret = 0;
}
+ if (ret)
+ break;
+ }
- if (mismatch == _gf_true)
- /* undo pending will be skipped */
- ret = -1;
- return ret;
+ if (!ret && source != -1) {
+ /* Impunge */
+ if (local->need_full_crawl)
+ ret = afr_selfheal_entry_do_subvol(frame, this, fd, source);
+ else
+ ret = afr_selfheal_entry_granular(frame, this, fd, source,
+ _gf_true);
+ }
+
+ if (mismatch == _gf_true)
+ /* undo pending will be skipped */
+ ret = -1;
+ return ret;
}
static int
-__afr_selfheal_entry (call_frame_t *frame, xlator_t *this, fd_t *fd,
- unsigned char *locked_on)
+__afr_selfheal_entry(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ unsigned char *locked_on)
{
- int ret = -1;
- int source = -1;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *data_lock = NULL;
- unsigned char *postop_lock = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t did_sh = _gf_true;
-
- priv = this->private;
- local = frame->local;
-
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
- data_lock = alloca0 (priv->child_count);
- postop_lock = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
- data_lock);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "entry self-heal as only %d sub-volumes could "
- "be locked in %s domain",
- uuid_utoa (fd->inode->gfid), ret,
- this->name);
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_entry_prepare (frame, this, fd->inode,
- data_lock, sources, sinks,
- healed_sinks,
- locked_replies, &source,
- NULL);
- if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
- did_sh = _gf_false;
- goto unlock;
- }
-
- local->need_full_crawl = afr_need_full_heal (this,
- locked_replies,
- source,
- healed_sinks,
- AFR_ENTRY_TRANSACTION);
- }
-unlock:
- afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
- data_lock, NULL);
- if (ret < 0)
- goto out;
+ int ret = -1;
+ int source = -1;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *data_lock = NULL;
+ unsigned char *postop_lock = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t did_sh = _gf_true;
+
+ priv = this->private;
+ local = frame->local;
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+ data_lock = alloca0(priv->child_count);
+ postop_lock = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = afr_selfheal_entrylk(frame, this, fd->inode, this->name, NULL,
+ data_lock);
+ {
+ if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "entry self-heal as only %d sub-volumes could "
+ "be locked in %s domain",
+ uuid_utoa(fd->inode->gfid), ret, this->name);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
- if (!did_sh)
- goto out;
+ ret = __afr_selfheal_entry_prepare(frame, this, fd->inode, data_lock,
+ sources, sinks, healed_sinks,
+ locked_replies, &source, NULL);
+ if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
+ did_sh = _gf_false;
+ goto unlock;
+ }
- ret = afr_selfheal_entry_do (frame, this, fd, source, sources,
- healed_sinks);
- if (ret)
- goto out;
-
- /* Take entrylks in xlator domain before doing post-op (undo-pending) in
- * entry self-heal. This is to prevent a parallel name self-heal on
- * an entry under @fd->inode from reading pending xattrs while it is
- * being modified by SHD after entry sh below, given that
- * name self-heal takes locks ONLY in xlator domain and is free to read
- * pending changelog in the absence of the following locking.
- */
- ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
- postop_lock);
- {
- if (AFR_CMP (data_lock, postop_lock, priv->child_count) != 0) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "post-op after entry self-heal as %d "
- "sub-volumes, as opposed to %d, "
- "could be locked in %s domain",
- uuid_utoa (fd->inode->gfid),
- ret, AFR_COUNT (data_lock,
- priv->child_count), this->name);
- ret = -ENOTCONN;
- goto postop_unlock;
- }
-
- ret = afr_selfheal_undo_pending (frame, this, fd->inode,
- sources, sinks, healed_sinks,
- undid_pending,
- AFR_ENTRY_TRANSACTION,
- locked_replies, postop_lock);
+ local->need_full_crawl = afr_need_full_heal(
+ this, locked_replies, source, healed_sinks, AFR_ENTRY_TRANSACTION);
+ }
+unlock:
+ afr_selfheal_unentrylk(frame, this, fd->inode, this->name, NULL, data_lock,
+ NULL);
+ if (ret < 0)
+ goto out;
+
+ if (!did_sh)
+ goto out;
+
+ ret = afr_selfheal_entry_do(frame, this, fd, source, sources, healed_sinks);
+ if (ret)
+ goto out;
+
+ /* Take entrylks in xlator domain before doing post-op (undo-pending) in
+ * entry self-heal. This is to prevent a parallel name self-heal on
+ * an entry under @fd->inode from reading pending xattrs while it is
+ * being modified by SHD after entry sh below, given that
+ * name self-heal takes locks ONLY in xlator domain and is free to read
+ * pending changelog in the absence of the following locking.
+ */
+ ret = afr_selfheal_entrylk(frame, this, fd->inode, this->name, NULL,
+ postop_lock);
+ {
+ if (AFR_CMP(data_lock, postop_lock, priv->child_count) != 0) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "post-op after entry self-heal as %d "
+ "sub-volumes, as opposed to %d, "
+ "could be locked in %s domain",
+ uuid_utoa(fd->inode->gfid), ret,
+ AFR_COUNT(data_lock, priv->child_count), this->name);
+ ret = -ENOTCONN;
+ goto postop_unlock;
}
+
+ ret = afr_selfheal_undo_pending(
+ frame, this, fd->inode, sources, sinks, healed_sinks, undid_pending,
+ AFR_ENTRY_TRANSACTION, locked_replies, postop_lock);
+ }
postop_unlock:
- afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
- postop_lock, NULL);
+ afr_selfheal_unentrylk(frame, this, fd->inode, this->name, NULL,
+ postop_lock, NULL);
out:
- if (did_sh)
- afr_log_selfheal (fd->inode->gfid, this, ret, "entry", source,
- sources, healed_sinks);
- else
- ret = 1;
-
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ if (did_sh)
+ afr_log_selfheal(fd->inode->gfid, this, ret, "entry", source, sources,
+ healed_sinks);
+ else
+ ret = 1;
+
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
+ return ret;
}
-
static fd_t *
-afr_selfheal_data_opendir (xlator_t *this, inode_t *inode)
+afr_selfheal_data_opendir(xlator_t *this, inode_t *inode)
{
- loc_t loc = {0,};
- int ret = 0;
- fd_t *fd = NULL;
-
- fd = fd_create (inode, 0);
- if (!fd)
- return NULL;
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- ret = syncop_opendir (this, &loc, fd, NULL, NULL);
- if (ret) {
- fd_unref (fd);
- fd = NULL;
- } else {
- fd_bind (fd);
- }
-
- loc_wipe (&loc);
- return fd;
+ loc_t loc = {
+ 0,
+ };
+ int ret = 0;
+ fd_t *fd = NULL;
+
+ fd = fd_create(inode, 0);
+ if (!fd)
+ return NULL;
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ ret = syncop_opendir(this, &loc, fd, NULL, NULL);
+ if (ret) {
+ fd_unref(fd);
+ fd = NULL;
+ } else {
+ fd_bind(fd);
+ }
+
+ loc_wipe(&loc);
+ return fd;
}
-
int
-afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode)
+afr_selfheal_entry(call_frame_t *frame, xlator_t *this, inode_t *inode)
{
- afr_private_t *priv = NULL;
- unsigned char *locked_on = NULL;
- fd_t *fd = NULL;
- int ret = 0;
-
- priv = this->private;
-
- fd = afr_selfheal_data_opendir (this, inode);
- if (!fd)
- return -EIO;
-
- locked_on = alloca0 (priv->child_count);
-
- ret = afr_selfheal_tie_breaker_entrylk (frame, this, inode,
- priv->sh_domain, NULL,
- locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "entry self-heal as only %d sub-volumes could "
- "be locked in %s domain",
- uuid_utoa (fd->inode->gfid), ret,
- priv->sh_domain);
- /* Either less than two subvols available, or another
- selfheal (from another server) is in progress. Skip
- for now in any case there isn't anything to do.
- */
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_entry (frame, this, fd, locked_on);
- }
+ afr_private_t *priv = NULL;
+ unsigned char *locked_on = NULL;
+ fd_t *fd = NULL;
+ int ret = 0;
+
+ priv = this->private;
+
+ fd = afr_selfheal_data_opendir(this, inode);
+ if (!fd)
+ return -EIO;
+
+ locked_on = alloca0(priv->child_count);
+
+ ret = afr_selfheal_tie_breaker_entrylk(frame, this, inode, priv->sh_domain,
+ NULL, locked_on);
+ {
+ if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "entry self-heal as only %d sub-volumes could "
+ "be locked in %s domain",
+ uuid_utoa(fd->inode->gfid), ret, priv->sh_domain);
+ /* Either less than two subvols available, or another
+ selfheal (from another server) is in progress. Skip
+ for now in any case there isn't anything to do.
+ */
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_entry(frame, this, fd, locked_on);
+ }
unlock:
- afr_selfheal_unentrylk (frame, this, inode, priv->sh_domain, NULL,
- locked_on, NULL);
+ afr_selfheal_unentrylk(frame, this, inode, priv->sh_domain, NULL, locked_on,
+ NULL);
- if (fd)
- fd_unref (fd);
+ if (fd)
+ fd_unref(fd);
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index 199f8961480..be6e574b6ca 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -8,109 +8,108 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
#include "byte-order.h"
#include "protocol-common.h"
#include "events.h"
-#define AFR_HEAL_ATTR (GF_SET_ATTR_UID|GF_SET_ATTR_GID|GF_SET_ATTR_MODE)
+#define AFR_HEAL_ATTR (GF_SET_ATTR_UID | GF_SET_ATTR_GID | GF_SET_ATTR_MODE)
static gf_boolean_t
-_afr_ignorable_key_match (dict_t *d, char *k, data_t *val, void *mdata)
+_afr_ignorable_key_match(dict_t *d, char *k, data_t *val, void *mdata)
{
- return afr_is_xattr_ignorable (k);
+ return afr_is_xattr_ignorable(k);
}
void
-afr_delete_ignorable_xattrs (dict_t *xattr)
+afr_delete_ignorable_xattrs(dict_t *xattr)
{
- dict_foreach_match (xattr, _afr_ignorable_key_match, NULL,
- dict_remove_foreach_fn, NULL);
+ dict_foreach_match(xattr, _afr_ignorable_key_match, NULL,
+ dict_remove_foreach_fn, NULL);
}
int
-__afr_selfheal_metadata_do (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int source, unsigned char *healed_sinks,
- struct afr_reply *locked_replies)
+__afr_selfheal_metadata_do(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int source, unsigned char *healed_sinks,
+ struct afr_reply *locked_replies)
{
- int ret = -1;
- loc_t loc = {0,};
- dict_t *xattr = NULL;
- dict_t *old_xattr = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- priv = this->private;
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_SELF_HEAL_INFO, "performing metadata selfheal on %s",
- uuid_utoa (inode->gfid));
-
- ret = syncop_getxattr (priv->children[source], &loc, &xattr, NULL,
- NULL, NULL);
- if (ret < 0) {
- ret = -EIO;
- goto out;
- }
-
- afr_delete_ignorable_xattrs (xattr);
-
- for (i = 0; i < priv->child_count; i++) {
- if (old_xattr) {
- dict_unref (old_xattr);
- old_xattr = NULL;
- }
-
- if (!healed_sinks[i])
- continue;
-
- ret = syncop_setattr (priv->children[i], &loc,
- &locked_replies[source].poststat,
- AFR_HEAL_ATTR, NULL, NULL, NULL, NULL);
- if (ret)
- healed_sinks[i] = 0;
-
- ret = syncop_getxattr (priv->children[i], &loc, &old_xattr, 0,
- NULL, NULL);
- if (old_xattr) {
- afr_delete_ignorable_xattrs (old_xattr);
- ret = syncop_removexattr (priv->children[i], &loc, "",
- old_xattr, NULL);
- if (ret)
- healed_sinks[i] = 0;
- }
-
- ret = syncop_setxattr (priv->children[i], &loc, xattr, 0, NULL,
- NULL);
- if (ret)
- healed_sinks[i] = 0;
- }
- ret = 0;
+ int ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ dict_t *old_xattr = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ priv = this->private;
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "performing metadata selfheal on %s", uuid_utoa(inode->gfid));
+
+ ret = syncop_getxattr(priv->children[source], &loc, &xattr, NULL, NULL,
+ NULL);
+ if (ret < 0) {
+ ret = -EIO;
+ goto out;
+ }
+
+ afr_delete_ignorable_xattrs(xattr);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (old_xattr) {
+ dict_unref(old_xattr);
+ old_xattr = NULL;
+ }
+
+ if (!healed_sinks[i])
+ continue;
+
+ ret = syncop_setattr(priv->children[i], &loc,
+ &locked_replies[source].poststat, AFR_HEAL_ATTR,
+ NULL, NULL, NULL, NULL);
+ if (ret)
+ healed_sinks[i] = 0;
+
+ ret = syncop_getxattr(priv->children[i], &loc, &old_xattr, 0, NULL,
+ NULL);
+ if (old_xattr) {
+ afr_delete_ignorable_xattrs(old_xattr);
+ ret = syncop_removexattr(priv->children[i], &loc, "", old_xattr,
+ NULL);
+ if (ret)
+ healed_sinks[i] = 0;
+ }
+
+ ret = syncop_setxattr(priv->children[i], &loc, xattr, 0, NULL, NULL);
+ if (ret)
+ healed_sinks[i] = 0;
+ }
+ ret = 0;
out:
- loc_wipe (&loc);
- if (xattr)
- dict_unref (xattr);
- if (old_xattr)
- dict_unref (old_xattr);
+ loc_wipe(&loc);
+ if (xattr)
+ dict_unref(xattr);
+ if (old_xattr)
+ dict_unref(old_xattr);
- return ret;
+ return ret;
}
static uint64_t
mtime_ns(struct iatt *ia)
{
- uint64_t ret;
+ uint64_t ret;
- ret = (((uint64_t)(ia->ia_mtime)) * 1000000000)
- + (uint64_t)(ia->ia_mtime_nsec);
+ ret = (((uint64_t)(ia->ia_mtime)) * 1000000000) +
+ (uint64_t)(ia->ia_mtime_nsec);
- return ret;
+ return ret;
}
/*
@@ -123,382 +122,367 @@ mtime_ns(struct iatt *ia)
* the source with the most recent modification date.
*/
static int
-afr_dirtime_splitbrain_source (call_frame_t *frame, xlator_t *this,
- struct afr_reply *replies,
- unsigned char *locked_on)
+afr_dirtime_splitbrain_source(call_frame_t *frame, xlator_t *this,
+ struct afr_reply *replies,
+ unsigned char *locked_on)
{
- afr_private_t *priv = NULL;
- int source = -1;
- struct iatt source_ia;
- struct iatt child_ia;
- uint64_t mtime = 0;
- int i;
- int ret = -1;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!locked_on[i])
- continue;
-
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret != 0)
- continue;
-
- if (mtime_ns(&replies[i].poststat) <= mtime)
- continue;
-
- mtime = mtime_ns(&replies[i].poststat);
- source = i;
- }
-
- if (source == -1)
- goto out;
-
- source_ia = replies[source].poststat;
- if (source_ia.ia_type != IA_IFDIR)
- goto out;
-
- for (i = 0; i < priv->child_count; i++) {
- if (i == source)
- continue;
-
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret != 0)
- continue;
-
- child_ia = replies[i].poststat;
-
- if (!IA_EQUAL(source_ia, child_ia, gfid) ||
- !IA_EQUAL(source_ia, child_ia, type) ||
- !IA_EQUAL(source_ia, child_ia, prot) ||
- !IA_EQUAL(source_ia, child_ia, uid) ||
- !IA_EQUAL(source_ia, child_ia, gid) ||
- !afr_xattrs_are_equal (replies[source].xdata,
- replies[i].xdata))
- goto out;
- }
-
- /*
- * Metadata split brain is just about [amc]time
- * We return our source.
- */
- ret = source;
+ afr_private_t *priv = NULL;
+ int source = -1;
+ struct iatt source_ia;
+ struct iatt child_ia;
+ uint64_t mtime = 0;
+ int i;
+ int ret = -1;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!locked_on[i])
+ continue;
+
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret != 0)
+ continue;
+
+ if (mtime_ns(&replies[i].poststat) <= mtime)
+ continue;
+
+ mtime = mtime_ns(&replies[i].poststat);
+ source = i;
+ }
+
+ if (source == -1)
+ goto out;
+
+ source_ia = replies[source].poststat;
+ if (source_ia.ia_type != IA_IFDIR)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source)
+ continue;
+
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret != 0)
+ continue;
+
+ child_ia = replies[i].poststat;
+
+ if (!IA_EQUAL(source_ia, child_ia, gfid) ||
+ !IA_EQUAL(source_ia, child_ia, type) ||
+ !IA_EQUAL(source_ia, child_ia, prot) ||
+ !IA_EQUAL(source_ia, child_ia, uid) ||
+ !IA_EQUAL(source_ia, child_ia, gid) ||
+ !afr_xattrs_are_equal(replies[source].xdata, replies[i].xdata))
+ goto out;
+ }
+
+ /*
+ * Metadata split brain is just about [amc]time
+ * We return our source.
+ */
+ ret = source;
out:
- return ret;
+ return ret;
}
-
/*
* Look for mismatching uid/gid or mode or user xattrs even if
* AFR xattrs don't say so, and pick one arbitrarily as winner. */
static int
-__afr_selfheal_metadata_finalize_source (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- unsigned char *locked_on,
- struct afr_reply *replies)
+__afr_selfheal_metadata_finalize_source(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, unsigned char *sources,
+ unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ unsigned char *locked_on,
+ struct afr_reply *replies)
{
- int i = 0;
- afr_private_t *priv = NULL;
- struct iatt srcstat = {0, };
- int source = -1;
- int sources_count = 0;
-
- priv = this->private;
-
- sources_count = AFR_COUNT (sources, priv->child_count);
-
- if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count) {
-
- source = afr_mark_split_brain_source_sinks (frame, this, inode,
- sources, sinks,
- healed_sinks,
- locked_on, replies,
- AFR_METADATA_TRANSACTION);
- if (source >= 0) {
- _afr_fav_child_reset_sink_xattrs (frame, this, inode,
- source, healed_sinks,
- undid_pending,
- AFR_METADATA_TRANSACTION,
- locked_on, replies);
- goto out;
- }
-
- /* If this is a directory mtime/ctime only split brain
- use the most recent */
- source = afr_dirtime_splitbrain_source (frame, this,
- replies, locked_on);
- if (source != -1) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_SPLIT_BRAIN, "clear time "
- "split brain on %s",
- uuid_utoa (replies[source].poststat.ia_gfid));
- sources[source] = 1;
- healed_sinks[source] = 0;
- goto out;
- }
-
- if (!priv->metadata_splitbrain_forced_heal) {
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;"
- "type=metadata;file=%s",
- this->name, uuid_utoa(inode->gfid));
- return -EIO;
- }
-
- /* Metadata split brain, select one subvol
- arbitrarily */
- for (i = 0; i < priv->child_count; i++) {
- if (locked_on[i] && healed_sinks[i]) {
- sources[i] = 1;
- healed_sinks[i] = 0;
- break;
- }
- }
- }
-
- /* No split brain at this point. If we were called from
- * afr_heal_splitbrain_file(), abort.*/
- if (afr_dict_contains_heal_op(frame))
- return -EIO;
-
- source = afr_choose_source_by_policy (priv, sources,
- AFR_METADATA_TRANSACTION);
- srcstat = replies[source].poststat;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i] || i == source)
- continue;
- if (!IA_EQUAL (srcstat, replies[i].poststat, type) ||
- !IA_EQUAL (srcstat, replies[i].poststat, uid) ||
- !IA_EQUAL (srcstat, replies[i].poststat, gid) ||
- !IA_EQUAL (srcstat, replies[i].poststat, prot)) {
- gf_msg_debug (this->name, 0, "%s: iatt mismatch "
- "for source(%d) vs (%d)",
- uuid_utoa
- (replies[source].poststat.ia_gfid),
- source, i);
- sources[i] = 0;
- healed_sinks[i] = 1;
- }
- }
-
- for (i =0; i < priv->child_count; i++) {
- if (!sources[i] || i == source)
- continue;
- if (!afr_xattrs_are_equal (replies[source].xdata,
- replies[i].xdata)) {
- gf_msg_debug (this->name, 0, "%s: xattr mismatch "
- "for source(%d) vs (%d)",
- uuid_utoa
- (replies[source].poststat.ia_gfid),
- source, i);
- sources[i] = 0;
- healed_sinks[i] = 1;
- }
+ int i = 0;
+ afr_private_t *priv = NULL;
+ struct iatt srcstat = {
+ 0,
+ };
+ int source = -1;
+ int sources_count = 0;
+
+ priv = this->private;
+
+ sources_count = AFR_COUNT(sources, priv->child_count);
+
+ if ((AFR_CMP(locked_on, healed_sinks, priv->child_count) == 0) ||
+ !sources_count) {
+ source = afr_mark_split_brain_source_sinks(
+ frame, this, inode, sources, sinks, healed_sinks, locked_on,
+ replies, AFR_METADATA_TRANSACTION);
+ if (source >= 0) {
+ _afr_fav_child_reset_sink_xattrs(
+ frame, this, inode, source, healed_sinks, undid_pending,
+ AFR_METADATA_TRANSACTION, locked_on, replies);
+ goto out;
+ }
+
+ /* If this is a directory mtime/ctime only split brain
+ use the most recent */
+ source = afr_dirtime_splitbrain_source(frame, this, replies, locked_on);
+ if (source != -1) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SPLIT_BRAIN,
+ "clear time "
+ "split brain on %s",
+ uuid_utoa(replies[source].poststat.ia_gfid));
+ sources[source] = 1;
+ healed_sinks[source] = 0;
+ goto out;
}
+ if (!priv->metadata_splitbrain_forced_heal) {
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "subvol=%s;"
+ "type=metadata;file=%s",
+ this->name, uuid_utoa(inode->gfid));
+ return -EIO;
+ }
+
+ /* Metadata split brain, select one subvol
+ arbitrarily */
+ for (i = 0; i < priv->child_count; i++) {
+ if (locked_on[i] && healed_sinks[i]) {
+ sources[i] = 1;
+ healed_sinks[i] = 0;
+ break;
+ }
+ }
+ }
+
+ /* No split brain at this point. If we were called from
+ * afr_heal_splitbrain_file(), abort.*/
+ if (afr_dict_contains_heal_op(frame))
+ return -EIO;
+
+ source = afr_choose_source_by_policy(priv, sources,
+ AFR_METADATA_TRANSACTION);
+ srcstat = replies[source].poststat;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i] || i == source)
+ continue;
+ if (!IA_EQUAL(srcstat, replies[i].poststat, type) ||
+ !IA_EQUAL(srcstat, replies[i].poststat, uid) ||
+ !IA_EQUAL(srcstat, replies[i].poststat, gid) ||
+ !IA_EQUAL(srcstat, replies[i].poststat, prot)) {
+ gf_msg_debug(this->name, 0,
+ "%s: iatt mismatch "
+ "for source(%d) vs (%d)",
+ uuid_utoa(replies[source].poststat.ia_gfid), source,
+ i);
+ sources[i] = 0;
+ healed_sinks[i] = 1;
+ }
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i] || i == source)
+ continue;
+ if (!afr_xattrs_are_equal(replies[source].xdata, replies[i].xdata)) {
+ gf_msg_debug(this->name, 0,
+ "%s: xattr mismatch "
+ "for source(%d) vs (%d)",
+ uuid_utoa(replies[source].poststat.ia_gfid), source,
+ i);
+ sources[i] = 0;
+ healed_sinks[i] = 1;
+ }
+ }
+
out:
- afr_mark_active_sinks (this, sources, locked_on, healed_sinks);
- return source;
+ afr_mark_active_sinks(this, sources, locked_on, healed_sinks);
+ return source;
}
-
int
-__afr_selfheal_metadata_prepare (call_frame_t *frame, xlator_t *this, inode_t *inode,
- unsigned char *locked_on, unsigned char *sources,
- unsigned char *sinks, unsigned char *healed_sinks,
- unsigned char *undid_pending,
- struct afr_reply *replies, gf_boolean_t *pflag)
+__afr_selfheal_metadata_prepare(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ struct afr_reply *replies, gf_boolean_t *pflag)
{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- int i = 0;
- uint64_t *witness = NULL;
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ uint64_t *witness = NULL;
- priv = this->private;
+ priv = this->private;
- ret = afr_selfheal_unlocked_discover (frame, inode, inode->gfid,
- replies);
- if (ret)
- return ret;
-
- witness = alloca0 (sizeof (*witness) * priv->child_count);
- ret = afr_selfheal_find_direction (frame, this, replies,
- AFR_METADATA_TRANSACTION,
- locked_on, sources, sinks, witness,
- pflag);
- if (ret)
- return ret;
-
- /* Initialize the healed_sinks[] array optimistically to
- the intersection of to-be-healed (i.e sinks[]) and
- the list of servers which are up (i.e locked_on[]).
-
- As we encounter failures in the healing process, we
- will unmark the respective servers in the healed_sinks[]
- array.
- */
- AFR_INTERSECT (healed_sinks, sinks, locked_on, priv->child_count);
-
- /* If any source has witness, pick first
- * witness source and make everybody else sinks */
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i] && witness[i]) {
- source = i;
- break;
- }
- }
+ ret = afr_selfheal_unlocked_discover(frame, inode, inode->gfid, replies);
+ if (ret)
+ return ret;
- if (source != -1) {
- for (i = 0; i < priv->child_count; i++) {
- if (i != source && sources[i]) {
- sources[i] = 0;
- healed_sinks[i] = 1;
- }
- }
+ witness = alloca0(sizeof(*witness) * priv->child_count);
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_METADATA_TRANSACTION, locked_on,
+ sources, sinks, witness, pflag);
+ if (ret)
+ return ret;
+
+ /* Initialize the healed_sinks[] array optimistically to
+ the intersection of to-be-healed (i.e sinks[]) and
+ the list of servers which are up (i.e locked_on[]).
+
+ As we encounter failures in the healing process, we
+ will unmark the respective servers in the healed_sinks[]
+ array.
+ */
+ AFR_INTERSECT(healed_sinks, sinks, locked_on, priv->child_count);
+
+ /* If any source has witness, pick first
+ * witness source and make everybody else sinks */
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i] && witness[i]) {
+ source = i;
+ break;
}
+ }
- source = __afr_selfheal_metadata_finalize_source (frame, this, inode,
- sources, sinks,
- healed_sinks,
- undid_pending,
- locked_on, replies);
+ if (source != -1) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (i != source && sources[i]) {
+ sources[i] = 0;
+ healed_sinks[i] = 1;
+ }
+ }
+ }
- if (source < 0)
- return -EIO;
+ source = __afr_selfheal_metadata_finalize_source(
+ frame, this, inode, sources, sinks, healed_sinks, undid_pending,
+ locked_on, replies);
- return source;
-}
+ if (source < 0)
+ return -EIO;
-int
-afr_selfheal_metadata (call_frame_t *frame, xlator_t *this, inode_t *inode)
-{
- afr_private_t *priv = NULL;
- int ret = -1;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *data_lock = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
- gf_boolean_t did_sh = _gf_true;
- int source = -1;
-
- priv = this->private;
-
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
- data_lock = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, inode, this->name,
- LLONG_MAX - 1, 0, data_lock);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_metadata_prepare (frame, this, inode,
- data_lock, sources,
- sinks, healed_sinks,
- undid_pending,
- locked_replies, NULL);
- if (ret < 0)
- goto unlock;
-
- source = ret;
-
- if (AFR_COUNT (healed_sinks, priv->child_count) == 0) {
- did_sh = _gf_false;
- goto unlock;
- }
-
- ret = __afr_selfheal_metadata_do (frame, this, inode, source,
- healed_sinks, locked_replies);
- if (ret)
- goto unlock;
-
- /* Restore atime/mtime for files that don't need data heal as
- * restoring timestamps happens only as a part of data-heal.
- */
- if (!IA_ISREG (locked_replies[source].poststat.ia_type))
- afr_selfheal_restore_time (frame, this, inode, source,
- healed_sinks, locked_replies);
-
- ret = afr_selfheal_undo_pending (frame, this, inode, sources,
- sinks, healed_sinks,
- undid_pending,
- AFR_METADATA_TRANSACTION,
- locked_replies, data_lock);
- }
-unlock:
- afr_selfheal_uninodelk (frame, this, inode, this->name,
- LLONG_MAX -1, 0, data_lock);
-
- if (did_sh)
- afr_log_selfheal (inode->gfid, this, ret, "metadata", source,
- sources, healed_sinks);
- else
- ret = 1;
-
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ return source;
}
int
-afr_selfheal_metadata_by_stbuf (xlator_t *this, struct iatt *stbuf)
+afr_selfheal_metadata(call_frame_t *frame, xlator_t *this, inode_t *inode)
{
- inode_t *inode = NULL;
- inode_t *link_inode = NULL;
- call_frame_t *frame = NULL;
- int ret = 0;
-
- if (gf_uuid_is_null (stbuf->ia_gfid)) {
- ret = -EINVAL;
- goto out;
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *data_lock = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+ gf_boolean_t did_sh = _gf_true;
+ int source = -1;
+
+ priv = this->private;
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+ data_lock = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = afr_selfheal_inodelk(frame, this, inode, this->name, LLONG_MAX - 1, 0,
+ data_lock);
+ {
+ if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ ret = -ENOTCONN;
+ goto unlock;
}
- inode = inode_new (this->itable);
- if (!inode) {
- ret = -ENOMEM;
- goto out;
- }
+ ret = __afr_selfheal_metadata_prepare(
+ frame, this, inode, data_lock, sources, sinks, healed_sinks,
+ undid_pending, locked_replies, NULL);
+ if (ret < 0)
+ goto unlock;
- link_inode = inode_link (inode, NULL, NULL, stbuf);
- if (!link_inode) {
- ret = -ENOMEM;
- goto out;
- }
+ source = ret;
- frame = afr_frame_create (this, &ret);
- if (!frame) {
- ret = -ret;
- goto out;
+ if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
+ did_sh = _gf_false;
+ goto unlock;
}
- ret = afr_selfheal_metadata (frame, this, link_inode);
+ ret = __afr_selfheal_metadata_do(frame, this, inode, source,
+ healed_sinks, locked_replies);
+ if (ret)
+ goto unlock;
+
+ /* Restore atime/mtime for files that don't need data heal as
+ * restoring timestamps happens only as a part of data-heal.
+ */
+ if (!IA_ISREG(locked_replies[source].poststat.ia_type))
+ afr_selfheal_restore_time(frame, this, inode, source, healed_sinks,
+ locked_replies);
+
+ ret = afr_selfheal_undo_pending(
+ frame, this, inode, sources, sinks, healed_sinks, undid_pending,
+ AFR_METADATA_TRANSACTION, locked_replies, data_lock);
+ }
+unlock:
+ afr_selfheal_uninodelk(frame, this, inode, this->name, LLONG_MAX - 1, 0,
+ data_lock);
+
+ if (did_sh)
+ afr_log_selfheal(inode->gfid, this, ret, "metadata", source, sources,
+ healed_sinks);
+ else
+ ret = 1;
+
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
+ return ret;
+}
+
+int
+afr_selfheal_metadata_by_stbuf(xlator_t *this, struct iatt *stbuf)
+{
+ inode_t *inode = NULL;
+ inode_t *link_inode = NULL;
+ call_frame_t *frame = NULL;
+ int ret = 0;
+
+ if (gf_uuid_is_null(stbuf->ia_gfid)) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ inode = inode_new(this->itable);
+ if (!inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ link_inode = inode_link(inode, NULL, NULL, stbuf);
+ if (!link_inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ frame = afr_frame_create(this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+
+ ret = afr_selfheal_metadata(frame, this, link_inode);
out:
- if (inode)
- inode_unref (inode);
- if (link_inode)
- inode_unref (link_inode);
- if (frame)
- AFR_STACK_DESTROY (frame);
- return ret;
+ if (inode)
+ inode_unref(inode);
+ if (link_inode)
+ inode_unref(link_inode);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-name.c b/xlators/cluster/afr/src/afr-self-heal-name.c
index 0a5be29d5ee..39aacee6ecf 100644
--- a/xlators/cluster/afr/src/afr-self-heal-name.c
+++ b/xlators/cluster/afr/src/afr-self-heal-name.c
@@ -8,696 +8,667 @@
cases as published by the Free Software Foundation.
