diff options
Diffstat (limited to 'xlators/features/bit-rot/src/bitd')
| -rw-r--r-- | xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c | 82 | ||||
| -rw-r--r-- | xlators/features/bit-rot/src/bitd/bit-rot-scrub.c | 3171 | ||||
| -rw-r--r-- | xlators/features/bit-rot/src/bitd/bit-rot-ssm.c | 129 | ||||
| -rw-r--r-- | xlators/features/bit-rot/src/bitd/bit-rot.c | 3282 |
4 files changed, 3357 insertions, 3307 deletions
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c b/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c index 210beca7e2f..34e20f9df11 100644 --- a/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c +++ b/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c @@ -14,67 +14,65 @@ #include "bit-rot-scrub-status.h" void -br_inc_unsigned_file_count (br_scrub_stats_t *scrub_stat) +br_inc_unsigned_file_count(br_scrub_stats_t *scrub_stat) { - if (!scrub_stat) - return; + if (!scrub_stat) + return; - pthread_mutex_lock (&scrub_stat->lock); - { - scrub_stat->unsigned_files++; - } - pthread_mutex_unlock (&scrub_stat->lock); + pthread_mutex_lock(&scrub_stat->lock); + { + scrub_stat->unsigned_files++; + } + pthread_mutex_unlock(&scrub_stat->lock); } void -br_inc_scrubbed_file (br_scrub_stats_t *scrub_stat) +br_inc_scrubbed_file(br_scrub_stats_t *scrub_stat) { - if (!scrub_stat) - return; + if (!scrub_stat) + return; - pthread_mutex_lock (&scrub_stat->lock); - { - scrub_stat->scrubbed_files++; - } - pthread_mutex_unlock (&scrub_stat->lock); + pthread_mutex_lock(&scrub_stat->lock); + { + scrub_stat->scrubbed_files++; + } + pthread_mutex_unlock(&scrub_stat->lock); } void -br_update_scrub_start_time (br_scrub_stats_t *scrub_stat, struct timeval *tv) +br_update_scrub_start_time(br_scrub_stats_t *scrub_stat, struct timeval *tv) { - if (!scrub_stat) - return; + if (!scrub_stat) + return; - pthread_mutex_lock (&scrub_stat->lock); - { - scrub_stat->scrub_start_tv.tv_sec = tv->tv_sec; - } - pthread_mutex_unlock (&scrub_stat->lock); + pthread_mutex_lock(&scrub_stat->lock); + { + scrub_stat->scrub_start_tv.tv_sec = tv->tv_sec; + } + pthread_mutex_unlock(&scrub_stat->lock); } void -br_update_scrub_finish_time (br_scrub_stats_t *scrub_stat, char *timestr, - struct timeval *tv) +br_update_scrub_finish_time(br_scrub_stats_t *scrub_stat, char *timestr, + struct timeval *tv) { - int lst_size = 0; + int lst_size = 0; - if (!scrub_stat) - return; + if (!scrub_stat) + return; - lst_size = sizeof (scrub_stat->last_scrub_time); - if (strlen (timestr) >= lst_size) - return; + lst_size = sizeof(scrub_stat->last_scrub_time); + if (strlen(timestr) >= lst_size) + return; - pthread_mutex_lock (&scrub_stat->lock); - { - scrub_stat->scrub_end_tv.tv_sec = tv->tv_sec; + pthread_mutex_lock(&scrub_stat->lock); + { + scrub_stat->scrub_end_tv.tv_sec = tv->tv_sec; - scrub_stat->scrub_duration = - scrub_stat->scrub_end_tv.tv_sec - - scrub_stat->scrub_start_tv.tv_sec; + scrub_stat->scrub_duration = scrub_stat->scrub_end_tv.tv_sec - + scrub_stat->scrub_start_tv.tv_sec; - snprintf (scrub_stat->last_scrub_time, lst_size, "%s", - timestr); - } - pthread_mutex_unlock (&scrub_stat->lock); + snprintf(scrub_stat->last_scrub_time, lst_size, "%s", timestr); + } + pthread_mutex_unlock(&scrub_stat->lock); } diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-scrub.c b/xlators/features/bit-rot/src/bitd/bit-rot-scrub.c index e2665acab98..b856c168eb7 100644 --- a/xlators/features/bit-rot/src/bitd/bit-rot-scrub.c +++ b/xlators/features/bit-rot/src/bitd/bit-rot-scrub.c @@ -23,21 +23,21 @@ #include "events.h" struct br_scrubbers { - pthread_t scrubthread; + pthread_t scrubthread; - struct list_head list; + struct list_head list; }; struct br_fsscan_entry { - void *data; + void *data; - loc_t parent; + loc_t parent; - gf_dirent_t *entry; + gf_dirent_t *entry; - struct br_scanfs *fsscan; /* backpointer to subvolume scanner */ + struct br_scanfs *fsscan; /* backpointer to subvolume scanner */ - struct list_head list; + struct list_head list; }; /** @@ -46,34 +46,32 @@ struct br_fsscan_entry { * to the dictionary value. */ static int32_t -bitd_fetch_signature (xlator_t *this, br_child_t *child, - fd_t *fd, dict_t **xattr, br_isignature_out_t **sign) +bitd_fetch_signature(xlator_t *this, br_child_t *child, fd_t *fd, + dict_t **xattr, br_isignature_out_t **sign) { - int32_t ret = -1; - - ret = syncop_fgetxattr (child->xl, fd, xattr, - GLUSTERFS_GET_OBJECT_SIGNATURE, NULL, NULL); - if (ret < 0) { - br_log_object (this, "fgetxattr", fd->inode->gfid, -ret); - goto out; - } - - ret = dict_get_ptr - (*xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, (void **) sign); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED, - "failed to extract signature info [GFID: %s]", - uuid_utoa (fd->inode->gfid)); - goto unref_dict; - } - - return 0; - - unref_dict: - dict_unref (*xattr); - out: - return -1; - + int32_t ret = -1; + + ret = syncop_fgetxattr(child->xl, fd, xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, + NULL, NULL); + if (ret < 0) { + br_log_object(this, "fgetxattr", fd->inode->gfid, -ret); + goto out; + } + + ret = dict_get_ptr(*xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, (void **)sign); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED, + "failed to extract signature info [GFID: %s]", + uuid_utoa(fd->inode->gfid)); + goto unref_dict; + } + + return 0; + +unref_dict: + dict_unref(*xattr); +out: + return -1; } /** @@ -86,88 +84,85 @@ bitd_fetch_signature (xlator_t *this, br_child_t *child, */ int32_t -bitd_scrub_post_compute_check (xlator_t *this, - br_child_t *child, - fd_t *fd, unsigned long version, - br_isignature_out_t **signature, - br_scrub_stats_t *scrub_stat, - gf_boolean_t skip_stat) +bitd_scrub_post_compute_check(xlator_t *this, br_child_t *child, fd_t *fd, + unsigned long version, + br_isignature_out_t **signature, + br_scrub_stats_t *scrub_stat, + gf_boolean_t skip_stat) { - int32_t ret = 0; - size_t signlen = 0; - dict_t *xattr = NULL; - br_isignature_out_t *signptr = NULL; - - ret = bitd_fetch_signature (this, child, fd, &xattr, &signptr); - if (ret < 0) { - if (!skip_stat) - br_inc_unsigned_file_count (scrub_stat); - goto out; - } - - /** - * Either the object got dirtied during the time the signature was - * calculated OR the version we saved during pre-compute check does - * not match now, implying that the object got dirtied and signed in - * between scrubs pre & post compute checks (checksum window). - * - * The log entry looks pretty ugly, but helps in debugging.. - */ - if (signptr->stale || (signptr->version != version)) { - if (!skip_stat) - br_inc_unsigned_file_count (scrub_stat); - gf_msg_debug (this->name, 0, "<STAGE: POST> Object [GFID: %s] " - "either has a stale signature OR underwent " - "signing during checksumming {Stale: %d | " - "Version: %lu,%lu}", uuid_utoa (fd->inode->gfid), - (signptr->stale) ? 