summaryrefslogtreecommitdiffstats
path: root/xlators/cluster/afr/src/afr-dir-read.c
diff options
context:
space:
mode:
authorPranith Kumar K <pkarampu@redhat.com>2012-11-05 21:42:22 +0530
committerVijay Bellur <vbellur@redhat.com>2012-12-03 00:11:02 -0800
commit2fd342a0c21d761d73bfee782717accbce819f24 (patch)
tree981acf4e2fe8f993b0bd520b7dba782ebd87b0de /xlators/cluster/afr/src/afr-dir-read.c
parent07c3801808db787e6c0cf0b2bf60a7ab62bc38b7 (diff)
cluster/afr: Provide option to disable readdir failover
In a replica pair unlike files, directories may not have their content in same order, so readdir for same (offset, size) may not give same entries on both the sobvolumes of replica pair. Switching over from one subvolume to another may not be a good idea sometimes. It may lead to duplicate entries or fewer entries or both. This patch provides a way to disable readdir-failover so that applications like rebalance can retry if they want to. Change-Id: I2b23eb224a2e84016a561362932613ac824c11a0 BUG: 859387 Signed-off-by: Pranith Kumar K <pkarampu@redhat.com> Reviewed-on: http://review.gluster.org/4159 Tested-by: Gluster Build System <jenkins@build.gluster.com> Reviewed-by: Vijay Bellur <vbellur@redhat.com>
Diffstat (limited to 'xlators/cluster/afr/src/afr-dir-read.c')
-rw-r--r--xlators/cluster/afr/src/afr-dir-read.c55
1 files changed, 30 insertions, 25 deletions
diff --git a/xlators/cluster/afr/src/afr-dir-read.c b/xlators/cluster/afr/src/afr-dir-read.c
index ce91ffba729..c201d45fd68 100644
--- a/xlators/cluster/afr/src/afr-dir-read.c
+++ b/xlators/cluster/afr/src/afr-dir-read.c
@@ -497,6 +497,9 @@ afr_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local = frame->local;
+ if ((priv->readdir_failover == _gf_false) && (op_ret < 0))
+ goto out;
+
read_child = (long) cookie;
last_index = &local->cont.readdir.last_index;
fresh_children = local->fresh_children;
@@ -593,15 +596,14 @@ int32_t
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_private_t * priv = NULL;
- xlator_t ** children = NULL;
- int call_child = 0;
- afr_local_t *local = NULL;
- uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- int ret = -1;
- int32_t op_errno = 0;
- uint64_t read_child = 0;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
+ uint64_t read_child = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -626,29 +628,33 @@ afr_do_readdir (call_frame_t *frame, xlator_t *this,
read_child = afr_inode_get_read_ctx (this, fd->inode,
local->fresh_children);
ret = afr_get_call_child (this, local->child_up, read_child,
- local->fresh_children,
- &call_child,
- &local->cont.readdir.last_index);
+ local->fresh_children,
+ &call_child,
+ &local->cont.readdir.last_index);
if (ret < 0) {
op_errno = -ret;
goto out;
}
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx) {
+ op_errno = EBADF;
+ goto out;
+ }
+
+ if ((offset == 0) || (fd_ctx->call_child == -1)) {
+ fd_ctx->call_child = call_child;
+ } else if ((priv->readdir_failover == _gf_false) &&
+ (call_child != fd_ctx->call_child)) {
+ op_errno = EBADF;
+ goto out;
+ }
+
local->fd = fd_ref (fd);
local->cont.readdir.size = size;
local->cont.readdir.dict = (dict)? dict_ref (dict) : NULL;
if (priv->strict_readdir) {
- ret = fd_ctx_get (fd, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_INFO,
- "could not get fd ctx for fd=%p", fd);
- op_errno = -ret;
- goto out;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
if (fd_ctx->last_tried != call_child) {
gf_log (this->name, GF_LOG_TRACE,
"first up child has changed from %d to %d, "
@@ -675,10 +681,9 @@ afr_do_readdir (call_frame_t *frame, xlator_t *this,
children[call_child]->fops->readdirp, fd,
size, offset, dict);
- ret = 0;
+ return 0;
out:
- if (ret < 0)
- AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
return 0;
}