*/
-
#include "events.h"
#include "afr.h"
#include "afr-self-heal.h"
#include "afr-messages.h"
int
-__afr_selfheal_assign_gfid (xlator_t *this, inode_t *parent, uuid_t pargfid,
- const char *bname, inode_t *inode,
- struct afr_reply *replies, void *gfid,
- unsigned char *locked_on, int source,
- unsigned char *sources, gf_boolean_t is_gfid_absent)
+__afr_selfheal_assign_gfid(xlator_t *this, inode_t *parent, uuid_t pargfid,
+ const char *bname, inode_t *inode,
+ struct afr_reply *replies, void *gfid,
+ unsigned char *locked_on, int source,
+ unsigned char *sources, gf_boolean_t is_gfid_absent)
{
- int ret = 0;
- int up_count = 0;
- int locked_count = 0;
- afr_private_t *priv = NULL;
+ int ret = 0;
+ int up_count = 0;
+ int locked_count = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- gf_uuid_copy (parent->gfid, pargfid);
+ gf_uuid_copy(parent->gfid, pargfid);
- if (is_gfid_absent) {
- /* Ensure all children of AFR are up before performing gfid heal, to
- * guard against the possibility of gfid split brain. */
+ if (is_gfid_absent) {
+ /* Ensure all children of AFR are up before performing gfid heal, to
+ * guard against the possibility of gfid split brain. */
- up_count = AFR_COUNT (priv->child_up, priv->child_count);
- if (up_count != priv->child_count) {
- ret = -EIO;
- goto out;
- }
+ up_count = AFR_COUNT(priv->child_up, priv->child_count);
+ if (up_count != priv->child_count) {
+ ret = -EIO;
+ goto out;
+ }
- locked_count = AFR_COUNT (locked_on, priv->child_count);
- if (locked_count != priv->child_count) {
- ret = -EIO;
- goto out;
- }
+ locked_count = AFR_COUNT(locked_on, priv->child_count);
+ if (locked_count != priv->child_count) {
+ ret = -EIO;
+ goto out;
}
+ }
- afr_lookup_and_heal_gfid (this, parent, bname, inode, replies, source,
- sources, gfid);
+ afr_lookup_and_heal_gfid(this, parent, bname, inode, replies, source,
+ sources, gfid);
out:
- return ret;
+ return ret;
}
int
-__afr_selfheal_name_impunge (call_frame_t *frame, xlator_t *this,
- inode_t *parent, uuid_t pargfid,
- const char *bname, inode_t *inode,
- struct afr_reply *replies, int gfid_idx)
+__afr_selfheal_name_impunge(call_frame_t *frame, xlator_t *this,
+ inode_t *parent, uuid_t pargfid, const char *bname,
+ inode_t *inode, struct afr_reply *replies,
+ int gfid_idx)
{
- int i = 0;
- afr_private_t *priv = NULL;
- int ret = 0;
- unsigned char *sources = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ unsigned char *sources = NULL;
- priv = this->private;
+ priv = this->private;
- sources = alloca0 (priv->child_count);
+ sources = alloca0(priv->child_count);
- gf_uuid_copy (parent->gfid, pargfid);
+ gf_uuid_copy(parent->gfid, pargfid);
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[gfid_idx].poststat.ia_gfid) == 0) {
- sources[i] = 1;
- continue;
- }
+ if (gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[gfid_idx].poststat.ia_gfid) == 0) {
+ sources[i] = 1;
+ continue;
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i])
+ continue;
- ret |= afr_selfheal_recreate_entry (frame, i, gfid_idx, sources,
- parent, bname, inode,
- replies);
- }
+ ret |= afr_selfheal_recreate_entry(frame, i, gfid_idx, sources, parent,
+ bname, inode, replies);
+ }
- return ret;
+ return ret;
}
-
int
-__afr_selfheal_name_expunge (xlator_t *this, inode_t *parent, uuid_t pargfid,
- const char *bname, inode_t *inode,
- struct afr_reply *replies)
+__afr_selfheal_name_expunge(xlator_t *this, inode_t *parent, uuid_t pargfid,
+ const char *bname, inode_t *inode,
+ struct afr_reply *replies)
{
- loc_t loc = {0, };
- int i = 0;
- afr_private_t *priv = NULL;
- char g[64];
- int ret = 0;
-
- priv = this->private;
-
- loc.parent = inode_ref (parent);
- gf_uuid_copy (loc.pargfid, pargfid);
- loc.name = bname;
- loc.inode = inode_ref (inode);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret)
- continue;
-
- switch (replies[i].poststat.ia_type) {
- case IA_IFDIR:
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- "expunging dir %s/%s (%s) on %s",
- uuid_utoa (pargfid), bname,
- uuid_utoa_r (replies[i].poststat.ia_gfid, g),
- priv->children[i]->name);
-
- ret |= syncop_rmdir (priv->children[i], &loc, 1, NULL,
- NULL);
- break;
- default:
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- "expunging file %s/%s (%s) on %s",
- uuid_utoa (pargfid), bname,
- uuid_utoa_r (replies[i].poststat.ia_gfid, g),
- priv->children[i]->name);
-
- ret |= syncop_unlink (priv->children[i], &loc, NULL,
- NULL);
- break;
- }
- }
-
- loc_wipe (&loc);
-
- return ret;
+ loc_t loc = {
+ 0,
+ };
+ int i = 0;
+ afr_private_t *priv = NULL;
+ char g[64];
+ int ret = 0;
+
+ priv = this->private;
+
+ loc.parent = inode_ref(parent);
+ gf_uuid_copy(loc.pargfid, pargfid);
+ loc.name = bname;
+ loc.inode = inode_ref(inode);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret)
+ continue;
+
+ switch (replies[i].poststat.ia_type) {
+ case IA_IFDIR:
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "expunging dir %s/%s (%s) on %s", uuid_utoa(pargfid),
+ bname, uuid_utoa_r(replies[i].poststat.ia_gfid, g),
+ priv->children[i]->name);
+
+ ret |= syncop_rmdir(priv->children[i], &loc, 1, NULL, NULL);
+ break;
+ default:
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "expunging file %s/%s (%s) on %s", uuid_utoa(pargfid),
+ bname, uuid_utoa_r(replies[i].poststat.ia_gfid, g),
+ priv->children[i]->name);
+
+ ret |= syncop_unlink(priv->children[i], &loc, NULL, NULL);
+ break;
+ }
+ }
+ loc_wipe(&loc);
+
+ return ret;
}
/* This function is to be called after ensuring that there is no gfid mismatch
* for the inode across multiple sources
*/
static int
-afr_selfheal_gfid_idx_get (xlator_t *this, struct afr_reply *replies,
- unsigned char *sources)
+afr_selfheal_gfid_idx_get(xlator_t *this, struct afr_reply *replies,
+ unsigned char *sources)
{
- int i = 0;
- int gfid_idx = -1;
- afr_private_t *priv = NULL;
+ int i = 0;
+ int gfid_idx = -1;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
- if (!sources[i])
- continue;
+ if (!sources[i])
+ continue;
- if (gf_uuid_is_null (replies[i].poststat.ia_gfid))
- continue;
+ if (gf_uuid_is_null(replies[i].poststat.ia_gfid))
+ continue;
- gfid_idx = i;
- break;
- }
- return gfid_idx;
+ gfid_idx = i;
+ break;
+ }
+ return gfid_idx;
}
static gf_boolean_t
-afr_selfheal_name_need_heal_check (xlator_t *this, struct afr_reply *replies)
+afr_selfheal_name_need_heal_check(xlator_t *this, struct afr_reply *replies)
{
- int i = 0;
- int first_idx = -1;
- gf_boolean_t need_heal = _gf_false;
- afr_private_t *priv = NULL;
-
- priv = this->private;
+ int i = 0;
+ int first_idx = -1;
+ gf_boolean_t need_heal = _gf_false;
+ afr_private_t *priv = NULL;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
+ priv = this->private;
- if ((replies[i].op_ret == -1) &&
- (replies[i].op_errno == ENODATA))
- need_heal = _gf_true;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- if (first_idx == -1) {
- first_idx = i;
- continue;
- }
+ if ((replies[i].op_ret == -1) && (replies[i].op_errno == ENODATA))
+ need_heal = _gf_true;
- if (replies[i].op_ret != replies[first_idx].op_ret)
- need_heal = _gf_true;
+ if (first_idx == -1) {
+ first_idx = i;
+ continue;
+ }
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[first_idx].poststat.ia_gfid))
- need_heal = _gf_true;
+ if (replies[i].op_ret != replies[first_idx].op_ret)
+ need_heal = _gf_true;
- if ((replies[i].op_ret == 0) &&
- (gf_uuid_is_null(replies[i].poststat.ia_gfid)))
- need_heal = _gf_true;
+ if (gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[first_idx].poststat.ia_gfid))
+ need_heal = _gf_true;
- }
+ if ((replies[i].op_ret == 0) &&
+ (gf_uuid_is_null(replies[i].poststat.ia_gfid)))
+ need_heal = _gf_true;
+ }
- return need_heal;
+ return need_heal;
}
static int
-afr_selfheal_name_type_mismatch_check (xlator_t *this, struct afr_reply *replies,
- int source, unsigned char *sources,
- uuid_t pargfid, const char *bname)
+afr_selfheal_name_type_mismatch_check(xlator_t *this, struct afr_reply *replies,
+ int source, unsigned char *sources,
+ uuid_t pargfid, const char *bname)
{
- int i = 0;
- int type_idx = -1;
- ia_type_t inode_type = IA_INVAL;
- ia_type_t inode_type1 = IA_INVAL;
- afr_private_t *priv = NULL;
+ int i = 0;
+ int type_idx = -1;
+ ia_type_t inode_type = IA_INVAL;
+ ia_type_t inode_type1 = IA_INVAL;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
- if (replies[i].poststat.ia_type == IA_INVAL)
- continue;
+ if (replies[i].poststat.ia_type == IA_INVAL)
+ continue;
- if (inode_type == IA_INVAL) {
- inode_type = replies[i].poststat.ia_type;
- type_idx = i;
- continue;
- }
- inode_type1 = replies[i].poststat.ia_type;
- if (sources[i] || source == -1) {
- if ((sources[type_idx] || source == -1) &&
- (inode_type != inode_type1)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SPLIT_BRAIN,
- "Type mismatch for <gfid:%s>/%s: "
- "%s on %s and %s on %s",
- uuid_utoa(pargfid), bname,
- gf_inode_type_to_str (inode_type1),
- priv->children[i]->name,
- gf_inode_type_to_str (inode_type),
- priv->children[type_idx]->name);
- gf_event (EVENT_AFR_SPLIT_BRAIN,
- "subvol=%s;type=file;"
- "file=<gfid:%s>/%s;count=2;"
- "child-%d=%s;type-%d=%s;child-%d=%s;"
- "type-%d=%s", this->name,
- uuid_utoa (pargfid), bname, i,
- priv->children[i]->name, i,
- gf_inode_type_to_str (inode_type1),
- type_idx,
- priv->children[type_idx]->name,
- type_idx,
- gf_inode_type_to_str (inode_type));
- return -EIO;
- }
- inode_type = replies[i].poststat.ia_type;
- type_idx = i;
- }
+ if (inode_type == IA_INVAL) {
+ inode_type = replies[i].poststat.ia_type;
+ type_idx = i;
+ continue;
}
- return 0;
+ inode_type1 = replies[i].poststat.ia_type;
+ if (sources[i] || source == -1) {
+ if ((sources[type_idx] || source == -1) &&
+ (inode_type != inode_type1)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SPLIT_BRAIN,
+ "Type mismatch for <gfid:%s>/%s: "
+ "%s on %s and %s on %s",
+ uuid_utoa(pargfid), bname,
+ gf_inode_type_to_str(inode_type1),
+ priv->children[i]->name,
+ gf_inode_type_to_str(inode_type),
+ priv->children[type_idx]->name);
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "subvol=%s;type=file;"
+ "file=<gfid:%s>/%s;count=2;"
+ "child-%d=%s;type-%d=%s;child-%d=%s;"
+ "type-%d=%s",
+ this->name, uuid_utoa(pargfid), bname, i,
+ priv->children[i]->name, i,
+ gf_inode_type_to_str(inode_type1), type_idx,
+ priv->children[type_idx]->name, type_idx,
+ gf_inode_type_to_str(inode_type));
+ return -EIO;
+ }
+ inode_type = replies[i].poststat.ia_type;
+ type_idx = i;
+ }
+ }
+ return 0;
}
static int
-afr_selfheal_name_gfid_mismatch_check (xlator_t *this, struct afr_reply *replies,
- int source, unsigned char *sources,
- int *gfid_idx, uuid_t pargfid,
- const char *bname, inode_t *inode,
- unsigned char *locked_on, dict_t *xdata)
+afr_selfheal_name_gfid_mismatch_check(xlator_t *this, struct afr_reply *replies,
+ int source, unsigned char *sources,
+ int *gfid_idx, uuid_t pargfid,
+ const char *bname, inode_t *inode,
+ unsigned char *locked_on, dict_t *xdata)
{
- int i = 0;
- int gfid_idx_iter = -1;
- int ret = -1;
- void *gfid = NULL;
- void *gfid1 = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
-
- if (gf_uuid_is_null (replies[i].poststat.ia_gfid))
- continue;
-
- if (!gfid) {
- gfid = &replies[i].poststat.ia_gfid;
- gfid_idx_iter = i;
- continue;
- }
-
- gfid1 = &replies[i].poststat.ia_gfid;
- if (sources[i] || source == -1) {
- if ((sources[gfid_idx_iter] || source == -1) &&
- gf_uuid_compare (gfid, gfid1)) {
- ret = afr_gfid_split_brain_source (this,
- replies,
- inode,
- pargfid,
- bname,
- gfid_idx_iter,
- i, locked_on,
- gfid_idx,
- xdata);
- if (!ret && *gfid_idx >= 0) {
- ret = dict_set_str (xdata,
- "gfid-heal-msg",
- "GFID split-brain "
- "resolved");
- if (ret)
- gf_msg (this->name,
- GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED,
- "Error setting gfid-"
- "heal-msg dict");
- }
- return ret;
- }
- gfid = &replies[i].poststat.ia_gfid;
- gfid_idx_iter = i;
- }
- }
-
- *gfid_idx = gfid_idx_iter;
- return 0;
+ int i = 0;
+ int gfid_idx_iter = -1;
+ int ret = -1;
+ void *gfid = NULL;
+ void *gfid1 = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
+
+ if (gf_uuid_is_null(replies[i].poststat.ia_gfid))
+ continue;
+
+ if (!gfid) {
+ gfid = &replies[i].poststat.ia_gfid;
+ gfid_idx_iter = i;
+ continue;
+ }
+
+ gfid1 = &replies[i].poststat.ia_gfid;
+ if (sources[i] || source == -1) {
+ if ((sources[gfid_idx_iter] || source == -1) &&
+ gf_uuid_compare(gfid, gfid1)) {
+ ret = afr_gfid_split_brain_source(this, replies, inode, pargfid,
+ bname, gfid_idx_iter, i,
+ locked_on, gfid_idx, xdata);
+ if (!ret && *gfid_idx >= 0) {
+ ret = dict_set_str(xdata, "gfid-heal-msg",
+ "GFID split-brain "
+ "resolved");
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_DICT_SET_FAILED,
+ "Error setting gfid-"
+ "heal-msg dict");
+ }
+ return ret;
+ }
+ gfid = &replies[i].poststat.ia_gfid;
+ gfid_idx_iter = i;
+ }
+ }
+
+ *gfid_idx = gfid_idx_iter;
+ return 0;
}
static gf_boolean_t
-afr_selfheal_name_source_empty_check (xlator_t *this, struct afr_reply *replies,
- unsigned char *sources, int source)
+afr_selfheal_name_source_empty_check(xlator_t *this, struct afr_reply *replies,
+ unsigned char *sources, int source)
{
- int i = 0;
- afr_private_t *priv = NULL;
- gf_boolean_t source_is_empty = _gf_true;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ gf_boolean_t source_is_empty = _gf_true;
- priv = this->private;
+ priv = this->private;
- if (source == -1) {
- source_is_empty = _gf_false;
- goto out;
- }
+ if (source == -1) {
+ source_is_empty = _gf_false;
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
- if (replies[i].op_ret == -1 && replies[i].op_errno == ENOENT)
- continue;
+ if (replies[i].op_ret == -1 && replies[i].op_errno == ENOENT)
+ continue;
- source_is_empty = _gf_false;
- break;
- }
+ source_is_empty = _gf_false;
+ break;
+ }
out:
- return source_is_empty;
+ return source_is_empty;
}
int
-__afr_selfheal_name_do (call_frame_t *frame, xlator_t *this, inode_t *parent,
- uuid_t pargfid, const char *bname, inode_t *inode,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks, int source,
- unsigned char *locked_on, struct afr_reply *replies,
- void *gfid_req, dict_t *xdata)
+__afr_selfheal_name_do(call_frame_t *frame, xlator_t *this, inode_t *parent,
+ uuid_t pargfid, const char *bname, inode_t *inode,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks, int source,
+ unsigned char *locked_on, struct afr_reply *replies,
+ void *gfid_req, dict_t *xdata)
{
- int gfid_idx = -1;
- int ret = -1;
- void *gfid = NULL;
- gf_boolean_t source_is_empty = _gf_true;
- gf_boolean_t need_heal = _gf_false;
- gf_boolean_t is_gfid_absent = _gf_false;
-
- need_heal = afr_selfheal_name_need_heal_check (this, replies);
- if (!need_heal)
- return 0;
-
- source_is_empty = afr_selfheal_name_source_empty_check (this, replies,
- sources,
- source);
- if (source_is_empty) {
- ret = __afr_selfheal_name_expunge (this, parent, pargfid,
- bname, inode, replies);
- if (ret == -EIO)
- ret = -1;
- return ret;
- }
+ int gfid_idx = -1;
+ int ret = -1;
+ void *gfid = NULL;
+ gf_boolean_t source_is_empty = _gf_true;
+ gf_boolean_t need_heal = _gf_false;
+ gf_boolean_t is_gfid_absent = _gf_false;
+
+ need_heal = afr_selfheal_name_need_heal_check(this, replies);
+ if (!need_heal)
+ return 0;
- ret = afr_selfheal_name_type_mismatch_check (this, replies, source,
- sources, pargfid, bname);
- if (ret)
- return ret;
+ source_is_empty = afr_selfheal_name_source_empty_check(this, replies,
+ sources, source);
+ if (source_is_empty) {
+ ret = __afr_selfheal_name_expunge(this, parent, pargfid, bname, inode,
+ replies);
+ if (ret == -EIO)
+ ret = -1;
+ return ret;
+ }
- ret = afr_selfheal_name_gfid_mismatch_check (this, replies, source,
- sources, &gfid_idx,
- pargfid, bname, inode,
- locked_on, xdata);
- if (ret)
- return ret;
+ ret = afr_selfheal_name_type_mismatch_check(this, replies, source, sources,
+ pargfid, bname);
+ if (ret)
+ return ret;
- if (gfid_idx == -1) {
- if (!gfid_req || gf_uuid_is_null (gfid_req))
- return -1;
- gfid = gfid_req;
- } else {
- gfid = &replies[gfid_idx].poststat.ia_gfid;
- }
+ ret = afr_selfheal_name_gfid_mismatch_check(this, replies, source, sources,
+ &gfid_idx, pargfid, bname,
+ inode, locked_on, xdata);
+ if (ret)
+ return ret;
- is_gfid_absent = (gfid_idx == -1) ? _gf_true : _gf_false;
- ret = __afr_selfheal_assign_gfid (this, parent, pargfid, bname, inode,
- replies, gfid, locked_on, source,
- sources, is_gfid_absent);
- if (ret)
- return ret;
+ if (gfid_idx == -1) {
+ if (!gfid_req || gf_uuid_is_null(gfid_req))
+ return -1;
+ gfid = gfid_req;
+ } else {
+ gfid = &replies[gfid_idx].poststat.ia_gfid;
+ }
+
+ is_gfid_absent = (gfid_idx == -1) ? _gf_true : _gf_false;
+ ret = __afr_selfheal_assign_gfid(this, parent, pargfid, bname, inode,
+ replies, gfid, locked_on, source, sources,
+ is_gfid_absent);
+ if (ret)
+ return ret;
- if (gfid_idx == -1) {
- gfid_idx = afr_selfheal_gfid_idx_get (this, replies, sources);
- if (gfid_idx == -1)
- return -1;
- }
+ if (gfid_idx == -1) {
+ gfid_idx = afr_selfheal_gfid_idx_get(this, replies, sources);
+ if (gfid_idx == -1)
+ return -1;
+ }
- ret = __afr_selfheal_name_impunge (frame, this, parent, pargfid,
- bname, inode, replies, gfid_idx);
- if (ret == -EIO)
- ret = -1;
+ ret = __afr_selfheal_name_impunge(frame, this, parent, pargfid, bname,
+ inode, replies, gfid_idx);
+ if (ret == -EIO)
+ ret = -1;
- return ret;
+ return ret;
}
-
int
-__afr_selfheal_name_finalize_source (xlator_t *this, unsigned char *sources,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- uint64_t *witness)
+__afr_selfheal_name_finalize_source(xlator_t *this, unsigned char *sources,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on, uint64_t *witness)
{
- int i = 0;
- afr_private_t *priv = NULL;
- int source = -1;
- int sources_count = 0;
-
- priv = this->private;
-
- sources_count = AFR_COUNT (sources, priv->child_count);
-
- if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count || afr_does_witness_exist (this, witness)) {
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- afr_mark_active_sinks (this, sources, locked_on, healed_sinks);
- return -1;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i]) {
- source = i;
- break;
- }
- }
-
- return source;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int source = -1;
+ int sources_count = 0;
+
+ priv = this->private;
+
+ sources_count = AFR_COUNT(sources, priv->child_count);
+
+ if ((AFR_CMP(locked_on, healed_sinks, priv->child_count) == 0) ||
+ !sources_count || afr_does_witness_exist(this, witness)) {
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ afr_mark_active_sinks(this, sources, locked_on, healed_sinks);
+ return -1;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i]) {
+ source = i;
+ break;
+ }
+ }
+
+ return source;
}
int
-__afr_selfheal_name_prepare (call_frame_t *frame, xlator_t *this, inode_t *parent,
- uuid_t pargfid, unsigned char *locked_on,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks, int *source_p)
+__afr_selfheal_name_prepare(call_frame_t *frame, xlator_t *this,
+ inode_t *parent, uuid_t pargfid,
+ unsigned char *locked_on, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks,
+ int *source_p)
{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- uint64_t *witness = NULL;
-
- priv = this->private;
-
- replies = alloca0 (priv->child_count * sizeof(*replies));
-
- ret = afr_selfheal_unlocked_discover (frame, parent, pargfid, replies);
- if (ret)
- goto out;
-
- witness = alloca0 (sizeof (*witness) * priv->child_count);
- ret = afr_selfheal_find_direction (frame, this, replies,
- AFR_ENTRY_TRANSACTION,
- locked_on, sources, sinks, witness,
- NULL);
- if (ret)
- goto out;
-
- /* Initialize the healed_sinks[] array optimistically to
- the intersection of to-be-healed (i.e sinks[]) and
- the list of servers which are up (i.e locked_on[]).
-
- As we encounter failures in the healing process, we
- will unmark the respective servers in the healed_sinks[]
- array.
- */
- AFR_INTERSECT (healed_sinks, sinks, locked_on, priv->child_count);
-
- source = __afr_selfheal_name_finalize_source (this, sources,
- healed_sinks,
- locked_on, witness);
- if (source < 0) {
- /* If source is < 0 (typically split-brain), we perform a
- conservative merge of entries rather than erroring out */
- }
- *source_p = source;
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ uint64_t *witness = NULL;
+
+ priv = this->private;
+
+ replies = alloca0(priv->child_count * sizeof(*replies));
+
+ ret = afr_selfheal_unlocked_discover(frame, parent, pargfid, replies);
+ if (ret)
+ goto out;
+
+ witness = alloca0(sizeof(*witness) * priv->child_count);
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_ENTRY_TRANSACTION, locked_on, sources,
+ sinks, witness, NULL);
+ if (ret)
+ goto out;
+
+ /* Initialize the healed_sinks[] array optimistically to
+ the intersection of to-be-healed (i.e sinks[]) and
+ the list of servers which are up (i.e locked_on[]).
+
+ As we encounter failures in the healing process, we
+ will unmark the respective servers in the healed_sinks[]
+ array.
+ */
+ AFR_INTERSECT(healed_sinks, sinks, locked_on, priv->child_count);
+
+ source = __afr_selfheal_name_finalize_source(this, sources, healed_sinks,
+ locked_on, witness);
+ if (source < 0) {
+ /* If source is < 0 (typically split-brain), we perform a
+ conservative merge of entries rather than erroring out */
+ }
+ *source_p = source;
out:
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
- return ret;
+ return ret;
}
-
int
-afr_selfheal_name_do (call_frame_t *frame, xlator_t *this, inode_t *parent,
- uuid_t pargfid, const char *bname, void *gfid_req,
- dict_t *xdata)
+afr_selfheal_name_do(call_frame_t *frame, xlator_t *this, inode_t *parent,
+ uuid_t pargfid, const char *bname, void *gfid_req,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *locked_on = NULL;
- int source = -1;
- struct afr_reply *replies = NULL;
- int ret = -1;
- inode_t *inode = NULL;
- dict_t *xattr = NULL;
-
- xattr = dict_new ();
- if (!xattr)
- return -ENOMEM;
-
- ret = dict_set_int32 (xattr, GF_GFIDLESS_LOOKUP, 1);
- if (ret) {
- dict_unref (xattr);
- return -1;
+ afr_private_t *priv = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *locked_on = NULL;
+ int source = -1;
+ struct afr_reply *replies = NULL;
+ int ret = -1;
+ inode_t *inode = NULL;
+ dict_t *xattr = NULL;
+
+ xattr = dict_new();
+ if (!xattr)
+ return -ENOMEM;
+
+ ret = dict_set_int32(xattr, GF_GFIDLESS_LOOKUP, 1);
+ if (ret) {
+ dict_unref(xattr);
+ return -1;
+ }
+
+ priv = this->private;
+
+ locked_on = alloca0(priv->child_count);
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+
+ replies = alloca0(priv->child_count * sizeof(*replies));
+
+ ret = afr_selfheal_entrylk(frame, this, parent, this->name, bname,
+ locked_on);
+ {
+ if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_name_prepare(frame, this, parent, pargfid,
+ locked_on, sources, sinks,
+ healed_sinks, &source);
+ if (ret)
+ goto unlock;
+
+ inode = afr_selfheal_unlocked_lookup_on(frame, parent, bname, replies,
+ locked_on, xattr);
+ if (!inode) {
+ ret = -ENOMEM;
+ goto unlock;
}
- priv = this->private;
-
- locked_on = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
-
- replies = alloca0 (priv->child_count * sizeof(*replies));
-
- ret = afr_selfheal_entrylk (frame, this, parent, this->name, bname,
- locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_name_prepare (frame, this, parent, pargfid,
- locked_on, sources, sinks,
- healed_sinks, &source);
- if (ret)
- goto unlock;
-
- inode = afr_selfheal_unlocked_lookup_on (frame, parent, bname,
- replies, locked_on,
- xattr);
- if (!inode) {
- ret = -ENOMEM;
- goto unlock;
- }
-
- ret = __afr_selfheal_name_do (frame, this, parent, pargfid,
- bname, inode, sources, sinks,
- healed_sinks, source, locked_on,
- replies, gfid_req, xdata);
- }
+ ret = __afr_selfheal_name_do(frame, this, parent, pargfid, bname, inode,
+ sources, sinks, healed_sinks, source,
+ locked_on, replies, gfid_req, xdata);
+ }
unlock:
- afr_selfheal_unentrylk (frame, this, parent, this->name, bname,
- locked_on, NULL);
- if (inode)
- inode_unref (inode);
+ afr_selfheal_unentrylk(frame, this, parent, this->name, bname, locked_on,
+ NULL);
+ if (inode)
+ inode_unref(inode);
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
- if (xattr)
- dict_unref (xattr);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
+ if (xattr)
+ dict_unref(xattr);
- return ret;
+ return ret;
}
-
int
-afr_selfheal_name_unlocked_inspect (call_frame_t *frame, xlator_t *this,
- inode_t *parent, uuid_t pargfid,
- const char *bname, gf_boolean_t *need_heal)
+afr_selfheal_name_unlocked_inspect(call_frame_t *frame, xlator_t *this,
+ inode_t *parent, uuid_t pargfid,
+ const char *bname, gf_boolean_t *need_heal)
{
- afr_private_t *priv = NULL;
- int i = 0;
- struct afr_reply *replies = NULL;
- inode_t *inode = NULL;
- int first_idx = -1;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ struct afr_reply *replies = NULL;
+ inode_t *inode = NULL;
+ int first_idx = -1;
- priv = this->private;
+ priv = this->private;
- replies = alloca0 (sizeof (*replies) * priv->child_count);
+ replies = alloca0(sizeof(*replies) * priv->child_count);
- inode = afr_selfheal_unlocked_lookup_on (frame, parent, bname,
- replies, priv->child_up, NULL);
- if (!inode)
- return -ENOMEM;
+ inode = afr_selfheal_unlocked_lookup_on(frame, parent, bname, replies,
+ priv->child_up, NULL);
+ if (!inode)
+ return -ENOMEM;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- if ((replies[i].op_ret == -1) &&
- (replies[i].op_errno == ENODATA)) {
- *need_heal = _gf_true;
- break;
- }
+ if ((replies[i].op_ret == -1) && (replies[i].op_errno == ENODATA)) {
+ *need_heal = _gf_true;
+ break;
+ }
- if (first_idx == -1) {
- first_idx = i;
- continue;
- }
+ if (first_idx == -1) {
+ first_idx = i;
+ continue;
+ }
- if (replies[i].op_ret != replies[first_idx].op_ret) {
- *need_heal = _gf_true;
- break;
- }
+ if (replies[i].op_ret != replies[first_idx].op_ret) {
+ *need_heal = _gf_true;
+ break;
+ }
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[first_idx].poststat.ia_gfid)) {
- *need_heal = _gf_true;
- break;
- }
- }
+ if (gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[first_idx].poststat.ia_gfid)) {
+ *need_heal = _gf_true;
+ break;
+ }
+ }
- if (inode)
- inode_unref (inode);
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
- return 0;
+ if (inode)
+ inode_unref(inode);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
+ return 0;
}
int
-afr_selfheal_name (xlator_t *this, uuid_t pargfid, const char *bname,
- void *gfid_req, dict_t *xdata)
+afr_selfheal_name(xlator_t *this, uuid_t pargfid, const char *bname,
+ void *gfid_req, dict_t *xdata)
{
- inode_t *parent = NULL;
- call_frame_t *frame = NULL;
- int ret = -1;
- gf_boolean_t need_heal = _gf_false;
-
- parent = afr_inode_find (this, pargfid);
- if (!parent)
- goto out;
-
- frame = afr_frame_create (this, NULL);
- if (!frame)
- goto out;
-
- ret = afr_selfheal_name_unlocked_inspect (frame, this, parent, pargfid,
- bname, &need_heal);
- if (ret)
- goto out;
-
- if (need_heal) {
- ret = afr_selfheal_name_do (frame, this, parent, pargfid, bname,
- gfid_req, xdata);
- if (ret)
- goto out;
- }
+ inode_t *parent = NULL;
+ call_frame_t *frame = NULL;
+ int ret = -1;
+ gf_boolean_t need_heal = _gf_false;
+
+ parent = afr_inode_find(this, pargfid);
+ if (!parent)
+ goto out;
+
+ frame = afr_frame_create(this, NULL);
+ if (!frame)
+ goto out;
+
+ ret = afr_selfheal_name_unlocked_inspect(frame, this, parent, pargfid,
+ bname, &need_heal);
+ if (ret)
+ goto out;
+
+ if (need_heal) {
+ ret = afr_selfheal_name_do(frame, this, parent, pargfid, bname,
+ gfid_req, xdata);
+ if (ret)
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- if (parent)
- inode_unref (parent);
- if (frame)
- AFR_STACK_DESTROY (frame);
+ if (parent)
+ inode_unref(parent);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
index 1f7ae7bb43d..0cf01a041b4 100644
--- a/xlators/cluster/afr/src/afr-self-heald.c
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
#include "afr-self-heald.h"
@@ -17,1378 +16,1334 @@
#include "afr-messages.h"
#include "byte-order.h"
-#define SHD_INODE_LRU_LIMIT 2048
-#define AFR_EH_SPLIT_BRAIN_LIMIT 1024
-#define AFR_STATISTICS_HISTORY_SIZE 50
-
-
-#define ASSERT_LOCAL(this, healer) \
- if (!afr_shd_is_subvol_local(this, healer->subvol)) { \
- healer->local = _gf_false; \
- if (safe_break (healer)) { \
- break; \
- } else { \
- continue; \
- } \
- } else { \
- healer->local = _gf_true; \
- }
-
-
-#define NTH_INDEX_HEALER(this, n) &((((afr_private_t *)this->private))->shd.index_healers[n])
-#define NTH_FULL_HEALER(this, n) &((((afr_private_t *)this->private))->shd.full_healers[n])
+#define SHD_INODE_LRU_LIMIT 2048
+#define AFR_EH_SPLIT_BRAIN_LIMIT 1024
+#define AFR_STATISTICS_HISTORY_SIZE 50
+
+#define ASSERT_LOCAL(this, healer) \
+ if (!afr_shd_is_subvol_local(this, healer->subvol)) { \
+ healer->local = _gf_false; \
+ if (safe_break(healer)) { \
+ break; \
+ } else { \
+ continue; \
+ } \
+ } else { \
+ healer->local = _gf_true; \
+ }
+
+#define NTH_INDEX_HEALER(this, n) \
+ &((((afr_private_t *)this->private))->shd.index_healers[n])
+#define NTH_FULL_HEALER(this, n) \
+ &((((afr_private_t *)this->private))->shd.full_healers[n])
char *
-afr_subvol_name (xlator_t *this, int subvol)
+afr_subvol_name(xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
- priv = this->private;
- if (subvol < 0 || subvol > priv->child_count)
- return NULL;
+ priv = this->private;
+ if (subvol < 0 || subvol > priv->child_count)
+ return NULL;
- return priv->children[subvol]->name;
+ return priv->children[subvol]->name;
}
-
void
-afr_destroy_crawl_event_data (void *data)
+afr_destroy_crawl_event_data(void *data)
{
- return;
+ return;
}
-
void
-afr_destroy_shd_event_data (void *data)
+afr_destroy_shd_event_data(void *data)
{
- shd_event_t *shd_event = data;
-
- if (!shd_event)
- return;
- GF_FREE (shd_event->path);
+ shd_event_t *shd_event = data;
+ if (!shd_event)
return;
-}
+ GF_FREE(shd_event->path);
+ return;
+}
gf_boolean_t
-afr_shd_is_subvol_local (xlator_t *this, int subvol)
+afr_shd_is_subvol_local(xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
- gf_boolean_t is_local = _gf_false;
- loc_t loc = {0, };
-
- loc.inode = this->itable->root;
- gf_uuid_copy (loc.gfid, loc.inode->gfid);
- priv = this->private;
- syncop_is_subvol_local(priv->children[subvol], &loc, &is_local);
- return is_local;
+ afr_private_t *priv = NULL;
+ gf_boolean_t is_local = _gf_false;
+ loc_t loc = {
+ 0,
+ };
+
+ loc.inode = this->itable->root;
+ gf_uuid_copy(loc.gfid, loc.inode->gfid);
+ priv = this->private;
+ syncop_is_subvol_local(priv->children[subvol], &loc, &is_local);
+ return is_local;
}
-
int
-__afr_shd_healer_wait (struct subvol_healer *healer)
+__afr_shd_healer_wait(struct subvol_healer *healer)
{
- afr_private_t *priv = NULL;
- struct timespec wait_till = {0, };
- int ret = 0;
+ afr_private_t *priv = NULL;
+ struct timespec wait_till = {
+ 0,
+ };
+ int ret = 0;
- priv = healer->this->private;
+ priv = healer->this->private;
disabled_loop:
- wait_till.tv_sec = time (NULL) + priv->shd.timeout;
+ wait_till.tv_sec = time(NULL) + priv->shd.timeout;
- while (!healer->rerun) {
- ret = pthread_cond_timedwait (&healer->cond,
- &healer->mutex,
- &wait_till);
- if (ret == ETIMEDOUT)
- break;
- }
+ while (!healer->rerun) {
+ ret = pthread_cond_timedwait(&healer->cond, &healer->mutex, &wait_till);
+ if (ret == ETIMEDOUT)
+ break;
+ }
- ret = healer->rerun;
- healer->rerun = 0;
+ ret = healer->rerun;
+ healer->rerun = 0;
- if (!priv->shd.enabled)
- goto disabled_loop;
+ if (!priv->shd.enabled)
+ goto disabled_loop;
- return ret;
+ return ret;
}
-
int
-afr_shd_healer_wait (struct subvol_healer *healer)
+afr_shd_healer_wait(struct subvol_healer *healer)
{
- int ret = 0;
+ int ret = 0;
- pthread_mutex_lock (&healer->mutex);
- {
- ret = __afr_shd_healer_wait (healer);
- }
- pthread_mutex_unlock (&healer->mutex);
+ pthread_mutex_lock(&healer->mutex);
+ {
+ ret = __afr_shd_healer_wait(healer);
+ }
+ pthread_mutex_unlock(&healer->mutex);
- return ret;
+ return ret;
}
-
gf_boolean_t
-safe_break (struct subvol_healer *healer)
+safe_break(struct subvol_healer *healer)
{
- gf_boolean_t ret = _gf_false;
+ gf_boolean_t ret = _gf_false;
- pthread_mutex_lock (&healer->mutex);
- {
- if (healer->rerun)
- goto unlock;
+ pthread_mutex_lock(&healer->mutex);
+ {
+ if (healer->rerun)
+ goto unlock;
- healer->running = _gf_false;
- ret = _gf_true;
- }
+ healer->running = _gf_false;
+ ret = _gf_true;
+ }
unlock:
- pthread_mutex_unlock (&healer->mutex);
+ pthread_mutex_unlock(&healer->mutex);
- return ret;
+ return ret;
}
-
inode_t *
-afr_shd_inode_find (xlator_t *this, xlator_t *subvol, uuid_t gfid)
+afr_shd_inode_find(xlator_t *this, xlator_t *subvol, uuid_t gfid)
{
- int ret = 0;
- uint64_t val = IA_INVAL;
- dict_t *xdata = NULL;
- dict_t *rsp_dict = NULL;
- inode_t *inode = NULL;
-
- xdata = dict_new ();
- if (!xdata)
- goto out;
-
- ret = dict_set_int8 (xdata, GF_INDEX_IA_TYPE_GET_REQ, 1);
+ int ret = 0;
+ uint64_t val = IA_INVAL;
+ dict_t *xdata = NULL;
+ dict_t *rsp_dict = NULL;
+ inode_t *inode = NULL;
+
+ xdata = dict_new();
+ if (!xdata)
+ goto out;
+
+ ret = dict_set_int8(xdata, GF_INDEX_IA_TYPE_GET_REQ, 1);
+ if (ret)
+ goto out;
+
+ ret = syncop_inode_find(this, subvol, gfid, &inode, xdata, &rsp_dict);
+ if (ret < 0)
+ goto out;
+
+ if (rsp_dict) {
+ ret = dict_get_uint64(rsp_dict, GF_INDEX_IA_TYPE_GET_RSP, &val);
if (ret)
- goto out;
-
- ret = syncop_inode_find (this, subvol, gfid, &inode,
- xdata, &rsp_dict);
- if (ret < 0)
- goto out;
-
- if (rsp_dict) {
- ret = dict_get_uint64 (rsp_dict, GF_INDEX_IA_TYPE_GET_RSP,
- &val);
- if (ret)
- goto out;
- }
- ret = inode_ctx_set2 (inode, subvol, 0, &val);
+ goto out;
+ }
+ ret = inode_ctx_set2(inode, subvol, 0, &val);
out:
- if (ret && inode) {
- inode_unref (inode);
- inode = NULL;
- }
- if (xdata)
- dict_unref (xdata);
- if (rsp_dict)
- dict_unref (rsp_dict);
- return inode;
+ if (ret && inode) {
+ inode_unref(inode);
+ inode = NULL;
+ }
+ if (xdata)
+ dict_unref(xdata);
+ if (rsp_dict)
+ dict_unref(rsp_dict);
+ return inode;
}
-inode_t*
-afr_shd_index_inode (xlator_t *this, xlator_t *subvol, char *vgfid)
+inode_t *
+afr_shd_index_inode(xlator_t *this, xlator_t *subvol, char *vgfid)
{
- loc_t rootloc = {0, };
- inode_t *inode = NULL;
- int ret = 0;
- dict_t *xattr = NULL;
- void *index_gfid = NULL;
+ loc_t rootloc = {
+ 0,
+ };
+ inode_t *inode = NULL;
+ int ret = 0;
+ dict_t *xattr = NULL;
+ void *index_gfid = NULL;
- rootloc.inode = inode_ref (this->itable->root);
- gf_uuid_copy (rootloc.gfid, rootloc.inode->gfid);
+ rootloc.inode = inode_ref(this->itable->root);
+ gf_uuid_copy(rootloc.gfid, rootloc.inode->gfid);
- ret = syncop_getxattr (subvol, &rootloc, &xattr,
- vgfid, NULL, NULL);
- if (ret || !xattr) {
- errno = -ret;
- goto out;
- }
+ ret = syncop_getxattr(subvol, &rootloc, &xattr, vgfid, NULL, NULL);
+ if (ret || !xattr) {
+ errno = -ret;
+ goto out;
+ }
- ret = dict_get_ptr (xattr, vgfid, &index_gfid);
- if (ret)
- goto out;
+ ret = dict_get_ptr(xattr, vgfid, &index_gfid);
+ if (ret)
+ goto out;
- gf_msg_debug (this->name, 0, "%s dir gfid for %s: %s",
- vgfid, subvol->name, uuid_utoa (index_gfid));
+ gf_msg_debug(this->name, 0, "%s dir gfid for %s: %s", vgfid, subvol->name,
+ uuid_utoa(index_gfid));
- inode = afr_shd_inode_find (this, subvol, index_gfid);
+ inode = afr_shd_inode_find(this, subvol, index_gfid);
out:
- loc_wipe (&rootloc);
+ loc_wipe(&rootloc);
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
- return inode;
+ return inode;
}
int
-afr_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name,
- ia_type_t type)
+afr_shd_index_purge(xlator_t *subvol, inode_t *inode, char *name,
+ ia_type_t type)
{
- int ret = 0;
- loc_t loc = {0,};
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
- loc.parent = inode_ref (inode);
- loc.name = name;
+ loc.parent = inode_ref(inode);
+ loc.name = name;
- if (IA_ISDIR (type))
- ret = syncop_rmdir (subvol, &loc, 1, NULL, NULL);
- else
- ret = syncop_unlink (subvol, &loc, NULL, NULL);
+ if (IA_ISDIR(type))
+ ret = syncop_rmdir(subvol, &loc, 1, NULL, NULL);
+ else
+ ret = syncop_unlink(subvol, &loc, NULL, NULL);
- loc_wipe (&loc);
- return ret;
+ loc_wipe(&loc);
+ return ret;
}
void
-afr_shd_zero_xattrop (xlator_t *this, uuid_t gfid)
+afr_shd_zero_xattrop(xlator_t *this, uuid_t gfid)
{
-
- call_frame_t *frame = NULL;
- inode_t *inode = NULL;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- int ret = 0;
- int i = 0;
- int raw[AFR_NUM_CHANGE_LOGS] = {0};
-
- priv = this->private;
- frame = afr_frame_create (this, NULL);
- if (!frame)
- goto out;
- inode = afr_inode_find (this, gfid);
- if (!inode)
- goto out;
- xattr = dict_new();
- if (!xattr)
- goto out;
- ret = dict_set_static_bin (xattr, AFR_DIRTY, raw,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ call_frame_t *frame = NULL;
+ inode_t *inode = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int ret = 0;
+ int i = 0;
+ int raw[AFR_NUM_CHANGE_LOGS] = {0};
+
+ priv = this->private;
+ frame = afr_frame_create(this, NULL);
+ if (!frame)
+ goto out;
+ inode = afr_inode_find(this, gfid);
+ if (!inode)
+ goto out;
+ xattr = dict_new();
+ if (!xattr)
+ goto out;
+ ret = dict_set_static_bin(xattr, AFR_DIRTY, raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret)
+ goto out;
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_set_static_bin(xattr, priv->pending_key[i], raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
if (ret)
- goto out;
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_static_bin (xattr, priv->pending_key[i], raw,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret)
- goto out;
- }
+ goto out;
+ }
- /*Send xattrop to all bricks. Doing a lookup to see if bricks are up or
- * has valid repies for this gfid seems a bit of an overkill.*/
- for (i = 0; i < priv->child_count; i++)
- afr_selfheal_post_op (frame, this, inode, i, xattr, NULL);
+ /*Send xattrop to all bricks. Doing a lookup to see if bricks are up or
+ * has valid repies for this gfid seems a bit of an overkill.*/
+ for (i = 0; i < priv->child_count; i++)
+ afr_selfheal_post_op(frame, this, inode, i, xattr, NULL);
out:
- if (frame)
- AFR_STACK_DESTROY (frame);
- if (inode)
- inode_unref (inode);
- if (xattr)
- dict_unref (xattr);
- return;
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ if (inode)
+ inode_unref(inode);
+ if (xattr)
+ dict_unref(xattr);
+ return;
}
int
-afr_shd_selfheal_name (struct subvol_healer *healer, int child, uuid_t parent,
- const char *bname)
+afr_shd_selfheal_name(struct subvol_healer *healer, int child, uuid_t parent,
+ const char *bname)
{
- int ret = -1;
+ int ret = -1;
- ret = afr_selfheal_name (THIS, parent, bname, NULL, NULL);
+ ret = afr_selfheal_name(THIS, parent, bname, NULL, NULL);
- return ret;
+ return ret;
}
int
-afr_shd_selfheal (struct subvol_healer *healer, int child, uuid_t gfid)
+afr_shd_selfheal(struct subvol_healer *healer, int child, uuid_t gfid)
{
- int ret = 0;
- eh_t *eh = NULL;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- shd_event_t *shd_event = NULL;
- char *path = NULL;
- xlator_t *subvol = NULL;
- xlator_t *this = NULL;
- crawl_event_t *crawl_event = NULL;
-
- this = healer->this;
- priv = this->private;
- shd = &priv->shd;
- crawl_event = &healer->crawl_event;
-
- subvol = priv->children[child];
-
- //If this fails with ENOENT/ESTALE index is stale
- ret = syncop_gfid_to_path (this->itable, subvol, gfid, &path);
- if (ret < 0)
- return ret;
-
- ret = afr_selfheal (this, gfid);
-
- LOCK (&priv->lock);
- {
- if (ret == -EIO) {
- eh = shd->split_brain;
- crawl_event->split_brain_count++;
- } else if (ret < 0) {
- crawl_event->heal_failed_count++;
- } else if (ret == 0) {
- crawl_event->healed_count++;
- }
+ int ret = 0;
+ eh_t *eh = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ shd_event_t *shd_event = NULL;
+ char *path = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *this = NULL;
+ crawl_event_t *crawl_event = NULL;
+
+ this = healer->this;
+ priv = this->private;
+ shd = &priv->shd;
+ crawl_event = &healer->crawl_event;
+
+ subvol = priv->children[child];
+
+ // If this fails with ENOENT/ESTALE index is stale
+ ret = syncop_gfid_to_path(this->itable, subvol, gfid, &path);
+ if (ret < 0)
+ return ret;
+
+ ret = afr_selfheal(this, gfid);
+
+ LOCK(&priv->lock);
+ {
+ if (ret == -EIO) {
+ eh = shd->split_brain;
+ crawl_event->split_brain_count++;
+ } else if (ret < 0) {
+ crawl_event->heal_failed_count++;
+ } else if (ret == 0) {
+ crawl_event->healed_count++;
}
- UNLOCK (&priv->lock);
+ }
+ UNLOCK(&priv->lock);
- if (eh) {
- shd_event = GF_CALLOC (1, sizeof(*shd_event),
- gf_afr_mt_shd_event_t);
- if (!shd_event)
- goto out;
+ if (eh) {
+ shd_event = GF_CALLOC(1, sizeof(*shd_event), gf_afr_mt_shd_event_t);
+ if (!shd_event)
+ goto out;
- shd_event->child = child;
- shd_event->path = path;
+ shd_event->child = child;
+ shd_event->path = path;
- if (eh_save_history (eh, shd_event) < 0)
- goto out;
+ if (eh_save_history(eh, shd_event) < 0)
+ goto out;
- shd_event = NULL;
- path = NULL;
- }
+ shd_event = NULL;
+ path = NULL;
+ }
out:
- GF_FREE (shd_event);
- GF_FREE (path);
- return ret;
+ GF_FREE(shd_event);
+ GF_FREE(path);
+ return ret;
}
-
void
-afr_shd_sweep_prepare (struct subvol_healer *healer)
+afr_shd_sweep_prepare(struct subvol_healer *healer)
{
- crawl_event_t *event = NULL;
+ crawl_event_t *event = NULL;
- event = &healer->crawl_event;
+ event = &healer->crawl_event;
- event->healed_count = 0;
- event->split_brain_count = 0;
- event->heal_failed_count = 0;
+ event->healed_count = 0;
+ event->split_brain_count = 0;
+ event->heal_failed_count = 0;
- time (&event->start_time);
- event->end_time = 0;
+ time(&event->start_time);
+ event->end_time = 0;
}
-
void
-afr_shd_sweep_done (struct subvol_healer *healer)
+afr_shd_sweep_done(struct subvol_healer *healer)
{
- crawl_event_t *event = NULL;
- crawl_event_t *history = NULL;
- afr_self_heald_t *shd = NULL;
+ crawl_event_t *event = NULL;
+ crawl_event_t *history = NULL;
+ afr_self_heald_t *shd = NULL;
- event = &healer->crawl_event;
- shd = &(((afr_private_t *)healer->this->private)->shd);
+ event = &healer->crawl_event;
+ shd = &(((afr_private_t *)healer->this->private)->shd);
- time (&event->end_time);
- history = memdup (event, sizeof (*event));
- event->start_time = 0;
+ time(&event->end_time);
+ history = memdup(event, sizeof(*event));
+ event->start_time = 0;
- if (!history)
- return;
+ if (!history)
+ return;
- if (eh_save_history (shd->statistics[healer->subvol], history) < 0)
- GF_FREE (history);
+ if (eh_save_history(shd->statistics[healer->subvol], history) < 0)
+ GF_FREE(history);
}
int
-afr_shd_index_heal (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
- void *data)
+afr_shd_index_heal(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
{
- struct subvol_healer *healer = data;
- afr_private_t *priv = NULL;
- uuid_t gfid = {0};
- int ret = 0;
- uint64_t val = IA_INVAL;
+ struct subvol_healer *healer = data;
+ afr_private_t *priv = NULL;
+ uuid_t gfid = {0};
+ int ret = 0;
+ uint64_t val = IA_INVAL;
- priv = healer->this->private;
- if (!priv->shd.enabled)
- return -EBUSY;
+ priv = healer->this->private;
+ if (!priv->shd.enabled)
+ return -EBUSY;
- gf_msg_debug (healer->this->name, 0, "got entry: %s from %s",
- entry->d_name, priv->children[healer->subvol]->name);
+ gf_msg_debug(healer->this->name, 0, "got entry: %s from %s", entry->d_name,
+ priv->children[healer->subvol]->name);
- ret = gf_uuid_parse (entry->d_name, gfid);
- if (ret)
- return 0;
+ ret = gf_uuid_parse(entry->d_name, gfid);
+ if (ret)
+ return 0;
- inode_ctx_get2 (parent->inode, subvol, NULL, &val);
+ inode_ctx_get2(parent->inode, subvol, NULL, &val);
- ret = afr_shd_selfheal (healer, healer->subvol, gfid);
+ ret = afr_shd_selfheal(healer, healer->subvol, gfid);
- if (ret == -ENOENT || ret == -ESTALE)
- afr_shd_index_purge (subvol, parent->inode, entry->d_name, val);
+ if (ret == -ENOENT || ret == -ESTALE)
+ afr_shd_index_purge(subvol, parent->inode, entry->d_name, val);
- if (ret == 2)
- /* If bricks crashed in pre-op after creating indices/xattrop
- * link but before setting afr changelogs, we end up with stale
- * xattrop links but zero changelogs. Remove such entries by
- * sending a post-op with zero changelogs.