1 : 0, version, - signptr->version); - ret = -1; - goto unref_dict; - } - - signlen = signptr->signaturelen; - *signature = GF_MALLOC (sizeof (br_isignature_out_t) + signlen, - gf_common_mt_char); - - (void) memcpy (*signature, signptr, - sizeof (br_isignature_out_t) + signlen); - - unref_dict: - dict_unref (xattr); - out: - return ret; + int32_t ret = 0; + size_t signlen = 0; + dict_t *xattr = NULL; + br_isignature_out_t *signptr = NULL; + ret = bitd_fetch_signature(this, child, fd, &xattr, &signptr); + if (ret < 0) { + if (!skip_stat) + br_inc_unsigned_file_count(scrub_stat); + goto out; + } + + /** + * Either the object got dirtied during the time the signature was + * calculated OR the version we saved during pre-compute check does + * not match now, implying that the object got dirtied and signed in + * between scrubs pre & post compute checks (checksum window). + * + * The log entry looks pretty ugly, but helps in debugging.. + */ + if (signptr->stale || (signptr->version != version)) { + if (!skip_stat) + br_inc_unsigned_file_count(scrub_stat); + gf_msg_debug(this->name, 0, + "<STAGE: POST> Object [GFID: %s] " + "either has a stale signature OR underwent " + "signing during checksumming {Stale: %d | " + "Version: %lu,%lu}", + uuid_utoa(fd->inode->gfid), (signptr->stale) ? 1 : 0, + version, signptr->version); + ret = -1; + goto unref_dict; + } + + signlen = signptr->signaturelen; + *signature = GF_MALLOC(sizeof(br_isignature_out_t) + signlen, + gf_common_mt_char); + + (void)memcpy(*signature, signptr, sizeof(br_isignature_out_t) + signlen); + +unref_dict: + dict_unref(xattr); +out: + return ret; } static int32_t -bitd_signature_staleness (xlator_t *this, - br_child_t *child, fd_t *fd, - int *stale, unsigned long *version, - br_scrub_stats_t *scrub_stat, gf_boolean_t skip_stat) +bitd_signature_staleness(xlator_t *this, br_child_t *child, fd_t *fd, + int *stale, unsigned long *version, + br_scrub_stats_t *scrub_stat, gf_boolean_t skip_stat) { - int32_t ret = -1; - dict_t *xattr = NULL; - br_isignature_out_t *signptr = NULL; + int32_t ret = -1; + dict_t *xattr = NULL; + br_isignature_out_t *signptr = NULL; - ret = bitd_fetch_signature (this, child, fd, &xattr, &signptr); - if (ret < 0) { - if (!skip_stat) - br_inc_unsigned_file_count (scrub_stat); - goto out; - } + ret = bitd_fetch_signature(this, child, fd, &xattr, &signptr); + if (ret < 0) { + if (!skip_stat) + br_inc_unsigned_file_count(scrub_stat); + goto out; + } - /** - * save version for validation in post compute stage - * c.f. bitd_scrub_post_compute_check() - */ - *stale = signptr->stale ? 1 : 0; - *version = signptr->version; + /** + * save version for validation in post compute stage + * c.f. bitd_scrub_post_compute_check() + */ + *stale = signptr->stale ? 1 : 0; + *version = signptr->version; - dict_unref (xattr); + dict_unref(xattr); - out: - return ret; +out: + return ret; } /** @@ -179,101 +174,102 @@ bitd_signature_staleness (xlator_t *this, * - has stale signature */ int32_t -bitd_scrub_pre_compute_check (xlator_t *this, br_child_t *child, - fd_t *fd, unsigned long *version, - br_scrub_stats_t *scrub_stat, - gf_boolean_t skip_stat) +bitd_scrub_pre_compute_check(xlator_t *this, br_child_t *child, fd_t *fd, + unsigned long *version, + br_scrub_stats_t *scrub_stat, + gf_boolean_t skip_stat) { - int stale = 0; - int32_t ret = -1; - - if (bitd_is_bad_file (this, child, NULL, fd)) { - gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SKIP_OBJECT, - "Object [GFID: %s] is marked corrupted, skipping..", - uuid_utoa (fd->inode->gfid)); - goto out; - } - - ret = bitd_signature_staleness (this, child, fd, &stale, version, - scrub_stat, skip_stat); - if (!ret && stale) { - if (!skip_stat) - br_inc_unsigned_file_count (scrub_stat); - gf_msg_debug (this->name, 0, "<STAGE: PRE> Object [GFID: %s] " - "has stale signature", - uuid_utoa (fd->inode->gfid)); - ret = -1; - } + int stale = 0; + int32_t ret = -1; + + if (bitd_is_bad_file(this, child, NULL, fd)) { + gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SKIP_OBJECT, + "Object [GFID: %s] is marked corrupted, skipping..", + uuid_utoa(fd->inode->gfid)); + goto out; + } + + ret = bitd_signature_staleness(this, child, fd, &stale, version, scrub_stat, + skip_stat); + if (!ret && stale) { + if (!skip_stat) + br_inc_unsigned_file_count(scrub_stat); + gf_msg_debug(this->name, 0, + "<STAGE: PRE> Object [GFID: %s] " + "has stale signature", + uuid_utoa(fd->inode->gfid)); + ret = -1; + } - out: - return ret; +out: + return ret; } /* static int */ int -bitd_compare_ckum (xlator_t *this, - br_isignature_out_t *sign, - unsigned char *md, inode_t *linked_inode, - gf_dirent_t *entry, fd_t *fd, br_child_t *child, loc_t *loc) +bitd_compare_ckum(xlator_t *this, br_isignature_out_t *sign, unsigned char *md, + inode_t *linked_inode, gf_dirent_t *entry, fd_t *fd, + br_child_t *child, loc_t *loc) { - int ret = -1; - dict_t *xattr = NULL; - - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, sign, out); - GF_VALIDATE_OR_GOTO (this->name, fd, out); - GF_VALIDATE_OR_GOTO (this->name, child, out); - GF_VALIDATE_OR_GOTO (this->name, linked_inode, out); - GF_VALIDATE_OR_GOTO (this->name, md, out); - GF_VALIDATE_OR_GOTO (this->name, entry, out); - - if (strncmp - (sign->signature, (char *) md, strlen (sign->signature)) == 0) { - gf_msg_debug (this->name, 0, "%s [GFID: %s | Brick: %s] " - "matches calculated checksum", loc->path, - uuid_utoa (linked_inode->gfid), - child->brick_path); - return 0; - } - - gf_msg (this->name, GF_LOG_DEBUG, 0, BRB_MSG_CHECKSUM_MISMATCH, - "Object checksum mismatch: %s [GFID: %s | Brick: %s]", - loc->path, uuid_utoa (linked_inode->gfid), child->brick_path); - gf_msg (this->name, GF_LOG_ALERT, 0, BRB_MSG_CHECKSUM_MISMATCH, - "CORRUPTION DETECTED: Object %s {Brick: %s | GFID: %s}", - loc->path, child->brick_path, uuid_utoa (linked_inode->gfid)); - - /* Perform bad-file marking */ - xattr = dict_new (); - if (!xattr) { - ret = -1; - goto out; - } - - ret = dict_set_int32 (xattr, BITROT_OBJECT_BAD_KEY, _gf_true); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_MARK_BAD_FILE, - "Error setting bad-file marker for %s [GFID: %s | " - "Brick: %s]", loc->path, uuid_utoa (linked_inode->gfid), - child->brick_path); - goto dictfree; - } - - gf_msg (this->name, GF_LOG_ALERT, 0, BRB_MSG_MARK_CORRUPTED, "Marking" - " %s [GFID: %s | Brick: %s] as corrupted..", loc->path, - uuid_utoa (linked_inode->gfid), child->brick_path); - gf_event (EVENT_BITROT_BAD_FILE, "gfid=%s;path=%s;brick=%s", - uuid_utoa (linked_inode->gfid), loc->path, child->brick_path); - ret = syncop_fsetxattr (child->xl, fd, xattr, 0, NULL, NULL); - if (ret) - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_MARK_BAD_FILE, - "Error marking object %s [GFID: %s] as corrupted", - loc->path, uuid_utoa (linked_inode->gfid)); - - dictfree: - dict_unref (xattr); - out: - return ret; + int ret = -1; + dict_t *xattr = NULL; + + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, sign, out); + GF_VALIDATE_OR_GOTO(this->name, fd, out); + GF_VALIDATE_OR_GOTO(this->name, child, out); + GF_VALIDATE_OR_GOTO(this->name, linked_inode, out); + GF_VALIDATE_OR_GOTO(this->name, md, out); + GF_VALIDATE_OR_GOTO(this->name, entry, out); + + if (strncmp(sign->signature, (char *)md, strlen(sign->signature)) == 0) { + gf_msg_debug(this->name, 0, + "%s [GFID: %s | Brick: %s] " + "matches calculated checksum", + loc->path, uuid_utoa(linked_inode->gfid), + child->brick_path); + return 0; + } + + gf_msg(this->name, GF_LOG_DEBUG, 0, BRB_MSG_CHECKSUM_MISMATCH, + "Object checksum mismatch: %s [GFID: %s | Brick: %s]", loc->path, + uuid_utoa(linked_inode->gfid), child->brick_path); + gf_msg(this->name, GF_LOG_ALERT, 0, BRB_MSG_CHECKSUM_MISMATCH, + "CORRUPTION DETECTED: Object %s {Brick: %s | GFID: %s}", loc->path, + child->brick_path, uuid_utoa(linked_inode->gfid)); + + /* Perform bad-file marking */ + xattr = dict_new(); + if (!xattr) { + ret = -1; + goto out; + } + + ret = dict_set_int32(xattr, BITROT_OBJECT_BAD_KEY, _gf_true); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_MARK_BAD_FILE, + "Error setting bad-file marker for %s [GFID: %s | " + "Brick: %s]", + loc->path, uuid_utoa(linked_inode->gfid), child->brick_path); + goto dictfree; + } + + gf_msg(this->name, GF_LOG_ALERT, 0, BRB_MSG_MARK_CORRUPTED, + "Marking" + " %s [GFID: %s | Brick: %s] as corrupted..", + loc->path, uuid_utoa(linked_inode->gfid), child->brick_path); + gf_event(EVENT_BITROT_BAD_FILE, "gfid=%s;path=%s;brick=%s", + uuid_utoa(linked_inode->gfid), loc->path, child->brick_path); + ret = syncop_fsetxattr(child->xl, fd, xattr, 0, NULL, NULL); + if (ret) + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_MARK_BAD_FILE, + "Error marking object %s [GFID: %s] as corrupted", loc->path, + uuid_utoa(linked_inode->gfid)); + +dictfree: + dict_unref(xattr); +out: + return ret; } /** @@ -284,539 +280,552 @@ bitd_compare_ckum (xlator_t *this, * signs with SHA256). */ int -br_scrubber_scrub_begin (xlator_t *this, struct br_fsscan_entry *fsentry) +br_scrubber_scrub_begin(xlator_t *this, struct br_fsscan_entry *fsentry) { - int32_t ret = -1; - fd_t *fd = NULL; - loc_t loc = {0, }; - struct iatt iatt = {0, }; - struct iatt parent_buf = {0, }; - pid_t pid = 0; - br_child_t *child = NULL; - unsigned char *md = NULL; - inode_t *linked_inode = NULL; - br_isignature_out_t *sign = NULL; - unsigned long signedversion = 0; - gf_dirent_t *entry = NULL; - br_private_t *priv = NULL; - loc_t *parent = NULL; - gf_boolean_t skip_stat = _gf_false; - uuid_t shard_root_gfid = {0,}; - - - GF_VALIDATE_OR_GOTO ("bit-rot", fsentry, out); - - entry = fsentry->entry; - parent = &fsentry->parent; - child = fsentry->data; - - priv = this->private; - - GF_VALIDATE_OR_GOTO ("bit-rot", entry, out); - GF_VALIDATE_OR_GOTO ("bit-rot", parent, out); - GF_VALIDATE_OR_GOTO ("bit-rot", child, out); - GF_VALIDATE_OR_GOTO ("bit-rot", priv, out); - - pid = GF_CLIENT_PID_SCRUB; - - ret = br_prepare_loc (this, child, parent, entry, &loc); - if (!ret) - goto out; - - syncopctx_setfspid (&pid); - - ret = syncop_lookup (child->xl, &loc, &iatt, &parent_buf, NULL, NULL); - if (ret) { - br_log_object_path (this, "lookup", loc.path, -ret); - goto out; - } - - linked_inode = inode_link (loc.inode, parent->inode, loc.name, &iatt); - if (linked_inode) - inode_lookup (linked_inode); - - gf_msg_debug (this->name, 0, "Scrubbing object %s [GFID: %s]", - entry->d_name, uuid_utoa (linked_inode->gfid)); - - if (iatt.ia_type != IA_IFREG) { - gf_msg_debug (this->name, 0, "%s is not a regular file", - entry->d_name); - ret = 0; - goto unref_inode; - } - - if (IS_DHT_LINKFILE_MODE ((&iatt))) { - gf_msg_debug (this->name, 0, "%s is a dht sticky bit file", - entry->d_name); - ret = 0; - goto unref_inode; - } - - /* skip updating scrub statistics for shard entries */ - gf_uuid_parse (SHARD_ROOT_GFID, shard_root_gfid); - if (gf_uuid_compare (loc.pargfid, shard_root_gfid) == 0) - skip_stat = _gf_true; - - /** - * open() an fd for subsequent operations - */ - fd = fd_create (linked_inode, 0); - if (!fd) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED, - "failed to create fd for inode %s", - uuid_utoa (linked_inode->gfid)); - goto unref_inode; - } - - ret = syncop_open (child->xl, &loc, O_RDWR, fd, NULL, NULL); - if (ret) { - br_log_object (this, "open", linked_inode->gfid, -ret); - ret = -1; - goto unrefd; - } - - fd_bind (fd); - - /** - * perform pre compute checks before initiating checksum - * computation - * - presence of bad object - * - signature staleness - */ - ret = bitd_scrub_pre_compute_check (this, child, fd, &signedversion, - &priv->scrub_stat, skip_stat); - if (ret) - goto unrefd; /* skip this object */ - - /* if all's good, proceed to calculate the hash */ - md = GF_MALLOC (SHA256_DIGEST_LENGTH, gf_common_mt_char); - if (!md) - goto unrefd; - - ret = br_calculate_obj_checksum (md, child, fd, &iatt); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_CALC_ERROR, - "error calculating hash for object [GFID: %s]", - uuid_utoa (fd->inode->gfid)); - ret = -1; - goto free_md; - } - - /** - * perform post compute checks as an object's signature may have - * become stale while scrubber calculated checksum. - */ - ret = bitd_scrub_post_compute_check (this, child, fd, signedversion, - &sign, &priv->scrub_stat, - skip_stat); - if (ret) - goto free_md; - - ret = bitd_compare_ckum (this, sign, md, - linked_inode, entry, fd, child, &loc); - - if (!skip_stat) - br_inc_scrubbed_file (&priv->scrub_stat); - - GF_FREE (sign); /* allocated on post-compute */ - - /** fd_unref() takes care of closing fd.. like syncop_close() */ + int32_t ret = -1; + fd_t *fd = NULL; + loc_t loc = { + 0, + }; + struct iatt iatt = { + 0, + }; + struct iatt parent_buf = { + 0, + }; + pid_t pid = 0; + br_child_t *child = NULL; + unsigned char *md = NULL; + inode_t *linked_inode = NULL; + br_isignature_out_t *sign = NULL; + unsigned long signedversion = 0; + gf_dirent_t *entry = NULL; + br_private_t *priv = NULL; + loc_t *parent = NULL; + gf_boolean_t skip_stat = _gf_false; + uuid_t shard_root_gfid = { + 0, + }; + + GF_VALIDATE_OR_GOTO("bit-rot", fsentry, out); + + entry = fsentry->entry; + parent = &fsentry->parent; + child = fsentry->data; + + priv = this->private; + + GF_VALIDATE_OR_GOTO("bit-rot", entry, out); + GF_VALIDATE_OR_GOTO("bit-rot", parent, out); + GF_VALIDATE_OR_GOTO("bit-rot", child, out); + GF_VALIDATE_OR_GOTO("bit-rot", priv, out); + + pid = GF_CLIENT_PID_SCRUB; + + ret = br_prepare_loc(this, child, parent, entry, &loc); + if (!ret) + goto out; + + syncopctx_setfspid(&pid); + + ret = syncop_lookup(child->xl, &loc, &iatt, &parent_buf, NULL, NULL); + if (ret) { + br_log_object_path(this, "lookup", loc.path, -ret); + goto out; + } + + linked_inode = inode_link(loc.inode, parent->inode, loc.name, &iatt); + if (linked_inode) + inode_lookup(linked_inode); + + gf_msg_debug(this->name, 0, "Scrubbing object %s [GFID: %s]", entry->d_name, + uuid_utoa(linked_inode->gfid)); + + if (iatt.ia_type != IA_IFREG) { + gf_msg_debug(this->name, 0, "%s is not a regular file", entry->d_name); + ret = 0; + goto unref_inode; + } - free_md: - GF_FREE (md); - unrefd: - fd_unref (fd); - unref_inode: - inode_unref (linked_inode); - out: - loc_wipe (&loc); - return ret; + if (IS_DHT_LINKFILE_MODE((&iatt))) { + gf_msg_debug(this->name, 0, "%s is a dht sticky bit file", + entry->d_name); + ret = 0; + goto unref_inode; + } + + /* skip updating scrub statistics for shard entries */ + gf_uuid_parse(SHARD_ROOT_GFID, shard_root_gfid); + if (gf_uuid_compare(loc.pargfid, shard_root_gfid) == 0) + skip_stat = _gf_true; + + /** + * open() an fd for subsequent operations + */ + fd = fd_create(linked_inode, 0); + if (!fd) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED, + "failed to create fd for inode %s", + uuid_utoa(linked_inode->gfid)); + goto unref_inode; + } + + ret = syncop_open(child->xl, &loc, O_RDWR, fd, NULL, NULL); + if (ret) { + br_log_object(this, "open", linked_inode->gfid, -ret); + ret = -1; + goto unrefd; + } + + fd_bind(fd); + + /** + * perform pre compute checks before initiating checksum + * computation + * - presence of bad object + * - signature staleness + */ + ret = bitd_scrub_pre_compute_check(this, child, fd, &signedversion, + &priv->scrub_stat, skip_stat); + if (ret) + goto unrefd; /* skip this object */ + + /* if all's good, proceed to calculate the hash */ + md = GF_MALLOC(SHA256_DIGEST_LENGTH, gf_common_mt_char); + if (!md) + goto unrefd; + + ret = br_calculate_obj_checksum(md, child, fd, &iatt); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_CALC_ERROR, + "error calculating hash for object [GFID: %s]", + uuid_utoa(fd->inode->gfid)); + ret = -1; + goto free_md; + } + + /** + * perform post compute checks as an object's signature may have + * become stale while scrubber calculated checksum. + */ + ret = bitd_scrub_post_compute_check(this, child, fd, signedversion, &sign, + &priv->scrub_stat, skip_stat); + if (ret) + goto free_md; + + ret = bitd_compare_ckum(this, sign, md, linked_inode, entry, fd, child, + &loc); + + if (!skip_stat) + br_inc_scrubbed_file(&priv->scrub_stat); + + GF_FREE(sign); /* allocated on post-compute */ + + /** fd_unref() takes care of closing fd.. like syncop_close() */ + +free_md: + GF_FREE(md); +unrefd: + fd_unref(fd); +unref_inode: + inode_unref(linked_inode); +out: + loc_wipe(&loc); + return ret; } static void -_br_lock_cleaner (void *arg) +_br_lock_cleaner(void *arg) { - pthread_mutex_t *mutex = arg; + pthread_mutex_t *mutex = arg; - pthread_mutex_unlock (mutex); + pthread_mutex_unlock(mutex); } static void -wait_for_scrubbing (xlator_t *this, struct br_scanfs *fsscan) +wait_for_scrubbing(xlator_t *this, struct br_scanfs *fsscan) { - br_private_t *priv = NULL; - struct br_scrubber *fsscrub = NULL; + br_private_t *priv = NULL; + struct br_scrubber *fsscrub = NULL; - priv = this->private; - fsscrub = &priv->fsscrub; + priv = this->private; + fsscrub = &priv->fsscrub; - pthread_cleanup_push (_br_lock_cleaner, &fsscan->waitlock); - pthread_mutex_lock (&fsscan->waitlock); + pthread_cleanup_push(_br_lock_cleaner, &fsscan->waitlock); + pthread_mutex_lock(&fsscan->waitlock); + { + pthread_cleanup_push(_br_lock_cleaner, &fsscrub->mutex); + pthread_mutex_lock(&fsscrub->mutex); { - pthread_cleanup_push (_br_lock_cleaner, &fsscrub->mutex); - pthread_mutex_lock (&fsscrub->mutex); - { - list_replace_init (&fsscan->queued, &fsscan->ready); - - /* wake up scrubbers */ - pthread_cond_broadcast (&fsscrub->cond); - } - pthread_mutex_unlock (&fsscrub->mutex); - pthread_cleanup_pop (0); - - while (fsscan->entries != 0) - pthread_cond_wait - (&fsscan->waitcond, &fsscan->waitlock); + list_replace_init(&fsscan->queued, &fsscan->ready); + + /* wake up scrubbers */ + pthread_cond_broadcast(&fsscrub->cond); } - pthread_mutex_unlock (&fsscan->waitlock); - pthread_cleanup_pop (0); + pthread_mutex_unlock(&fsscrub->mutex); + pthread_cleanup_pop(0); + + while (fsscan->entries != 0) + pthread_cond_wait(&fsscan->waitcond, &fsscan->waitlock); + } + pthread_mutex_unlock(&fsscan->waitlock); + pthread_cleanup_pop(0); } static void -_br_fsscan_inc_entry_count (struct br_scanfs *fsscan) +_br_fsscan_inc_entry_count(struct br_scanfs *fsscan) { - fsscan->entries++; + fsscan->entries++; } static void -_br_fsscan_dec_entry_count (struct br_scanfs *fsscan) +_br_fsscan_dec_entry_count(struct br_scanfs *fsscan) { - if (--fsscan->entries == 0) { - pthread_mutex_lock (&fsscan->waitlock); - { - pthread_cond_signal (&fsscan->waitcond); - } - pthread_mutex_unlock (&fsscan->waitlock); + if (--fsscan->entries == 0) { + pthread_mutex_lock(&fsscan->waitlock); + { + pthread_cond_signal(&fsscan->waitcond); } + pthread_mutex_unlock(&fsscan->waitlock); + } } static void -_br_fsscan_collect_entry (struct br_scanfs *fsscan, - struct br_fsscan_entry *fsentry) +_br_fsscan_collect_entry(struct br_scanfs *fsscan, + struct br_fsscan_entry *fsentry) { - list_add_tail (&fsentry->list, &fsscan->queued); - _br_fsscan_inc_entry_count (fsscan); + list_add_tail(&fsentry->list, &fsscan->queued); + _br_fsscan_inc_entry_count(fsscan); } -#define NR_ENTRIES (1<<7) /* ..bulk scrubbing */ +#define NR_ENTRIES (1 << 7) /* ..bulk scrubbing */ int -br_fsscanner_handle_entry (xlator_t *subvol, - gf_dirent_t *entry, loc_t *parent, void *data) +br_fsscanner_handle_entry(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent, + void *data) { - int32_t ret = -1; - int scrub = 0; - br_child_t *child = NULL; - xlator_t *this = NULL; - struct br_scanfs *fsscan = NULL; - struct br_fsscan_entry *fsentry = NULL; + int32_t ret = -1; + int scrub = 0; + br_child_t *child = NULL; + xlator_t *this = NULL; + struct br_scanfs *fsscan = NULL; + struct br_fsscan_entry *fsentry = NULL; - GF_VALIDATE_OR_GOTO ("bit-rot", subvol, error_return); - GF_VALIDATE_OR_GOTO ("bit-rot", data, error_return); + GF_VALIDATE_OR_GOTO("bit-rot", subvol, error_return); + GF_VALIDATE_OR_GOTO("bit-rot", data, error_return); - child = data; - this = child->this; - fsscan = &child->fsscan; + child = data; + this = child->this; + fsscan = &child->fsscan; - _mask_cancellation (); + _mask_cancellation(); - fsentry = GF_CALLOC (1, sizeof (*fsentry), gf_br_mt_br_fsscan_entry_t); - if (!fsentry) - goto error_return; + fsentry = GF_CALLOC(1, sizeof(*fsentry), gf_br_mt_br_fsscan_entry_t); + if (!fsentry) + goto error_return; - { - fsentry->data = data; - fsentry->fsscan = &child->fsscan; + { + fsentry->data = data; + fsentry->fsscan = &child->fsscan; - /* copy parent loc */ - ret = loc_copy (&fsentry->parent, parent); - if (ret) - goto dealloc; + /* copy parent loc */ + ret = loc_copy(&fsentry->parent, parent); + if (ret) + goto dealloc; - /* copy child entry */ - fsentry->entry = entry_copy (entry); - if (!fsentry->entry) - goto locwipe; + /* copy child entry */ + fsentry->entry = entry_copy(entry); + if (!fsentry->entry) + goto locwipe; - INIT_LIST_HEAD (&fsentry->list); - } + INIT_LIST_HEAD(&fsentry->list); + } - LOCK (&fsscan->entrylock); - { - _br_fsscan_collect_entry (fsscan, fsentry); - - /** - * need not be a equality check as entries may be pushed - * back onto the scanned queue when thread(s) are cleaned. - */ - if (fsscan->entries >= NR_ENTRIES) - scrub = 1; - } - UNLOCK (&fsscan->entrylock); + LOCK(&fsscan->entrylock); + { + _br_fsscan_collect_entry(fsscan, fsentry); - _unmask_cancellation (); + /** + * need not be a equality check as entries may be pushed + * back onto the scanned queue when thread(s) are cleaned. + */ + if (fsscan->entries >= NR_ENTRIES) + scrub = 1; + } + UNLOCK(&fsscan->entrylock); - if (scrub) - wait_for_scrubbing (this, fsscan); + _unmask_cancellation(); - return 0; + if (scrub) + wait_for_scrubbing(this, fsscan); - locwipe: - loc_wipe (&fsentry->parent); - dealloc: - GF_FREE (fsentry); - error_return: - return -1; + return 0; + +locwipe: + loc_wipe(&fsentry->parent); +dealloc: + GF_FREE(fsentry); +error_return: + return -1; } int32_t -br_fsscan_deactivate (xlator_t *this) +br_fsscan_deactivate(xlator_t *this) { - int ret = 0; - br_private_t *priv = NULL; - br_scrub_state_t nstate = 0; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - scrub_monitor = &priv->scrub_monitor; - - ret = gf_tw_del_timer (priv->timer_wheel, scrub_monitor->timer); - if (ret == 0) { - nstate = BR_SCRUB_STATE_STALLED; - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, - "Volume is under active scrubbing. Pausing scrub.."); - } else { - nstate = BR_SCRUB_STATE_PAUSED; - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, - "Scrubber paused"); - } - - _br_monitor_set_scrub_state (scrub_monitor, nstate); - - return 0; + int ret = 0; + br_private_t *priv = NULL; + br_scrub_state_t nstate = 0; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + scrub_monitor = &priv->scrub_monitor; + + ret = gf_tw_del_timer(priv->timer_wheel, scrub_monitor->timer); + if (ret == 0) { + nstate = BR_SCRUB_STATE_STALLED; + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Volume is under active scrubbing. Pausing scrub.."); + } else { + nstate = BR_SCRUB_STATE_PAUSED; + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Scrubber paused"); + } + + _br_monitor_set_scrub_state(scrub_monitor, nstate); + + return 0; } static void -br_scrubber_log_time (xlator_t *this, const char *sfx) +br_scrubber_log_time(xlator_t *this, const char *sfx) { - char timestr[1024] = {0,}; - struct timeval tv = {0,}; - br_private_t *priv = NULL; - - priv = this->private; - - gettimeofday (&tv, NULL); - gf_time_fmt (timestr, sizeof (timestr), tv.tv_sec, gf_timefmt_FT); - - if (strcasecmp (sfx, "started") == 0) { - br_update_scrub_start_time (&priv->scrub_stat, &tv); - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_START, - "Scrubbing %s at %s", sfx, timestr); - } else { - br_update_scrub_finish_time (&priv->scrub_stat, timestr, &tv); - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_FINISH, - "Scrubbing %s at %s", sfx, timestr); - } + char timestr[1024] = { + 0, + }; + struct timeval tv = { + 0, + }; + br_private_t *priv = NULL; + + priv = this->private; + + gettimeofday(&tv, NULL); + gf_time_fmt(timestr, sizeof(timestr), tv.tv_sec, gf_timefmt_FT); + + if (strcasecmp(sfx, "started") == 0) { + br_update_scrub_start_time(&priv->scrub_stat, &tv); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_START, + "Scrubbing %s at %s", sfx, timestr); + } else { + br_update_scrub_finish_time(&priv->scrub_stat, timestr, &tv); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_FINISH, + "Scrubbing %s at %s", sfx, timestr); + } } static void -br_fsscanner_log_time (xlator_t *this, br_child_t *child, const char *sfx) +br_fsscanner_log_time(xlator_t *this, br_child_t *child, const char *sfx) { - char timestr[1024] = {0,}; - struct timeval tv = {0,}; - - gettimeofday (&tv, NULL); - gf_time_fmt (timestr, sizeof (timestr), tv.tv_sec, gf_timefmt_FT); - - if (strcasecmp (sfx, "started") == 0) { - gf_msg_debug (this->name, 0, "Scrubbing \"%s\" %s at %s", - child->brick_path, sfx, timestr); - } else { - gf_msg_debug (this->name, 0, "Scrubbing \"%s\" %s at %s", - child->brick_path, sfx, timestr); - } + char timestr[1024] = { + 0, + }; + struct timeval tv = { + 0, + }; + + gettimeofday(&tv, NULL); + gf_time_fmt(timestr, sizeof(timestr), tv.tv_sec, gf_timefmt_FT); + + if (strcasecmp(sfx, "started") == 0) { + gf_msg_debug(this->name, 0, "Scrubbing \"%s\" %s at %s", + child->brick_path, sfx, timestr); + } else { + gf_msg_debug(this->name, 0, "Scrubbing \"%s\" %s at %s", + child->brick_path, sfx, timestr); + } } void -br_child_set_scrub_state (br_child_t *child, gf_boolean_t state) +br_child_set_scrub_state(br_child_t *child, gf_boolean_t state) { - child->active_scrubbing = state; + child->active_scrubbing = state; } static void -br_fsscanner_wait_until_kicked (xlator_t *this, br_child_t *child) +br_fsscanner_wait_until_kicked(xlator_t *this, br_child_t *child) { - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - scrub_monitor = &priv->scrub_monitor; - - pthread_cleanup_push (_br_lock_cleaner, &scrub_monitor->wakelock); - pthread_mutex_lock (&scrub_monitor->wakelock); + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + scrub_monitor = &priv->scrub_monitor; + + pthread_cleanup_push(_br_lock_cleaner, &scrub_monitor->wakelock); + pthread_mutex_lock(&scrub_monitor->wakelock); + { + while (!scrub_monitor->kick) + pthread_cond_wait(&scrub_monitor->wakecond, + &scrub_monitor->wakelock); + + /* Child lock is to synchronize with disconnect events */ + pthread_cleanup_push(_br_lock_cleaner, &child->lock); + pthread_mutex_lock(&child->lock); { - while (!scrub_monitor->kick) - pthread_cond_wait (&scrub_monitor->wakecond, - &scrub_monitor->wakelock); - - /* Child lock is to synchronize with disconnect events */ - pthread_cleanup_push (_br_lock_cleaner, &child->lock); - pthread_mutex_lock (&child->lock); - { - scrub_monitor->active_child_count++; - br_child_set_scrub_state (child, _gf_true); - } - pthread_mutex_unlock (&child->lock); - pthread_cleanup_pop (0); - } - pthread_mutex_unlock (&scrub_monitor->wakelock); - pthread_cleanup_pop (0); + scrub_monitor->active_child_count++; + br_child_set_scrub_state(child, _gf_true); + } + pthread_mutex_unlock(&child->lock); + pthread_cleanup_pop(0); + } + pthread_mutex_unlock(&scrub_monitor->wakelock); + pthread_cleanup_pop(0); } static void -br_scrubber_entry_control (xlator_t *this) +br_scrubber_entry_control(xlator_t *this) { - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - scrub_monitor = &priv->scrub_monitor; - - LOCK (&scrub_monitor->lock); - { - /* Move the state to BR_SCRUB_STATE_ACTIVE */ - if (scrub_monitor->state == BR_SCRUB_STATE_PENDING) - scrub_monitor->state = BR_SCRUB_STATE_ACTIVE; - br_scrubber_log_time (this, "started"); - priv->scrub_stat.scrub_running = 1; - } - UNLOCK (&scrub_monitor->lock); + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + scrub_monitor = &priv->scrub_monitor; + + LOCK(&scrub_monitor->lock); + { + /* Move the state to BR_SCRUB_STATE_ACTIVE */ + if (scrub_monitor->state == BR_SCRUB_STATE_PENDING) + scrub_monitor->state = BR_SCRUB_STATE_ACTIVE; + br_scrubber_log_time(this, "started"); + priv->scrub_stat.scrub_running = 1; + } + UNLOCK(&scrub_monitor->lock); } static void -br_scrubber_exit_control (xlator_t *this) +br_scrubber_exit_control(xlator_t *this) { - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; - priv = this->private; - scrub_monitor = &priv->scrub_monitor; + priv = this->private; + scrub_monitor = &priv->scrub_monitor; - LOCK (&scrub_monitor->lock); - { - br_scrubber_log_time (this, "finished"); - priv->scrub_stat.scrub_running = 0; - - if (scrub_monitor->state == BR_SCRUB_STATE_ACTIVE) { - (void) br_fsscan_activate (this); - } else { - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, - "Volume waiting to get rescheduled.."); - } + LOCK(&scrub_monitor->lock); + { + br_scrubber_log_time(this, "finished"); + priv->scrub_stat.scrub_running = 0; + + if (scrub_monitor->state == BR_SCRUB_STATE_ACTIVE) { + (void)br_fsscan_activate(this); + } else { + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Volume waiting to get rescheduled.."); } - UNLOCK (&scrub_monitor->lock); + } + UNLOCK(&scrub_monitor->lock); } static void -br_fsscanner_entry_control (xlator_t *this, br_child_t *child) +br_fsscanner_entry_control(xlator_t *this, br_child_t *child) { - br_fsscanner_log_time (this, child, "started"); + br_fsscanner_log_time(this, child, "started"); } static void -br_fsscanner_exit_control (xlator_t *this, br_child_t *child) +br_fsscanner_exit_control(xlator_t *this, br_child_t *child) { - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - scrub_monitor = &priv->scrub_monitor; - - if (!