- */
- afr_shd_zero_xattrop (healer->this, gfid);
+ if (ret == 2)
+ /* If bricks crashed in pre-op after creating indices/xattrop
+ * link but before setting afr changelogs, we end up with stale
+ * xattrop links but zero changelogs. Remove such entries by
+ * sending a post-op with zero changelogs.
+ */
+ afr_shd_zero_xattrop(healer->this, gfid);
- return 0;
+ return 0;
}
int
-afr_shd_index_sweep (struct subvol_healer *healer, char *vgfid)
+afr_shd_index_sweep(struct subvol_healer *healer, char *vgfid)
{
- loc_t loc = {0};
- afr_private_t *priv = NULL;
- int ret = 0;
- xlator_t *subvol = NULL;
- dict_t *xdata = NULL;
- call_frame_t *frame = NULL;
-
- priv = healer->this->private;
- subvol = priv->children[healer->subvol];
-
- frame = afr_frame_create (healer->this, &ret);
- if (!frame) {
- ret = -ret;
- goto out;
- }
-
- loc.inode = afr_shd_index_inode (healer->this, subvol, vgfid);
- if (!loc.inode) {
- gf_msg (healer->this->name, GF_LOG_WARNING,
- 0, AFR_MSG_INDEX_DIR_GET_FAILED,
- "unable to get index-dir on %s", subvol->name);
- ret = -errno;
- goto out;
- }
-
- xdata = dict_new ();
- if (!xdata || dict_set_int32 (xdata, "get-gfid-type", 1)) {
- ret = -ENOMEM;
- goto out;
- }
-
- ret = syncop_mt_dir_scan (frame, subvol, &loc, GF_CLIENT_PID_SELF_HEALD,
- healer, afr_shd_index_heal, xdata,
- priv->shd.max_threads, priv->shd.wait_qlength);
-
- if (ret == 0)
- ret = healer->crawl_event.healed_count;
+ loc_t loc = {0};
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ xlator_t *subvol = NULL;
+ dict_t *xdata = NULL;
+ call_frame_t *frame = NULL;
+
+ priv = healer->this->private;
+ subvol = priv->children[healer->subvol];
+
+ frame = afr_frame_create(healer->this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+
+ loc.inode = afr_shd_index_inode(healer->this, subvol, vgfid);
+ if (!loc.inode) {
+ gf_msg(healer->this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_INDEX_DIR_GET_FAILED, "unable to get index-dir on %s",
+ subvol->name);
+ ret = -errno;
+ goto out;
+ }
+
+ xdata = dict_new();
+ if (!xdata || dict_set_int32(xdata, "get-gfid-type", 1)) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ ret = syncop_mt_dir_scan(frame, subvol, &loc, GF_CLIENT_PID_SELF_HEALD,
+ healer, afr_shd_index_heal, xdata,
+ priv->shd.max_threads, priv->shd.wait_qlength);
+
+ if (ret == 0)
+ ret = healer->crawl_event.healed_count;
out:
- loc_wipe (&loc);
+ loc_wipe(&loc);
- if (xdata)
- dict_unref (xdata);
- if (frame)
- AFR_STACK_DESTROY (frame);
- return ret;
+ if (xdata)
+ dict_unref(xdata);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ return ret;
}
int
-afr_shd_index_sweep_all (struct subvol_healer *healer)
+afr_shd_index_sweep_all(struct subvol_healer *healer)
{
- int ret = 0;
- int count = 0;
-
- ret = afr_shd_index_sweep (healer, GF_XATTROP_INDEX_GFID);
- if (ret < 0)
- goto out;
- count = ret;
-
- ret = afr_shd_index_sweep (healer, GF_XATTROP_DIRTY_GFID);
- if (ret < 0)
- goto out;
- count += ret;
-
- ret = afr_shd_index_sweep (healer, GF_XATTROP_ENTRY_CHANGES_GFID);
- if (ret < 0)
- goto out;
- count += ret;
+ int ret = 0;
+ int count = 0;
+
+ ret = afr_shd_index_sweep(healer, GF_XATTROP_INDEX_GFID);
+ if (ret < 0)
+ goto out;
+ count = ret;
+
+ ret = afr_shd_index_sweep(healer, GF_XATTROP_DIRTY_GFID);
+ if (ret < 0)
+ goto out;
+ count += ret;
+
+ ret = afr_shd_index_sweep(healer, GF_XATTROP_ENTRY_CHANGES_GFID);
+ if (ret < 0)
+ goto out;
+ count += ret;
out:
- if (ret < 0)
- return ret;
- else
- return count;
+ if (ret < 0)
+ return ret;
+ else
+ return count;
}
int
-afr_shd_full_heal (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
- void *data)
+afr_shd_full_heal(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
{
- struct subvol_healer *healer = data;
- xlator_t *this = healer->this;
- afr_private_t *priv = NULL;
+ struct subvol_healer *healer = data;
+ xlator_t *this = healer->this;
+ afr_private_t *priv = NULL;
- priv = this->private;
- if (!priv->shd.enabled)
- return -EBUSY;
+ priv = this->private;
+ if (!priv->shd.enabled)
+ return -EBUSY;
- afr_shd_selfheal_name (healer, healer->subvol,
- parent->inode->gfid, entry->d_name);
+ afr_shd_selfheal_name(healer, healer->subvol, parent->inode->gfid,
+ entry->d_name);
- afr_shd_selfheal (healer, healer->subvol, entry->d_stat.ia_gfid);
+ afr_shd_selfheal(healer, healer->subvol, entry->d_stat.ia_gfid);
- return 0;
+ return 0;
}
int
-afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
+afr_shd_full_sweep(struct subvol_healer *healer, inode_t *inode)
{
- afr_private_t *priv = NULL;
- loc_t loc = {0};
-
- priv = healer->this->private;
- loc.inode = inode;
- return syncop_ftw (priv->children[healer->subvol], &loc,
- GF_CLIENT_PID_SELF_HEALD, healer,
- afr_shd_full_heal);
+ afr_private_t *priv = NULL;
+ loc_t loc = {0};
+
+ priv = healer->this->private;
+ loc.inode = inode;
+ return syncop_ftw(priv->children[healer->subvol], &loc,
+ GF_CLIENT_PID_SELF_HEALD, healer, afr_shd_full_heal);
}
void
-afr_shd_ta_set_xattrs (xlator_t *this, loc_t *loc, dict_t **xdata,
- int healer)
+afr_shd_ta_set_xattrs(xlator_t *this, loc_t *loc, dict_t **xdata, int healer)
{
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- struct gf_flock flock = {0, };
- gf_boolean_t need_xattrop = _gf_false;
- void *pending_raw = NULL;
- int *raw = NULL;
- int pending[AFR_NUM_CHANGE_LOGS] = {0,};
- int i = 0;
- int j = 0;
- int val = 0;
- int ret = 0;
-
- priv = this->private;
-
- xattr = dict_new ();
- if (!xattr) {
- goto out;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ gf_boolean_t need_xattrop = _gf_false;
+ void *pending_raw = NULL;
+ int *raw = NULL;
+ int pending[AFR_NUM_CHANGE_LOGS] = {
+ 0,
+ };
+ int i = 0;
+ int j = 0;
+ int val = 0;
+ int ret = 0;
+
+ priv = this->private;
+
+ xattr = dict_new();
+ if (!xattr) {
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ raw = GF_CALLOC(AFR_NUM_CHANGE_LOGS, sizeof(int), gf_afr_mt_int32_t);
+ if (!raw) {
+ goto out;
}
- for (i = 0; i < priv->child_count; i++) {
- raw = GF_CALLOC (AFR_NUM_CHANGE_LOGS, sizeof(int),
- gf_afr_mt_int32_t);
- if (!raw) {
- goto out;
- }
-
- ret = dict_get_ptr (*xdata, priv->pending_key[i], &pending_raw);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_GET_FAILED, "Error getting value "
- "of pending key %s", priv->pending_key[i]);
- GF_FREE (raw);
- goto out;
- }
-
- memcpy (pending, pending_raw, sizeof (pending));
- for (j = 0; j < AFR_NUM_CHANGE_LOGS; j++) {
- val = ntoh32 (pending[j]);
- if (val) {
- if (i == healer) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_THIN_ARB, "I am "
- "not the good shd. Skipping. "
- "SHD = %d.", healer);
- GF_FREE (raw);
- goto out;
- }
- need_xattrop = _gf_true;
- raw[j] = hton32 (-val);
- }
- }
+ ret = dict_get_ptr(*xdata, priv->pending_key[i], &pending_raw);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "Error getting value "
+ "of pending key %s",
+ priv->pending_key[i]);
+ GF_FREE(raw);
+ goto out;
+ }
- ret = dict_set_bin (xattr, priv->pending_key[i], raw,
- AFR_NUM_CHANGE_LOGS * sizeof (int));
- if (ret) {
- GF_FREE (raw);
- goto out;
+ memcpy(pending, pending_raw, sizeof(pending));
+ for (j = 0; j < AFR_NUM_CHANGE_LOGS; j++) {
+ val = ntoh32(pending[j]);
+ if (val) {
+ if (i == healer) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_THIN_ARB,
+ "I am "
+ "not the good shd. Skipping. "
+ "SHD = %d.",
+ healer);
+ GF_FREE(raw);
+ goto out;
}
-
- memset (pending, 0, sizeof (pending));
+ need_xattrop = _gf_true;
+ raw[j] = hton32(-val);
+ }
}
- if (!need_xattrop) {
- goto out;
+ ret = dict_set_bin(xattr, priv->pending_key[i], raw,
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ if (ret) {
+ GF_FREE(raw);
+ goto out;
}
- flock.l_type = F_WRLCK;
- flock.l_start = 0;
- flock.l_len = 0;
+ memset(pending, 0, sizeof(pending));
+ }
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_NOTIFY, loc, F_SETLKW, &flock,
- NULL, NULL);
- if (ret)
- goto out;
+ if (!need_xattrop) {
+ goto out;
+ }
- ret = syncop_xattrop (priv->children[THIN_ARBITER_BRICK_INDEX], loc,
- GF_XATTROP_ADD_ARRAY, xattr, NULL, NULL, NULL);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Xattrop failed.");
+ flock.l_type = F_WRLCK;
+ flock.l_start = 0;
+ flock.l_len = 0;
+
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, loc, F_SETLKW, &flock, NULL, NULL);
+ if (ret)
+ goto out;
+
+ ret = syncop_xattrop(priv->children[THIN_ARBITER_BRICK_INDEX], loc,
+ GF_XATTROP_ADD_ARRAY, xattr, NULL, NULL, NULL);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Xattrop failed.");
- flock.l_type = F_UNLCK;
- syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_NOTIFY, loc, F_SETLKW, &flock, NULL, NULL);
+ flock.l_type = F_UNLCK;
+ syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX], AFR_TA_DOM_NOTIFY,
+ loc, F_SETLKW, &flock, NULL, NULL);
out:
- if (xattr)
- dict_unref (xattr);
- return;
+ if (xattr)
+ dict_unref(xattr);
+ return;
}
void
-afr_shd_ta_get_xattrs (xlator_t *this, loc_t *loc, dict_t **xdata)
+afr_shd_ta_get_xattrs(xlator_t *this, loc_t *loc, dict_t **xdata)
{
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- struct iatt stbuf = {0,};
- int *raw = NULL;
- int ret = 0;
- int i = 0;
-
- priv = this->private;
-
- loc->parent = inode_ref (this->itable->root);
- gf_uuid_copy (loc->pargfid, loc->parent->gfid);
- loc->name = priv->pending_key[THIN_ARBITER_BRICK_INDEX];
- loc->inode = inode_new (loc->parent->table);
- if (!loc->inode) {
- goto out;
- }
-
- ret = syncop_lookup (priv->children[THIN_ARBITER_BRICK_INDEX], loc,
- &stbuf, 0, 0, 0);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed lookup on file %s.", loc->name);
- goto out;
- }
-
- gf_uuid_copy (priv->ta_gfid, stbuf.ia_gfid);
- gf_uuid_copy (loc->gfid, priv->ta_gfid);
-
- xattr = dict_new ();
- if (!xattr) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_GET_FAILED,
- "Failed to create dict.");
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- raw = GF_CALLOC (AFR_NUM_CHANGE_LOGS, sizeof(int),
- gf_afr_mt_int32_t);
- if (!raw) {
- goto out;
- }
-
- ret = dict_set_bin (xattr, priv->pending_key[i], raw,
- AFR_NUM_CHANGE_LOGS * sizeof (int));
- if (ret) {
- GF_FREE (raw);
- goto out;
- }
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+ int *raw = NULL;
+ int ret = 0;
+ int i = 0;
+
+ priv = this->private;
+
+ loc->parent = inode_ref(this->itable->root);
+ gf_uuid_copy(loc->pargfid, loc->parent->gfid);
+ loc->name = priv->pending_key[THIN_ARBITER_BRICK_INDEX];
+ loc->inode = inode_new(loc->parent->table);
+ if (!loc->inode) {
+ goto out;
+ }
+
+ ret = syncop_lookup(priv->children[THIN_ARBITER_BRICK_INDEX], loc, &stbuf,
+ 0, 0, 0);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed lookup on file %s.", loc->name);
+ goto out;
+ }
+
+ gf_uuid_copy(priv->ta_gfid, stbuf.ia_gfid);
+ gf_uuid_copy(loc->gfid, priv->ta_gfid);
+
+ xattr = dict_new();
+ if (!xattr) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_GET_FAILED,
+ "Failed to create dict.");
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ raw = GF_CALLOC(AFR_NUM_CHANGE_LOGS, sizeof(int), gf_afr_mt_int32_t);
+ if (!raw) {
+ goto out;
}
- ret = syncop_xattrop (priv->children[THIN_ARBITER_BRICK_INDEX],
- loc, GF_XATTROP_ADD_ARRAY, xattr, NULL, xdata,
- NULL);
+ ret = dict_set_bin(xattr, priv->pending_key[i], raw,
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Xattrop failed.");
- goto out;
+ GF_FREE(raw);
+ goto out;
}
- if (!(*xdata))
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_GET_FAILED,
- "Xdata response is empty.");
+ }
+
+ ret = syncop_xattrop(priv->children[THIN_ARBITER_BRICK_INDEX], loc,
+ GF_XATTROP_ADD_ARRAY, xattr, NULL, xdata, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Xattrop failed.");
+ goto out;
+ }
+ if (!(*xdata))
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_GET_FAILED,
+ "Xdata response is empty.");
out:
- if (xattr)
- dict_unref (xattr);
- return;
+ if (xattr)
+ dict_unref(xattr);
+ return;
}
void *
-afr_shd_index_healer (void *data)
+afr_shd_index_healer(void *data)
{
- struct subvol_healer *healer = NULL;
- xlator_t *this = NULL;
- int ret = 0;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- loc_t loc = {0, };
-
- healer = data;
- THIS = this = healer->this;
- priv = this->private;
-
- for (;;) {
- afr_shd_healer_wait (healer);
-
- ASSERT_LOCAL(this, healer);
- priv->local[healer->subvol] = healer->local;
-
- if (priv->thin_arbiter_count) {
- loc_wipe (&loc);
- afr_shd_ta_get_xattrs (this, &loc, &xdata);
- }
-
+ struct subvol_healer *healer = NULL;
+ xlator_t *this = NULL;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ loc_t loc = {
+ 0,
+ };
+
+ healer = data;
+ THIS = this = healer->this;
+ priv = this->private;
+
+ for (;;) {
+ afr_shd_healer_wait(healer);
+
+ ASSERT_LOCAL(this, healer);
+ priv->local[healer->subvol] = healer->local;
+
+ if (priv->thin_arbiter_count) {
+ loc_wipe(&loc);
+ afr_shd_ta_get_xattrs(this, &loc, &xdata);
+ }
- do {
- gf_msg_debug (this->name, 0,
- "starting index sweep on subvol %s",
- afr_subvol_name (this, healer->subvol));
-
- afr_shd_sweep_prepare (healer);
-
- ret = afr_shd_index_sweep_all (healer);
-
- afr_shd_sweep_done (healer);
- /*
- As long as at least one gfid was
- healed, keep retrying. We may have
- just healed a directory and thereby
- created entries for other gfids which
- could not be healed thus far.
- */
-
- gf_msg_debug (this->name, 0,
- "finished index sweep on subvol %s",
- afr_subvol_name (this, healer->subvol));
- /*
- Give a pause before retrying to avoid a busy loop
- in case the only entry in index is because of
- an ongoing I/O.
- */
- sleep (1);
- } while (ret > 0);
-
- if (xdata && !healer->crawl_event.heal_failed_count) {
- afr_shd_ta_set_xattrs (this, &loc, &xdata,
- healer->subvol);
- dict_unref (xdata);
- xdata = NULL;
- }
- }
+ do {
+ gf_msg_debug(this->name, 0, "starting index sweep on subvol %s",
+ afr_subvol_name(this, healer->subvol));
+
+ afr_shd_sweep_prepare(healer);
+
+ ret = afr_shd_index_sweep_all(healer);
+
+ afr_shd_sweep_done(healer);
+ /*
+ As long as at least one gfid was
+ healed, keep retrying. We may have
+ just healed a directory and thereby
+ created entries for other gfids which
+ could not be healed thus far.
+ */
+
+ gf_msg_debug(this->name, 0, "finished index sweep on subvol %s",
+ afr_subvol_name(this, healer->subvol));
+ /*
+ Give a pause before retrying to avoid a busy loop
+ in case the only entry in index is because of
+ an ongoing I/O.
+ */
+ sleep(1);
+ } while (ret > 0);
+
+ if (xdata && !healer->crawl_event.heal_failed_count) {
+ afr_shd_ta_set_xattrs(this, &loc, &xdata, healer->subvol);
+ dict_unref(xdata);
+ xdata = NULL;
+ }
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return NULL;
+ return NULL;
}
-
void *
-afr_shd_full_healer (void *data)
+afr_shd_full_healer(void *data)
{
- struct subvol_healer *healer = NULL;
- xlator_t *this = NULL;
- int run = 0;
+ struct subvol_healer *healer = NULL;
+ xlator_t *this = NULL;
+ int run = 0;
- healer = data;
- THIS = this = healer->this;
+ healer = data;
+ THIS = this = healer->this;
- for (;;) {
- pthread_mutex_lock (&healer->mutex);
- {
- run = __afr_shd_healer_wait (healer);
- if (!run)
- healer->running = _gf_false;
- }
- pthread_mutex_unlock (&healer->mutex);
+ for (;;) {
+ pthread_mutex_lock(&healer->mutex);
+ {
+ run = __afr_shd_healer_wait(healer);
+ if (!run)
+ healer->running = _gf_false;
+ }
+ pthread_mutex_unlock(&healer->mutex);
- if (!run)
- break;
+ if (!run)
+ break;
- ASSERT_LOCAL(this, healer);
+ ASSERT_LOCAL(this, healer);
- gf_msg (this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
- "starting full sweep on subvol %s",
- afr_subvol_name (this, healer->subvol));
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "starting full sweep on subvol %s",
+ afr_subvol_name(this, healer->subvol));
- afr_shd_sweep_prepare (healer);
+ afr_shd_sweep_prepare(healer);
- afr_shd_full_sweep (healer, this->itable->root);
+ afr_shd_full_sweep(healer, this->itable->root);
- afr_shd_sweep_done (healer);
+ afr_shd_sweep_done(healer);
- gf_msg (this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
- "finished full sweep on subvol %s",
- afr_subvol_name (this, healer->subvol));
- }
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "finished full sweep on subvol %s",
+ afr_subvol_name(this, healer->subvol));
+ }
- return NULL;
+ return NULL;
}
-
int
-afr_shd_healer_init (xlator_t *this, struct subvol_healer *healer)
+afr_shd_healer_init(xlator_t *this, struct subvol_healer *healer)
{
- int ret = 0;
+ int ret = 0;
- ret = pthread_mutex_init (&healer->mutex, NULL);
- if (ret)
- goto out;
+ ret = pthread_mutex_init(&healer->mutex, NULL);
+ if (ret)
+ goto out;
- ret = pthread_cond_init (&healer->cond, NULL);
- if (ret)
- goto out;
+ ret = pthread_cond_init(&healer->cond, NULL);
+ if (ret)
+ goto out;
- healer->this = this;
- healer->running = _gf_false;
- healer->rerun = _gf_false;
- healer->local = _gf_false;
+ healer->this = this;
+ healer->running = _gf_false;
+ healer->rerun = _gf_false;
+ healer->local = _gf_false;
out:
- return ret;
+ return ret;
}
-
int
-afr_shd_healer_spawn (xlator_t *this, struct subvol_healer *healer,
- void *(threadfn)(void *))
+afr_shd_healer_spawn(xlator_t *this, struct subvol_healer *healer,
+ void *(threadfn)(void *))
{
- int ret = 0;
-
- pthread_mutex_lock (&healer->mutex);
- {
- if (healer->running) {
- pthread_cond_signal (&healer->cond);
- } else {
- ret = gf_thread_create (&healer->thread, NULL,
- threadfn, healer, "shdheal");
- if (ret)
- goto unlock;
- healer->running = 1;
- }
-
- healer->rerun = 1;
- }
+ int ret = 0;
+
+ pthread_mutex_lock(&healer->mutex);
+ {
+ if (healer->running) {
+ pthread_cond_signal(&healer->cond);
+ } else {
+ ret = gf_thread_create(&healer->thread, NULL, threadfn, healer,
+ "shdheal");
+ if (ret)
+ goto unlock;
+ healer->running = 1;
+ }
+
+ healer->rerun = 1;
+ }
unlock:
- pthread_mutex_unlock (&healer->mutex);
+ pthread_mutex_unlock(&healer->mutex);
- return ret;
+ return ret;
}
-
int
-afr_shd_full_healer_spawn (xlator_t *this, int subvol)
+afr_shd_full_healer_spawn(xlator_t *this, int subvol)
{
- return afr_shd_healer_spawn (this, NTH_FULL_HEALER (this, subvol),
- afr_shd_full_healer);
+ return afr_shd_healer_spawn(this, NTH_FULL_HEALER(this, subvol),
+ afr_shd_full_healer);
}
-
int
-afr_shd_index_healer_spawn (xlator_t *this, int subvol)
+afr_shd_index_healer_spawn(xlator_t *this, int subvol)
{
- return afr_shd_healer_spawn (this, NTH_INDEX_HEALER (this, subvol),
- afr_shd_index_healer);
+ return afr_shd_healer_spawn(this, NTH_INDEX_HEALER(this, subvol),
+ afr_shd_index_healer);
}
-
int
-afr_shd_dict_add_crawl_event (xlator_t *this, dict_t *output,
- crawl_event_t *crawl_event)
+afr_shd_dict_add_crawl_event(xlator_t *this, dict_t *output,
+ crawl_event_t *crawl_event)
{
- int ret = 0;
- uint64_t count = 0;
- char key[256] = {0};
- int xl_id = 0;
- uint64_t healed_count = 0;
- uint64_t split_brain_count = 0;
- uint64_t heal_failed_count = 0;
- char *start_time_str = 0;
- char *end_time_str = NULL;
- char *crawl_type = NULL;
- int progress = -1;
- int child = -1;
-
- child = crawl_event->child;
- healed_count = crawl_event->healed_count;
- split_brain_count = crawl_event->split_brain_count;
- heal_failed_count = crawl_event->heal_failed_count;
- crawl_type = crawl_event->crawl_type;
-
- if (!crawl_event->start_time)
- goto out;
-
- start_time_str = gf_strdup (ctime (&crawl_event->start_time));
-
- if (crawl_event->end_time)
- end_time_str = gf_strdup (ctime (&crawl_event->end_time));
-
- ret = dict_get_int32 (output, this->name, &xl_id);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_GET_FAILED, "xl does not have id");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics-%d-%d-count", xl_id, child);
- ret = dict_get_uint64 (output, key, &count);
-
-
- snprintf (key, sizeof (key), "statistics_healed_cnt-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_uint64(output, key, healed_count);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_healed_count to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics_sb_cnt-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_uint64 (output, key, split_brain_count);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_split_brain_count to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics_crawl_type-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_str (output, key, crawl_type);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_crawl_type to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics_heal_failed_cnt-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_uint64 (output, key, heal_failed_count);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_healed_failed_count to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics_strt_time-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_dynstr (output, key, start_time_str);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_crawl_start_time to output");
- goto out;
- } else {
- start_time_str = NULL;
- }
-
- if (!end_time_str)
- progress = 1;
- else
- progress = 0;
-
- snprintf (key, sizeof (key), "statistics_end_time-%d-%d-%"PRIu64,
- xl_id, child, count);
- if (!end_time_str)
- end_time_str = gf_strdup ("Could not determine the end time");
- ret = dict_set_dynstr (output, key, end_time_str);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_crawl_end_time to output");
- goto out;
- } else {
- end_time_str = NULL;
- }
-
- snprintf (key, sizeof (key), "statistics_inprogress-%d-%d-%"PRIu64,
- xl_id, child, count);
-
- ret = dict_set_int32 (output, key, progress);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_inprogress to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics-%d-%d-count", xl_id, child);
- ret = dict_set_uint64 (output, key, count + 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not increment the counter.");
- goto out;
- }
+ int ret = 0;
+ uint64_t count = 0;
+ char key[256] = {0};
+ int xl_id = 0;
+ uint64_t healed_count = 0;
+ uint64_t split_brain_count = 0;
+ uint64_t heal_failed_count = 0;
+ char *start_time_str = 0;
+ char *end_time_str = NULL;
+ char *crawl_type = NULL;
+ int progress = -1;
+ int child = -1;
+
+ child = crawl_event->child;
+ healed_count = crawl_event->healed_count;
+ split_brain_count = crawl_event->split_brain_count;
+ heal_failed_count = crawl_event->heal_failed_count;
+ crawl_type = crawl_event->crawl_type;
+
+ if (!crawl_event->start_time)
+ goto out;
+
+ start_time_str = gf_strdup(ctime(&crawl_event->start_time));
+
+ if (crawl_event->end_time)
+ end_time_str = gf_strdup(ctime(&crawl_event->end_time));
+
+ ret = dict_get_int32(output, this->name, &xl_id);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "xl does not have id");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics-%d-%d-count", xl_id, child);
+ ret = dict_get_uint64(output, key, &count);
+
+ snprintf(key, sizeof(key), "statistics_healed_cnt-%d-%d-%" PRIu64, xl_id,
+ child, count);
+ ret = dict_set_uint64(output, key, healed_count);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_healed_count to output");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics_sb_cnt-%d-%d-%" PRIu64, xl_id, child,
+ count);
+ ret = dict_set_uint64(output, key, split_brain_count);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_split_brain_count to output");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics_crawl_type-%d-%d-%" PRIu64, xl_id,
+ child, count);
+ ret = dict_set_str(output, key, crawl_type);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_crawl_type to output");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics_heal_failed_cnt-%d-%d-%" PRIu64,
+ xl_id, child, count);
+ ret = dict_set_uint64(output, key, heal_failed_count);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_healed_failed_count to output");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics_strt_time-%d-%d-%" PRIu64, xl_id,
+ child, count);
+ ret = dict_set_dynstr(output, key, start_time_str);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_crawl_start_time to output");
+ goto out;
+ } else {
+ start_time_str = NULL;
+ }
+
+ if (!end_time_str)
+ progress = 1;
+ else
+ progress = 0;
+
+ snprintf(key, sizeof(key), "statistics_end_time-%d-%d-%" PRIu64, xl_id,
+ child, count);
+ if (!end_time_str)
+ end_time_str = gf_strdup("Could not determine the end time");
+ ret = dict_set_dynstr(output, key, end_time_str);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_crawl_end_time to output");
+ goto out;
+ } else {
+ end_time_str = NULL;
+ }
+
+ snprintf(key, sizeof(key), "statistics_inprogress-%d-%d-%" PRIu64, xl_id,
+ child, count);
+
+ ret = dict_set_int32(output, key, progress);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_inprogress to output");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics-%d-%d-count", xl_id, child);
+ ret = dict_set_uint64(output, key, count + 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not increment the counter.");
+ goto out;
+ }
out:
- GF_FREE (start_time_str);
- GF_FREE (end_time_str);
- return ret;
+ GF_FREE(start_time_str);
+ GF_FREE(end_time_str);
+ return ret;
}
-
int
-afr_shd_dict_add_path (xlator_t *this, dict_t *output, int child, char *path,
- struct timeval *tv)
+afr_shd_dict_add_path(xlator_t *this, dict_t *output, int child, char *path,
+ struct timeval *tv)
{
- int ret = -1;
- uint64_t count = 0;
- char key[256] = {0};
- int xl_id = 0;
-
- ret = dict_get_int32 (output, this->name, &xl_id);
+ int ret = -1;
+ uint64_t count = 0;
+ char key[256] = {0};
+ int xl_id = 0;
+
+ ret = dict_get_int32(output, this->name, &xl_id);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "xl does not have id");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "%d-%d-count", xl_id, child);
+ ret = dict_get_uint64(output, key, &count);
+
+ snprintf(key, sizeof(key), "%d-%d-%" PRIu64, xl_id, child, count);
+ ret = dict_set_dynstr(output, key, path);
+
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Could not add to output", path);
+ goto out;
+ }
+
+ if (tv) {
+ snprintf(key, sizeof(key), "%d-%d-%" PRIu64 "-time", xl_id, child,
+ count);
+ ret = dict_set_uint32(output, key, tv->tv_sec);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_GET_FAILED, "xl does not have id");
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Could not set time", path);
+ goto out;
}
+ }
- snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
- ret = dict_get_uint64 (output, key, &count);
+ snprintf(key, sizeof(key), "%d-%d-count", xl_id, child);
- snprintf (key, sizeof (key), "%d-%d-%"PRIu64, xl_id, child, count);
- ret = dict_set_dynstr (output, key, path);
+ ret = dict_set_uint64(output, key, count + 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not increment count");
+ goto out;
+ }
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_SET_FAILED, "%s: Could not add to output",
- path);
- goto out;
- }
-
- if (tv) {
- snprintf (key, sizeof (key), "%d-%d-%"PRIu64"-time", xl_id,
- child, count);
- ret = dict_set_uint32 (output, key, tv->tv_sec);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "%s: Could not set time",
- path);
- goto out;
- }
- }
-
- snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
-
- ret = dict_set_uint64 (output, key, count + 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not increment count");
- goto out;
- }
-
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-afr_add_shd_event (circular_buffer_t *cb, void *data)
+afr_add_shd_event(circular_buffer_t *cb, void *data)
{
- dict_t *output = NULL;
- xlator_t *this = THIS;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- shd_event_t *shd_event = NULL;
- char *path = NULL;
-
- output = data;
- priv = this->private;
- shd = &priv->shd;
- shd_event = cb->data;
-
- if (!shd->index_healers[shd_event->child].local)
- return 0;
-
- path = gf_strdup (shd_event->path);
- if (!path)
- return -ENOMEM;
-
- afr_shd_dict_add_path (this, output, shd_event->child, path,
- &cb->tv);
- return 0;
+ dict_t *output = NULL;
+ xlator_t *this = THIS;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ shd_event_t *shd_event = NULL;
+ char *path = NULL;
+
+ output = data;
+ priv = this->private;
+ shd = &priv->shd;
+ shd_event = cb->data;
+
+ if (!shd->index_healers[shd_event->child].local)
+ return 0;
+
+ path = gf_strdup(shd_event->path);
+ if (!path)
+ return -ENOMEM;
+
+ afr_shd_dict_add_path(this, output, shd_event->child, path, &cb->tv);
+ return 0;
}
int
-afr_add_crawl_event (circular_buffer_t *cb, void *data)
+afr_add_crawl_event(circular_buffer_t *cb, void *data)
{
- dict_t *output = NULL;
- xlator_t *this = THIS;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- crawl_event_t *crawl_event = NULL;
-
- output = data;
- priv = this->private;
- shd = &priv->shd;
- crawl_event = cb->data;
-
- if (!shd->index_healers[crawl_event->child].local)
- return 0;
+ dict_t *output = NULL;
+ xlator_t *this = THIS;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ crawl_event_t *crawl_event = NULL;
+
+ output = data;
+ priv = this->private;
+ shd = &priv->shd;
+ crawl_event = cb->data;
+
+ if (!shd->index_healers[crawl_event->child].local)
+ return 0;
- afr_shd_dict_add_crawl_event (this, output, crawl_event);
+ afr_shd_dict_add_crawl_event(this, output, crawl_event);
- return 0;
+ return 0;
}
-
int
-afr_selfheal_daemon_init (xlator_t *this)
+afr_selfheal_daemon_init(xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- int ret = -1;
- int i = 0;
-
- priv = this->private;
- shd = &priv->shd;
-
- this->itable = inode_table_new (SHD_INODE_LRU_LIMIT, this);
- if (!this->itable)
- goto out;
-
- shd->index_healers = GF_CALLOC (sizeof(*shd->index_healers),
- priv->child_count,
- gf_afr_mt_subvol_healer_t);
- if (!shd->index_healers)
- goto out;
-
- for (i = 0; i < priv->child_count; i++) {
- shd->index_healers[i].subvol = i;
- ret = afr_shd_healer_init (this, &shd->index_healers[i]);
- if (ret)
- goto out;
- }
-
- shd->full_healers = GF_CALLOC (sizeof(*shd->full_healers),
- priv->child_count,
- gf_afr_mt_subvol_healer_t);
- if (!shd->full_healers)
- goto out;
- for (i = 0; i < priv->child_count; i++) {
- shd->full_healers[i].subvol = i;
- ret = afr_shd_healer_init (this, &shd->full_healers[i]);
- if (ret)
- goto out;
- }
-
- shd->split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
- afr_destroy_shd_event_data);
- if (!shd->split_brain)
- goto out;
-
- shd->statistics = GF_CALLOC (sizeof(eh_t *), priv->child_count,
- gf_common_mt_eh_t);
- if (!shd->statistics)
- goto out;
-
- for (i = 0; i < priv->child_count ; i++) {
- shd->statistics[i] = eh_new (AFR_STATISTICS_HISTORY_SIZE,
- _gf_false,
- afr_destroy_crawl_event_data);
- if (!shd->statistics[i])
- goto out;
- shd->full_healers[i].crawl_event.child = i;
- shd->full_healers[i].crawl_event.crawl_type = "FULL";
- shd->index_healers[i].crawl_event.child = i;
- shd->index_healers[i].crawl_event.crawl_type = "INDEX";
- }
-
- ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int ret = -1;
+ int i = 0;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ this->itable = inode_table_new(SHD_INODE_LRU_LIMIT, this);
+ if (!this->itable)
+ goto out;
+
+ shd->index_healers = GF_CALLOC(sizeof(*shd->index_healers),
+ priv->child_count,
+ gf_afr_mt_subvol_healer_t);
+ if (!shd->index_healers)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ shd->index_healers[i].subvol = i;
+ ret = afr_shd_healer_init(this, &shd->index_healers[i]);
+ if (ret)
+ goto out;
+ }
+
+ shd->full_healers = GF_CALLOC(sizeof(*shd->full_healers), priv->child_count,
+ gf_afr_mt_subvol_healer_t);
+ if (!shd->full_healers)
+ goto out;
+ for (i = 0; i < priv->child_count; i++) {
+ shd->full_healers[i].subvol = i;
+ ret = afr_shd_healer_init(this, &shd->full_healers[i]);
+ if (ret)
+ goto out;
+ }
+
+ shd->split_brain = eh_new(AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
+ afr_destroy_shd_event_data);
+ if (!shd->split_brain)
+ goto out;
+
+ shd->statistics = GF_CALLOC(sizeof(eh_t *), priv->child_count,
+ gf_common_mt_eh_t);
+ if (!shd->statistics)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ shd->statistics[i] = eh_new(AFR_STATISTICS_HISTORY_SIZE, _gf_false,
+ afr_destroy_crawl_event_data);
+ if (!shd->statistics[i])
+ goto out;
+ shd->full_healers[i].crawl_event.child = i;
+ shd->full_healers[i].crawl_event.crawl_type = "FULL";
+ shd->index_healers[i].crawl_event.child = i;
+ shd->index_healers[i].crawl_event.crawl_type = "INDEX";
+ }
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
int
-afr_selfheal_childup (xlator_t *this, int subvol)
+afr_selfheal_childup(xlator_t *this, int subvol)
{
- afr_shd_index_healer_spawn (this, subvol);
+ afr_shd_index_healer_spawn(this, subvol);
- return 0;
+ return 0;
}
-
int
-afr_shd_get_index_count (xlator_t *this, int i, uint64_t *count)
+afr_shd_get_index_count(xlator_t *this, int i, uint64_t *count)
{
- afr_private_t *priv = NULL;
- xlator_t *subvol = NULL;
- loc_t rootloc = {0, };
- dict_t *xattr = NULL;
- int ret = -1;
+ afr_private_t *priv = NULL;
+ xlator_t *subvol = NULL;
+ loc_t rootloc = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ int ret = -1;
- priv = this->private;
- subvol = priv->children[i];
+ priv = this->private;
+ subvol = priv->children[i];
- rootloc.inode = inode_ref (this->itable->root);
- gf_uuid_copy (rootloc.gfid, rootloc.inode->gfid);
+ rootloc.inode = inode_ref(this->itable->root);
+ gf_uuid_copy(rootloc.gfid, rootloc.inode->gfid);
- ret = syncop_getxattr (subvol, &rootloc, &xattr,
- GF_XATTROP_INDEX_COUNT, NULL, NULL);
- if (ret < 0)
- goto out;
+ ret = syncop_getxattr(subvol, &rootloc, &xattr, GF_XATTROP_INDEX_COUNT,
+ NULL, NULL);
+ if (ret < 0)
+ goto out;
- ret = dict_get_uint64 (xattr, GF_XATTROP_INDEX_COUNT, count);
- if (ret)
- goto out;
+ ret = dict_get_uint64(xattr, GF_XATTROP_INDEX_COUNT, count);
+ if (ret)
+ goto out;
- ret = 0;
+ ret = 0;
out:
- if (xattr)
- dict_unref (xattr);
- loc_wipe (&rootloc);
+ if (xattr)
+ dict_unref(xattr);
+ loc_wipe(&rootloc);
- return ret;
+ return ret;
}
-
int
-afr_xl_op (xlator_t *this, dict_t *input, dict_t *output)
+afr_xl_op(xlator_t *this, dict_t *input, dict_t *output)
{
- gf_xl_afr_op_t op = GF_SHD_OP_INVALID;
- int ret = 0;
- int xl_id = 0;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- struct subvol_healer *healer = NULL;
- int i = 0;
- char key[64];
- int op_ret = 0;
- uint64_t cnt = 0;
-
- priv = this->private;
- shd = &priv->shd;
-
- ret = dict_get_int32 (input, "xl-op", (int32_t*)&op);
- if (ret)
- goto out;
- ret = dict_get_int32 (input, this->name, &xl_id);
- if (ret)
- goto out;
- ret = dict_set_int32 (output, this->name, xl_id);
- if (ret)
- goto out;
- switch (op) {
+ gf_xl_afr_op_t op = GF_SHD_OP_INVALID;
+ int ret = 0;
+ int xl_id = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ struct subvol_healer *healer = NULL;
+ int i = 0;
+ char key[64];
+ int op_ret = 0;
+ uint64_t cnt = 0;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ ret = dict_get_int32(input, "xl-op", (int32_t *)&op);
+ if (ret)
+ goto out;
+ ret = dict_get_int32(input, this->name, &xl_id);
+ if (ret)
+ goto out;
+ ret = dict_set_int32(output, this->name, xl_id);
+ if (ret)
+ goto out;
+ switch (op) {
case GF_SHD_OP_HEAL_INDEX:
- op_ret = 0;
-
- for (i = 0; i < priv->child_count; i++) {
- healer = &shd->index_healers[i];
- snprintf (key, sizeof (key), "%d-%d-status", xl_id, i);
-
- if (!priv->child_up[i]) {
- ret = dict_set_str (output, key,
- "Brick is not connected");
- op_ret = -1;
- } else if (AFR_COUNT (priv->child_up,
- priv->child_count) < 2) {
- ret = dict_set_str (output, key,
- "< 2 bricks in replica are up");
- op_ret = -1;
- } else if (!afr_shd_is_subvol_local (this, healer->subvol)) {
- ret = dict_set_str (output, key,
- "Brick is remote");
- } else {
- ret = dict_set_str (output, key,
- "Started self-heal");
- afr_shd_index_healer_spawn (this, i);
- }
- }
- break;
+ op_ret = 0;
+
+ for (i = 0; i < priv->child_count; i++) {
+ healer = &shd->index_healers[i];
+ snprintf(key, sizeof(key), "%d-%d-status", xl_id, i);
+
+ if (!priv->child_up[i]) {
+ ret = dict_set_str(output, key, "Brick is not connected");
+ op_ret = -1;
+ } else if (AFR_COUNT(priv->child_up, priv->child_count) < 2) {
+ ret = dict_set_str(output, key,
+ "< 2 bricks in replica are up");
+ op_ret = -1;
+ } else if (!afr_shd_is_subvol_local(this, healer->subvol)) {
+ ret = dict_set_str(output, key, "Brick is remote");
+ } else {
+ ret = dict_set_str(output, key, "Started self-heal");
+ afr_shd_index_healer_spawn(this, i);
+ }
+ }
+ break;
case GF_SHD_OP_HEAL_FULL:
- op_ret = -1;
-
- for (i = 0; i < priv->child_count; i++) {
- healer = &shd->full_healers[i];
- snprintf (key, sizeof (key), "%d-%d-status", xl_id, i);
-
- if (!priv->child_up[i]) {
- ret = dict_set_str (output, key,
- "Brick is not connected");
- } else if (AFR_COUNT (priv->child_up,
- priv->child_count) < 2) {
- ret = dict_set_str (output, key,
- "< 2 bricks in replica are up");
- } else if (!afr_shd_is_subvol_local (this, healer->subvol)) {
- ret = dict_set_str (output, key,
- "Brick is remote");
- } else {
- ret = dict_set_str (output, key,
- "Started self-heal");
- afr_shd_full_healer_spawn (this, i);
- op_ret = 0;
- }
- }
- break;
+ op_ret = -1;
+
+ for (i = 0; i < priv->child_count; i++) {
+ healer = &shd->full_healers[i];
+ snprintf(key, sizeof(key), "%d-%d-status", xl_id, i);
+
+ if (!priv->child_up[i]) {
+ ret = dict_set_str(output, key, "Brick is not connected");
+ } else if (AFR_COUNT(priv->child_up, priv->child_count) < 2) {
+ ret = dict_set_str(output, key,
+ "< 2 bricks in replica are up");
+ } else if (!afr_shd_is_subvol_local(this, healer->subvol)) {
+ ret = dict_set_str(output, key, "Brick is remote");
+ } else {
+ ret = dict_set_str(output, key, "Started self-heal");
+ afr_shd_full_healer_spawn(this, i);
+ op_ret = 0;
+ }
+ }
+ break;
case GF_SHD_OP_INDEX_SUMMARY:
- /* this case has been handled in glfs-heal.c */
- break;
+ /* this case has been handled in glfs-heal.c */
+ break;
case GF_SHD_OP_HEALED_FILES:
case GF_SHD_OP_HEAL_FAILED_FILES:
- for (i = 0; i < priv->child_count; i++) {
- snprintf (key, sizeof (key), "%d-%d-status", xl_id, i);
- ret = dict_set_str (output, key, "Operation Not "
- "Supported");
- }
- break;
+ for (i = 0; i < priv->child_count; i++) {
+ snprintf(key, sizeof(key), "%d-%d-status", xl_id, i);
+ ret = dict_set_str(output, key,
+ "Operation Not "
+ "Supported");
+ }
+ break;
case GF_SHD_OP_SPLIT_BRAIN_FILES:
- eh_dump (shd->split_brain, output, afr_add_shd_event);
- break;
+ eh_dump(shd->split_brain, output, afr_add_shd_event);
+ break;
case GF_SHD_OP_STATISTICS:
- for (i = 0; i < priv->child_count; i++) {
- eh_dump (shd->statistics[i], output,
- afr_add_crawl_event);
- afr_shd_dict_add_crawl_event (this, output,
- &shd->index_healers[i].crawl_event);
- afr_shd_dict_add_crawl_event (this, output,
- &shd->full_healers[i].crawl_event);
- }
- break;
+ for (i = 0; i < priv->child_count; i++) {
+ eh_dump(shd->statistics[i], output, afr_add_crawl_event);
+ afr_shd_dict_add_crawl_event(
+ this, output, &shd->index_healers[i].crawl_event);
+ afr_shd_dict_add_crawl_event(this, output,
+ &shd->full_healers[i].crawl_event);
+ }
+ break;
case GF_SHD_OP_STATISTICS_HEAL_COUNT:
case GF_SHD_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
- op_ret = -1;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!priv->child_up[i]) {
- snprintf (key, sizeof (key), "%d-%d-status",
- xl_id, i);
- ret = dict_set_str (output, key,
- "Brick is not connected");
- } else {
- snprintf (key, sizeof (key), "%d-%d-hardlinks",
- xl_id, i);
- ret = afr_shd_get_index_count (this, i, &cnt);
- if (ret == 0) {
- ret = dict_set_uint64 (output, key, cnt);
- }
- op_ret = 0;
- }
- }
-
- break;
+ op_ret = -1;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!priv->child_up[i]) {
+ snprintf(key, sizeof(key), "%d-%d-status", xl_id, i);
+ ret = dict_set_str(output, key, "Brick is not connected");
+ } else {
+ snprintf(key, sizeof(key), "%d-%d-hardlinks", xl_id, i);
+ ret = afr_shd_get_index_count(this, i, &cnt);
+ if (ret == 0) {
+ ret = dict_set_uint64(output, key, cnt);
+ }
+ op_ret = 0;
+ }
+ }
+
+ break;
default:
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_ARG, "Unknown set op %d", op);
- break;
- }
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_ARG,
+ "Unknown set op %d", op);
+ break;
+ }
out:
- dict_del (output, this->name);
- return op_ret;
+ dict_del(output, this->name);
+ return op_ret;
}
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index 9c587db0562..3a542ceca43 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -21,57 +21,57 @@
#include <signal.h>
typedef enum {
- AFR_TRANSACTION_PRE_OP,
- AFR_TRANSACTION_POST_OP,
+ AFR_TRANSACTION_PRE_OP,
+ AFR_TRANSACTION_POST_OP,
} afr_xattrop_type_t;
static void
-afr_lock_resume_shared (struct list_head *list);
+afr_lock_resume_shared(struct list_head *list);
void
-__afr_transaction_wake_shared (afr_local_t *local, struct list_head *shared);
+__afr_transaction_wake_shared(afr_local_t *local, struct list_head *shared);
void
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this);
+afr_changelog_post_op(call_frame_t *frame, xlator_t *this);
int
-afr_changelog_post_op_safe (call_frame_t *frame, xlator_t *this);
+afr_changelog_post_op_safe(call_frame_t *frame, xlator_t *this);
gf_boolean_t
-afr_changelog_pre_op_uninherit (call_frame_t *frame, xlator_t *this);
+afr_changelog_pre_op_uninherit(call_frame_t *frame, xlator_t *this);
gf_boolean_t
-afr_changelog_pre_op_update (call_frame_t *frame, xlator_t *this);
+afr_changelog_pre_op_update(call_frame_t *frame, xlator_t *this);
int
-afr_changelog_call_count (afr_transaction_type type,
- unsigned char *pre_op_subvols,
- unsigned char *failed_subvols,
- unsigned int child_count);
+afr_changelog_call_count(afr_transaction_type type,
+ unsigned char *pre_op_subvols,
+ unsigned char *failed_subvols,
+ unsigned int child_count);
int
-afr_changelog_do (call_frame_t *frame, xlator_t *this, dict_t *xattr,
- afr_changelog_resume_t changelog_resume,
- afr_xattrop_type_t op);
+afr_changelog_do(call_frame_t *frame, xlator_t *this, dict_t *xattr,
+ afr_changelog_resume_t changelog_resume,
+ afr_xattrop_type_t op);
void
-afr_zero_fill_stat (afr_local_t *local)
-{
- if (!local)
- return;
- if (local->transaction.type == AFR_DATA_TRANSACTION ||
- local->transaction.type == AFR_METADATA_TRANSACTION) {
- gf_zero_fill_stat (&local->cont.inode_wfop.prebuf);
- gf_zero_fill_stat (&local->cont.inode_wfop.postbuf);
- } else if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
- local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
- gf_zero_fill_stat (&local->cont.dir_fop.buf);
- gf_zero_fill_stat (&local->cont.dir_fop.preparent);
- gf_zero_fill_stat (&local->cont.dir_fop.postparent);
- if (local->transaction.type == AFR_ENTRY_TRANSACTION)
- return;
- gf_zero_fill_stat (&local->cont.dir_fop.prenewparent);
- gf_zero_fill_stat (&local->cont.dir_fop.postnewparent);
- }
+afr_zero_fill_stat(afr_local_t *local)
+{
+ if (!local)
+ return;
+ if (local->transaction.type == AFR_DATA_TRANSACTION ||
+ local->transaction.type == AFR_METADATA_TRANSACTION) {
+ gf_zero_fill_stat(&local->cont.inode_wfop.prebuf);
+ gf_zero_fill_stat(&local->cont.inode_wfop.postbuf);
+ } else if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+ local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+ gf_zero_fill_stat(&local->cont.dir_fop.buf);
+ gf_zero_fill_stat(&local->cont.dir_fop.preparent);
+ gf_zero_fill_stat(&local->cont.dir_fop.postparent);
+ if (local->transaction.type == AFR_ENTRY_TRANSACTION)
+ return;
+ gf_zero_fill_stat(&local->cont.dir_fop.prenewparent);
+ gf_zero_fill_stat(&local->cont.dir_fop.postnewparent);
+ }
}
/* In case of errors afr needs to choose which xdata from lower xlators it needs
@@ -79,2483 +79,2402 @@ afr_zero_fill_stat (afr_local_t *local)
* any good subvols which failed. Give preference to errnos other than
* ENOTCONN even if the child is source */
void
-afr_pick_error_xdata (afr_local_t *local, afr_private_t *priv,
- inode_t *inode1, unsigned char *readable1,
- inode_t *inode2, unsigned char *readable2)
-{
- int s = -1;/*selection*/
- int i = 0;
- unsigned char *readable = NULL;
-
- if (local->xdata_rsp) {
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = NULL;
- }
-
- readable = alloca0 (priv->child_count * sizeof (*readable));
- if (inode2 && readable2) {/*rename fop*/
- AFR_INTERSECT (readable, readable1, readable2,
- priv->child_count);
- } else {
- memcpy (readable, readable1,
- sizeof (*readable) * priv->child_count);
- }
-
+afr_pick_error_xdata(afr_local_t *local, afr_private_t *priv, inode_t *inode1,
+ unsigned char *readable1, inode_t *inode2,
+ unsigned char *readable2)
+{
+ int s = -1; /*selection*/
+ int i = 0;
+ unsigned char *readable = NULL;
+
+ if (local->xdata_rsp) {
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = NULL;
+ }
+
+ readable = alloca0(priv->child_count * sizeof(*readable));
+ if (inode2 && readable2) { /*rename fop*/
+ AFR_INTERSECT(readable, readable1, readable2, priv->child_count);
+ } else {
+ memcpy(readable, readable1, sizeof(*readable) * priv->child_count);
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+
+ if (local->replies[i].op_ret >= 0)
+ continue;
+
+ if (local->replies[i].op_errno == ENOTCONN)
+ continue;
+
+ /*Order is important in the following condition*/
+ if ((s < 0) || (!readable[s] && readable[i]))
+ s = i;
+ }
+
+ if (s != -1 && local->replies[s].xdata) {
+ local->xdata_rsp = dict_ref(local->replies[s].xdata);
+ } else if (s == -1) {
for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
-
- if (local->replies[i].op_ret >= 0)
- continue;
+ if (!local->replies[i].valid)
+ continue;
- if (local->replies[i].op_errno == ENOTCONN)
- continue;
+ if (local->replies[i].op_ret >= 0)
+ continue;
- /*Order is important in the following condition*/
- if ((s < 0) || (!readable[s] && readable[i]))
- s = i;
- }
-
- if (s != -1 && local->replies[s].xdata) {
- local->xdata_rsp = dict_ref (local->replies[s].xdata);
- } else if (s == -1) {
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
-
- if (local->replies[i].op_ret >= 0)
- continue;
-
- if (!local->replies[i].xdata)
- continue;
- local->xdata_rsp = dict_ref (local->replies[i].xdata);
- break;
- }
+ if (!local->replies[i].xdata)
+ continue;
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ break;
}
+ }
}
gf_boolean_t
-afr_needs_changelog_update (afr_local_t *local)
+afr_needs_changelog_update(afr_local_t *local)
{
- if (local->transaction.type == AFR_DATA_TRANSACTION)
- return _gf_true;
- if (!local->optimistic_change_log)
- return _gf_true;
- return _gf_false;
+ if (local->transaction.type == AFR_DATA_TRANSACTION)
+ return _gf_true;
+ if (!local->optimistic_change_log)
+ return _gf_true;
+ return _gf_false;
}
gf_boolean_t
-afr_changelog_has_quorum (afr_local_t *local, xlator_t *this)
+afr_changelog_has_quorum(afr_local_t *local, xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = 0;
- unsigned char *success_children = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ unsigned char *success_children = NULL;
- priv = this->private;
- success_children = alloca0 (priv->child_count);
+ priv = this->private;
+ success_children = alloca0(priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.failed_subvols[i]) {
- success_children[i] = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.failed_subvols[i]) {
+ success_children[i] = 1;
}
+ }
- if (afr_has_quorum (success_children, this)) {
- return _gf_true;
- }
+ if (afr_has_quorum(success_children, this)) {
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
-
gf_boolean_t
-afr_is_write_subvol_valid (call_frame_t *frame, xlator_t *this)
+afr_is_write_subvol_valid(call_frame_t *frame, xlator_t *this)
{
- int i = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- uint64_t write_subvol = 0;
- unsigned char *writable = NULL;
- uint16_t datamap = 0;
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ uint64_t write_subvol = 0;
+ unsigned char *writable = NULL;
+ uint16_t datamap = 0;
- local = frame->local;
- priv = this->private;
- writable = alloca0 (priv->child_count);
+ local = frame->local;
+ priv = this->private;
+ writable = alloca0(priv->child_count);
- write_subvol = afr_write_subvol_get (frame, this);
- datamap = (write_subvol & 0x00000000ffff0000) >> 16;
- for (i = 0; i < priv->child_count; i++) {
- if (datamap & (1 << i))
- writable[i] = 1;
+ write_subvol = afr_write_subvol_get(frame, this);
+ datamap = (write_subvol & 0x00000000ffff0000) >> 16;
+ for (i = 0; i < priv->child_count; i++) {
+ if (datamap & (1 << i))
+ writable[i] = 1;
- if (writable[i] && !local->transaction.failed_subvols[i])
- return _gf_true;
- }
+ if (writable[i] && !local->transaction.failed_subvols[i])
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
int
-afr_transaction_fop (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = -1;
- unsigned char *failed_subvols = NULL;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- failed_subvols = local->transaction.failed_subvols;
- call_count = priv->child_count - AFR_COUNT (failed_subvols,
- priv->child_count);
- /* Fail if pre-op did not succeed on quorum no. of bricks. */
- if (!afr_changelog_has_quorum (local, this) || !call_count) {
- local->op_ret = -1;
- /* local->op_errno is already captured in changelog cbk. */
- afr_transaction_resume (frame, this);
- return 0;
- }
+afr_transaction_fop(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ unsigned char *failed_subvols = NULL;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ failed_subvols = local->transaction.failed_subvols;
+ call_count = priv->child_count -
+ AFR_COUNT(failed_subvols, priv->child_count);
+ /* Fail if pre-op did not succeed on quorum no. of bricks. */
+ if (!afr_changelog_has_quorum(local, this) || !call_count) {
+ local->op_ret = -1;
+ /* local->op_errno is already captured in changelog cbk. */
+ afr_transaction_resume(frame, this);
+ return 0;
+ }
- /* Fail if at least one writeable brick isn't up.*/
- if (local->transaction.type == AFR_DATA_TRANSACTION &&
- !afr_is_write_subvol_valid (frame, this)) {
- local->op_ret = -1;
- local->op_errno = EIO;
- afr_transaction_resume (frame, this);
- return 0;
- }
+ /* Fail if at least one writeable brick isn't up.*/
+ if (local->transaction.type == AFR_DATA_TRANSACTION &&
+ !afr_is_write_subvol_valid(frame, this)) {
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ afr_transaction_resume(frame, this);
+ return 0;
+ }
- local->call_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] && !failed_subvols[i]) {
- local->transaction.wind (frame, this, i);
+ local->call_count = call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] && !failed_subvols[i]) {
+ local->transaction.wind(frame, this, i);
- if (!--call_count)
- break;
- }
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
}
int
-afr_transaction_done (call_frame_t *frame, xlator_t *this)
+afr_transaction_done(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t unwind = _gf_false;
- afr_lock_t *lock = NULL;
- afr_local_t *lock_local = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t unwind = _gf_false;
+ afr_lock_t *lock = NULL;
+ afr_local_t *lock_local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- if (priv->consistent_metadata) {
- LOCK (&frame->lock);
- {
- unwind = (local->transaction.main_frame != NULL);
- }
- UNLOCK (&frame->lock);
- if (unwind)/*It definitely did post-op*/
- afr_zero_fill_stat (local);
+ if (priv->consistent_metadata) {
+ LOCK(&frame->lock);
+ {
+ unwind = (local->transaction.main_frame != NULL);
}
+ UNLOCK(&frame->lock);
+ if (unwind) /*It definitely did post-op*/
+ afr_zero_fill_stat(local);
+ }
- if (local->transaction.do_eager_unlock) {
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- lock->acquired = _gf_false;
- lock->release = _gf_false;
- list_splice_init (&lock->frozen,
- &lock->waiting);
- if (list_empty (&lock->waiting))
- goto unlock;
- lock_local = list_entry (lock->waiting.next,
- afr_local_t,
- transaction.wait_list);
- list_del_init (&lock_local->transaction.wait_list);
- list_add (&lock_local->transaction.owner_list,
- &lock->owners);
- }
-unlock:
- UNLOCK (&local->inode->lock);
- }
- if (lock_local) {
- afr_lock (lock_local->transaction.frame,
- lock_local->transaction.frame->this);
- }
- local->transaction.unwind (frame, this);
+ if (local->transaction.do_eager_unlock) {
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ lock->acquired = _gf_false;
+ lock->release = _gf_false;
+ list_splice_init(&lock->frozen, &lock->waiting);
+ if (list_empty(&lock->waiting))
+ goto unlock;
+ lock_local = list_entry(lock->waiting.next, afr_local_t,
+ transaction.wait_list);
+ list_del_init(&lock_local->transaction.wait_list);
+ list_add(&lock_local->transaction.owner_list, &lock->owners);
+ }
+ unlock:
+ UNLOCK(&local->inode->lock);
+ }
+ if (lock_local) {
+ afr_lock(lock_local->transaction.frame,
+ lock_local->transaction.frame->this);
+ }
+ local->transaction.unwind(frame, this);
- AFR_STACK_DESTROY (frame);
+ AFR_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
static void
-afr_lock_fail_shared (afr_local_t *local, struct list_head *list)
-{
- afr_local_t *each = NULL;
-
- while (!list_empty(list)) {
- each = list_entry (list->next, afr_local_t,
- transaction.wait_list);
- list_del_init(&each->transaction.wait_list);
- each->op_ret = -1;
- each->op_errno = local->op_errno;
- afr_transaction_done (each->transaction.frame,
- each->transaction.frame->this);
- }
+afr_lock_fail_shared(afr_local_t *local, struct list_head *list)
+{
+ afr_local_t *each = NULL;
+
+ while (!list_empty(list)) {
+ each = list_entry(list->next, afr_local_t, transaction.wait_list);
+ list_del_init(&each->transaction.wait_list);
+ each->op_ret = -1;
+ each->op_errno = local->op_errno;
+ afr_transaction_done(each->transaction.frame,
+ each->transaction.frame->this);
+ }
}
static void
-afr_handle_lock_acquire_failure (afr_local_t *local, gf_boolean_t locked)
+afr_handle_lock_acquire_failure(afr_local_t *local, gf_boolean_t locked)
{
- struct list_head shared;
- afr_lock_t *lock = NULL;
+ struct list_head shared;
+ afr_lock_t *lock = NULL;
- if (!local->transaction.eager_lock_on)
- goto out;
+ if (!local->transaction.eager_lock_on)
+ goto out;
- lock = &local->inode_ctx->lock[local->transaction.type];
+ lock = &local->inode_ctx->lock[local->transaction.type];
- INIT_LIST_HEAD (&shared);
- LOCK (&local->inode->lock);
- {
- lock->release = _gf_true;
- list_splice_init (&lock->waiting, &shared);
- }
- UNLOCK (&local->inode->lock);
+ INIT_LIST_HEAD(&shared);
+ LOCK(&local->inode->lock);
+ {
+ lock->release = _gf_true;
+ list_splice_init(&lock->waiting, &shared);
+ }
+ UNLOCK(&local->inode->lock);
- afr_lock_fail_shared (local, &shared);
- local->transaction.do_eager_unlock = _gf_true;
+ afr_lock_fail_shared(local, &shared);
+ local->transaction.do_eager_unlock = _gf_true;
out:
- if (locked) {
- local->internal_lock.lock_cbk = afr_transaction_done;
- afr_unlock (local->transaction.frame,
- local->transaction.frame->this);
- } else {
- afr_transaction_done (local->transaction.frame,
- local->transaction.frame->this);
- }
+ if (locked) {
+ local->internal_lock.lock_cbk = afr_transaction_done;
+ afr_unlock(local->transaction.frame, local->transaction.frame->this);
+ } else {
+ afr_transaction_done(local->transaction.frame,
+ local->transaction.frame->this);
+ }
}
-call_frame_t*
-afr_transaction_detach_fop_frame (call_frame_t *frame)
+call_frame_t *
+afr_transaction_detach_fop_frame(call_frame_t *frame)
{
- afr_local_t * local = NULL;
- call_frame_t *fop_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *fop_frame = NULL;
- local = frame->local;
+ local = frame->local;
- afr_handle_inconsistent_fop (frame, &local->op_ret, &local->op_errno);
- LOCK (&frame->lock);
- {
- fop_frame = local->transaction.main_frame;
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
+ afr_handle_inconsistent_fop(frame, &local->op_ret, &local->op_errno);
+ LOCK(&frame->lock);
+ {
+ fop_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK(&frame->lock);
- return fop_frame;
+ return fop_frame;
}
-
static void
-afr_save_lk_owner (call_frame_t *frame)
+afr_save_lk_owner(call_frame_t *frame)
{
- afr_local_t * local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->saved_lk_owner = frame->root->lk_owner;
+ local->saved_lk_owner = frame->root->lk_owner;
}
-
static void
-afr_restore_lk_owner (call_frame_t *frame)
+afr_restore_lk_owner(call_frame_t *frame)
{
- afr_local_t * local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- frame->root->lk_owner = local->saved_lk_owner;
+ frame->root->lk_owner = local->saved_lk_owner;
}
void
-__mark_all_success (call_frame_t *frame, xlator_t *this)
+__mark_all_success(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- local->transaction.failed_subvols[i] = 0;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ local->transaction.failed_subvols[i] = 0;
+ }
}
void
-afr_compute_pre_op_sources (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_transaction_type type = -1;
- dict_t *xdata = NULL;
- int **matrix = NULL;
- int idx = -1;
- int i = 0;
- int j = 0;
-
- priv = this->private;
- local = frame->local;
- type = local->transaction.type;
- idx = afr_index_for_transaction_type (type);
- matrix = ALLOC_MATRIX (priv->child_count, int);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.changelog_xdata[i])
- continue;
- xdata = local->transaction.changelog_xdata[i];
- afr_selfheal_fill_matrix (this, matrix, i, idx, xdata);
- }
-
- memset (local->transaction.pre_op_sources, 1, priv->child_count);
-
- /*If lock or pre-op failed on a brick, it is not a source. */
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.failed_subvols[i])
- local->transaction.pre_op_sources[i] = 0;
- }
-
- /* If brick is blamed by others, it is not a source. */
- for (i = 0; i < priv->child_count; i++)
- for (j = 0; j < priv->child_count; j++)
- if (matrix[i][j] != 0)
- local->transaction.pre_op_sources[j] = 0;
+afr_compute_pre_op_sources(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_transaction_type type = -1;
+ dict_t *xdata = NULL;
+ int **matrix = NULL;
+ int idx = -1;
+ int i = 0;
+ int j = 0;
+
+ priv = this->private;
+ local = frame->local;
+ type = local->transaction.type;
+ idx = afr_index_for_transaction_type(type);
+ matrix = ALLOC_MATRIX(priv->child_count, int);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.changelog_xdata[i])
+ continue;
+ xdata = local->transaction.changelog_xdata[i];
+ afr_selfheal_fill_matrix(this, matrix, i, idx, xdata);
+ }
+
+ memset(local->transaction.pre_op_sources, 1, priv->child_count);
+
+ /*If lock or pre-op failed on a brick, it is not a source. */
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.failed_subvols[i])
+ local->transaction.pre_op_sources[i] = 0;
+ }
+
+ /* If brick is blamed by others, it is not a source. */
+ for (i = 0; i < priv->child_count; i++)
+ for (j = 0; j < priv->child_count; j++)
+ if (matrix[i][j] != 0)
+ local->transaction.pre_op_sources[j] = 0;
}
gf_boolean_t
-afr_has_arbiter_fop_cbk_quorum (call_frame_t *frame)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- xlator_t *this = NULL;
- gf_boolean_t fop_failed = _gf_false;
- unsigned char *pre_op_sources = NULL;
- int i = 0;
-
- local = frame->local;
- this = frame->this;
- priv = this->private;
- pre_op_sources = local->transaction.pre_op_sources;
-
- /* If the fop failed on the brick, it is not a source. */
- for (i = 0; i < priv->child_count; i++)
- if (local->transaction.failed_subvols[i])
- pre_op_sources[i] = 0;
-
- switch (AFR_COUNT (pre_op_sources, priv->child_count)) {
+afr_has_arbiter_fop_cbk_quorum(call_frame_t *frame)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t *this = NULL;
+ gf_boolean_t fop_failed = _gf_false;
+ unsigned char *pre_op_sources = NULL;
+ int i = 0;
+
+ local = frame->local;
+ this = frame->this;
+ priv = this->private;
+ pre_op_sources = local->transaction.pre_op_sources;
+
+ /* If the fop failed on the brick, it is not a source. */
+ for (i = 0; i < priv->child_count; i++)
+ if (local->transaction.failed_subvols[i])
+ pre_op_sources[i] = 0;
+
+ switch (AFR_COUNT(pre_op_sources, priv->child_count)) {
case 1:
- if (pre_op_sources[ARBITER_BRICK_INDEX])
- fop_failed = _gf_true;
- break;
- case 0:
+ if (pre_op_sources[ARBITER_BRICK_INDEX])
fop_failed = _gf_true;
- break;
- }
+ break;
+ case 0:
+ fop_failed = _gf_true;
+ break;
+ }
- if (fop_failed)
- return _gf_false;
+ if (fop_failed)
+ return _gf_false;
- return _gf_true;
+ return _gf_true;
}
void
-afr_txn_arbitrate_fop (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int pre_op_sources_count = 0;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
-
- afr_compute_pre_op_sources (frame, this);
- pre_op_sources_count = AFR_COUNT (local->transaction.pre_op_sources,
- priv->child_count);
-
- /* If arbiter is the only source, do not proceed. */
- if (pre_op_sources_count < 2 &&
- local->transaction.pre_op_sources[ARBITER_BRICK_INDEX]) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- for (i = 0; i < priv->child_count; i++)
- local->transaction.failed_subvols[i] = 1;
- }
+afr_txn_arbitrate_fop(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int pre_op_sources_count = 0;
+ int i = 0;
- afr_transaction_fop (frame, this);
+ priv = this->private;
+ local = frame->local;
- return;
+ afr_compute_pre_op_sources(frame, this);
+ pre_op_sources_count = AFR_COUNT(local->transaction.pre_op_sources,
+ priv->child_count);
+
+ /* If arbiter is the only source, do not proceed. */
+ if (pre_op_sources_count < 2 &&
+ local->transaction.pre_op_sources[ARBITER_BRICK_INDEX]) {
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ for (i = 0; i < priv->child_count; i++)
+ local->transaction.failed_subvols[i] = 1;
+ }
+
+ afr_transaction_fop(frame, this);
+
+ return;
}
int
-afr_transaction_perform_fop (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int ret = 0;
- int failure_count = 0;
- struct list_head shared;
- afr_lock_t *lock = NULL;
-
- local = frame->local;
- priv = this->private;
-
- INIT_LIST_HEAD (&shared);
- if (local->transaction.type == AFR_DATA_TRANSACTION &&
- !local->transaction.inherited) {
- ret = afr_write_subvol_set (frame, this);
- if (ret) {
- /*act as if operation failed on all subvols*/
- local->op_ret = -1;
- local->op_errno = -ret;
- for (i = 0; i < priv->child_count; i++)
- local->transaction.failed_subvols[i] = 1;
- }
- }
-
- if (local->pre_op_compat)
- /* old mode, pre-op was done as afr_changelog_do()
- just now, before OP */
- afr_changelog_pre_op_update (frame, this);
-
- if (!local->transaction.eager_lock_on ||
- local->transaction.inherited)
- goto fop;
- failure_count = AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count);
- if (failure_count == priv->child_count) {
- afr_handle_lock_acquire_failure (local, _gf_true);
- return 0;
- } else {
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- lock->acquired = _gf_true;
- __afr_transaction_wake_shared (local, &shared);
- }
- UNLOCK (&local->inode->lock);
+afr_transaction_perform_fop(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int ret = 0;
+ int failure_count = 0;
+ struct list_head shared;
+ afr_lock_t *lock = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ INIT_LIST_HEAD(&shared);
+ if (local->transaction.type == AFR_DATA_TRANSACTION &&
+ !local->transaction.inherited) {
+ ret = afr_write_subvol_set(frame, this);
+ if (ret) {
+ /*act as if operation failed on all subvols*/
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ for (i = 0; i < priv->child_count; i++)
+ local->transaction.failed_subvols[i] = 1;
+ }
+ }
+
+ if (local->pre_op_compat)
+ /* old mode, pre-op was done as afr_changelog_do()
+ just now, before OP */
+ afr_changelog_pre_op_update(frame, this);
+
+ if (!local->transaction.eager_lock_on || local->transaction.inherited)
+ goto fop;
+ failure_count = AFR_COUNT(local->transaction.failed_subvols,
+ priv->child_count);
+ if (failure_count == priv->child_count) {
+ afr_handle_lock_acquire_failure(local, _gf_true);
+ return 0;
+ } else {
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ lock->acquired = _gf_true;
+ __afr_transaction_wake_shared(local, &shared);
}
+ UNLOCK(&local->inode->lock);
+ }
fop:
- /* Perform fops with the lk-owner from top xlator.
- * Eg: lk-owner of posix-lk and flush should be same,
- * flush cant clear the posix-lks without that lk-owner.
- */
- afr_save_lk_owner (frame);
- frame->root->lk_owner =
- local->transaction.main_frame->root->lk_owner;
+ /* Perform fops with the lk-owner from top xlator.
+ * Eg: lk-owner of posix-lk and flush should be same,
+ * flush cant clear the posix-lks without that lk-owner.
+ */
+ afr_save_lk_owner(frame);
+ frame->root->lk_owner = local->transaction.main_frame->root->lk_owner;
- if (priv->arbiter_count == 1) {
- afr_txn_arbitrate_fop (frame, this);
- } else {
- afr_transaction_fop (frame, this);
- }
+ if (priv->arbiter_count == 1) {
+ afr_txn_arbitrate_fop(frame, this);
+ } else {
+ afr_transaction_fop(frame, this);
+ }
- afr_lock_resume_shared (&shared);
- return 0;
+ afr_lock_resume_shared(&shared);
+ return 0;
}
int
-afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int **pending)
+afr_set_pending_dict(afr_private_t *priv, dict_t *xattr, int **pending)
{
- int i = 0;
- int ret = 0;
-
- for (i = 0; i < priv->child_count; i++) {
+ int i = 0;
+ int ret = 0;
- ret = dict_set_static_bin (xattr, priv->pending_key[i],
- pending[i],
- AFR_NUM_CHANGE_LOGS * sizeof (int));
- /* 3 = data+metadata+entry */
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_set_static_bin(xattr, priv->pending_key[i], pending[i],
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ /* 3 = data+metadata+entry */
- if (ret)
- break;
- }
+ if (ret)
+ break;
+ }
- return ret;
+ return ret;
}
/* {{{ pending */
int
-afr_changelog_post_op_done (call_frame_t *frame, xlator_t *this)
+afr_changelog_post_op_done(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- /* Fail the FOP if post-op did not succeed on quorum no. of bricks. */
- if (!afr_changelog_has_quorum (local, this)) {
- local->op_ret = -1;
- /*local->op_errno is already captured in changelog cbk*/
- }
+ /* Fail the FOP if post-op did not succeed on quorum no. of bricks. */
+ if (!afr_changelog_has_quorum(local, this)) {
+ local->op_ret = -1;
+ /*local->op_errno is already captured in changelog cbk*/
+ }
- if (local->transaction.resume_stub) {
- call_resume (local->transaction.resume_stub);
- local->transaction.resume_stub = NULL;
- }
+ if (local->transaction.resume_stub) {
+ call_resume(local->transaction.resume_stub);
+ local->transaction.resume_stub = NULL;
+ }
- int_lock->lock_cbk = afr_transaction_done;
- afr_unlock (frame, this);
+ int_lock->lock_cbk = afr_transaction_done;
+ afr_unlock(frame, this);
- return 0;
+ return 0;
}
-
-unsigned char*
-afr_locked_nodes_get (afr_transaction_type type, afr_internal_lock_t *int_lock)
+unsigned char *
+afr_locked_nodes_get(afr_transaction_type type, afr_internal_lock_t *int_lock)
{
- unsigned char *locked_nodes = NULL;
- switch (type) {
+ unsigned char *locked_nodes = NULL;
+ switch (type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
- locked_nodes = int_lock->locked_nodes;
- break;
+ locked_nodes = int_lock->locked_nodes;
+ break;
case AFR_ENTRY_TRANSACTION:
case AFR_ENTRY_RENAME_TRANSACTION:
- /*Because same set of subvols participate in all lockee
- * entities*/
- locked_nodes = int_lock->lockee[0].locked_nodes;
- break;
- }
- return locked_nodes;
+ /*Because same set of subvols participate in all lockee
+ * entities*/
+ locked_nodes = int_lock->lockee[0].locked_nodes;
+ break;
+ }
+ return locked_nodes;
}
-
int
-afr_changelog_call_count (afr_transaction_type type,
- unsigned char *pre_op_subvols,
- unsigned char *failed_subvols,
- unsigned int child_count)
+afr_changelog_call_count(afr_transaction_type type,
+ unsigned char *pre_op_subvols,
+ unsigned char *failed_subvols,
+ unsigned int child_count)
{
- int i = 0;
- int call_count = 0;
+ int i = 0;
+ int call_count = 0;
- for (i = 0; i < child_count; i++) {
- if (pre_op_subvols[i] && !failed_subvols[i]) {
- call_count++;
- }
+ for (i = 0; i < child_count; i++) {
+ if (pre_op_subvols[i] && !failed_subvols[i]) {
+ call_count++;
}
+ }
- if (type == AFR_ENTRY_RENAME_TRANSACTION)
- call_count *= 2;
+ if (type == AFR_ENTRY_RENAME_TRANSACTION)
+ call_count *= 2;
- return call_count;
+ return call_count;
}
-
gf_boolean_t
-afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this)
+afr_txn_nothing_failed(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (priv->thin_arbiter_count) {
- /* We need to perform post-op even if 1 data brick was down
- * before the txn started.*/
- if (AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count))
- return _gf_false;
- }
+ if (priv->thin_arbiter_count) {
+ /* We need to perform post-op even if 1 data brick was down
+ * before the txn started.*/
+ if (AFR_COUNT(local->transaction.failed_subvols, priv->child_count))
+ return _gf_false;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] &&
- local->transaction.failed_subvols[i])
- return _gf_false;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] &&
+ local->transaction.failed_subvols[i])
+ return _gf_false;
+ }
- return _gf_true;
+ return _gf_true;
}
void
-afr_handle_symmetric_errors (call_frame_t *frame, xlator_t *this)
+afr_handle_symmetric_errors(call_frame_t *frame, xlator_t *this)
{
- if (afr_is_symmetric_error (frame, this))
- __mark_all_success (frame, this);
+ if (afr_is_symmetric_error(frame, this))
+ __mark_all_success(frame, this);
}
gf_boolean_t
-afr_has_quorum (unsigned char *subvols, xlator_t *this)
-{
- unsigned int quorum_count = 0;
- afr_private_t *priv = NULL;
- unsigned int up_children_count = 0;
-
- priv = this->private;
- up_children_count = AFR_COUNT (subvols, priv->child_count);
-
- if (priv->quorum_count == AFR_QUORUM_AUTO) {
- /*
- * Special case for auto-quorum with an even number of nodes.
- *
- * A replica set with even count N can only handle the same
- * number of failures as odd N-1 before losing "vanilla"
- * quorum, and the probability of more simultaneous failures is
- * actually higher. For example, with a 1% chance of failure
- * we'd have a 0.03% chance of two simultaneous failures with
- * N=3 but a 0.06% chance with N=4. However, the special case
- * is necessary for N=2 because there's no real quorum in that
- * case (i.e. can't normally survive *any* failures). In that
- * case, we treat the first node as a tie-breaker, allowing
- * quorum to be retained in some cases while still honoring the
- * all-important constraint that there can not simultaneously
- * be two partitioned sets of nodes each believing they have
- * quorum. Of two equally sized sets, the one without that
- * first node will lose.
- *
- * It turns out that the special case is beneficial for higher
- * values of N as well. Continuing the example above, the
- * probability of losing quorum with N=4 and this type of
- * quorum is (very) slightly lower than with N=3 and vanilla
- * quorum. The difference becomes even more pronounced with
- * higher N. Therefore, even though such replica counts are
- * unlikely to be seen in practice, we might as well use the
- * "special" quorum then as well.
- */
- if ((up_children_count * 2) == priv->child_count) {
- return subvols[0];
- }
- }
+afr_has_quorum(unsigned char *subvols, xlator_t *this)
+{
+ unsigned int quorum_count = 0;
+ afr_private_t *priv = NULL;
+ unsigned int up_children_count = 0;
- if (priv->quorum_count == AFR_QUORUM_AUTO) {
- quorum_count = priv->child_count/2 + 1;
- } else {
- quorum_count = priv->quorum_count;
+ priv = this->private;
+ up_children_count = AFR_COUNT(subvols, priv->child_count);
+
+ if (priv->quorum_count == AFR_QUORUM_AUTO) {
+ /*
+ * Special case for auto-quorum with an even number of nodes.
+ *
+ * A replica set with even count N can only handle the same
+ * number of failures as odd N-1 before losing "vanilla"
+ * quorum, and the probability of more simultaneous failures is
+ * actually higher. For example, with a 1% chance of failure
+ * we'd have a 0.03% chance of two simultaneous failures with
+ * N=3 but a 0.06% chance with N=4. However, the special case
+ * is necessary for N=2 because there's no real quorum in that
+ * case (i.e. can't normally survive *any* failures). In that
+ * case, we treat the first node as a tie-breaker, allowing
+ * quorum to be retained in some cases while still honoring the
+ * all-important constraint that there can not simultaneously
+ * be two partitioned sets of nodes each believing they have
+ * quorum. Of two equally sized sets, the one without that
+ * first node will lose.
+ *
+ * It turns out that the special case is beneficial for higher
+ * values of N as well. Continuing the example above, the
+ * probability of losing quorum with N=4 and this type of
+ * quorum is (very) slightly lower than with N=3 and vanilla
+ * quorum. The difference becomes even more pronounced with
+ * higher N. Therefore, even though such replica counts are
+ * unlikely to be seen in practice, we might as well use the
+ * "special" quorum then as well.
+ */
+ if ((up_children_count * 2) == priv->child_count) {
+ return subvols[0];
}
+ }
- if (up_children_count >= quorum_count)
- return _gf_true;
+ if (priv->quorum_count == AFR_QUORUM_AUTO) {
+ quorum_count = priv->child_count / 2 + 1;
+ } else {
+ quorum_count = priv->quorum_count;
+ }
- return _gf_false;
+ if (up_children_count >= quorum_count)
+ return _gf_true;
+
+ return _gf_false;
}
static gf_boolean_t
-afr_has_fop_quorum (call_frame_t *frame)
+afr_has_fop_quorum(call_frame_t *frame)
{
- xlator_t *this = frame->this;
- afr_local_t *local = frame->local;
- unsigned char *locked_nodes = NULL;
+ xlator_t *this = frame->this;
+ afr_local_t *local = frame->local;
+ unsigned char *locked_nodes = NULL;
- locked_nodes = afr_locked_nodes_get (local->transaction.type,
- &local->internal_lock);
- return afr_has_quorum (locked_nodes, this);
+ locked_nodes = afr_locked_nodes_get(local->transaction.type,
+ &local->internal_lock);
+ return afr_has_quorum(locked_nodes, this);
}
static gf_boolean_t
-afr_has_fop_cbk_quorum (call_frame_t *frame)
+afr_has_fop_cbk_quorum(call_frame_t *frame)
{
- afr_local_t *local = frame->local;
- xlator_t *this = frame->this;
- afr_private_t *priv = this->private;
- unsigned char *success = alloca0(priv->child_count);
- int i = 0;
+ afr_local_t *local = frame->local;
+ xlator_t *this = frame->this;
+ afr_private_t *priv = this->private;
+ unsigned char *success = alloca0(priv->child_count);
+ int i = 0;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i])
- if (!local->transaction.failed_subvols[i])
- success[i] = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i])
+ if (!local->transaction.failed_subvols[i])
+ success[i] = 1;
+ }
- return afr_has_quorum (success, this);
+ return afr_has_quorum(success, this);
}
gf_boolean_t
-afr_need_dirty_marking (call_frame_t *frame, xlator_t *this)
+afr_need_dirty_marking(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = this->private;
- afr_local_t *local = NULL;
- gf_boolean_t need_dirty = _gf_false;
+ afr_private_t *priv = this->private;
+ afr_local_t *local = NULL;
+ gf_boolean_t need_dirty = _gf_false;
- local = frame->local;
+ local = frame->local;
- if (!priv->quorum_count || !local->optimistic_change_log)
- return _gf_false;
+ if (!priv->quorum_count || !local->optimistic_change_log)
+ return _gf_false;
- if (local->transaction.type == AFR_DATA_TRANSACTION ||
- local->transaction.type == AFR_METADATA_TRANSACTION)
- return _gf_false;
+ if (local->transaction.type == AFR_DATA_TRANSACTION ||
+ local->transaction.type == AFR_METADATA_TRANSACTION)
+ return _gf_false;
- if (AFR_COUNT (local->transaction.failed_subvols, priv->child_count) ==
- priv->child_count)
- return _gf_false;
+ if (AFR_COUNT(local->transaction.failed_subvols, priv->child_count) ==
+ priv->child_count)
+ return _gf_false;
- if (priv->arbiter_count) {
- if (!afr_has_arbiter_fop_cbk_quorum (frame))
- need_dirty = _gf_true;
- } else if (!afr_has_fop_cbk_quorum (frame)) {
- need_dirty = _gf_true;
- }
+ if (priv->arbiter_count) {
+ if (!afr_has_arbiter_fop_cbk_quorum(frame))
+ need_dirty = _gf_true;
+ } else if (!afr_has_fop_cbk_quorum(frame)) {
+ need_dirty = _gf_true;
+ }
- return need_dirty;
+ return need_dirty;
}
void
-afr_handle_quorum (call_frame_t *frame, xlator_t *this)
+afr_handle_quorum(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- const char *file = NULL;
- uuid_t gfid = {0};
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ const char *file = NULL;
+ uuid_t gfid = {0};
- local = frame->local;
- priv = frame->this->private;
+ local = frame->local;
+ priv = frame->this->private;
- if (priv->quorum_count == 0)
- return;
-
- /* If the fop already failed return right away to preserve errno */
- if (local->op_ret == -1)
- return;
+ if (priv->quorum_count == 0)
+ return;
- /*
- * Network split may happen just after the fops are unwound, so check
- * if the fop succeeded in a way it still follows quorum. If it doesn't,
- * mark the fop as failure, mark the changelogs so it reflects that
- * failure.
- *
- * Scenario:
- * There are 3 mounts on 3 machines(node1, node2, node3) all writing to
- * single file. Network split happened in a way that node1 can't see
- * node2, node3. Node2, node3 both of them can't see node1. Now at the
- * time of sending write all the bricks are up. Just after write fop is
- * wound on node1, network split happens. Node1 thinks write fop failed
- * on node2, node3 so marks pending changelog for those 2 extended
- * attributes on node1. Node2, node3 thinks writes failed on node1 so
- * they mark pending changelog for node1. When the network is stable
- * again the file already is in split-brain. These checks prevent
- * marking pending changelog on other subvolumes if the fop doesn't
- * succeed in a way it is still following quorum. So with this fix what
- * is happening is, node1 will have all pending changelog(FOOL) because
- * the write succeeded only on node1 but failed on node2, node3 so
- * instead of marking pending changelogs on node2, node3 it just treats
- * the fop as failure and goes into DIRTY state. Where as node2, node3
- * say they are sources and have pending changelog to node1 so there is
- * no split-brain with the fix. The problem is eliminated completely.
- */
+ /* If the fop already failed return right away to preserve errno */
+ if (local->op_ret == -1)
+ return;
- if (priv->arbiter_count) {
- if (afr_has_arbiter_fop_cbk_quorum (frame))
- return;
- } else if (afr_has_fop_cbk_quorum (frame)) {
- return;
- }
+ /*
+ * Network split may happen just after the fops are unwound, so check
+ * if the fop succeeded in a way it still follows quorum. If it doesn't,
+ * mark the fop as failure, mark the changelogs so it reflects that
+ * failure.
+ *
+ * Scenario:
+ * There are 3 mounts on 3 machines(node1, node2, node3) all writing to
+ * single file. Network split happened in a way that node1 can't see
+ * node2, node3. Node2, node3 both of them can't see node1. Now at the
+ * time of sending write all the bricks are up. Just after write fop is
+ * wound on node1, network split happens. Node1 thinks write fop failed
+ * on node2, node3 so marks pending changelog for those 2 extended
+ * attributes on node1. Node2, node3 thinks writes failed on node1 so
+ * they mark pending changelog for node1. When the network is stable
+ * again the file already is in split-brain. These checks prevent
+ * marking pending changelog on other subvolumes if the fop doesn't
+ * succeed in a way it is still following quorum. So with this fix what
+ * is happening is, node1 will have all pending changelog(FOOL) because
+ * the write succeeded only on node1 but failed on node2, node3 so
+ * instead of marking pending changelogs on node2, node3 it just treats
+ * the fop as failure and goes into DIRTY state. Where as node2, node3
+ * say they are sources and have pending changelog to node1 so there is
+ * no split-brain with the fix. The problem is eliminated completely.
+ */
+
+ if (priv->arbiter_count) {
+ if (afr_has_arbiter_fop_cbk_quorum(frame))
+ return;
+ } else if (afr_has_fop_cbk_quorum(frame)) {
+ return;
+ }
- if (afr_need_dirty_marking (frame, this))
- goto set_response;
+ if (afr_need_dirty_marking(frame, this))
+ goto set_response;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i])
- afr_transaction_fop_failed (frame, frame->this, i);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i])
+ afr_transaction_fop_failed(frame, frame->this, i);
+ }
set_response:
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
- if (local->op_errno == 0)
- local->op_errno = afr_quorum_errno (priv);
-
- if (local->fd) {
- gf_uuid_copy (gfid, local->fd->inode->gfid);
- file = uuid_utoa (gfid);
- } else {
- loc_path (&local->loc, local->loc.name);
- file = local->loc.path;
- }
-
- gf_msg (frame->this->name, GF_LOG_WARNING, local->op_errno,
- AFR_MSG_QUORUM_FAIL, "%s: Failing %s as quorum is not met",
- file, gf_fop_list[local->op]);
-
- switch (local->transaction.type) {
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ if (local->op_errno == 0)
+ local->op_errno = afr_quorum_errno(priv);
+
+ if (local->fd) {
+ gf_uuid_copy(gfid, local->fd->inode->gfid);
+ file = uuid_utoa(gfid);
+ } else {
+ loc_path(&local->loc, local->loc.name);
+ file = local->loc.path;
+ }
+
+ gf_msg(frame->this->name, GF_LOG_WARNING, local->op_errno,
+ AFR_MSG_QUORUM_FAIL, "%s: Failing %s as quorum is not met", file,
+ gf_fop_list[local->op]);
+
+ switch (local->transaction.type) {
case AFR_ENTRY_TRANSACTION:
case AFR_ENTRY_RENAME_TRANSACTION:
- afr_pick_error_xdata (local, priv, local->parent,
- local->readable, local->parent2,
- local->readable2);
- break;
+ afr_pick_error_xdata(local, priv, local->parent, local->readable,
+ local->parent2, local->readable2);
+ break;
default:
- afr_pick_error_xdata (local, priv, local->inode,
- local->readable, NULL, NULL);
- break;
- }
+ afr_pick_error_xdata(local, priv, local->inode, local->readable,
+ NULL, NULL);
+ break;
+ }
}
int
-afr_fill_ta_loc (xlator_t *this, loc_t *loc)
-{
- afr_private_t *priv = NULL;
-
- priv = this->private;
- loc->parent = inode_ref (priv->root_inode);
- gf_uuid_copy (loc->pargfid, loc->parent->gfid);
- loc->name = priv->pending_key[THIN_ARBITER_BRICK_INDEX];
- gf_uuid_copy (loc->gfid, priv->ta_gfid);
- loc->inode = inode_new (loc->parent->table);
- if (!loc->inode) {
- loc_wipe(loc);
- return -ENOMEM;
- }
- return 0;
+afr_fill_ta_loc(xlator_t *this, loc_t *loc)
+{
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ loc->parent = inode_ref(priv->root_inode);
+ gf_uuid_copy(loc->pargfid, loc->parent->gfid);
+ loc->name = priv->pending_key[THIN_ARBITER_BRICK_INDEX];
+ gf_uuid_copy(loc->gfid, priv->ta_gfid);
+ loc->inode = inode_new(loc->parent->table);
+ if (!loc->inode) {
+ loc_wipe(loc);
+ return -ENOMEM;
+ }
+ return 0;
}
int
-afr_changelog_thin_arbiter_post_op (xlator_t *this, afr_local_t *local)
-{
- int ret = 0;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- int failed_count = 0;
- struct gf_flock flock = {0, };
- loc_t loc = {0,};
- int i = 0;
+afr_changelog_thin_arbiter_post_op(xlator_t *this, afr_local_t *local)
+{
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int failed_count = 0;
+ struct gf_flock flock = {
+ 0,
+ };
+ loc_t loc = {
+ 0,
+ };
+ int i = 0;
+
+ priv = this->private;
+ if (!priv->thin_arbiter_count)
+ return 0;
- priv = this->private;
- if (!priv->thin_arbiter_count)
- return 0;
+ failed_count = AFR_COUNT(local->transaction.failed_subvols,
+ priv->child_count);
+ if (!failed_count)
+ return 0;
+ GF_ASSERT(failed_count == 1);
+ ret = afr_fill_ta_loc(this, &loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate thin-arbiter loc for: %s.", loc.name);
+ goto out;
+ }
+
+ xattr = dict_new();
+ if (!xattr) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_set_static_bin(xattr, priv->pending_key[i],
+ local->pending[i],
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ if (ret)
+ goto out;
+ }
+
+ flock.l_type = F_WRLCK;
+ flock.l_start = 0;
+ flock.l_len = 0;
+
+ /*TODO: Convert to two domain locking. */
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, &loc, F_SETLKW, &flock, NULL, NULL);
+ if (ret)
+ goto out;
+
+ ret = syncop_xattrop(priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
+ GF_XATTROP_ADD_ARRAY, xattr, NULL, NULL, NULL);
+
+ if (ret == -EINVAL) {
+ gf_msg(this->name, GF_LOG_INFO, -ret, AFR_MSG_THIN_ARB,
+ "Thin-arbiter has denied post-op on %s for gfid %s.",
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX],
+ uuid_utoa(local->inode->gfid));
+
+ } else if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Post-op on thin-arbiter id file %s failed for gfid %s.",
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX],
+ uuid_utoa(local->inode->gfid));
+ }
+ flock.l_type = F_UNLCK;
+ syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX], AFR_TA_DOM_NOTIFY,
+ &loc, F_SETLK, &flock, NULL, NULL);
+out:
+ if (xattr)
+ dict_unref(xattr);
- failed_count = AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count);
- if (!failed_count)
- return 0;
+ return ret;
+}
- GF_ASSERT (failed_count == 1);
- ret = afr_fill_ta_loc (this, &loc);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed to populate thin-arbiter loc for: %s.",
- loc.name);
- goto out;
- }
+int
+afr_changelog_post_op_now(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ afr_local_t *local = NULL;
+ dict_t *xattr = NULL;
+ int i = 0;
+ int ret = 0;
+ int idx = 0;
+ int nothing_failed = 1;
+ gf_boolean_t need_undirty = _gf_false;
+
+ afr_handle_quorum(frame, this);
+ local = frame->local;
+ idx = afr_index_for_transaction_type(local->transaction.type);
+
+ xattr = dict_new();
+ if (!xattr) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
- xattr = dict_new ();
- if (!xattr) {
- ret = -ENOMEM;
- goto out;
- }
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_static_bin (xattr, priv->pending_key[i],
- local->pending[i],
- AFR_NUM_CHANGE_LOGS * sizeof (int));
- if (ret)
- goto out;
+ nothing_failed = afr_txn_nothing_failed(frame, this);
+
+ if (afr_changelog_pre_op_uninherit(frame, this))
+ need_undirty = _gf_false;
+ else
+ need_undirty = _gf_true;
+
+ if (local->op_ret < 0 && !nothing_failed) {
+ if (afr_need_dirty_marking(frame, this)) {
+ local->dirty[idx] = hton32(1);
+ goto set_dirty;
}
- flock.l_type = F_WRLCK;
- flock.l_start = 0;
- flock.l_len = 0;
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
- /*TODO: Convert to two domain locking. */
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_NOTIFY, &loc, F_SETLKW, &flock,
- NULL, NULL);
- if (ret)
- goto out;
+ if (nothing_failed && !need_undirty) {
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
+
+ if (local->transaction.in_flight_sb) {
+ local->op_ret = -1;
+ local->op_errno = local->transaction.in_flight_sb_errno;
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.failed_subvols[i])
+ local->pending[i][idx] = hton32(1);
+ }
+
+ ret = afr_set_pending_dict(priv, xattr, local->pending);
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
- ret = syncop_xattrop (priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
- GF_XATTROP_ADD_ARRAY, xattr, NULL, NULL, NULL);
+ ret = afr_changelog_thin_arbiter_post_op(this, local);
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
- if (ret == -EINVAL) {
- gf_msg (this->name, GF_LOG_INFO, -ret, AFR_MSG_THIN_ARB,
- "Thin-arbiter has denied post-op on %s for gfid %s.",
- priv->pending_key[THIN_ARBITER_BRICK_INDEX],
- uuid_utoa (local->inode->gfid));
+ if (need_undirty)
+ local->dirty[idx] = hton32(-1);
+ else
+ local->dirty[idx] = hton32(0);
- } else if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Post-op on thin-arbiter id file %s failed for gfid %s.",
- priv->pending_key[THIN_ARBITER_BRICK_INDEX],
- uuid_utoa (local->inode->gfid));
- }
- flock.l_type = F_UNLCK;
- syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- AFR_TA_DOM_NOTIFY, &loc, F_SETLK, &flock, NULL, NULL);
+set_dirty:
+ ret = dict_set_static_bin(xattr, AFR_DIRTY, local->dirty,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
+
+ afr_changelog_do(frame, this, xattr, afr_changelog_post_op_done,
+ AFR_TRANSACTION_POST_OP);
out:
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
- return ret;
+ return 0;
}
-int
-afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t *priv = this->private;
- afr_local_t *local = NULL;
- dict_t *xattr = NULL;
- int i = 0;
- int ret = 0;
- int idx = 0;
- int nothing_failed = 1;
- gf_boolean_t need_undirty = _gf_false;
-
- afr_handle_quorum (frame, this);
- local = frame->local;
- idx = afr_index_for_transaction_type (local->transaction.type);
-
- xattr = dict_new ();
- if (!xattr) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
+gf_boolean_t
+afr_changelog_pre_op_uninherit(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_inode_ctx_t *ctx = NULL;
+ int i = 0;
+ gf_boolean_t ret = _gf_false;
+ int type = 0;
- nothing_failed = afr_txn_nothing_failed (frame, this);
+ local = frame->local;
+ priv = this->private;
+ ctx = local->inode_ctx;
- if (afr_changelog_pre_op_uninherit (frame, this))
- need_undirty = _gf_false;
- else
- need_undirty = _gf_true;
+ type = afr_index_for_transaction_type(local->transaction.type);
+ if (type != AFR_DATA_TRANSACTION)
+ return !local->transaction.dirtied;
- if (local->op_ret < 0 && !nothing_failed) {
- if (afr_need_dirty_marking (frame, this)) {
- local->dirty[idx] = hton32(1);
- goto set_dirty;
- }
+ if (local->transaction.no_uninherit)
+ return _gf_false;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
+ /* This function must be idempotent. So check if we
+ were called before and return the same answer again.
- if (nothing_failed && !need_undirty) {
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
+ It is important to keep this function idempotent for
+ the call in afr_changelog_post_op_safe() to not have
+ side effects on the call from afr_changelog_post_op_now()
+ */
+ if (local->transaction.uninherit_done)
+ return local->transaction.uninherit_value;
- if (local->transaction.in_flight_sb) {
- local->op_ret = -1;
- local->op_errno = local->transaction.in_flight_sb_errno;
- afr_changelog_post_op_done (frame, this);
- goto out;
+ LOCK(&local->inode->lock);
+ {
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] != ctx->pre_op_done[type][i]) {
+ ret = !local->transaction.dirtied;
+ goto unlock;
+ }
+ }
+
+ if (ctx->inherited[type]) {
+ ret = _gf_true;
+ ctx->inherited[type]--;
+ } else if (ctx->on_disk[type]) {
+ ret = _gf_false;
+ ctx->on_disk[type]--;
+ } else {
+ /* ASSERT */
+ ret = _gf_false;
}
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.failed_subvols[i])
- local->pending[i][idx] = hton32(1);
- }
-
- ret = afr_set_pending_dict (priv, xattr, local->pending);
- if (ret < 0) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
-
- ret = afr_changelog_thin_arbiter_post_op (this, local);
- if (ret < 0) {
- local->op_ret = -1;
- local->op_errno = -ret;
- afr_changelog_post_op_done (frame, this);
- goto out;
+ if (!ctx->inherited[type] && !ctx->on_disk[type]) {
+ for (i = 0; i < priv->child_count; i++)
+ ctx->pre_op_done[type][i] = 0;
}
-
- if (need_undirty)
- local->dirty[idx] = hton32(-1);
- else
- local->dirty[idx] = hton32(0);
-
-set_dirty:
- ret = dict_set_static_bin (xattr, AFR_DIRTY, local->dirty,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
-
- afr_changelog_do (frame, this, xattr, afr_changelog_post_op_done,
- AFR_TRANSACTION_POST_OP);
-out:
- if (xattr)
- dict_unref (xattr);
-
- return 0;
-}
-
-
-gf_boolean_t
-afr_changelog_pre_op_uninherit (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_inode_ctx_t *ctx = NULL;
- int i = 0;
- gf_boolean_t ret = _gf_false;
- int type = 0;
-
- local = frame->local;
- priv = this->private;
- ctx = local->inode_ctx;
-
- type = afr_index_for_transaction_type (local->transaction.type);
- if (type != AFR_DATA_TRANSACTION)
- return !local->transaction.dirtied;
-
- if (local->transaction.no_uninherit)
- return _gf_false;
-
- /* This function must be idempotent. So check if we
- were called before and return the same answer again.
-
- It is important to keep this function idempotent for
- the call in afr_changelog_post_op_safe() to not have
- side effects on the call from afr_changelog_post_op_now()
- */
- if (local->transaction.uninherit_done)
- return local->transaction.uninherit_value;
-
- LOCK(&local->inode->lock);
- {
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] !=
- ctx->pre_op_done[type][i]) {
- ret = !local->transaction.dirtied;
- goto unlock;
- }
- }
-
- if (ctx->inherited[type]) {
- ret = _gf_true;
- ctx->inherited[type]--;
- } else if (ctx->on_disk[type]) {
- ret = _gf_false;
- ctx->on_disk[type]--;
- } else {
- /* ASSERT */
- ret = _gf_false;
- }
-
- if (!ctx->inherited[type] && !ctx->on_disk[type]) {
- for (i = 0; i < priv->child_count; i++)
- ctx->pre_op_done[type][i] = 0;
- }
- }
+ }
unlock:
- UNLOCK(&local->inode->lock);
+ UNLOCK(&local->inode->lock);
- local->transaction.uninherit_done = _gf_true;
- local->transaction.uninherit_value = ret;
+ local->transaction.uninherit_done = _gf_true;
+ local->transaction.uninherit_value = ret;
- return ret;
+ return ret;
}
-
gf_boolean_t
-afr_changelog_pre_op_inherit (call_frame_t *frame, xlator_t *this)
+afr_changelog_pre_op_inherit(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- gf_boolean_t ret = _gf_false;
- int type = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ gf_boolean_t ret = _gf_false;
+ int type = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (local->transaction.type != AFR_DATA_TRANSACTION)
- return _gf_false;
+ if (local->transaction.type != AFR_DATA_TRANSACTION)
+ return _gf_false;
- type = afr_index_for_transaction_type (local->transaction.type);
+ type = afr_index_for_transaction_type(local->transaction.type);
- LOCK(&local->inode->lock);
- {
- if (!local->inode_ctx->on_disk[type]) {
- /* nothing to inherit yet */
- ret = _gf_false;
- goto unlock;
- }
+ LOCK(&local->inode->lock);
+ {
+ if (!local->inode_ctx->on_disk[type]) {
+ /* nothing to inherit yet */
+ ret = _gf_false;
+ goto unlock;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] !=
- local->inode_ctx->pre_op_done[type][i]) {
- /* either inherit exactly, or don't */
- ret = _gf_false;
- goto unlock;
- }
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] !=
+ local->inode_ctx->pre_op_done[type][i]) {
+ /* either inherit exactly, or don't */
+ ret = _gf_false;
+ goto unlock;
+ }
+ }
- local->inode_ctx->inherited[type]++;
+ local->inode_ctx->inherited[type]++;
- ret = _gf_true;
+ ret = _gf_true;
- local->transaction.inherited = _gf_true;
- }
+ local->transaction.inherited = _gf_true;
+ }
unlock:
- UNLOCK(&local->inode->lock);
+ UNLOCK(&local->inode->lock);
- return ret;
+ return ret;
}
-
gf_boolean_t
-afr_changelog_pre_op_update (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- gf_boolean_t ret = _gf_false;
- int type = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
- local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION)
- return _gf_false;
-
- if (local->transaction.inherited)
- /* was already inherited in afr_changelog_pre_op */
- return _gf_false;
-
- if (!local->transaction.dirtied)
- return _gf_false;
-
- if (!afr_txn_nothing_failed (frame, this))
- return _gf_false;
-
- type = afr_index_for_transaction_type (local->transaction.type);
-
- ret = _gf_false;
-
- LOCK(&local->inode->lock);
- {
- if (!local->inode_ctx->on_disk[type]) {
- for (i = 0; i < priv->child_count; i++)
- local->inode_ctx->pre_op_done[type][i] =
- (!local->transaction.failed_subvols[i]);
- } else {
- for (i = 0; i < priv->child_count; i++)
- if (local->inode_ctx->pre_op_done[type][i] !=
- (!local->transaction.failed_subvols[i])) {
- local->transaction.no_uninherit = 1;
- goto unlock;
- }
- }
- local->inode_ctx->on_disk[type]++;
-
- ret = _gf_true;
- }
-unlock:
- UNLOCK(&local->inode->lock);
+afr_changelog_pre_op_update(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ gf_boolean_t ret = _gf_false;
+ int type = 0;
- return ret;
-}
+ local = frame->local;
+ priv = this->private;
+ if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+ local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION)
+ return _gf_false;
-int
-afr_changelog_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
-{
- afr_local_t *local = NULL;
- int call_count = -1;
- int child_index = -1;
+ if (local->transaction.inherited)
+ /* was already inherited in afr_changelog_pre_op */
+ return _gf_false;
- local = frame->local;
- child_index = (long) cookie;
+ if (!local->transaction.dirtied)
+ return _gf_false;
- if (op_ret == -1) {
- local->op_errno = op_errno;
- afr_transaction_fop_failed (frame, this, child_index);
- }
+ if (!afr_txn_nothing_failed(frame, this))
+ return _gf_false;
- if (xattr)
- local->transaction.changelog_xdata[child_index] = dict_ref (xattr);
+ type = afr_index_for_transaction_type(local->transaction.type);
- call_count = afr_frame_return (frame);
+ ret = _gf_false;
- if (call_count == 0) {
- local->transaction.changelog_resume (frame, this);
+ LOCK(&local->inode->lock);
+ {
+ if (!local->inode_ctx->on_disk[type]) {
+ for (i = 0; i < priv->child_count; i++)
+ local->inode_ctx->pre_op_done[type][i] =
+ (!local->transaction.failed_subvols[i]);
+ } else {
+ for (i = 0; i < priv->child_count; i++)
+ if (local->inode_ctx->pre_op_done[type][i] !=
+ (!local->transaction.failed_subvols[i])) {
+ local->transaction.no_uninherit = 1;
+ goto unlock;
+ }
}
+ local->inode_ctx->on_disk[type]++;
- return 0;
+ ret = _gf_true;
+ }
+unlock:
+ UNLOCK(&local->inode->lock);
+
+ return ret;
}
-void
-afr_changelog_populate_xdata (call_frame_t *frame, afr_xattrop_type_t op,
- dict_t **xdata, dict_t **newloc_xdata)
-{
- int i = 0;
- int ret = 0;
- char *key = NULL;
- const char *name = NULL;
- dict_t *xdata1 = NULL;
- dict_t *xdata2 = NULL;
- xlator_t *this = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t need_entry_key_set = _gf_true;
-
- local = frame->local;
- this = THIS;
- priv = this->private;
-
- if (local->transaction.type == AFR_DATA_TRANSACTION ||
- local->transaction.type == AFR_METADATA_TRANSACTION)
- goto out;
+int
+afr_changelog_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xattr, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int child_index = -1;
- if (!priv->esh_granular)
- goto out;
+ local = frame->local;
+ child_index = (long)cookie;
- xdata1 = dict_new();
- if (!xdata1)
- goto out;
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ afr_transaction_fop_failed(frame, this, child_index);
+ }
+
+ if (xattr)
+ local->transaction.changelog_xdata[child_index] = dict_ref(xattr);
- name = local->loc.name;
- if (local->op == GF_FOP_LINK)
- name = local->newloc.name;
+ call_count = afr_frame_return(frame);
- switch (op) {
+ if (call_count == 0) {
+ local->transaction.changelog_resume(frame, this);
+ }
+
+ return 0;
+}
+
+void
+afr_changelog_populate_xdata(call_frame_t *frame, afr_xattrop_type_t op,
+ dict_t **xdata, dict_t **newloc_xdata)
+{
+ int i = 0;
+ int ret = 0;
+ char *key = NULL;
+ const char *name = NULL;
+ dict_t *xdata1 = NULL;
+ dict_t *xdata2 = NULL;
+ xlator_t *this = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t need_entry_key_set = _gf_true;
+
+ local = frame->local;
+ this = THIS;
+ priv = this->private;
+
+ if (local->transaction.type == AFR_DATA_TRANSACTION ||
+ local->transaction.type == AFR_METADATA_TRANSACTION)
+ goto out;
+
+ if (!priv->esh_granular)
+ goto out;
+
+ xdata1 = dict_new();
+ if (!xdata1)
+ goto out;
+
+ name = local->loc.name;
+ if (local->op == GF_FOP_LINK)
+ name = local->newloc.name;
+
+ switch (op) {
case AFR_TRANSACTION_PRE_OP:
- key = GF_XATTROP_ENTRY_IN_KEY;
- break;
+ key = GF_XATTROP_ENTRY_IN_KEY;
+ break;
case AFR_TRANSACTION_POST_OP:
- if (afr_txn_nothing_failed (frame, this)) {
- key = GF_XATTROP_ENTRY_OUT_KEY;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.failed_subvols[i])
- continue;
- need_entry_key_set = _gf_false;
- break;
- }
- /* If the transaction itself did not fail and there
- * are no failed subvolumes, check whether the fop
- * failed due to a symmetric error. If it did, do
- * not set the ENTRY_OUT xattr which would end up
- * deleting a name index which was created possibly by
- * an earlier entry txn that may have failed on some
- * of the sub-volumes.
- */
- if (local->op_ret)
- need_entry_key_set = _gf_false;
- } else {
- key = GF_XATTROP_ENTRY_IN_KEY;
+ if (afr_txn_nothing_failed(frame, this)) {
+ key = GF_XATTROP_ENTRY_OUT_KEY;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.failed_subvols[i])
+ continue;
+ need_entry_key_set = _gf_false;
+ break;
}
- break;
- }
+ /* If the transaction itself did not fail and there
+ * are no failed subvolumes, check whether the fop
+ * failed due to a symmetric error. If it did, do
+ * not set the ENTRY_OUT xattr which would end up
+ * deleting a name index which was created possibly by
+ * an earlier entry txn that may have failed on some
+ * of the sub-volumes.
+ */
+ if (local->op_ret)
+ need_entry_key_set = _gf_false;
+ } else {
+ key = GF_XATTROP_ENTRY_IN_KEY;
+ }
+ break;
+ }
- if (need_entry_key_set) {
- ret = dict_set_str (xdata1, key, (char *)name);
- if (ret)
- gf_msg (THIS->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED,
- "%s/%s: Could not set %s key during xattrop",
- uuid_utoa (local->loc.pargfid), local->loc.name,
- key);
- if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
- xdata2 = dict_new ();
- if (!xdata2)
- goto out;
-
- ret = dict_set_str (xdata2, key,
- (char *)local->newloc.name);
- if (ret)
- gf_msg (THIS->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED,
- "%s/%s: Could not set %s key during "
- "xattrop",
- uuid_utoa (local->newloc.pargfid),
- local->newloc.name, key);
- }
+ if (need_entry_key_set) {
+ ret = dict_set_str(xdata1, key, (char *)name);
+ if (ret)
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_SET_FAILED,
+ "%s/%s: Could not set %s key during xattrop",
+ uuid_utoa(local->loc.pargfid), local->loc.name, key);
+ if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+ xdata2 = dict_new();
+ if (!xdata2)
+ goto out;
+
+ ret = dict_set_str(xdata2, key, (char *)local->newloc.name);
+ if (ret)
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_SET_FAILED,
+ "%s/%s: Could not set %s key during "
+ "xattrop",
+ uuid_utoa(local->newloc.pargfid), local->newloc.name,
+ key);
}
+ }
- *xdata = xdata1;
- *newloc_xdata = xdata2;
- xdata1 = xdata2 = NULL;
+ *xdata = xdata1;
+ *newloc_xdata = xdata2;
+ xdata1 = xdata2 = NULL;
out:
- if (xdata1)
- dict_unref (xdata1);
- return;
+ if (xdata1)
+ dict_unref(xdata1);
+ return;
}
int
-afr_changelog_prepare (xlator_t *this, call_frame_t *frame, int *call_count,
- afr_changelog_resume_t changelog_resume,
- afr_xattrop_type_t op, dict_t **xdata,
- dict_t **newloc_xdata)
+afr_changelog_prepare(xlator_t *this, call_frame_t *frame, int *call_count,
+ afr_changelog_resume_t changelog_resume,
+ afr_xattrop_type_t op, dict_t **xdata,
+ dict_t **newloc_xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- *call_count = afr_changelog_call_count (local->transaction.type,
- local->transaction.pre_op,
- local->transaction.failed_subvols,
- priv->child_count);
+ *call_count = afr_changelog_call_count(
+ local->transaction.type, local->transaction.pre_op,
+ local->transaction.failed_subvols, priv->child_count);
- if (*call_count == 0) {
- changelog_resume (frame, this);
- return -1;
- }
+ if (*call_count == 0) {
+ changelog_resume(frame, this);
+ return -1;
+ }
- afr_changelog_populate_xdata (frame, op, xdata, newloc_xdata);
- local->call_count = *call_count;
+ afr_changelog_populate_xdata(frame, op, xdata, newloc_xdata);
+ local->call_count = *call_count;
- local->transaction.changelog_resume = changelog_resume;
- return 0;
+ local->transaction.changelog_resume = changelog_resume;
+ return 0;
}
int
-afr_changelog_do (call_frame_t *frame, xlator_t *this, dict_t *xattr,
- afr_changelog_resume_t changelog_resume,
- afr_xattrop_type_t op)
+afr_changelog_do(call_frame_t *frame, xlator_t *this, dict_t *xattr,
+ afr_changelog_resume_t changelog_resume, afr_xattrop_type_t op)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- dict_t *newloc_xdata = NULL;
- int i = 0;
- int call_count = 0;
- int ret = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ dict_t *newloc_xdata = NULL;
+ int i = 0;
+ int call_count = 0;
+ int ret = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.changelog_xdata[i]) {
- dict_unref (local->transaction.changelog_xdata[i]);
- local->transaction.changelog_xdata[i] = NULL;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.changelog_xdata[i]) {
+ dict_unref(local->transaction.changelog_xdata[i]);
+ local->transaction.changelog_xdata[i] = NULL;
}
+ }
- ret = afr_changelog_prepare (this, frame, &call_count, changelog_resume,
- op, &xdata, &newloc_xdata);
+ ret = afr_changelog_prepare(this, frame, &call_count, changelog_resume, op,
+ &xdata, &newloc_xdata);
- if (ret)
- return 0;
+ if (ret)
+ return 0;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.pre_op[i] ||
- local->transaction.failed_subvols[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i] ||
+ local->transaction.failed_subvols[i])
+ continue;
- switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
- if (!local->fd) {
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->loc,
- GF_XATTROP_ADD_ARRAY, xattr,
- xdata);
- } else {
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fxattrop,
- local->fd,
- GF_XATTROP_ADD_ARRAY, xattr,
- xdata);
- }
- break;
- case AFR_ENTRY_RENAME_TRANSACTION:
-
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->transaction.new_parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr,
- newloc_xdata);
- call_count--;
+ switch (local->transaction.type) {
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
+ if (!local->fd) {
+ STACK_WIND_COOKIE(
+ frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->xattrop,
+ &local->loc, GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ } else {
+ STACK_WIND_COOKIE(
+ frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->fxattrop,
+ local->fd, GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ }
+ break;
+ case AFR_ENTRY_RENAME_TRANSACTION:
+
+ STACK_WIND_COOKIE(frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &local->transaction.new_parent_loc,
+ GF_XATTROP_ADD_ARRAY, xattr, newloc_xdata);
+ call_count--;
/* fall through */
- case AFR_ENTRY_TRANSACTION:
- if (local->fd)
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fxattrop,
- local->fd,
- GF_XATTROP_ADD_ARRAY, xattr,
- xdata);
- else
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->transaction.parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr,
- xdata);
- break;
- }
-
- if (!--call_count)
- break;
+ case AFR_ENTRY_TRANSACTION:
+ if (local->fd)
+ STACK_WIND_COOKIE(
+ frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->fxattrop,
+ local->fd, GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ else
+ STACK_WIND_COOKIE(frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &local->transaction.parent_loc,
+ GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ break;
}
- if (xdata)
- dict_unref (xdata);
- if (newloc_xdata)
- dict_unref (newloc_xdata);
- return 0;
+ if (!--call_count)
+ break;
+ }
+
+ if (xdata)
+ dict_unref(xdata);
+ if (newloc_xdata)
+ dict_unref(newloc_xdata);
+ return 0;
}
static void
-afr_init_optimistic_changelog_for_txn (xlator_t *this, afr_local_t *local)
+afr_init_optimistic_changelog_for_txn(xlator_t *this, afr_local_t *local)
{
- int locked_count = 0;
- afr_private_t *priv = NULL;
+ int locked_count = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- locked_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
- if (priv->optimistic_change_log && locked_count == priv->child_count)
- local->optimistic_change_log = 1;
+ locked_count = AFR_COUNT(local->transaction.pre_op, priv->child_count);
+ if (priv->optimistic_change_log && locked_count == priv->child_count)
+ local->optimistic_change_log = 1;
- return;
+ return;
}
int
-afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t * priv = this->private;
- int i = 0;
- int ret = 0;
- int call_count = 0;
- int op_errno = 0;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- unsigned char *locked_nodes = NULL;
- int idx = -1;
- gf_boolean_t pre_nop = _gf_true;
- dict_t *xdata_req = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- idx = afr_index_for_transaction_type (local->transaction.type);
-
- locked_nodes = afr_locked_nodes_get (local->transaction.type, int_lock);
-
- for (i = 0; i < priv->child_count; i++) {
- if (locked_nodes[i]) {
- local->transaction.pre_op[i] = 1;
- call_count++;
- } else {
- local->transaction.failed_subvols[i] = 1;
- }
- }
-
- afr_init_optimistic_changelog_for_txn (this, local);
-
- if (afr_changelog_pre_op_inherit (frame, this))
- goto next;
-
- /* This condition should not be met with present code, as
- * transaction.done will be called if locks are not acquired on even a
- * single node.
- */
- if (call_count == 0) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Check if the fop can be performed on at least
- * quorum number of nodes.
- */
- if (priv->quorum_count && !afr_has_fop_quorum (frame)) {
- op_errno = int_lock->lock_op_errno;
- if (op_errno == 0)
- op_errno = afr_quorum_errno (priv);
- goto err;
+afr_changelog_pre_op(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ int i = 0;
+ int ret = 0;
+ int call_count = 0;
+ int op_errno = 0;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ unsigned char *locked_nodes = NULL;
+ int idx = -1;
+ gf_boolean_t pre_nop = _gf_true;
+ dict_t *xdata_req = NULL;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ idx = afr_index_for_transaction_type(local->transaction.type);
+
+ locked_nodes = afr_locked_nodes_get(local->transaction.type, int_lock);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (locked_nodes[i]) {
+ local->transaction.pre_op[i] = 1;
+ call_count++;
+ } else {
+ local->transaction.failed_subvols[i] = 1;
+ }
+ }
+
+ afr_init_optimistic_changelog_for_txn(this, local);
+
+ if (afr_changelog_pre_op_inherit(frame, this))
+ goto next;
+
+ /* This condition should not be met with present code, as
+ * transaction.done will be called if locks are not acquired on even a
+ * single node.
+ */
+ if (call_count == 0) {
+ op_errno = ENOTCONN;
+ goto err;
+ }
+
+ /* Check if the fop can be performed on at least
+ * quorum number of nodes.
+ */
+ if (priv->quorum_count && !afr_has_fop_quorum(frame)) {
+ op_errno = int_lock->lock_op_errno;
+ if (op_errno == 0)
+ op_errno = afr_quorum_errno(priv);
+ goto err;
+ }
+
+ xdata_req = dict_new();
+ if (!xdata_req) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ if (call_count < priv->child_count)
+ pre_nop = _gf_false;
+
+ /* Set an all-zero pending changelog so that in the cbk, we can get the
+ * current on-disk values. In a replica 3 volume with arbiter enabled,
+ * these values are needed to arrive at a go/ no-go of the fop phase to
+ * avoid ending up in split-brain.*/
+
+ ret = afr_set_pending_dict(priv, xdata_req, local->pending);
+ if (ret < 0) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ if (afr_needs_changelog_update(local)) {
+ local->dirty[idx] = hton32(1);
+
+ ret = dict_set_static_bin(xdata_req, AFR_DIRTY, local->dirty,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ op_errno = ENOMEM;
+ goto err;
}
- xdata_req = dict_new();
- if (!xdata_req) {
- op_errno = ENOMEM;
- goto err;
- }
-
- if (call_count < priv->child_count)
- pre_nop = _gf_false;
-
- /* Set an all-zero pending changelog so that in the cbk, we can get the
- * current on-disk values. In a replica 3 volume with arbiter enabled,
- * these values are needed to arrive at a go/ no-go of the fop phase to
- * avoid ending up in split-brain.*/
-
- ret = afr_set_pending_dict (priv, xdata_req, local->pending);
- if (ret < 0) {
- op_errno = ENOMEM;
- goto err;
- }
-
- if (afr_needs_changelog_update (local)) {
-
- local->dirty[idx] = hton32(1);
-
- ret = dict_set_static_bin (xdata_req, AFR_DIRTY, local->dirty,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret) {
- op_errno = ENOMEM;
- goto err;
- }
+ pre_nop = _gf_false;
+ local->transaction.dirtied = 1;
+ }
- pre_nop = _gf_false;
- local->transaction.dirtied = 1;
- }
+ if (pre_nop)
+ goto next;
- if (pre_nop)
- goto next;
+ if (!local->pre_op_compat) {
+ dict_copy(xdata_req, local->xdata_req);
+ goto next;
+ }
- if (!local->pre_op_compat) {
- dict_copy (xdata_req, local->xdata_req);
- goto next;
- }
+ afr_changelog_do(frame, this, xdata_req, afr_transaction_perform_fop,
+ AFR_TRANSACTION_PRE_OP);
- afr_changelog_do (frame, this, xdata_req, afr_transaction_perform_fop,
- AFR_TRANSACTION_PRE_OP);
+ if (xdata_req)
+ dict_unref(xdata_req);
- if (xdata_req)
- dict_unref (xdata_req);
-
- return 0;
+ return 0;
next:
- afr_transaction_perform_fop (frame, this);
+ afr_transaction_perform_fop(frame, this);
- if (xdata_req)
- dict_unref (xdata_req);
+ if (xdata_req)
+ dict_unref(xdata_req);
- return 0;
+ return 0;
err:
- local->internal_lock.lock_cbk = afr_transaction_done;
- local->op_ret = -1;
- local->op_errno = op_errno;
+ local->internal_lock.lock_cbk = afr_transaction_done;
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- afr_handle_lock_acquire_failure (local, _gf_true);
+ afr_handle_lock_acquire_failure(local, _gf_true);
- if (xdata_req)
- dict_unref (xdata_req);
+ if (xdata_req)
+ dict_unref(xdata_req);
- return 0;
+ return 0;
}
-
int
-afr_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
+afr_post_nonblocking_inodelk_cbk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- /* Initiate blocking locks if non-blocking has failed */
- if (int_lock->lock_op_ret < 0) {
- gf_msg_debug (this->name, 0,
- "Non blocking inodelks failed. Proceeding to blocking");
- int_lock->lock_cbk = afr_internal_lock_finish;
- afr_blocking_lock (frame, this);
- } else {
+ /* Initiate blocking locks if non-blocking has failed */
+ if (int_lock->lock_op_ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "Non blocking inodelks failed. Proceeding to blocking");
+ int_lock->lock_cbk = afr_internal_lock_finish;
+ afr_blocking_lock(frame, this);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Non blocking inodelks done. Proceeding to FOP");
+ afr_internal_lock_finish(frame, this);
+ }
- gf_msg_debug (this->name, 0,
- "Non blocking inodelks done. Proceeding to FOP");
- afr_internal_lock_finish (frame, this);
- }
-
- return 0;
+ return 0;
}
-
int
-afr_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
+afr_post_nonblocking_entrylk_cbk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- /* Initiate blocking locks if non-blocking has failed */
- if (int_lock->lock_op_ret < 0) {
- gf_msg_debug (this->name, 0,
- "Non blocking entrylks failed. Proceeding to blocking");
- int_lock->lock_cbk = afr_internal_lock_finish;
- afr_blocking_lock (frame, this);
- } else {
+ local = frame->local;
+ int_lock = &local->internal_lock;
- gf_msg_debug (this->name, 0,
- "Non blocking entrylks done. Proceeding to FOP");
+ /* Initiate blocking locks if non-blocking has failed */
+ if (int_lock->lock_op_ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "Non blocking entrylks failed. Proceeding to blocking");
+ int_lock->lock_cbk = afr_internal_lock_finish;
+ afr_blocking_lock(frame, this);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Non blocking entrylks done. Proceeding to FOP");
- afr_internal_lock_finish (frame, this);
- }
+ afr_internal_lock_finish(frame, this);
+ }
- return 0;
+ return 0;
}
-
int
-afr_post_blocking_rename_cbk (call_frame_t *frame, xlator_t *this)
+afr_post_blocking_rename_cbk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- if (int_lock->lock_op_ret < 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_BLOCKING_LKS_FAILED,
- "Blocking entrylks failed.");
-
- afr_transaction_done (frame, this);
- } else {
+ if (int_lock->lock_op_ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_BLOCKING_LKS_FAILED,
+ "Blocking entrylks failed.");
- gf_msg_debug (this->name, 0,
- "Blocking entrylks done. Proceeding to FOP");
+ afr_transaction_done(frame, this);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Blocking entrylks done. Proceeding to FOP");
- afr_internal_lock_finish (frame, this);
- }
- return 0;
+ afr_internal_lock_finish(frame, this);
+ }
+ return 0;
}
int
-afr_post_lower_unlock_cbk (call_frame_t *frame, xlator_t *this)
+afr_post_lower_unlock_cbk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- GF_ASSERT (!int_lock->higher_locked);
+ GF_ASSERT(!int_lock->higher_locked);
- int_lock->lock_cbk = afr_post_blocking_rename_cbk;
- afr_blocking_lock (frame, this);
+ int_lock->lock_cbk = afr_post_blocking_rename_cbk;
+ afr_blocking_lock(frame, this);
- return 0;
+ return 0;
}
-
int
-afr_set_transaction_flock (xlator_t *this, afr_local_t *local)
+afr_set_transaction_flock(xlator_t *this, afr_local_t *local)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_private_t *priv = NULL;
- int_lock = &local->internal_lock;
- priv = this->private;
+ int_lock = &local->internal_lock;
+ priv = this->private;
- if ((priv->arbiter_count || local->transaction.eager_lock_on ||
- priv->full_lock) &&
- local->transaction.type == AFR_DATA_TRANSACTION) {
- /*Lock entire file to avoid network split brains.*/
- int_lock->flock.l_len = 0;
- int_lock->flock.l_start = 0;
- } else {
- int_lock->flock.l_len = local->transaction.len;
- int_lock->flock.l_start = local->transaction.start;
- }
- int_lock->flock.l_type = F_WRLCK;
+ if ((priv->arbiter_count || local->transaction.eager_lock_on ||
+ priv->full_lock) &&
+ local->transaction.type == AFR_DATA_TRANSACTION) {
+ /*Lock entire file to avoid network split brains.*/
+ int_lock->flock.l_len = 0;
+ int_lock->flock.l_start = 0;
+ } else {
+ int_lock->flock.l_len = local->transaction.len;
+ int_lock->flock.l_start = local->transaction.start;
+ }
+ int_lock->flock.l_type = F_WRLCK;
- return 0;
+ return 0;
}
int
-afr_lock (call_frame_t *frame, xlator_t *this)
+afr_lock(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- int_lock->domain = this->name;
+ int_lock->domain = this->name;
- switch (local->transaction.type) {
+ switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
- afr_set_transaction_flock (this, local);
+ afr_set_transaction_flock(this, local);
- int_lock->lock_cbk = afr_post_nonblocking_inodelk_cbk;
+ int_lock->lock_cbk = afr_post_nonblocking_inodelk_cbk;
- afr_nonblocking_inodelk (frame, this);
- break;
+ afr_nonblocking_inodelk(frame, this);
+ break;
case AFR_ENTRY_RENAME_TRANSACTION:
- int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
- afr_nonblocking_entrylk (frame, this);
- break;
+ int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+ afr_nonblocking_entrylk(frame, this);
+ break;
case AFR_ENTRY_TRANSACTION:
- int_lock->lk_basename = local->transaction.basename;
- if (local->transaction.parent_loc.path)
- int_lock->lk_loc = &local->transaction.parent_loc;
- else
- GF_ASSERT (local->fd);
+ int_lock->lk_basename = local->transaction.basename;
+ if (local->transaction.parent_loc.path)
+ int_lock->lk_loc = &local->transaction.parent_loc;
+ else
+ GF_ASSERT(local->fd);
- int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
- afr_nonblocking_entrylk (frame, this);
- break;
- }
+ int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+ afr_nonblocking_entrylk(frame, this);
+ break;
+ }
- return 0;
+ return 0;
}
static gf_boolean_t
-afr_locals_overlap (afr_local_t *local1, afr_local_t *local2)
+afr_locals_overlap(afr_local_t *local1, afr_local_t *local2)
{
- uint64_t start1 = local1->transaction.start;
- uint64_t start2 = local2->transaction.start;
- uint64_t end1 = 0;
- uint64_t end2 = 0;
+ uint64_t start1 = local1->transaction.start;
+ uint64_t start2 = local2->transaction.start;
+ uint64_t end1 = 0;
+ uint64_t end2 = 0;
- if (local1->transaction.len)
- end1 = start1 + local1->transaction.len - 1;
- else
- end1 = ULLONG_MAX;
+ if (local1->transaction.len)
+ end1 = start1 + local1->transaction.len - 1;
+ else
+ end1 = ULLONG_MAX;
- if (local2->transaction.len)
- end2 = start2 + local2->transaction.len - 1;
- else
- end2 = ULLONG_MAX;
+ if (local2->transaction.len)
+ end2 = start2 + local2->transaction.len - 1;
+ else
+ end2 = ULLONG_MAX;
- return ((end1 >= start2) && (end2 >= start1));
+ return ((end1 >= start2) && (end2 >= start1));
}
gf_boolean_t
-afr_has_lock_conflict (afr_local_t *local, gf_boolean_t waitlist_check)
-{
- afr_local_t *each = NULL;
- afr_lock_t *lock = NULL;
-
- lock = &local->inode_ctx->lock[local->transaction.type];
- /*
- * Once full file lock is acquired in eager-lock phase, overlapping
- * writes do not compete for inode-locks, instead are transferred to the
- * next writes. Because of this overlapping writes are not ordered.
- * This can cause inconsistencies in replication.
- * Example:
- * Two overlapping writes w1, w2 are sent in parallel on same fd
- * in two threads t1, t2.
- * Both threads can execute afr_writev_wind in the following manner.
- * t1 winds w1 on brick-0
- * t2 winds w2 on brick-0
- * t2 winds w2 on brick-1
- * t1 winds w1 on brick-1
- *
- * This check makes sure the locks are not transferred for
- * overlapping writes.
- */
- list_for_each_entry (each, &lock->owners, transaction.owner_list) {
- if (afr_locals_overlap (each, local)) {
- return _gf_true;
- }
- }
-
- if (!waitlist_check)
- return _gf_false;
- list_for_each_entry (each, &lock->waiting, transaction.wait_list) {
- if (afr_locals_overlap (each, local)) {
- return _gf_true;
- }
- }
+afr_has_lock_conflict(afr_local_t *local, gf_boolean_t waitlist_check)
+{
+ afr_local_t *each = NULL;
+ afr_lock_t *lock = NULL;
+
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ /*
+ * Once full file lock is acquired in eager-lock phase, overlapping
+ * writes do not compete for inode-locks, instead are transferred to the
+ * next writes. Because of this overlapping writes are not ordered.
+ * This can cause inconsistencies in replication.
+ * Example:
+ * Two overlapping writes w1, w2 are sent in parallel on same fd
+ * in two threads t1, t2.
+ * Both threads can execute afr_writev_wind in the following manner.
+ * t1 winds w1 on brick-0
+ * t2 winds w2 on brick-0
+ * t2 winds w2 on brick-1
+ * t1 winds w1 on brick-1
+ *
+ * This check makes sure the locks are not transferred for
+ * overlapping writes.
+ */
+ list_for_each_entry(each, &lock->owners, transaction.owner_list)
+ {
+ if (afr_locals_overlap(each, local)) {
+ return _gf_true;
+ }
+ }
+
+ if (!waitlist_check)
return _gf_false;
+ list_for_each_entry(each, &lock->waiting, transaction.wait_list)
+ {
+ if (afr_locals_overlap(each, local)) {
+ return _gf_true;
+ }
+ }
+ return _gf_false;
}
-
/* }}} */
static void
-afr_copy_inodelk_vars (afr_internal_lock_t *dst, afr_internal_lock_t *src,
- xlator_t *this)
+afr_copy_inodelk_vars(afr_internal_lock_t *dst, afr_internal_lock_t *src,
+ xlator_t *this)
{
- afr_private_t *priv = this->private;
+ afr_private_t *priv = this->private;
- dst->domain = src->domain;
- dst->flock.l_len = src->flock.l_len;
- dst->flock.l_start = src->flock.l_start;
- dst->flock.l_type = src->flock.l_type;
- dst->lock_count = src->lock_count;
- memcpy (dst->locked_nodes, src->locked_nodes,
- priv->child_count * sizeof (*dst->locked_nodes));
+ dst->domain = src->domain;
+ dst->flock.l_len = src->flock.l_len;
+ dst->flock.l_start = src->flock.l_start;
+ dst->flock.l_type = src->flock.l_type;
+ dst->lock_count = src->lock_count;
+ memcpy(dst->locked_nodes, src->locked_nodes,
+ priv->child_count * sizeof(*dst->locked_nodes));
}
void
-__afr_transaction_wake_shared (afr_local_t *local, struct list_head *shared)
-{
- gf_boolean_t conflict = _gf_false;
- afr_local_t *each = NULL;
- afr_lock_t *lock = &local->inode_ctx->lock[local->transaction.type];
-
- while (!conflict) {
- if (list_empty (&lock->waiting))
- return;
- each = list_entry(lock->waiting.next, afr_local_t,
- transaction.wait_list);
- if (afr_has_lock_conflict (each, _gf_false)) {
- conflict = _gf_true;
- }
- if (conflict && !list_empty (&lock->owners))
- return;
- afr_copy_inodelk_vars (&each->internal_lock,
- &local->internal_lock,
- each->transaction.frame->this);
- list_move_tail (&each->transaction.wait_list, shared);
- list_add_tail(&each->transaction.owner_list, &lock->owners);
+__afr_transaction_wake_shared(afr_local_t *local, struct list_head *shared)
+{
+ gf_boolean_t conflict = _gf_false;
+ afr_local_t *each = NULL;
+ afr_lock_t *lock = &local->inode_ctx->lock[local->transaction.type];
+
+ while (!conflict) {
+ if (list_empty(&lock->waiting))
+ return;
+ each = list_entry(lock->waiting.next, afr_local_t,
+ transaction.wait_list);
+ if (afr_has_lock_conflict(each, _gf_false)) {
+ conflict = _gf_true;
}
+ if (conflict && !list_empty(&lock->owners))
+ return;
+ afr_copy_inodelk_vars(&each->internal_lock, &local->internal_lock,
+ each->transaction.frame->this);
+ list_move_tail(&each->transaction.wait_list, shared);
+ list_add_tail(&each->transaction.owner_list, &lock->owners);
+ }
}
static void
-afr_lock_resume_shared (struct list_head *list)
+afr_lock_resume_shared(struct list_head *list)
{
- afr_local_t *each = NULL;
+ afr_local_t *each = NULL;
- while (!list_empty(list)) {
- each = list_entry(list->next, afr_local_t,
- transaction.wait_list);
- list_del_init(&each->transaction.wait_list);
- afr_changelog_pre_op (each->transaction.frame,
- each->transaction.frame->this);
- }
+ while (!list_empty(list)) {
+ each = list_entry(list->next, afr_local_t, transaction.wait_list);
+ list_del_init(&each->transaction.wait_list);
+ afr_changelog_pre_op(each->transaction.frame,
+ each->transaction.frame->this);
+ }
}
int
-afr_internal_lock_finish (call_frame_t *frame, xlator_t *this)
+afr_internal_lock_finish(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = frame->local;
- afr_lock_t *lock = NULL;
+ afr_local_t *local = frame->local;
+ afr_lock_t *lock = NULL;
-
- local->internal_lock.lock_cbk = NULL;
- if (!local->transaction.eager_lock_on) {
- if (local->internal_lock.lock_op_ret < 0) {
- afr_transaction_done (frame, this);
- return 0;
- }
- afr_changelog_pre_op (frame, this);
+ local->internal_lock.lock_cbk = NULL;
+ if (!local->transaction.eager_lock_on) {
+ if (local->internal_lock.lock_op_ret < 0) {
+ afr_transaction_done(frame, this);
+ return 0;
+ }
+ afr_changelog_pre_op(frame, this);
+ } else {
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ if (local->internal_lock.lock_op_ret < 0) {
+ afr_handle_lock_acquire_failure(local, _gf_false);
} else {
- lock = &local->inode_ctx->lock[local->transaction.type];
- if (local->internal_lock.lock_op_ret < 0) {
- afr_handle_lock_acquire_failure (local, _gf_false);
- } else {
- lock->event_generation = local->event_generation;
- afr_changelog_pre_op (frame, this);
- }
+ lock->event_generation = local->event_generation;
+ afr_changelog_pre_op(frame, this);
}
+ }
- return 0;
+ return 0;
}
gf_boolean_t
-afr_are_multiple_fds_opened (afr_local_t *local, xlator_t *this)
+afr_are_multiple_fds_opened(afr_local_t *local, xlator_t *this)
{
- /* Lets say mount1 has eager-lock(full-lock) and after the eager-lock
- * is taken mount2 opened the same file, it won't be able to
- * perform any data operations until mount1 releases eager-lock.
- * To avoid such scenario do not enable eager-lock for this transaction
- * if open-fd-count is > 1
- */
+ /* Lets say mount1 has eager-lock(full-lock) and after the eager-lock
+ * is taken mount2 opened the same file, it won't be able to
+ * perform any data operations until mount1 releases eager-lock.
+ * To avoid such scenario do not enable eager-lock for this transaction
+ * if open-fd-count is > 1
+ */
- if (local->inode_ctx->open_fd_count > 1)
- return _gf_true;
+ if (local->inode_ctx->open_fd_count > 1)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
-
gf_boolean_t
-afr_is_delayed_changelog_post_op_needed (call_frame_t *frame, xlator_t *this,
- int delay)
-{
- afr_local_t *local = NULL;
- afr_lock_t *lock = NULL;
- gf_boolean_t res = _gf_false;
-
- local = frame->local;
- lock = &local->inode_ctx->lock[local->transaction.type];
-
- if (!afr_txn_nothing_failed (frame, this)) {
- lock->release = _gf_true;
- goto out;
- }
-
- if (afr_are_multiple_fds_opened (local, this)) {
- lock->release = _gf_true;
- goto out;
- }
-
- if (!list_empty (&lock->owners))
- goto out;
- else
- GF_ASSERT (list_empty (&lock->waiting));
-
- if (lock->release) {
- goto out;
- }
-
- if (!delay) {
- goto out;
- }
-
- if ((local->op != GF_FOP_WRITE) &&
- (local->op != GF_FOP_FXATTROP)) {
- /*Only allow writes but shard does [f]xattrops on writes, so
- * they are fine too*/
- goto out;
- }
-
- res = _gf_true;
+afr_is_delayed_changelog_post_op_needed(call_frame_t *frame, xlator_t *this,
+ int delay)
+{
+ afr_local_t *local = NULL;
+ afr_lock_t *lock = NULL;
+ gf_boolean_t res = _gf_false;
+
+ local = frame->local;
+ lock = &local->inode_ctx->lock[local->transaction.type];
+
+ if (!afr_txn_nothing_failed(frame, this)) {
+ lock->release = _gf_true;
+ goto out;
+ }
+
+ if (afr_are_multiple_fds_opened(local, this)) {
+ lock->release = _gf_true;
+ goto out;
+ }
+
+ if (!list_empty(&lock->owners))
+ goto out;
+ else
+ GF_ASSERT(list_empty(&lock->waiting));
+
+ if (lock->release) {
+ goto out;
+ }
+
+ if (!delay) {
+ goto out;
+ }
+
+ if ((local->op != GF_FOP_WRITE) && (local->op != GF_FOP_FXATTROP)) {
+ /*Only allow writes but shard does [f]xattrops on writes, so
+ * they are fine too*/
+ goto out;
+ }
+
+ res = _gf_true;
out:
- return res;
+ return res;
}
-
void
-afr_delayed_changelog_wake_up_cbk (void *data)
-{
- afr_lock_t *lock = NULL;
- afr_local_t *local = data;
- afr_local_t *timer_local = NULL;
- struct list_head shared;
-
- INIT_LIST_HEAD (&shared);
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- timer_local = list_entry(lock->post_op.next,
- afr_local_t,
- transaction.owner_list);
- if (list_empty (&lock->owners) && (local == timer_local)) {
- GF_ASSERT (list_empty (&lock->waiting));
- /*Last owner*/
- lock->release = _gf_true;
- lock->delay_timer = NULL;
- }
- }
- UNLOCK (&local->inode->lock);
- afr_changelog_post_op_now (local->transaction.frame,
- local->transaction.frame->this);
+afr_delayed_changelog_wake_up_cbk(void *data)
+{
+ afr_lock_t *lock = NULL;
+ afr_local_t *local = data;
+ afr_local_t *timer_local = NULL;
+ struct list_head shared;
+
+ INIT_LIST_HEAD(&shared);
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ timer_local = list_entry(lock->post_op.next, afr_local_t,
+ transaction.owner_list);
+ if (list_empty(&lock->owners) && (local == timer_local)) {
+ GF_ASSERT(list_empty(&lock->waiting));
+ /*Last owner*/
+ lock->release = _gf_true;
+ lock->delay_timer = NULL;
+ }
+ }
+ UNLOCK(&local->inode->lock);
+ afr_changelog_post_op_now(local->transaction.frame,
+ local->transaction.frame->this);
}
-
/* SET operation */
int
-afr_fd_report_unstable_write (xlator_t *this, afr_local_t *local)
+afr_fd_report_unstable_write(xlator_t *this, afr_local_t *local)
{
- LOCK(&local->inode->lock);
- {
- local->inode_ctx->witnessed_unstable_write = _gf_true;
- }
- UNLOCK(&local->inode->lock);
+ LOCK(&local->inode->lock);
+ {
+ local->inode_ctx->witnessed_unstable_write = _gf_true;
+ }
+ UNLOCK(&local->inode->lock);
- return 0;
+ return 0;
}
/* TEST and CLEAR operation */
gf_boolean_t
-afr_fd_has_witnessed_unstable_write (xlator_t *this, inode_t *inode)
+afr_fd_has_witnessed_unstable_write(xlator_t *this, inode_t *inode)
{
- afr_inode_ctx_t *ctx = NULL;
- gf_boolean_t witness = _gf_false;
+ afr_inode_ctx_t *ctx = NULL;
+ gf_boolean_t witness = _gf_false;
- LOCK(&inode->lock);
- {
- (void)__afr_inode_ctx_get (this, inode, &ctx);
+ LOCK(&inode->lock);
+ {
+ (void)__afr_inode_ctx_get(this, inode, &ctx);
- if (ctx->witnessed_unstable_write) {
- witness = _gf_true;
- ctx->witnessed_unstable_write = _gf_false;
- }
+ if (ctx->witnessed_unstable_write) {
+ witness = _gf_true;
+ ctx->witnessed_unstable_write = _gf_false;
}
- UNLOCK (&inode->lock);
+ }
+ UNLOCK(&inode->lock);
- return witness;
+ return witness;
}
-
int
-afr_changelog_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *pre,
- struct iatt *post, dict_t *xdata)
+afr_changelog_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- int child_index = (long) cookie;
- int call_count = -1;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = (long)cookie;
+ int call_count = -1;
+ afr_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- if (op_ret != 0) {
- /* Failure of fsync() is as good as failure of previous
- write(). So treat it like one.
- */
- gf_msg (this->name, GF_LOG_WARNING,
- op_errno, AFR_MSG_FSYNC_FAILED,
- "fsync(%s) failed on subvolume %s. Transaction was %s",
- uuid_utoa (local->fd->inode->gfid),
- priv->children[child_index]->name,
- gf_fop_list[local->op]);
+ if (op_ret != 0) {
+ /* Failure of fsync() is as good as failure of previous
+ write(). So treat it like one.
+ */
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, AFR_MSG_FSYNC_FAILED,
+ "fsync(%s) failed on subvolume %s. Transaction was %s",
+ uuid_utoa(local->fd->inode->gfid),
+ priv->children[child_index]->name, gf_fop_list[local->op]);
- afr_transaction_fop_failed (frame, this, child_index);
- }
+ afr_transaction_fop_failed(frame, this, child_index);
+ }
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- afr_changelog_post_op_now (frame, this);
+ if (call_count == 0)
+ afr_changelog_post_op_now(frame, this);
- return 0;
+ return 0;
}
-
int
-afr_changelog_fsync (call_frame_t *frame, xlator_t *this)
+afr_changelog_fsync(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- int i = 0;
- int call_count = 0;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- GF_UNUSED int ret = -1;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ GF_UNUSED int ret = -1;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- call_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
+ call_count = AFR_COUNT(local->transaction.pre_op, priv->child_count);
- if (!call_count) {
- /* will go straight to unlock */
- afr_changelog_post_op_now (frame, this);
- return 0;
- }
+ if (!call_count) {
+ /* will go straight to unlock */
+ afr_changelog_post_op_now(frame, this);
+ return 0;
+ }
- local->call_count = call_count;
+ local->call_count = call_count;
- xdata = dict_new();
- if (xdata)
- ret = dict_set_int32 (xdata, "batch-fsync", 1);
+ xdata = dict_new();
+ if (xdata)
+ ret = dict_set_int32(xdata, "batch-fsync", 1);
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.pre_op[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i])
+ continue;
- STACK_WIND_COOKIE (frame, afr_changelog_fsync_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->fsync, local->fd,
- 1, xdata);
- if (!--call_count)
- break;
- }
+ STACK_WIND_COOKIE(frame, afr_changelog_fsync_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->fsync,
+ local->fd, 1, xdata);
+ if (!--call_count)
+ break;
+ }
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- return 0;
+ return 0;
}
-
int
-afr_changelog_post_op_safe (call_frame_t *frame, xlator_t *this)
+afr_changelog_post_op_safe(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- if (!local->fd || local->transaction.type != AFR_DATA_TRANSACTION) {
- afr_changelog_post_op_now (frame, this);
- return 0;
- }
+ local = frame->local;
+ priv = this->private;
- if (afr_changelog_pre_op_uninherit (frame, this) &&
- afr_txn_nothing_failed (frame, this)) {
- /* just detected that this post-op is about to
- be optimized away as a new write() has
- already piggybacked on this frame's changelog.
- */
- afr_changelog_post_op_now (frame, this);
- return 0;
- }
+ if (!local->fd || local->transaction.type != AFR_DATA_TRANSACTION) {
+ afr_changelog_post_op_now(frame, this);
+ return 0;
+ }
- /* Calling afr_changelog_post_op_now() now will result in
- issuing ->[f]xattrop().
-
- Performing a hard POST-OP (->[f]xattrop() FOP) is a more
- responsible operation that what it might appear on the surface.
-
- The changelog of a file (in the xattr of the file on the server)
- stores information (pending count) about the state of the file
- on the OTHER server. This changelog is blindly trusted, and must
- therefore be updated in such a way it remains trustworthy. This
- implies that decrementing the pending count (essentially "clearing
- the dirty flag") must be done STRICTLY after we are sure that the
- operation on the other server has reached stable storage.
-
- While the backend filesystem on that server will eventually flush
- it to stable storage, we (being in userspace) have no mechanism
- to get notified when the write became "stable".
-
- This means we need take matter into our own hands and issue an
- fsync() EVEN IF THE APPLICATION WAS PERFORMING UNSTABLE WRITES,
- and get an acknowledgement for it. And we need to wait for the
- fsync() acknowledgement before initiating the hard POST-OP.
-
- However if the FD itself was opened in O_SYNC or O_DSYNC then
- we are already guaranteed that the writes were made stable as
- part of the FOP itself. The same holds true for NFS stable
- writes which happen on an anonymous FD with O_DSYNC or O_SYNC
- flag set in the writev() @flags param. For all other write types,
- mark a flag in the fdctx whenever an unstable write is witnessed.
+ if (afr_changelog_pre_op_uninherit(frame, this) &&
+ afr_txn_nothing_failed(frame, this)) {
+ /* just detected that this post-op is about to
+ be optimized away as a new write() has
+ already piggybacked on this frame's changelog.
*/
+ afr_changelog_post_op_now(frame, this);
+ return 0;
+ }
+
+ /* Calling afr_changelog_post_op_now() now will result in
+ issuing ->[f]xattrop().
+
+ Performing a hard POST-OP (->[f]xattrop() FOP) is a more
+ responsible operation that what it might appear on the surface.
+
+ The changelog of a file (in the xattr of the file on the server)
+ stores information (pending count) about the state of the file
+ on the OTHER server. This changelog is blindly trusted, and must
+ therefore be updated in such a way it remains trustworthy. This
+ implies that decrementing the pending count (essentially "clearing
+ the dirty flag") must be done STRICTLY after we are sure that the
+ operation on the other server has reached stable storage.
+
+ While the backend filesystem on that server will eventually flush
+ it to stable storage, we (being in userspace) have no mechanism
+ to get notified when the write became "stable".
+
+ This means we need take matter into our own hands and issue an
+ fsync() EVEN IF THE APPLICATION WAS PERFORMING UNSTABLE WRITES,
+ and get an acknowledgement for it. And we need to wait for the
+ fsync() acknowledgement before initiating the hard POST-OP.
+
+ However if the FD itself was opened in O_SYNC or O_DSYNC then
+ we are already guaranteed that the writes were made stable as
+ part of the FOP itself. The same holds true for NFS stable
+ writes which happen on an anonymous FD with O_DSYNC or O_SYNC
+ flag set in the writev() @flags param. For all other write types,
+ mark a flag in the fdctx whenever an unstable write is witnessed.
+ */
+
+ if (!afr_fd_has_witnessed_unstable_write(this, local->inode)) {
+ afr_changelog_post_op_now(frame, this);
+ return 0;
+ }
- if (!afr_fd_has_witnessed_unstable_write (this, local->inode)) {
- afr_changelog_post_op_now (frame, this);
- return 0;
- }
-
- /* Check whether users want durability and perform fsync/post-op
- * accordingly.
- */
- if (priv->ensure_durability) {
- /* Time to fsync() */
- afr_changelog_fsync (frame, this);
- } else {
- afr_changelog_post_op_now (frame, this);
- }
+ /* Check whether users want durability and perform fsync/post-op
+ * accordingly.
+ */
+ if (priv->ensure_durability) {
+ /* Time to fsync() */
+ afr_changelog_fsync(frame, this);
+ } else {
+ afr_changelog_post_op_now(frame, this);
+ }
- return 0;
+ return 0;
}
void
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
-{
- struct timespec delta = {0, };
- afr_private_t *priv = NULL;
- afr_local_t *local = frame->local;
- afr_lock_t *lock = NULL;
- gf_boolean_t post_op = _gf_true;
- struct list_head shared;
-
- priv = this->private;
- delta.tv_sec = priv->post_op_delay_secs;
- delta.tv_nsec = 0;
-
- INIT_LIST_HEAD (&shared);
- if (!local->transaction.eager_lock_on)
- goto out;
-
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- list_del_init (&local->transaction.owner_list);
- list_add (&local->transaction.owner_list, &lock->post_op);
- __afr_transaction_wake_shared (local, &shared);
-
- if (!afr_is_delayed_changelog_post_op_needed (frame, this,
- delta.tv_sec)) {
- if (list_empty (&lock->owners))
- lock->release = _gf_true;
- goto unlock;
- }
-
- GF_ASSERT (lock->delay_timer == NULL);
- lock->delay_timer = gf_timer_call_after (this->ctx, delta,
- afr_delayed_changelog_wake_up_cbk,
- local);
- if (!lock->delay_timer) {
- lock->release = _gf_true;
- } else {
- post_op = _gf_false;
- }
+afr_changelog_post_op(call_frame_t *frame, xlator_t *this)
+{
+ struct timespec delta = {
+ 0,
+ };
+ afr_private_t *priv = NULL;
+ afr_local_t *local = frame->local;
+ afr_lock_t *lock = NULL;
+ gf_boolean_t post_op = _gf_true;
+ struct list_head shared;
+
+ priv = this->private;
+ delta.tv_sec = priv->post_op_delay_secs;
+ delta.tv_nsec = 0;
+
+ INIT_LIST_HEAD(&shared);
+ if (!local->transaction.eager_lock_on)
+ goto out;
+
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ list_del_init(&local->transaction.owner_list);
+ list_add(&local->transaction.owner_list, &lock->post_op);
+ __afr_transaction_wake_shared(local, &shared);
+
+ if (!afr_is_delayed_changelog_post_op_needed(frame, this,
+ delta.tv_sec)) {
+ if (list_empty(&lock->owners))
+ lock->release = _gf_true;
+ goto unlock;
+ }
- }
+ GF_ASSERT(lock->delay_timer == NULL);
+ lock->delay_timer = gf_timer_call_after(
+ this->ctx, delta, afr_delayed_changelog_wake_up_cbk, local);
+ if (!lock->delay_timer) {
+ lock->release = _gf_true;
+ } else {
+ post_op = _gf_false;
+ }
+ }
unlock:
- UNLOCK (&local->inode->lock);
+ UNLOCK(&local->inode->lock);
- if (!list_empty (&shared)) {
- afr_lock_resume_shared (&shared);
- }
+ if (!list_empty(&shared)) {
+ afr_lock_resume_shared(&shared);
+ }
out:
- if (post_op) {
- if (!local->transaction.eager_lock_on || lock->release) {
- afr_changelog_post_op_safe (frame, this);
- } else {
- afr_changelog_post_op_now (frame, this);
- }
+ if (post_op) {
+ if (!local->transaction.eager_lock_on || lock->release) {
+ afr_changelog_post_op_safe(frame, this);
+ } else {
+ afr_changelog_post_op_now(frame, this);
}
+ }
}
int
-afr_transaction_resume (call_frame_t *frame, xlator_t *this)
+afr_transaction_resume(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- afr_restore_lk_owner (frame);
+ afr_restore_lk_owner(frame);
- afr_handle_symmetric_errors (frame, this);
+ afr_handle_symmetric_errors(frame, this);
- if (!local->pre_op_compat)
- /* new mode, pre-op was done along
- with OP */
- afr_changelog_pre_op_update (frame, this);
+ if (!local->pre_op_compat)
+ /* new mode, pre-op was done along
+ with OP */
+ afr_changelog_pre_op_update(frame, this);
- afr_changelog_post_op (frame, this);
+ afr_changelog_post_op(frame, this);
- return 0;
+ return 0;
}
-
/**
* afr_transaction_fop_failed - inform that an fop failed
*/
void
-afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
- int child_index)
+afr_transaction_fop_failed(call_frame_t *frame, xlator_t *this, int child_index)
{
- afr_local_t * local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->transaction.failed_subvols[child_index] = 1;
+ local->transaction.failed_subvols[child_index] = 1;
}
static gf_boolean_t
-__need_previous_lock_unlocked (afr_local_t *local)
+__need_previous_lock_unlocked(afr_local_t *local)
{
- afr_lock_t *lock = NULL;
+ afr_lock_t *lock = NULL;
- if (!local->transaction.eager_lock_on)
- return _gf_true;
+ if (!local->transaction.eager_lock_on)
+ return _gf_true;
- lock = &local->inode_ctx->lock[local->transaction.type];
- if (!lock->acquired)
- return _gf_false;
- if (lock->acquired && lock->event_generation != local->event_generation)
- return _gf_true;
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ if (!lock->acquired)
return _gf_false;
+ if (lock->acquired && lock->event_generation != local->event_generation)
+ return _gf_true;
+ return _gf_false;
}
void
-__afr_eager_lock_handle (afr_local_t *local, gf_boolean_t *take_lock,
- gf_boolean_t *do_pre_op, afr_local_t **timer_local)
-{
- afr_lock_t *lock = NULL;
- afr_local_t *owner_local = NULL;
- xlator_t *this = local->transaction.frame->this;
-
- if (local->fd && !afr_are_multiple_fds_opened (local, this)) {
- local->transaction.eager_lock_on = _gf_true;
- afr_set_lk_owner (local->transaction.frame, this, local->inode);
- }
-
- lock = &local->inode_ctx->lock[local->transaction.type];
- if (__need_previous_lock_unlocked (local)) {
- if (!list_empty (&lock->owners)) {
- lock->release = _gf_true;
- } else if (lock->delay_timer) {
- lock->release = _gf_true;
- if (gf_timer_call_cancel (this->ctx,
- lock->delay_timer)) {
- /* It will be put in frozen list
- * in the code flow below*/
- } else {
- *timer_local = list_entry(lock->post_op.next,
- afr_local_t,
- transaction.owner_list);
- lock->delay_timer = NULL;
- }
- }
- if (!local->transaction.eager_lock_on)
- goto out;
- }
-
- if (lock->release) {
- list_add_tail (&local->transaction.wait_list,
- &lock->frozen);
- *take_lock = _gf_false;
- goto out;
- }
-
- if (lock->delay_timer) {
- *take_lock = _gf_false;
- if (gf_timer_call_cancel (this->ctx,
- lock->delay_timer)) {
- list_add_tail (&local->transaction.wait_list,
- &lock->frozen);
- } else {
- *timer_local = list_entry(lock->post_op.next,
- afr_local_t,
- transaction.owner_list);
- afr_copy_inodelk_vars (&local->internal_lock,
- &(*timer_local)->internal_lock,
- this);
- lock->delay_timer = NULL;
- *do_pre_op = _gf_true;
- list_add_tail (&local->transaction.owner_list,
- &lock->owners);
- }
- goto out;
- }
-
- if (!list_empty (&lock->owners)) {
- if (!lock->acquired ||
- afr_has_lock_conflict (local, _gf_true)) {
- list_add_tail (&local->transaction.wait_list,
- &lock->waiting);
- *take_lock = _gf_false;
- goto out;
- }
- owner_local = list_entry (lock->owners.next,
- afr_local_t,
+__afr_eager_lock_handle(afr_local_t *local, gf_boolean_t *take_lock,
+ gf_boolean_t *do_pre_op, afr_local_t **timer_local)
+{
+ afr_lock_t *lock = NULL;
+ afr_local_t *owner_local = NULL;
+ xlator_t *this = local->transaction.frame->this;
+
+ if (local->fd && !afr_are_multiple_fds_opened(local, this)) {
+ local->transaction.eager_lock_on = _gf_true;
+ afr_set_lk_owner(local->transaction.frame, this, local->inode);
+ }
+
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ if (__need_previous_lock_unlocked(local)) {
+ if (!list_empty(&lock->owners)) {
+ lock->release = _gf_true;
+ } else if (lock->delay_timer) {
+ lock->release = _gf_true;
+ if (gf_timer_call_cancel(this->ctx, lock->delay_timer)) {
+ /* It will be put in frozen list
+ * in the code flow below*/
+ } else {
+ *timer_local = list_entry(lock->post_op.next, afr_local_t,
transaction.owner_list);
- afr_copy_inodelk_vars (&local->internal_lock,
- &owner_local->internal_lock,
- this);
- *take_lock = _gf_false;
- *do_pre_op = _gf_true;
+ lock->delay_timer = NULL;
+ }
}
-
- if (lock->acquired)
- GF_ASSERT (!(*take_lock));
- list_add_tail (&local->transaction.owner_list, &lock->owners);
+ if (!local->transaction.eager_lock_on)
+ goto out;
+ }
+
+ if (lock->release) {
+ list_add_tail(&local->transaction.wait_list, &lock->frozen);
+ *take_lock = _gf_false;
+ goto out;
+ }
+
+ if (lock->delay_timer) {
+ *take_lock = _gf_false;
+ if (gf_timer_call_cancel(this->ctx, lock->delay_timer)) {
+ list_add_tail(&local->transaction.wait_list, &lock->frozen);
+ } else {
+ *timer_local = list_entry(lock->post_op.next, afr_local_t,
+ transaction.owner_list);
+ afr_copy_inodelk_vars(&local->internal_lock,
+ &(*timer_local)->internal_lock, this);
+ lock->delay_timer = NULL;
+ *do_pre_op = _gf_true;
+ list_add_tail(&local->transaction.owner_list, &lock->owners);
+ }
+ goto out;
+ }
+
+ if (!list_empty(&lock->owners)) {
+ if (!lock->acquired || afr_has_lock_conflict(local, _gf_true)) {
+ list_add_tail(&local->transaction.wait_list, &lock->waiting);
+ *take_lock = _gf_false;
+ goto out;
+ }
+ owner_local = list_entry(lock->owners.next, afr_local_t,
+ transaction.owner_list);
+ afr_copy_inodelk_vars(&local->internal_lock,
+ &owner_local->internal_lock, this);
+ *take_lock = _gf_false;
+ *do_pre_op = _gf_true;
+ }
+
+ if (lock->acquired)
+ GF_ASSERT(!(*take_lock));
+ list_add_tail(&local->transaction.owner_list, &lock->owners);
out:
- return;
+ return;
}
void
-afr_transaction_start (afr_local_t *local, xlator_t *this)
+afr_transaction_start(afr_local_t *local, xlator_t *this)
{
- afr_private_t *priv = NULL;
- gf_boolean_t take_lock = _gf_true;
- gf_boolean_t do_pre_op = _gf_false;
- afr_local_t *timer_local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t take_lock = _gf_true;
+ gf_boolean_t do_pre_op = _gf_false;
+ afr_local_t *timer_local = NULL;
- priv = this->private;
+ priv = this->private;
- if (local->transaction.type != AFR_DATA_TRANSACTION &&
- local->transaction.type != AFR_METADATA_TRANSACTION)
- goto lock_phase;
+ if (local->transaction.type != AFR_DATA_TRANSACTION &&
+ local->transaction.type != AFR_METADATA_TRANSACTION)
+ goto lock_phase;
- if (!priv->eager_lock)
- goto lock_phase;
+ if (!priv->eager_lock)
+ goto lock_phase;
- LOCK (&local->inode->lock);
- {
- __afr_eager_lock_handle (local, &take_lock, &do_pre_op,
- &timer_local);
- }
- UNLOCK (&local->inode->lock);
+ LOCK(&local->inode->lock);
+ {
+ __afr_eager_lock_handle(local, &take_lock, &do_pre_op, &timer_local);
+ }
+ UNLOCK(&local->inode->lock);
lock_phase:
- if (!local->transaction.eager_lock_on) {
- afr_set_lk_owner (local->transaction.frame, this,
- local->transaction.frame->root);
- }
-
-
- if (take_lock) {
- afr_lock (local->transaction.frame, this);
- } else if (do_pre_op) {
- afr_changelog_pre_op (local->transaction.frame, this);
- }
- /*Always call delayed_changelog_wake_up_cbk after calling pre-op above
- * so that any inheriting can happen*/
- if (timer_local)
- afr_delayed_changelog_wake_up_cbk (timer_local);
+ if (!local->transaction.eager_lock_on) {
+ afr_set_lk_owner(local->transaction.frame, this,
+ local->transaction.frame->root);
+ }
+
+ if (take_lock) {
+ afr_lock(local->transaction.frame, this);
+ } else if (do_pre_op) {
+ afr_changelog_pre_op(local->transaction.frame, this);
+ }
+ /*Always call delayed_changelog_wake_up_cbk after calling pre-op above
+ * so that any inheriting can happen*/
+ if (timer_local)
+ afr_delayed_changelog_wake_up_cbk(timer_local);
}
int
-afr_write_txn_refresh_done (call_frame_t *frame, xlator_t *this, int err)
+afr_write_txn_refresh_done(call_frame_t *frame, xlator_t *this, int err)
{
- afr_local_t *local = frame->local;
+ afr_local_t *local = frame->local;
- if (err) {
- AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN(-1, err);
- goto fail;
- }
+ if (err) {
+ AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN(-1, err);
+ goto fail;
+ }
- afr_transaction_start (local, this);
- return 0;
+ afr_transaction_start(local, this);
+ return 0;
fail:
- local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
- return 0;
+ local->transaction.unwind(frame, this);
+ AFR_STACK_DESTROY(frame);
+ return 0;
}
int
-afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int ret = -1;
- int event_generation = 0;
-
- local = frame->local;
- priv = this->private;
- local->transaction.frame = frame;
-
- local->transaction.type = type;
-
- if (priv->quorum_count && !afr_has_quorum (local->child_up, this)) {
- ret = -afr_quorum_errno(priv);
- goto out;
- }
-
- if (!afr_is_consistent_io_possible (local, priv, &ret)) {
- ret = -ret; /*op_errno to ret conversion*/
- goto out;
- }
-
- ret = afr_transaction_local_init (local, this);
- if (ret < 0)
- goto out;
-
-
- if (type != AFR_METADATA_TRANSACTION) {
- goto txn_start;
- }
-
- ret = afr_inode_get_readable (frame, local->inode, this,
- local->readable, &event_generation, type);
- if (ret < 0 || afr_is_inode_refresh_reqd (local->inode, this,
- priv->event_generation,
- event_generation)) {
- afr_inode_refresh (frame, this, local->inode, local->loc.gfid,
- afr_write_txn_refresh_done);
- ret = 0;
- goto out;
- }
+afr_transaction(call_frame_t *frame, xlator_t *this, afr_transaction_type type)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ int event_generation = 0;
+
+ local = frame->local;
+ priv = this->private;
+ local->transaction.frame = frame;
+
+ local->transaction.type = type;
+
+ if (priv->quorum_count && !afr_has_quorum(local->child_up, this)) {
+ ret = -afr_quorum_errno(priv);
+ goto out;
+ }
+
+ if (!afr_is_consistent_io_possible(local, priv, &ret)) {
+ ret = -ret; /*op_errno to ret conversion*/
+ goto out;
+ }
+
+ ret = afr_transaction_local_init(local, this);
+ if (ret < 0)
+ goto out;
+
+ if (type != AFR_METADATA_TRANSACTION) {
+ goto txn_start;
+ }
+
+ ret = afr_inode_get_readable(frame, local->inode, this, local->readable,
+ &event_generation, type);
+ if (ret < 0 ||
+ afr_is_inode_refresh_reqd(local->inode, this, priv->event_generation,
+ event_generation)) {
+ afr_inode_refresh(frame, this, local->inode, local->loc.gfid,
+ afr_write_txn_refresh_done);
+ ret = 0;
+ goto out;
+ }
txn_start:
- ret = 0;
- afr_transaction_start (local, this);
+ ret = 0;
+ afr_transaction_start(local, this);
out:
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c
index bde9a97a4e7..568293cdf2c 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -21,1196 +21,1135 @@
struct volume_options options[];
static char *afr_favorite_child_policies[AFR_FAV_CHILD_POLICY_MAX + 1] = {
- [AFR_FAV_CHILD_NONE] = "none",
- [AFR_FAV_CHILD_BY_SIZE] = "size",
- [AFR_FAV_CHILD_BY_CTIME] = "ctime",
- [AFR_FAV_CHILD_BY_MTIME] = "mtime",
- [AFR_FAV_CHILD_BY_MAJORITY] = "majority",
- [AFR_FAV_CHILD_POLICY_MAX] = NULL,
+ [AFR_FAV_CHILD_NONE] = "none",
+ [AFR_FAV_CHILD_BY_SIZE] = "size",
+ [AFR_FAV_CHILD_BY_CTIME] = "ctime",
+ [AFR_FAV_CHILD_BY_MTIME] = "mtime",
+ [AFR_FAV_CHILD_BY_MAJORITY] = "majority",
+ [AFR_FAV_CHILD_POLICY_MAX] = NULL,
};
int32_t
-notify (xlator_t *this, int32_t event,
- void *data, ...)
+notify(xlator_t *this, int32_t event, void *data, ...)
{
- int ret = -1;
- va_list ap;
- void *data2 = NULL;
+ int ret = -1;
+ va_list ap;
+ void *data2 = NULL;
- va_start (ap, data);
- data2 = va_arg (ap, dict_t*);
- va_end (ap);
- ret = afr_notify (this, event, data, data2);
+ va_start(ap, data);
+ data2 = va_arg(ap, dict_t *);
+ va_end(ap);
+ ret = afr_notify(this, event, data, data2);
- return ret;
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_afr_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_afr_mt_end + 1);
+ if (ret != 0) {
return ret;
-}
+ }
+ return ret;
+}
int
-xlator_subvolume_index (xlator_t *this, xlator_t *subvol)
+xlator_subvolume_index(xlator_t *this, xlator_t *subvol)
{
- int index = -1;
- int i = 0;
- xlator_list_t *list = NULL;
-
- list = this->children;
-
- while (list) {
- if (subvol == list->xlator ||
- strcmp (subvol->name, list->xlator->name) == 0) {
- index = i;
- break;
- }
- list = list->next;
- i++;
+ int index = -1;
+ int i = 0;
+ xlator_list_t *list = NULL;
+
+ list = this->children;
+
+ while (list) {
+ if (subvol == list->xlator ||
+ strcmp(subvol->name, list->xlator->name) == 0) {
+ index = i;
+ break;
}
+ list = list->next;
+ i++;
+ }
- return index;
+ return index;
}
static void
-fix_quorum_options (xlator_t *this, afr_private_t *priv, char *qtype,
- dict_t *options)
+fix_quorum_options(xlator_t *this, afr_private_t *priv, char *qtype,
+ dict_t *options)
{
-
- if (dict_get (options, "quorum-type") == NULL) {
- /* If user doesn't configure anything enable auto-quorum if the
- * replica has more than two subvolumes */
- if (priv->child_count > 2)
- qtype = "auto";
- }
-
- if (priv->quorum_count && strcmp (qtype, "fixed")) {
- gf_msg (this->name,GF_LOG_WARNING, 0, AFR_MSG_QUORUM_OVERRIDE,
- "quorum-type %s overriding quorum-count %u",
- qtype, priv->quorum_count);
- }
-
- if (!strcmp (qtype, "none")) {
- priv->quorum_count = 0;
- } else if (!strcmp (qtype, "auto")) {
- priv->quorum_count = AFR_QUORUM_AUTO;
- }
-
+ if (dict_get(options, "quorum-type") == NULL) {
+ /* If user doesn't configure anything enable auto-quorum if the
+ * replica has more than two subvolumes */
+ if (priv->child_count > 2)
+ qtype = "auto";
+ }
+
+ if (priv->quorum_count && strcmp(qtype, "fixed")) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_OVERRIDE,
+ "quorum-type %s overriding quorum-count %u", qtype,
+ priv->quorum_count);
+ }
+
+ if (!strcmp(qtype, "none")) {
+ priv->quorum_count = 0;
+ } else if (!strcmp(qtype, "auto")) {
+ priv->quorum_count = AFR_QUORUM_AUTO;
+ }
}
int
-afr_set_favorite_child_policy (afr_private_t *priv, char *policy)
+afr_set_favorite_child_policy(afr_private_t *priv, char *policy)
{
- int index = -1;
+ int index = -1;
- index = gf_get_index_by_elem (afr_favorite_child_policies, policy);
- if (index < 0 || index >= AFR_FAV_CHILD_POLICY_MAX)
- return -1;
+ index = gf_get_index_by_elem(afr_favorite_child_policies, policy);
+ if (index < 0 || index >= AFR_FAV_CHILD_POLICY_MAX)
+ return -1;
- priv->fav_child_policy = index;
+ priv->fav_child_policy = index;
- return 0;
+ return 0;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- afr_private_t *priv = NULL;
- xlator_t *read_subvol = NULL;
- int read_subvol_index = -1;
- int ret = -1;
- int index = -1;
- char *qtype = NULL;
- char *fav_child_policy = NULL;
- gf_boolean_t consistent_io = _gf_false;
- gf_boolean_t choose_local_old = _gf_false;
-
- priv = this->private;
+ afr_private_t *priv = NULL;
+ xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
+ int ret = -1;
+ int index = -1;
+ char *qtype = NULL;
+ char *fav_child_policy = NULL;
+ gf_boolean_t consistent_io = _gf_false;
+ gf_boolean_t choose_local_old = _gf_false;
- GF_OPTION_RECONF ("afr-dirty-xattr",
- priv->afr_dirty, options, str,
- out);
+ priv = this->private;
- GF_OPTION_RECONF ("metadata-splitbrain-forced-heal",
- priv->metadata_splitbrain_forced_heal, options, bool,
- out);
+ GF_OPTION_RECONF("afr-dirty-xattr", priv->afr_dirty, options, str, out);
- GF_OPTION_RECONF ("background-self-heal-count",
- priv->background_self_heal_count, options, uint32,
- out);
+ GF_OPTION_RECONF("metadata-splitbrain-forced-heal",
+ priv->metadata_splitbrain_forced_heal, options, bool, out);
- GF_OPTION_RECONF ("heal-wait-queue-length",
- priv->heal_wait_qlen, options, uint32, out);
+ GF_OPTION_RECONF("background-self-heal-count",
+ priv->background_self_heal_count, options, uint32, out);
+ GF_OPTION_RECONF("heal-wait-queue-length", priv->heal_wait_qlen, options,
+ uint32, out);
- GF_OPTION_RECONF ("metadata-self-heal",
- priv->metadata_self_heal, options, bool, out);
+ GF_OPTION_RECONF("metadata-self-heal", priv->metadata_self_heal, options,
+ bool, out);
- GF_OPTION_RECONF ("data-self-heal", priv->data_self_heal, options, str,
- out);
+ GF_OPTION_RECONF("data-self-heal", priv->data_self_heal, options, str, out);
- GF_OPTION_RECONF ("entry-self-heal", priv->entry_self_heal, options,
- bool, out);
+ GF_OPTION_RECONF("entry-self-heal", priv->entry_self_heal, options, bool,
+ out);
- GF_OPTION_RECONF ("data-self-heal-window-size",
- priv->data_self_heal_window_size, options,
- uint32, out);
+ GF_OPTION_RECONF("data-self-heal-window-size",
+ priv->data_self_heal_window_size, options, uint32, out);
- GF_OPTION_RECONF ("data-self-heal-algorithm",
- priv->data_self_heal_algorithm, options, str, out);
+ GF_OPTION_RECONF("data-self-heal-algorithm", priv->data_self_heal_algorithm,
+ options, str, out);
- GF_OPTION_RECONF ("halo-enabled",
- priv->halo_enabled, options, bool,
- out);
+ GF_OPTION_RECONF("halo-enabled", priv->halo_enabled, options, bool, out);
- GF_OPTION_RECONF ("halo-shd-max-latency",
- priv->shd.halo_max_latency_msec, options, uint32,
- out);
+ GF_OPTION_RECONF("halo-shd-max-latency", priv->shd.halo_max_latency_msec,
+ options, uint32, out);
- GF_OPTION_RECONF ("halo-nfsd-max-latency",
- priv->nfsd.halo_max_latency_msec, options, uint32,
- out);
+ GF_OPTION_RECONF("halo-nfsd-max-latency", priv->nfsd.halo_max_latency_msec,
+ options, uint32, out);
- GF_OPTION_RECONF ("halo-max-latency", priv->halo_max_latency_msec,
- options, uint32, out);
+ GF_OPTION_RECONF("halo-max-latency", priv->halo_max_latency_msec, options,
+ uint32, out);
- GF_OPTION_RECONF ("halo-max-replicas", priv->halo_max_replicas, options,
- uint32, out);
+ GF_OPTION_RECONF("halo-max-replicas", priv->halo_max_replicas, options,
+ uint32, out);
- GF_OPTION_RECONF ("halo-min-replicas", priv->halo_min_replicas, options,
- uint32, out);
+ GF_OPTION_RECONF("halo-min-replicas", priv->halo_min_replicas, options,
+ uint32, out);
- GF_OPTION_RECONF ("read-subvolume", read_subvol, options, xlator, out);
+ GF_OPTION_RECONF("read-subvolume", read_subvol, options, xlator, out);
- choose_local_old = priv->choose_local;
- GF_OPTION_RECONF ("choose-local", priv->choose_local, options, bool,
- out);
+ choose_local_old = priv->choose_local;
+ GF_OPTION_RECONF("choose-local", priv->choose_local, options, bool, out);
- if (choose_local_old != priv->choose_local) {
- priv->read_child = -1;
- if (choose_local_old == _gf_false)
- priv->did_discovery = _gf_false;
- }
-
- GF_OPTION_RECONF ("read-hash-mode", priv->hash_mode,
- options, uint32, out);
-
- if (read_subvol) {
- index = xlator_subvolume_index (this, read_subvol);
- if (index == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL, "%s not a subvolume",
- read_subvol->name);
- goto out;
- }
- priv->read_child = index;
+ if (choose_local_old != priv->choose_local) {
+ priv->read_child = -1;
+ if (choose_local_old == _gf_false)
+ priv->did_discovery = _gf_false;
+ }
+
+ GF_OPTION_RECONF("read-hash-mode", priv->hash_mode, options, uint32, out);
+
+ if (read_subvol) {
+ index = xlator_subvolume_index(this, read_subvol);
+ if (index == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "%s not a subvolume", read_subvol->name);
+ goto out;
}
-
- GF_OPTION_RECONF ("read-subvolume-index",read_subvol_index, options,int32,out);
-
- if (read_subvol_index >-1) {
- index=read_subvol_index;
- if (index >= priv->child_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL,
- "%d not a subvolume-index", index);
- goto out;
- }
- priv->read_child = index;
+ priv->read_child = index;
+ }
+
+ GF_OPTION_RECONF("read-subvolume-index", read_subvol_index, options, int32,
+ out);
+
+ if (read_subvol_index > -1) {
+ index = read_subvol_index;
+ if (index >= priv->child_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "%d not a subvolume-index", index);
+ goto out;
}
+ priv->read_child = index;
+ }
- GF_OPTION_RECONF ("pre-op-compat", priv->pre_op_compat, options, bool,
- out);
- GF_OPTION_RECONF ("locking-scheme", priv->locking_scheme, options, str,
- out);
- GF_OPTION_RECONF ("full-lock", priv->full_lock, options, bool, out);
- GF_OPTION_RECONF ("granular-entry-heal", priv->esh_granular, options,
- bool, out);
+ GF_OPTION_RECONF("pre-op-compat", priv->pre_op_compat, options, bool, out);
+ GF_OPTION_RECONF("locking-scheme", priv->locking_scheme, options, str, out);
+ GF_OPTION_RECONF("full-lock", priv->full_lock, options, bool, out);
+ GF_OPTION_RECONF("granular-entry-heal", priv->esh_granular, options, bool,
+ out);
- GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
- GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
- GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
- uint32, out);
- fix_quorum_options (this, priv, qtype, options);
- if (priv->quorum_count && !afr_has_quorum (priv->child_up, this))
- gf_msg (this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_FAIL,
- "Client-quorum is not met");
+ GF_OPTION_RECONF("eager-lock", priv->eager_lock, options, bool, out);
+ GF_OPTION_RECONF("quorum-type", qtype, options, str, out);
+ GF_OPTION_RECONF("quorum-count", priv->quorum_count, options, uint32, out);
+ fix_quorum_options(this, priv, qtype, options);
+ if (priv->quorum_count && !afr_has_quorum(priv->child_up, this))
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_FAIL,
+ "Client-quorum is not met");
+ GF_OPTION_RECONF("post-op-delay-secs", priv->post_op_delay_secs, options,
+ uint32, out);
- GF_OPTION_RECONF ("post-op-delay-secs", priv->post_op_delay_secs, options,
- uint32, out);
+ GF_OPTION_RECONF(AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, options,
+ size_uint64, out);
+ /* Reset this so we re-discover in case the topology changed. */
+ GF_OPTION_RECONF("ensure-durability", priv->ensure_durability, options,
+ bool, out);
- GF_OPTION_RECONF (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size,
- options, size_uint64, out);
- /* Reset this so we re-discover in case the topology changed. */
- GF_OPTION_RECONF ("ensure-durability", priv->ensure_durability, options,
- bool, out);
+ GF_OPTION_RECONF("self-heal-daemon", priv->shd.enabled, options, bool, out);
- GF_OPTION_RECONF ("self-heal-daemon", priv->shd.enabled, options,
- bool, out);
+ GF_OPTION_RECONF("iam-self-heal-daemon", priv->shd.iamshd, options, bool,
+ out);
- GF_OPTION_RECONF ("iam-self-heal-daemon", priv->shd.iamshd, options,
- bool, out);
+ GF_OPTION_RECONF("heal-timeout", priv->shd.timeout, options, int32, out);
- GF_OPTION_RECONF ("heal-timeout", priv->shd.timeout, options,
- int32, out);
+ GF_OPTION_RECONF("consistent-metadata", priv->consistent_metadata, options,
+ bool, out);
- GF_OPTION_RECONF ("consistent-metadata", priv->consistent_metadata,
- options, bool, out);
+ GF_OPTION_RECONF("shd-max-threads", priv->shd.max_threads, options, uint32,
+ out);
- GF_OPTION_RECONF ("shd-max-threads", priv->shd.max_threads,
- options, uint32, out);
+ GF_OPTION_RECONF("shd-wait-qlength", priv->shd.wait_qlength, options,
+ uint32, out);
- GF_OPTION_RECONF ("shd-wait-qlength", priv->shd.wait_qlength,
- options, uint32, out);
-
- GF_OPTION_RECONF ("favorite-child-policy", fav_child_policy, options,
- str, out);
- if (afr_set_favorite_child_policy (priv, fav_child_policy) == -1)
- goto out;
+ GF_OPTION_RECONF("favorite-child-policy", fav_child_policy, options, str,
+ out);
+ if (afr_set_favorite_child_policy(priv, fav_child_policy) == -1)
+ goto out;
- priv->did_discovery = _gf_false;
+ priv->did_discovery = _gf_false;
- GF_OPTION_RECONF ("consistent-io", consistent_io, options, bool, out);
- if (priv->quorum_count != 0)
- consistent_io = _gf_false;
- priv->consistent_io = consistent_io;
+ GF_OPTION_RECONF("consistent-io", consistent_io, options, bool, out);
+ if (priv->quorum_count != 0)
+ consistent_io = _gf_false;
+ priv->consistent_io = consistent_io;
- ret = 0;
+ ret = 0;
out:
- return ret;
-
+ return ret;
}
static int
-afr_pending_xattrs_init (afr_private_t *priv, xlator_t *this)
+afr_pending_xattrs_init(afr_private_t *priv, xlator_t *this)
{
- int ret = -1;
- int i = 0;
- char *ptr = NULL;
- char *ptr1 = NULL;
- char *xattrs_list = NULL;
- xlator_list_t *trav = NULL;
- int child_count = -1;
-
- trav = this->children;
- child_count = priv->child_count;
- if (priv->thin_arbiter_count) {
- /* priv->pending_key[THIN_ARBITER_BRICK_INDEX] is used as the
- * name of the thin arbiter file for persistence across add/
- * removal of DHT subvols.*/
- child_count++;
- }
+ int ret = -1;
+ int i = 0;
+ char *ptr = NULL;
+ char *ptr1 = NULL;
+ char *xattrs_list = NULL;
+ xlator_list_t *trav = NULL;
+ int child_count = -1;
+
+ trav = this->children;
+ child_count = priv->child_count;
+ if (priv->thin_arbiter_count) {
+ /* priv->pending_key[THIN_ARBITER_BRICK_INDEX] is used as the
+ * name of the thin arbiter file for persistence across add/
+ * removal of DHT subvols.*/
+ child_count++;
+ }
+
+ GF_OPTION_INIT("afr-pending-xattr", xattrs_list, str, out);
+ priv->pending_key = GF_CALLOC(sizeof(*priv->pending_key), child_count,
+ gf_afr_mt_char);
+ if (!priv->pending_key) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ if (!xattrs_list) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_NO_CHANGELOG,
+ "Unable to fetch afr-pending-xattr option from volfile."
+ " Falling back to using client translator names. ");
- GF_OPTION_INIT ("afr-pending-xattr", xattrs_list, str, out);
- priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key),
- child_count, gf_afr_mt_char);
- if (!priv->pending_key) {
- ret = -ENOMEM;
- goto out;
- }
- if (!xattrs_list) {
- gf_msg (this->name, GF_LOG_WARNING, 0, AFR_MSG_NO_CHANGELOG,
- "Unable to fetch afr-pending-xattr option from volfile."
- " Falling back to using client translator names. ");
-
- while (i < child_count) {
- ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
- AFR_XATTR_PREFIX,
- trav->xlator->name);
- if (ret == -1) {
- ret = -ENOMEM;
- goto out;
- }
- trav = trav->next;
- i++;
- }
- ret = 0;
- goto out;
- }
-
- ptr = ptr1 = gf_strdup (xattrs_list);
- if (!ptr) {
+ while (i < child_count) {
+ ret = gf_asprintf(&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
+ trav->xlator->name);
+ if (ret == -1) {
ret = -ENOMEM;
goto out;
- }
- for (i = 0, ptr = strtok (ptr, ","); ptr; ptr = strtok (NULL, ",")) {
- ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
- AFR_XATTR_PREFIX, ptr);
- if (ret == -1) {
- ret = -ENOMEM;
- goto out;
- }
- i++;
+ }
+ trav = trav->next;
+ i++;
}
ret = 0;
+ goto out;
+ }
+
+ ptr = ptr1 = gf_strdup(xattrs_list);
+ if (!ptr) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ for (i = 0, ptr = strtok(ptr, ","); ptr; ptr = strtok(NULL, ",")) {
+ ret = gf_asprintf(&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
+ ptr);
+ if (ret == -1) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ i++;
+ }
+ ret = 0;
out:
- GF_FREE (ptr1);
- return ret;
-
+ GF_FREE(ptr1);
+ return ret;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- afr_private_t *priv = NULL;
- int child_count = 0;
- xlator_list_t *trav = NULL;
- int i = 0;
- int ret = -1;
- GF_UNUSED int op_errno = 0;
- xlator_t *read_subvol = NULL;
- int read_subvol_index = -1;
- char *qtype = NULL;
- char *fav_child_policy = NULL;
- char *thin_arbiter = NULL;
-
- if (!this->children) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_CHILD_MISCONFIGURED,
- "replicate translator needs more than one "
- "subvolume defined.");
- return -1;
- }
-
- if (!this->parents) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_VOL_MISCONFIGURED, "Volume is dangling.");
- }
-
- this->private = GF_CALLOC (1, sizeof (afr_private_t),
- gf_afr_mt_afr_private_t);
- if (!this->private)
- goto out;
-
- priv = this->private;
- LOCK_INIT (&priv->lock);
-
- child_count = xlator_subvolume_count (this);
-
- priv->child_count = child_count;
-
- priv->read_child = -1;
-
- GF_OPTION_INIT ("arbiter-count", priv->arbiter_count, uint32, out);
- GF_OPTION_INIT ("thin-arbiter", thin_arbiter, str, out);
- if (thin_arbiter && strlen(thin_arbiter) > 0) {
- priv->thin_arbiter_count = 1;
- priv->child_count--;
- priv->ta_bad_child_index = AFR_CHILD_UNKNOWN;
- priv->ta_notify_dom_lock_offset = 0;
+ afr_private_t *priv = NULL;
+ int child_count = 0;
+ xlator_list_t *trav = NULL;
+ int i = 0;
+ int ret = -1;
+ GF_UNUSED int op_errno = 0;
+ xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
+ char *qtype = NULL;
+ char *fav_child_policy = NULL;
+ char *thin_arbiter = NULL;
+
+ if (!this->children) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_CHILD_MISCONFIGURED,
+ "replicate translator needs more than one "
+ "subvolume defined.");
+ return -1;
+ }
+
+ if (!this->parents) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_VOL_MISCONFIGURED,
+ "Volume is dangling.");
+ }
+
+ this->private = GF_CALLOC(1, sizeof(afr_private_t),
+ gf_afr_mt_afr_private_t);
+ if (!this->private)
+ goto out;
+
+ priv = this->private;
+ LOCK_INIT(&priv->lock);
+
+ child_count = xlator_subvolume_count(this);
+
+ priv->child_count = child_count;
+
+ priv->read_child = -1;
+
+ GF_OPTION_INIT("arbiter-count", priv->arbiter_count, uint32, out);
+ GF_OPTION_INIT("thin-arbiter", thin_arbiter, str, out);
+ if (thin_arbiter && strlen(thin_arbiter) > 0) {
+ priv->thin_arbiter_count = 1;
+ priv->child_count--;
+ priv->ta_bad_child_index = AFR_CHILD_UNKNOWN;
+ priv->ta_notify_dom_lock_offset = 0;
+ }
+ INIT_LIST_HEAD(&priv->healing);
+ INIT_LIST_HEAD(&priv->heal_waiting);
+
+ priv->spb_choice_timeout = AFR_DEFAULT_SPB_CHOICE_TIMEOUT;
+
+ GF_OPTION_INIT("afr-dirty-xattr", priv->afr_dirty, str, out);
+
+ GF_OPTION_INIT("metadata-splitbrain-forced-heal",
+ priv->metadata_splitbrain_forced_heal, bool, out);
+
+ GF_OPTION_INIT("read-subvolume", read_subvol, xlator, out);
+ if (read_subvol) {
+ priv->read_child = xlator_subvolume_index(this, read_subvol);
+ if (priv->read_child == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "%s not a subvolume", read_subvol->name);
+ goto out;
}
- INIT_LIST_HEAD (&priv->healing);
- INIT_LIST_HEAD (&priv->heal_waiting);
-
- priv->spb_choice_timeout = AFR_DEFAULT_SPB_CHOICE_TIMEOUT;
-
- GF_OPTION_INIT ("afr-dirty-xattr", priv->afr_dirty, str, out);
-
- GF_OPTION_INIT ("metadata-splitbrain-forced-heal",
- priv->metadata_splitbrain_forced_heal, bool, out);
-
- GF_OPTION_INIT ("read-subvolume", read_subvol, xlator, out);
- if (read_subvol) {
- priv->read_child = xlator_subvolume_index (this, read_subvol);
- if (priv->read_child == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL, "%s not a subvolume",
- read_subvol->name);
- goto out;
- }
- }
- GF_OPTION_INIT ("read-subvolume-index",read_subvol_index,int32,out);
- if (read_subvol_index > -1) {
- if (read_subvol_index >= priv->child_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL,
- "%d not a subvolume-index", read_subvol_index);
- goto out;
- }
- priv->read_child = read_subvol_index;
+ }
+ GF_OPTION_INIT("read-subvolume-index", read_subvol_index, int32, out);
+ if (read_subvol_index > -1) {
+ if (read_subvol_index >= priv->child_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "%d not a subvolume-index", read_subvol_index);
+ goto out;
}
- GF_OPTION_INIT ("choose-local", priv->choose_local, bool, out);
-
- priv->pending_reads = GF_CALLOC (sizeof(*priv->pending_reads),
- priv->child_count, gf_afr_mt_atomic_t);
-
- GF_OPTION_INIT ("read-hash-mode", priv->hash_mode, uint32, out);
+ priv->read_child = read_subvol_index;
+ }
+ GF_OPTION_INIT("choose-local", priv->choose_local, bool, out);
- priv->favorite_child = -1;
+ priv->pending_reads = GF_CALLOC(sizeof(*priv->pending_reads),
+ priv->child_count, gf_afr_mt_atomic_t);
- GF_OPTION_INIT ("favorite-child-policy", fav_child_policy, str, out);
- if (afr_set_favorite_child_policy(priv, fav_child_policy) == -1)
- goto out;
-
- GF_OPTION_INIT ("shd-max-threads", priv->shd.max_threads,
- uint32, out);
-
- GF_OPTION_INIT ("shd-wait-qlength", priv->shd.wait_qlength,
- uint32, out);
+ GF_OPTION_INIT("read-hash-mode", priv->hash_mode, uint32, out);
- GF_OPTION_INIT ("background-self-heal-count",
- priv->background_self_heal_count, uint32, out);
+ priv->favorite_child = -1;
- GF_OPTION_INIT ("heal-wait-queue-length",
- priv->heal_wait_qlen, uint32, out);
+ GF_OPTION_INIT("favorite-child-policy", fav_child_policy, str, out);
+ if (afr_set_favorite_child_policy(priv, fav_child_policy) == -1)
+ goto out;
- GF_OPTION_INIT ("data-self-heal", priv->data_self_heal, str, out);
+ GF_OPTION_INIT("shd-max-threads", priv->shd.max_threads, uint32, out);
- GF_OPTION_INIT ("data-self-heal-algorithm",
- priv->data_self_heal_algorithm, str, out);
+ GF_OPTION_INIT("shd-wait-qlength", priv->shd.wait_qlength, uint32, out);
- GF_OPTION_INIT ("data-self-heal-window-size",
- priv->data_self_heal_window_size, uint32, out);
+ GF_OPTION_INIT("background-self-heal-count",
+ priv->background_self_heal_count, uint32, out);
- GF_OPTION_INIT ("metadata-self-heal", priv->metadata_self_heal, bool,
- out);
+ GF_OPTION_INIT("heal-wait-queue-length", priv->heal_wait_qlen, uint32, out);
- GF_OPTION_INIT ("entry-self-heal", priv->entry_self_heal, bool, out);
+ GF_OPTION_INIT("data-self-heal", priv->data_self_heal, str, out);
- GF_OPTION_INIT ("halo-shd-max-latency", priv->shd.halo_max_latency_msec,
- uint32, out);
+ GF_OPTION_INIT("data-self-heal-algorithm", priv->data_self_heal_algorithm,
+ str, out);
- GF_OPTION_INIT ("halo-max-latency", priv->halo_max_latency_msec,
- uint32, out);
- GF_OPTION_INIT ("halo-max-replicas", priv->halo_max_replicas, uint32,
- out);
- GF_OPTION_INIT ("halo-min-replicas", priv->halo_min_replicas, uint32,
- out);
+ GF_OPTION_INIT("data-self-heal-window-size",
+ priv->data_self_heal_window_size, uint32, out);
- GF_OPTION_INIT ("halo-enabled",
- priv->halo_enabled, bool, out);
+ GF_OPTION_INIT("metadata-self-heal", priv->metadata_self_heal, bool, out);
- GF_OPTION_INIT ("halo-nfsd-max-latency",
- priv->nfsd.halo_max_latency_msec, uint32, out);
+ GF_OPTION_INIT("entry-self-heal", priv->entry_self_heal, bool, out);
- GF_OPTION_INIT ("iam-nfs-daemon", priv->nfsd.iamnfsd, bool, out);
+ GF_OPTION_INIT("halo-shd-max-latency", priv->shd.halo_max_latency_msec,
+ uint32, out);
- GF_OPTION_INIT ("optimistic-change-log", priv->optimistic_change_log,
- bool, out);
+ GF_OPTION_INIT("halo-max-latency", priv->halo_max_latency_msec, uint32,
+ out);
+ GF_OPTION_INIT("halo-max-replicas", priv->halo_max_replicas, uint32, out);
+ GF_OPTION_INIT("halo-min-replicas", priv->halo_min_replicas, uint32, out);
- GF_OPTION_INIT ("pre-op-compat", priv->pre_op_compat, bool, out);
- GF_OPTION_INIT ("locking-scheme", priv->locking_scheme, str, out);
- GF_OPTION_INIT ("full-lock", priv->full_lock, bool, out);
- GF_OPTION_INIT ("granular-entry-heal", priv->esh_granular, bool, out);
+ GF_OPTION_INIT("halo-enabled", priv->halo_enabled, bool, out);
- GF_OPTION_INIT ("eager-lock", priv->eager_lock, bool, out);
- GF_OPTION_INIT ("quorum-type", qtype, str, out);
- GF_OPTION_INIT ("quorum-count", priv->quorum_count, uint32, out);
- GF_OPTION_INIT (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size_uint64,
- out);
- fix_quorum_options (this, priv, qtype, this->options);
+ GF_OPTION_INIT("halo-nfsd-max-latency", priv->nfsd.halo_max_latency_msec,
+ uint32, out);
- GF_OPTION_INIT ("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
- GF_OPTION_INIT ("ensure-durability", priv->ensure_durability, bool,
- out);
+ GF_OPTION_INIT("iam-nfs-daemon", priv->nfsd.iamnfsd, bool, out);
- GF_OPTION_INIT ("self-heal-daemon", priv->shd.enabled, bool, out);
+ GF_OPTION_INIT("optimistic-change-log", priv->optimistic_change_log, bool,
+ out);
- GF_OPTION_INIT ("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
- GF_OPTION_INIT ("heal-timeout", priv->shd.timeout, int32, out);
+ GF_OPTION_INIT("pre-op-compat", priv->pre_op_compat, bool, out);
+ GF_OPTION_INIT("locking-scheme", priv->locking_scheme, str, out);
+ GF_OPTION_INIT("full-lock", priv->full_lock, bool, out);
+ GF_OPTION_INIT("granular-entry-heal", priv->esh_granular, bool, out);
- GF_OPTION_INIT ("consistent-metadata", priv->consistent_metadata, bool,
- out);
- GF_OPTION_INIT ("consistent-io", priv->consistent_io, bool, out);
+ GF_OPTION_INIT("eager-lock", priv->eager_lock, bool, out);
+ GF_OPTION_INIT("quorum-type", qtype, str, out);
+ GF_OPTION_INIT("quorum-count", priv->quorum_count, uint32, out);
+ GF_OPTION_INIT(AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size_uint64,
+ out);
+ fix_quorum_options(this, priv, qtype, this->options);
- if (priv->quorum_count != 0)
- priv->consistent_io = _gf_false;
+ GF_OPTION_INIT("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
+ GF_OPTION_INIT("ensure-durability", priv->ensure_durability, bool, out);
- priv->wait_count = 1;
+ GF_OPTION_INIT("self-heal-daemon", priv->shd.enabled, bool, out);
- priv->local = GF_CALLOC (sizeof (unsigned char), child_count,
- gf_afr_mt_char);
- if (!priv->local) {
- ret = -ENOMEM;
- goto out;
- }
+ GF_OPTION_INIT("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
+ GF_OPTION_INIT("heal-timeout", priv->shd.timeout, int32, out);
- priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
- gf_afr_mt_char);
+ GF_OPTION_INIT("consistent-metadata", priv->consistent_metadata, bool, out);
+ GF_OPTION_INIT("consistent-io", priv->consistent_io, bool, out);
- priv->child_latency = GF_MALLOC (sizeof (*priv->child_latency)
- * child_count,
- gf_afr_mt_child_latency_t);
+ if (priv->quorum_count != 0)
+ priv->consistent_io = _gf_false;
- if (!priv->child_up || !priv->child_latency) {
- ret = -ENOMEM;
- goto out;
- }
- /*Initialize to -ve ping timeout so that they are not considered
- * in child-up events until ping-event comes*/
- for (i = 0; i < child_count; i++)
- priv->child_latency[i] = -1;
-
- priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
- gf_afr_mt_xlator_t);
- if (!priv->children) {
- ret = -ENOMEM;
- goto out;
- }
+ priv->wait_count = 1;
- ret = afr_pending_xattrs_init (priv, this);
- if (ret)
- goto out;
+ priv->local = GF_CALLOC(sizeof(unsigned char), child_count, gf_afr_mt_char);
+ if (!priv->local) {
+ ret = -ENOMEM;
+ goto out;
+ }
- trav = this->children;
- i = 0;
- while (i < child_count) {
- priv->children[i] = trav->xlator;
- trav = trav->next;
- i++;
- }
+ priv->child_up = GF_CALLOC(sizeof(unsigned char), child_count,
+ gf_afr_mt_char);
- ret = gf_asprintf (&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT,
- this->name);
- if (-1 == ret) {
- ret = -ENOMEM;
- goto out;
- }
+ priv->child_latency = GF_MALLOC(sizeof(*priv->child_latency) * child_count,
+ gf_afr_mt_child_latency_t);
- priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
- gf_afr_mt_int32_t);
- if (!priv->last_event) {
- ret = -ENOMEM;
- goto out;
- }
+ if (!priv->child_up || !priv->child_latency) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ /*Initialize to -ve ping timeout so that they are not considered
+ * in child-up events until ping-event comes*/
+ for (i = 0; i < child_count; i++)
+ priv->child_latency[i] = -1;
- ret = afr_selfheal_daemon_init (this);
- if (ret) {
- ret = -ENOMEM;
- goto out;
- }
+ priv->children = GF_CALLOC(sizeof(xlator_t *), child_count,
+ gf_afr_mt_xlator_t);
+ if (!priv->children) {
+ ret = -ENOMEM;
+ goto out;
+ }
- /* keep more local here as we may need them for self-heal etc */
- this->local_pool = mem_pool_new (afr_local_t, 512);
- if (!this->local_pool) {
- ret = -1;
- goto out;
- }
+ ret = afr_pending_xattrs_init(priv, this);
+ if (ret)
+ goto out;
- priv->root_inode = NULL;
+ trav = this->children;
+ i = 0;
+ while (i < child_count) {
+ priv->children[i] = trav->xlator;
+ trav = trav->next;
+ i++;
+ }
- ret = 0;
+ ret = gf_asprintf(&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT, this->name);
+ if (-1 == ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ priv->last_event = GF_CALLOC(child_count, sizeof(*priv->last_event),
+ gf_afr_mt_int32_t);
+ if (!priv->last_event) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ ret = afr_selfheal_daemon_init(this);
+ if (ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ /* keep more local here as we may need them for self-heal etc */
+ this->local_pool = mem_pool_new(afr_local_t, 512);
+ if (!this->local_pool) {
+ ret = -1;
+ goto out;
+ }
+
+ priv->root_inode = NULL;
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
int
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- afr_private_t *priv = NULL;
-
- priv = this->private;
- LOCK (&priv->lock);
- if (priv->timer != NULL) {
- gf_timer_call_cancel(this->ctx, priv->timer);
- priv->timer = NULL;
- }
- UNLOCK (&priv->lock);
- this->private = NULL;
- afr_priv_destroy (priv);
- //if (this->itable);//I don't see any destroy func
-
- return 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ LOCK(&priv->lock);
+ if (priv->timer != NULL) {
+ gf_timer_call_cancel(this->ctx, priv->timer);
+ priv->timer = NULL;
+ }
+ UNLOCK(&priv->lock);
+ this->private = NULL;
+ afr_priv_destroy(priv);
+ // if (this->itable);//I don't see any destroy func
+
+ return 0;
}
-
struct xlator_fops fops = {
- .lookup = afr_lookup,
- .lk = afr_lk,
- .flush = afr_flush,
- .statfs = afr_statfs,
- .fsyncdir = afr_fsyncdir,
- .inodelk = afr_inodelk,
- .finodelk = afr_finodelk,
- .entrylk = afr_entrylk,
- .fentrylk = afr_fentrylk,
- .ipc = afr_ipc,
- .lease = afr_lease,
-
- /* inode read */
- .access = afr_access,
- .stat = afr_stat,
- .fstat = afr_fstat,
- .readlink = afr_readlink,
- .getxattr = afr_getxattr,
- .fgetxattr = afr_fgetxattr,
- .readv = afr_readv,
-
- /* inode write */
- .writev = afr_writev,
- .truncate = afr_truncate,
- .ftruncate = afr_ftruncate,
- .setxattr = afr_setxattr,
- .fsetxattr = afr_fsetxattr,
- .setattr = afr_setattr,
- .fsetattr = afr_fsetattr,
- .removexattr = afr_removexattr,
- .fremovexattr = afr_fremovexattr,
- .fallocate = afr_fallocate,
- .discard = afr_discard,
- .zerofill = afr_zerofill,
- .xattrop = afr_xattrop,
- .fxattrop = afr_fxattrop,
- .fsync = afr_fsync,
-
- /*inode open*/
- .opendir = afr_opendir,
- .open = afr_open,
-
- /* dir read */
- .readdir = afr_readdir,
- .readdirp = afr_readdirp,
-
- /* dir write */
- .create = afr_create,
- .mknod = afr_mknod,
- .mkdir = afr_mkdir,
- .unlink = afr_unlink,
- .rmdir = afr_rmdir,
- .link = afr_link,
- .symlink = afr_symlink,
- .rename = afr_rename,
+ .lookup = afr_lookup,
+ .lk = afr_lk,
+ .flush = afr_flush,
+ .statfs = afr_statfs,
+ .fsyncdir = afr_fsyncdir,
+ .inodelk = afr_inodelk,
+ .finodelk = afr_finodelk,
+ .entrylk = afr_entrylk,
+ .fentrylk = afr_fentrylk,
+ .ipc = afr_ipc,
+ .lease = afr_lease,
+
+ /* inode read */
+ .access = afr_access,
+ .stat = afr_stat,
+ .fstat = afr_fstat,
+ .readlink = afr_readlink,
+ .getxattr = afr_getxattr,
+ .fgetxattr = afr_fgetxattr,
+ .readv = afr_readv,
+
+ /* inode write */
+ .writev = afr_writev,
+ .truncate = afr_truncate,
+ .ftruncate = afr_ftruncate,
+ .setxattr = afr_setxattr,
+ .fsetxattr = afr_fsetxattr,
+ .setattr = afr_setattr,
+ .fsetattr = afr_fsetattr,
+ .removexattr = afr_removexattr,
+ .fremovexattr = afr_fremovexattr,
+ .fallocate = afr_fallocate,
+ .discard = afr_discard,
+ .zerofill = afr_zerofill,
+ .xattrop = afr_xattrop,
+ .fxattrop = afr_fxattrop,
+ .fsync = afr_fsync,
+
+ /*inode open*/
+ .opendir = afr_opendir,
+ .open = afr_open,
+
+ /* dir read */
+ .readdir = afr_readdir,
+ .readdirp = afr_readdirp,
+
+ /* dir write */
+ .create = afr_create,
+ .mknod = afr_mknod,
+ .mkdir = afr_mkdir,
+ .unlink = afr_unlink,
+ .rmdir = afr_rmdir,
+ .link = afr_link,
+ .symlink = afr_symlink,
+ .rename = afr_rename,
};
-
struct xlator_dumpops dumpops = {
- .priv = afr_priv_dump,
+ .priv = afr_priv_dump,
};
-
struct xlator_cbks cbks = {
- .release = afr_release,
- .releasedir = afr_releasedir,
- .forget = afr_forget,
+ .release = afr_release,
+ .releasedir = afr_releasedir,
+ .forget = afr_forget,
};
-
struct volume_options options[] = {
- { .key = {"read-subvolume" },
- .type = GF_OPTION_TYPE_XLATOR,
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "inode-read fops happen only on one of the bricks in "
- "replicate. Afr will prefer the one specified using "
- "this option if it is not stale. Option value must be "
- "one of the xlator names of the children. "
- "Ex: <volname>-client-0 till "
- "<volname>-client-<number-of-bricks - 1>"
- },
- { .key = {"read-subvolume-index" },
- .type = GF_OPTION_TYPE_INT,
- .default_value = "-1",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "inode-read fops happen only on one of the bricks in "
- "replicate. AFR will prefer the one specified using "
- "this option if it is not stale. allowed options"
- " include -1 till replica-count - 1"
- },
- { .key = {"read-hash-mode" },
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = 3,
- .default_value = "1",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "inode-read fops happen only on one of the bricks in "
- "replicate. AFR will prefer the one computed using "
- "the method specified using this option.\n"
- "0 = first readable child of AFR, starting from 1st child.\n"
- "1 = hash by GFID of file (all clients use "
- "same subvolume).\n"
- "2 = hash by GFID of file and client PID.\n"
- "3 = brick having the least outstanding read requests."
- },
- { .key = {"choose-local" },
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "true",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Choose a local subvolume (i.e. Brick) to read from"
- " if read-subvolume is not explicitly set.",
- },
- { .key = {"background-self-heal-count"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = 256,
- .default_value = "8",
- .validate = GF_OPT_VALIDATE_MIN,
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This specifies the number of per client self-heal "
- "jobs that can perform parallel heals in the "
- "background."
- },
- { .key = {"halo-shd-max-latency"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 99999,
- .default_value = "99999",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "Maximum latency for shd halo replication in msec."
- },
- { .key = {"halo-enabled"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "False",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "Enable Halo (geo) replication mode."
- },
- { .key = {"halo-nfsd-max-latency"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 99999,
- .default_value = "5",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "Maximum latency for nfsd halo replication in msec."
- },
- { .key = {"halo-max-latency"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = AFR_HALO_MAX_LATENCY,
- .default_value = "5",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "Maximum latency for halo replication in msec."
- },
- { .key = {"halo-max-replicas"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 99999,
- .default_value = "99999",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "The maximum number of halo replicas; replicas"
- " beyond this value will be written asynchronously"
- "via the SHD."
- },
- { .key = {"halo-min-replicas"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 99999,
- .default_value = "2",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "The minimmum number of halo replicas, before adding "
- "out of region replicas."
- },
- { .key = {"heal-wait-queue-length"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = 10000, /*Around 100MB with sizeof(afr_local_t)= 10496 bytes*/
- .default_value = "128",
- .validate = GF_OPT_VALIDATE_MIN,
- .op_version = {GD_OP_VERSION_3_7_10},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This specifies the number of heals that can be queued"
- " for the parallel background self heal jobs."
- },
- { .key = {"data-self-heal"},
- .type = GF_OPTION_TYPE_STR,
- .value = {"1", "on", "yes", "true", "enable",
- "0", "off", "no", "false", "disable",
- "open"},
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Using this option we can enable/disable data "
- "self-heal on the file. \"open\" means data "
- "self-heal action will only be triggered by file "
- "open operations."
- },
- { .key = {"data-self-heal-algorithm"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Select between \"full\", \"diff\". The "
- "\"full\" algorithm copies the entire file from "
- "source to sink. The \"diff\" algorithm copies to "
- "sink only those blocks whose checksums don't match "
- "with those of source. If no option is configured "
- "the option is chosen dynamically as follows: "
- "If the file does not exist on one of the sinks "
- "or empty file exists or if the source file size is "
- "about the same as page size the entire file will "
- "be read and written i.e \"full\" algo, "
- "otherwise \"diff\" algo is chosen.",
- .value = { "diff", "full"}
- },
- { .key = {"data-self-heal-window-size"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 1024,
- .default_value = "1",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Maximum number blocks per file for which self-heal "
- "process would be applied simultaneously."
- },
- { .key = {"metadata-self-heal"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- /*.validate_fn = validate_replica*/
- .description = "Using this option we can enable/disable metadata "
- "i.e. Permissions, ownerships, xattrs self-heal on "
- "the file/directory."
- },
- { .key = {"entry-self-heal"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- /*.validate_fn = validate_replica*/
- .description = "Using this option we can enable/disable entry "
- "self-heal on the directory."
- },
- { .key = {"data-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option exists only for backward compatibility "
- "and configuring it doesn't have any effect"
- },
- { .key = {"metadata-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option exists only for backward compatibility "
- "and configuring it doesn't have any effect"
- },
- { .key = {"entry-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option exists only for backward compatibility "
- "and configuring it doesn't have any effect"
- },
- { .key = {"optimistic-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "Entry/Metadata fops will not perform "
- "pre fop changelog operations in afr transaction "
- "if this option is enabled."
- },
- { .key = {"inodelk-trace"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "Enabling this option logs inode lock/unlocks"
- },
- { .key = {"entrylk-trace"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "Enabling this option logs entry lock/unlocks"
- },
- { .key = {"pre-op-compat"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "Use separate pre-op xattrop() FOP rather than "
- "overloading xdata of the OP"
- },
- { .key = {"eager-lock"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Enable/Disable eager lock for replica volume. "
- "Lock phase of a transaction has two sub-phases. "
- "First is an attempt to acquire locks in parallel by "
- "broadcasting non-blocking lock requests. If lock "
- "acquisition fails on any server, then the held locks "
- "are unlocked and we revert to a blocking locks mode "
- "sequentially on one server after another. If this "
- "option is enabled the initial broadcasting lock "
- "request attempts to acquire a full lock on the entire file. "
- "If this fails, we revert back to the sequential "
- "\"regional\" blocking locks as before. In the case "
- "where such an \"eager\" lock is granted in the "
- "non-blocking phase, it gives rise to an opportunity "
- "for optimization. i.e, if the next write transaction "
- "on the same FD arrives before the unlock phase of "
- "the first transaction, it \"takes over\" the full "
- "file lock. Similarly if yet another data transaction "
- "arrives before the unlock phase of the \"optimized\" "
- "transaction, that in turn \"takes over\" the lock as "
- "well. The actual unlock now happens at the end of "
- "the last \"optimized\" transaction."
-
- },
- { .key = {"self-heal-daemon"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- .tags = {"replicate"},
- /*.validate_fn = validate_replica_heal_enable_disable*/
- .description = "This option applies to only self-heal-daemon. "
- "Index directory crawl and automatic healing of files "
- "will not be performed if this option is turned off."
- },
- { .key = {"iam-self-heal-daemon"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "This option differentiates if the replicate "
- "translator is running as part of self-heal-daemon "
- "or not."
- },
- { .key = {"iam-nfs-daemon"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "This option differentiates if the replicate "
- "translator is running as part of an NFS daemon "
- "or not."
- },
- { .key = {"quorum-type"},
- .type = GF_OPTION_TYPE_STR,
- .value = { "none", "auto", "fixed"},
- .default_value = "none",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- /*.option = quorum-type*/
- .description = "If value is \"fixed\" only allow writes if "
- "quorum-count bricks are present. If value is "
- "\"auto\" only allow writes if more than half of "
- "bricks, or exactly half including the first, are "
- "present.",
- },
- { .key = {"quorum-count"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = INT_MAX,
- .default_value = 0,
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- /*.option = quorum-count*/
- /*.validate_fn = validate_quorum_count*/
- .description = "If quorum-type is \"fixed\" only allow writes if "
- "this many bricks are present. Other quorum types "
- "will OVERWRITE this value.",
- },
- { .key = {"quorum-reads"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option has been removed. Reads are not allowed "
- "if quorum is not met.",
- },
- { .key = {"node-uuid"},
- .type = GF_OPTION_TYPE_STR,
- .description = "Local glusterd uuid string, used in starting "
- "self-heal-daemon so that it can crawl only on "
- "local index directories.",
- },
- { .key = {"post-op-delay-secs"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = INT_MAX,
- .default_value = "1",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Time interval induced artificially before "
- "post-operation phase of the transaction to "
- "enhance overlap of adjacent write operations.",
- },
- { .key = {AFR_SH_READDIR_SIZE_KEY},
- .type = GF_OPTION_TYPE_SIZET,
- .description = "readdirp size for performing entry self-heal",
- .min = 1024,
- .max = 131072,
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- .tags = {"replicate"},
- .default_value = "1KB",
- },
- { .key = {"ensure-durability"},
- .type = GF_OPTION_TYPE_BOOL,
- .op_version = {3},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Afr performs fsyncs for transactions if this "
- "option is on to make sure the changelogs/data is "
- "written to the disk",
- .default_value = "on",
- },
- { .key = {"afr-dirty-xattr"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = AFR_DIRTY_DEFAULT,
- },
- { .key = {"afr-pending-xattr"},
- .type = GF_OPTION_TYPE_STR,
- .description = "Comma separated list of xattrs that are used to "
- "capture information on pending heals."
- },
- { .key = {"metadata-splitbrain-forced-heal"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- },
- { .key = {"heal-timeout"},
- .type = GF_OPTION_TYPE_INT,
- .min = 5,
- .max = INT_MAX,
- .default_value = "600",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "time interval for checking the need to self-heal "
- "in self-heal-daemon"
- },
- { .key = {"consistent-metadata"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "If this option is enabled, readdirp will force "
- "lookups on those entries read whose read child is "
- "not the same as that of the parent. This will "
- "guarantee that all read operations on a file serve "
- "attributes from the same subvol as long as it holds "
- " a good copy of the file/dir.",
- },
- { .key = {"arbiter-count"},
- .type = GF_OPTION_TYPE_INT,
- .description = "subset of child_count. Has to be 0 or 1."
- },
- { .key = {"thin-arbiter"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {GD_OP_VERSION_4_1_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"replicate"},
- .description = "contains host:path of thin abriter brick",
- },
- { .key = {"shd-max-threads"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 64,
- .default_value = "1",
- .op_version = {GD_OP_VERSION_3_7_12},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Maximum number of parallel heals SHD can do per "
- "local brick. This can substantially lower heal times"
- ", but can also crush your bricks if you don't have "
- "the storage hardware to support this."
- },
- { .key = {"shd-wait-qlength"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 655536,
- .default_value = "1024",
- .op_version = {GD_OP_VERSION_3_7_12},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option can be used to control number of heals"
- " that can wait in SHD per subvolume",
- },
- { .key = {"locking-scheme"},
- .type = GF_OPTION_TYPE_STR,
- .value = { "full", "granular"},
- .default_value = "full",
- .op_version = {GD_OP_VERSION_3_7_12},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "If this option is set to granular, self-heal will "
- "stop being compatible with afr-v1, which helps afr "
- "be more granular while self-healing",
- },
- { .key = {"full-lock"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "yes",
- .op_version = {GD_OP_VERSION_3_13_2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- .tags = {"replicate"},
- .description = "If this option is disabled, then the IOs will take "
- "range locks same as versions till 3.13.1."
- },
- { .key = {"granular-entry-heal"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "If this option is enabled, self-heal will resort to "
- "granular way of recording changelogs and doing entry "
- "self-heal.",
- },
- { .key = {"favorite-child-policy"},
- .type = GF_OPTION_TYPE_STR,
- .value = {"none", "size", "ctime", "mtime", "majority"},
- .default_value = "none",
- .op_version = {GD_OP_VERSION_3_7_12},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option can be used to automatically resolve "
- "split-brains using various policies without user "
- "intervention. \"size\" picks the file with the "
- "biggest size as the source. \"ctime\" and \"mtime\" "
- "pick the file with the latest ctime and mtime "
- "respectively as the source. \"majority\" picks a file"
- " with identical mtime and size in more than half the "
- "number of bricks in the replica.",
- },
- { .key = {"consistent-io"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .description = "If this option is enabled, i/o will fail even if "
- "one of the bricks is down in the replicas",
- },
- { .key = {"use-compound-fops"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .op_version = {GD_OP_VERSION_3_8_4},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option exists only for backward compatibility "
- "and configuring it doesn't have any effect"
- },
- { .key = {NULL} },
+ {.key = {"read-subvolume"},
+ .type = GF_OPTION_TYPE_XLATOR,
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. Afr will prefer the one specified using "
+ "this option if it is not stale. Option value must be "
+ "one of the xlator names of the children. "
+ "Ex: <volname>-client-0 till "
+ "<volname>-client-<number-of-bricks - 1>"},
+ {.key = {"read-subvolume-index"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "-1",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one specified using "
+ "this option if it is not stale. allowed options"
+ " include -1 till replica-count - 1"},
+ {.key = {"read-hash-mode"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 3,
+ .default_value = "1",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description =
+ "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one computed using "
+ "the method specified using this option.\n"
+ "0 = first readable child of AFR, starting from 1st child.\n"
+ "1 = hash by GFID of file (all clients use "
+ "same subvolume).\n"
+ "2 = hash by GFID of file and client PID.\n"
+ "3 = brick having the least outstanding read requests."},
+ {
+ .key = {"choose-local"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "true",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Choose a local subvolume (i.e. Brick) to read from"
+ " if read-subvolume is not explicitly set.",
+ },
+ {.key = {"background-self-heal-count"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 256,
+ .default_value = "8",
+ .validate = GF_OPT_VALIDATE_MIN,
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This specifies the number of per client self-heal "
+ "jobs that can perform parallel heals in the "
+ "background."},
+ {.key = {"halo-shd-max-latency"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 99999,
+ .default_value = "99999",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "Maximum latency for shd halo replication in msec."},
+ {.key = {"halo-enabled"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "False",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "Enable Halo (geo) replication mode."},
+ {.key = {"halo-nfsd-max-latency"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 99999,
+ .default_value = "5",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "Maximum latency for nfsd halo replication in msec."},
+ {.key = {"halo-max-latency"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = AFR_HALO_MAX_LATENCY,
+ .default_value = "5",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "Maximum latency for halo replication in msec."},
+ {.key = {"halo-max-replicas"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 99999,
+ .default_value = "99999",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "The maximum number of halo replicas; replicas"
+ " beyond this value will be written asynchronously"
+ "via the SHD."},
+ {.key = {"halo-min-replicas"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 99999,
+ .default_value = "2",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "The minimmum number of halo replicas, before adding "
+ "out of region replicas."},
+ {.key = {"heal-wait-queue-length"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 10000, /*Around 100MB with sizeof(afr_local_t)= 10496 bytes*/
+ .default_value = "128",
+ .validate = GF_OPT_VALIDATE_MIN,
+ .op_version = {GD_OP_VERSION_3_7_10},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This specifies the number of heals that can be queued"
+ " for the parallel background self heal jobs."},
+ {.key = {"data-self-heal"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"1", "on", "yes", "true", "enable", "0", "off", "no", "false",
+ "disable", "open"},
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Using this option we can enable/disable data "
+ "self-heal on the file. \"open\" means data "
+ "self-heal action will only be triggered by file "
+ "open operations."},
+ {.key = {"data-self-heal-algorithm"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Select between \"full\", \"diff\". The "
+ "\"full\" algorithm copies the entire file from "
+ "source to sink. The \"diff\" algorithm copies to "
+ "sink only those blocks whose checksums don't match "
+ "with those of source. If no option is configured "
+ "the option is chosen dynamically as follows: "
+ "If the file does not exist on one of the sinks "
+ "or empty file exists or if the source file size is "
+ "about the same as page size the entire file will "
+ "be read and written i.e \"full\" algo, "
+ "otherwise \"diff\" algo is chosen.",
+ .value = {"diff", "full"}},
+ {.key = {"data-self-heal-window-size"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 1024,
+ .default_value = "1",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Maximum number blocks per file for which self-heal "
+ "process would be applied simultaneously."},
+ {.key = {"metadata-self-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ /*.validate_fn = validate_replica*/
+ .description = "Using this option we can enable/disable metadata "
+ "i.e. Permissions, ownerships, xattrs self-heal on "
+ "the file/directory."},
+ {.key = {"entry-self-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ /*.validate_fn = validate_replica*/
+ .description = "Using this option we can enable/disable entry "
+ "self-heal on the directory."},
+ {.key = {"data-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option exists only for backward compatibility "
+ "and configuring it doesn't have any effect"},
+ {.key = {"metadata-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option exists only for backward compatibility "
+ "and configuring it doesn't have any effect"},
+ {.key = {"entry-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option exists only for backward compatibility "
+ "and configuring it doesn't have any effect"},
+ {.key = {"optimistic-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Entry/Metadata fops will not perform "
+ "pre fop changelog operations in afr transaction "
+ "if this option is enabled."},
+ {.key = {"inodelk-trace"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "Enabling this option logs inode lock/unlocks"},
+ {.key = {"entrylk-trace"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "Enabling this option logs entry lock/unlocks"},
+ {.key = {"pre-op-compat"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Use separate pre-op xattrop() FOP rather than "
+ "overloading xdata of the OP"},
+ {.key = {"eager-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description =
+ "Enable/Disable eager lock for replica volume. "
+ "Lock phase of a transaction has two sub-phases. "
+ "First is an attempt to acquire locks in parallel by "
+ "broadcasting non-blocking lock requests. If lock "
+ "acquisition fails on any server, then the held locks "
+ "are unlocked and we revert to a blocking locks mode "
+ "sequentially on one server after another. If this "
+ "option is enabled the initial broadcasting lock "
+ "request attempts to acquire a full lock on the entire file. "
+ "If this fails, we revert back to the sequential "
+ "\"regional\" blocking locks as before. In the case "
+ "where such an \"eager\" lock is granted in the "
+ "non-blocking phase, it gives rise to an opportunity "
+ "for optimization. i.e, if the next write transaction "
+ "on the same FD arrives before the unlock phase of "
+ "the first transaction, it \"takes over\" the full "
+ "file lock. Similarly if yet another data transaction "
+ "arrives before the unlock phase of the \"optimized\" "
+ "transaction, that in turn \"takes over\" the lock as "
+ "well. The actual unlock now happens at the end of "
+ "the last \"optimized\" transaction."
+
+ },
+ {.key = {"self-heal-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ /*.validate_fn = validate_replica_heal_enable_disable*/
+ .description = "This option applies to only self-heal-daemon. "
+ "Index directory crawl and automatic healing of files "
+ "will not be performed if this option is turned off."},
+ {.key = {"iam-self-heal-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option differentiates if the replicate "
+ "translator is running as part of self-heal-daemon "
+ "or not."},
+ {.key = {"iam-nfs-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option differentiates if the replicate "
+ "translator is running as part of an NFS daemon "
+ "or not."},
+ {
+ .key = {"quorum-type"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"none", "auto", "fixed"},
+ .default_value = "none",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ /*.option = quorum-type*/
+ .description = "If value is \"fixed\" only allow writes if "
+ "quorum-count bricks are present. If value is "
+ "\"auto\" only allow writes if more than half of "
+ "bricks, or exactly half including the first, are "
+ "present.",
+ },
+ {
+ .key = {"quorum-count"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = INT_MAX,
+ .default_value = 0,
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ /*.option = quorum-count*/
+ /*.validate_fn = validate_quorum_count*/
+ .description = "If quorum-type is \"fixed\" only allow writes if "
+ "this many bricks are present. Other quorum types "
+ "will OVERWRITE this value.",
+ },
+ {
+ .key = {"quorum-reads"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option has been removed. Reads are not allowed "
+ "if quorum is not met.",
+ },
+ {
+ .key = {"node-uuid"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Local glusterd uuid string, used in starting "
+ "self-heal-daemon so that it can crawl only on "
+ "local index directories.",
+ },
+ {
+ .key = {"post-op-delay-secs"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = INT_MAX,
+ .default_value = "1",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Time interval induced artificially before "
+ "post-operation phase of the transaction to "
+ "enhance overlap of adjacent write operations.",
+ },
+ {
+ .key = {AFR_SH_READDIR_SIZE_KEY},
+ .type = GF_OPTION_TYPE_SIZET,
+ .description = "readdirp size for performing entry self-heal",
+ .min = 1024,
+ .max = 131072,
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ .default_value = "1KB",
+ },
+ {
+ .key = {"ensure-durability"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .op_version = {3},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Afr performs fsyncs for transactions if this "
+ "option is on to make sure the changelogs/data is "
+ "written to the disk",
+ .default_value = "on",
+ },
+ {
+ .key = {"afr-dirty-xattr"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = AFR_DIRTY_DEFAULT,
+ },
+ {.key = {"afr-pending-xattr"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Comma separated list of xattrs that are used to "
+ "capture information on pending heals."},
+ {
+ .key = {"metadata-splitbrain-forced-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ },
+ {.key = {"heal-timeout"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 5,
+ .max = INT_MAX,
+ .default_value = "600",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "time interval for checking the need to self-heal "
+ "in self-heal-daemon"},
+ {
+ .key = {"consistent-metadata"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "If this option is enabled, readdirp will force "
+ "lookups on those entries read whose read child is "
+ "not the same as that of the parent. This will "
+ "guarantee that all read operations on a file serve "
+ "attributes from the same subvol as long as it holds "
+ " a good copy of the file/dir.",
+ },
+ {.key = {"arbiter-count"},
+ .type = GF_OPTION_TYPE_INT,
+ .description = "subset of child_count. Has to be 0 or 1."},
+ {
+ .key = {"thin-arbiter"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {GD_OP_VERSION_4_1_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ .description = "contains host:path of thin abriter brick",
+ },
+ {.key = {"shd-max-threads"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 64,
+ .default_value = "1",
+ .op_version = {GD_OP_VERSION_3_7_12},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Maximum number of parallel heals SHD can do per "
+ "local brick. This can substantially lower heal times"
+ ", but can also crush your bricks if you don't have "
+ "the storage hardware to support this."},
+ {
+ .key = {"shd-wait-qlength"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 655536,
+ .default_value = "1024",
+ .op_version = {GD_OP_VERSION_3_7_12},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option can be used to control number of heals"
+ " that can wait in SHD per subvolume",
+ },
+ {
+ .key = {"locking-scheme"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"full", "granular"},
+ .default_value = "full",
+ .op_version = {GD_OP_VERSION_3_7_12},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "If this option is set to granular, self-heal will "
+ "stop being compatible with afr-v1, which helps afr "
+ "be more granular while self-healing",
+ },
+ {.key = {"full-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "yes",
+ .op_version = {GD_OP_VERSION_3_13_2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ .description = "If this option is disabled, then the IOs will take "
+ "range locks same as versions till 3.13.1."},
+ {
+ .key = {"granular-entry-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "If this option is enabled, self-heal will resort to "
+ "granular way of recording changelogs and doing entry "
+ "self-heal.",
+ },
+ {
+ .key = {"favorite-child-policy"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"none", "size", "ctime", "mtime", "majority"},
+ .default_value = "none",
+ .op_version = {GD_OP_VERSION_3_7_12},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option can be used to automatically resolve "
+ "split-brains using various policies without user "
+ "intervention. \"size\" picks the file with the "
+ "biggest size as the source. \"ctime\" and \"mtime\" "
+ "pick the file with the latest ctime and mtime "
+ "respectively as the source. \"majority\" picks a file"
+ " with identical mtime and size in more than half the "
+ "number of bricks in the replica.",
+ },
+ {
+ .key = {"consistent-io"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .description = "If this option is enabled, i/o will fail even if "
+ "one of the bricks is down in the replicas",
+ },
+ {.key = {"use-compound-fops"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_3_8_4},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option exists only for backward compatibility "
+ "and configuring it doesn't have any effect"},
+ {.key = {NULL}},
};