_br_is_child_connected (child)) { - gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SCRUB_INFO, - "Brick [%s] disconnected while scrubbing. Scrubbing " - "might be incomplete", child->brick_path); - } - - br_fsscanner_log_time (this, child, "finished"); - - pthread_cleanup_push (_br_lock_cleaner, &scrub_monitor->wakelock); - pthread_mutex_lock (&scrub_monitor->wakelock); + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + scrub_monitor = &priv->scrub_monitor; + + if (!_br_is_child_connected(child)) { + gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SCRUB_INFO, + "Brick [%s] disconnected while scrubbing. Scrubbing " + "might be incomplete", + child->brick_path); + } + + br_fsscanner_log_time(this, child, "finished"); + + pthread_cleanup_push(_br_lock_cleaner, &scrub_monitor->wakelock); + pthread_mutex_lock(&scrub_monitor->wakelock); + { + scrub_monitor->active_child_count--; + pthread_cleanup_push(_br_lock_cleaner, &child->lock); + pthread_mutex_lock(&child->lock); { - scrub_monitor->active_child_count--; - pthread_cleanup_push (_br_lock_cleaner, &child->lock); - pthread_mutex_lock (&child->lock); - { - br_child_set_scrub_state (child, _gf_false); - } - pthread_mutex_unlock (&child->lock); - pthread_cleanup_pop (0); - - if (scrub_monitor->active_child_count == 0) { - /* The last child has finished scrubbing. - * Set the kick to false and wake up other - * children who are waiting for the last - * child to complete scrubbing. - */ - scrub_monitor->kick = _gf_false; - pthread_cond_broadcast (&scrub_monitor->wakecond); - - /* Signal monitor thread waiting for the all - * the children to finish scrubbing. - */ - pthread_cleanup_push (_br_lock_cleaner, - &scrub_monitor->donelock); - pthread_mutex_lock (&scrub_monitor->donelock); - { - scrub_monitor->done = _gf_true; - pthread_cond_signal (&scrub_monitor->donecond); - } - pthread_mutex_unlock (&scrub_monitor->donelock); - pthread_cleanup_pop (0); - } else { - while (scrub_monitor->active_child_count) - pthread_cond_wait (&scrub_monitor->wakecond, - &scrub_monitor->wakelock); - } + br_child_set_scrub_state(child, _gf_false); + } + pthread_mutex_unlock(&child->lock); + pthread_cleanup_pop(0); + + if (scrub_monitor->active_child_count == 0) { + /* The last child has finished scrubbing. + * Set the kick to false and wake up other + * children who are waiting for the last + * child to complete scrubbing. + */ + scrub_monitor->kick = _gf_false; + pthread_cond_broadcast(&scrub_monitor->wakecond); + + /* Signal monitor thread waiting for the all + * the children to finish scrubbing. + */ + pthread_cleanup_push(_br_lock_cleaner, &scrub_monitor->donelock); + pthread_mutex_lock(&scrub_monitor->donelock); + { + scrub_monitor->done = _gf_true; + pthread_cond_signal(&scrub_monitor->donecond); + } + pthread_mutex_unlock(&scrub_monitor->donelock); + pthread_cleanup_pop(0); + } else { + while (scrub_monitor->active_child_count) + pthread_cond_wait(&scrub_monitor->wakecond, + &scrub_monitor->wakelock); } - pthread_mutex_unlock (&scrub_monitor->wakelock); - pthread_cleanup_pop (0); + } + pthread_mutex_unlock(&scrub_monitor->wakelock); + pthread_cleanup_pop(0); } void * -br_fsscanner (void *arg) +br_fsscanner(void *arg) { - loc_t loc = {0,}; - br_child_t *child = NULL; - xlator_t *this = NULL; - struct br_scanfs *fsscan = NULL; - - child = arg; - this = child->this; - fsscan = &child->fsscan; - - THIS = this; - loc.inode = child->table->root; - - while (1) { - br_fsscanner_wait_until_kicked (this, child); - { - /* precursor for scrub */ - br_fsscanner_entry_control (this, child); - - /* scrub */ - (void) syncop_ftw (child->xl, - &loc, GF_CLIENT_PID_SCRUB, - child, br_fsscanner_handle_entry); - if (!list_empty (&fsscan->queued)) - wait_for_scrubbing (this, fsscan); - - /* scrub exit criteria */ - br_fsscanner_exit_control (this, child); - } + loc_t loc = { + 0, + }; + br_child_t *child = NULL; + xlator_t *this = NULL; + struct br_scanfs *fsscan = NULL; + + child = arg; + this = child->this; + fsscan = &child->fsscan; + + THIS = this; + loc.inode = child->table->root; + + while (1) { + br_fsscanner_wait_until_kicked(this, child); + { + /* precursor for scrub */ + br_fsscanner_entry_control(this, child); + + /* scrub */ + (void)syncop_ftw(child->xl, &loc, GF_CLIENT_PID_SCRUB, child, + br_fsscanner_handle_entry); + if (!list_empty(&fsscan->queued)) + wait_for_scrubbing(this, fsscan); + + /* scrub exit criteria */ + br_fsscanner_exit_control(this, child); } + } - return NULL; + return NULL; } /** @@ -827,263 +836,280 @@ br_fsscanner (void *arg) * non-pending timer. */ void -br_kickstart_scanner (struct gf_tw_timer_list *timer, - void *data, unsigned long calltime) +br_kickstart_scanner(struct gf_tw_timer_list *timer, void *data, + unsigned long calltime) { - xlator_t *this = NULL; - struct br_monitor *scrub_monitor = data; - br_private_t *priv = NULL; - - THIS = this = scrub_monitor->this; - priv = this->private; - - /* Reset scrub statistics */ - priv->scrub_stat.scrubbed_files = 0; - priv->scrub_stat.unsigned_files = 0; - - /* Moves state from PENDING to ACTIVE */ - (void) br_scrubber_entry_control (this); - - /* kickstart scanning.. */ - pthread_mutex_lock (&scrub_monitor->wakelock); - { - scrub_monitor->kick = _gf_true; - GF_ASSERT (scrub_monitor->active_child_count == 0); - pthread_cond_broadcast (&scrub_monitor->wakecond); - } - pthread_mutex_unlock (&scrub_monitor->wakelock); - - return; + xlator_t *this = NULL; + struct br_monitor *scrub_monitor = data; + br_private_t *priv = NULL; + + THIS = this = scrub_monitor->this; + priv = this->private; + + /* Reset scrub statistics */ + priv->scrub_stat.scrubbed_files = 0; + priv->scrub_stat.unsigned_files = 0; + + /* Moves state from PENDING to ACTIVE */ + (void)br_scrubber_entry_control(this); + + /* kickstart scanning.. */ + pthread_mutex_lock(&scrub_monitor->wakelock); + { + scrub_monitor->kick = _gf_true; + GF_ASSERT(scrub_monitor->active_child_count == 0); + pthread_cond_broadcast(&scrub_monitor->wakecond); + } + pthread_mutex_unlock(&scrub_monitor->wakelock); + + return; } static uint32_t -br_fsscan_calculate_delta (uint32_t times) +br_fsscan_calculate_delta(uint32_t times) { - return times; + return times; } -#define BR_SCRUB_ONDEMAND (1) -#define BR_SCRUB_MINUTE (60) -#define BR_SCRUB_HOURLY (60 * 60) -#define BR_SCRUB_DAILY (1 * 24 * 60 * 60) -#define BR_SCRUB_WEEKLY (7 * 24 * 60 * 60) -#define BR_SCRUB_BIWEEKLY (14 * 24 * 60 * 60) -#define BR_SCRUB_MONTHLY (30 * 24 * 60 * 60) +#define BR_SCRUB_ONDEMAND (1) +#define BR_SCRUB_MINUTE (60) +#define BR_SCRUB_HOURLY (60 * 60) +#define BR_SCRUB_DAILY (1 * 24 * 60 * 60) +#define BR_SCRUB_WEEKLY (7 * 24 * 60 * 60) +#define BR_SCRUB_BIWEEKLY (14 * 24 * 60 * 60) +#define BR_SCRUB_MONTHLY (30 * 24 * 60 * 60) static unsigned int -br_fsscan_calculate_timeout (scrub_freq_t freq) +br_fsscan_calculate_timeout(scrub_freq_t freq) { - uint32_t timo = 0; + uint32_t timo = 0; - switch (freq) { + switch (freq) { case BR_FSSCRUB_FREQ_MINUTE: - timo = br_fsscan_calculate_delta (BR_SCRUB_MINUTE); - break; + timo = br_fsscan_calculate_delta(BR_SCRUB_MINUTE); + break; case BR_FSSCRUB_FREQ_HOURLY: - timo = br_fsscan_calculate_delta (BR_SCRUB_HOURLY); - break; + timo = br_fsscan_calculate_delta(BR_SCRUB_HOURLY); + break; case BR_FSSCRUB_FREQ_DAILY: - timo = br_fsscan_calculate_delta (BR_SCRUB_DAILY); - break; + timo = br_fsscan_calculate_delta(BR_SCRUB_DAILY); + break; case BR_FSSCRUB_FREQ_WEEKLY: - timo = br_fsscan_calculate_delta (BR_SCRUB_WEEKLY); - break; + timo = br_fsscan_calculate_delta(BR_SCRUB_WEEKLY); + break; case BR_FSSCRUB_FREQ_BIWEEKLY: - timo = br_fsscan_calculate_delta (BR_SCRUB_BIWEEKLY); - break; + timo = br_fsscan_calculate_delta(BR_SCRUB_BIWEEKLY); + break; case BR_FSSCRUB_FREQ_MONTHLY: - timo = br_fsscan_calculate_delta (BR_SCRUB_MONTHLY); - break; + timo = br_fsscan_calculate_delta(BR_SCRUB_MONTHLY); + break; default: - timo = 0; - } + timo = 0; + } - return timo; + return timo; } int32_t -br_fsscan_schedule (xlator_t *this) +br_fsscan_schedule(xlator_t *this) { - uint32_t timo = 0; - br_private_t *priv = NULL; - struct timeval tv = {0,}; - char timestr[1024] = {0,}; - struct br_scrubber *fsscrub = NULL; - struct gf_tw_timer_list *timer = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - fsscrub = &priv->fsscrub; - scrub_monitor = &priv->scrub_monitor; - - (void) gettimeofday (&tv, NULL); - scrub_monitor->boot = tv.tv_sec; - - timo = br_fsscan_calculate_timeout (fsscrub->frequency); - if (timo == 0) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG, - "BUG: Zero schedule timeout"); - goto error_return; - } - - scrub_monitor->timer = GF_CALLOC (1, sizeof (*scrub_monitor->timer), - gf_br_stub_mt_br_scanner_freq_t); - if (!scrub_monitor->timer) - goto error_return; - - timer = scrub_monitor->timer; - INIT_LIST_HEAD (&timer->entry); - - timer->data = scrub_monitor; - timer->expires = timo; - timer->function = br_kickstart_scanner; - - gf_tw_add_timer (priv->timer_wheel, timer); - _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_PENDING); - - gf_time_fmt (timestr, sizeof (timestr), - (scrub_monitor->boot + timo), gf_timefmt_FT); - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, "Scrubbing is " - "scheduled to run at %s", timestr); - - return 0; - - error_return: - return -1; + uint32_t timo = 0; + br_private_t *priv = NULL; + struct timeval tv = { + 0, + }; + char timestr[1024] = { + 0, + }; + struct br_scrubber *fsscrub = NULL; + struct gf_tw_timer_list *timer = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + fsscrub = &priv->fsscrub; + scrub_monitor = &priv->scrub_monitor; + + (void)gettimeofday(&tv, NULL); + scrub_monitor->boot = tv.tv_sec; + + timo = br_fsscan_calculate_timeout(fsscrub->frequency); + if (timo == 0) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG, + "BUG: Zero schedule timeout"); + goto error_return; + } + + scrub_monitor->timer = GF_CALLOC(1, sizeof(*scrub_monitor->timer), + gf_br_stub_mt_br_scanner_freq_t); + if (!scrub_monitor->timer) + goto error_return; + + timer = scrub_monitor->timer; + INIT_LIST_HEAD(&timer->entry); + + timer->data = scrub_monitor; + timer->expires = timo; + timer->function = br_kickstart_scanner; + + gf_tw_add_timer(priv->timer_wheel, timer); + _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PENDING); + + gf_time_fmt(timestr, sizeof(timestr), (scrub_monitor->boot + timo), + gf_timefmt_FT); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Scrubbing is " + "scheduled to run at %s", + timestr); + + return 0; + +error_return: + return -1; } int32_t -br_fsscan_activate (xlator_t *this) +br_fsscan_activate(xlator_t *this) { - uint32_t timo = 0; - char timestr[1024] = {0,}; - struct timeval now = {0,}; - br_private_t *priv = NULL; - struct br_scrubber *fsscrub = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - fsscrub = &priv->fsscrub; - scrub_monitor = &priv->scrub_monitor; - - (void) gettimeofday (&now, NULL); - timo = br_fsscan_calculate_timeout (fsscrub->frequency); - if (timo == 0) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG, - "BUG: Zero schedule timeout"); - return -1; - } + uint32_t timo = 0; + char timestr[1024] = { + 0, + }; + struct timeval now = { + 0, + }; + br_private_t *priv = NULL; + struct br_scrubber *fsscrub = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + fsscrub = &priv->fsscrub; + scrub_monitor = &priv->scrub_monitor; + + (void)gettimeofday(&now, NULL); + timo = br_fsscan_calculate_timeout(fsscrub->frequency); + if (timo == 0) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG, + "BUG: Zero schedule timeout"); + return -1; + } - pthread_mutex_lock (&scrub_monitor->donelock); - { - scrub_monitor->done = _gf_false; - } - pthread_mutex_unlock (&scrub_monitor->donelock); + pthread_mutex_lock(&scrub_monitor->donelock); + { + scrub_monitor->done = _gf_false; + } + pthread_mutex_unlock(&scrub_monitor->donelock); - gf_time_fmt (timestr, sizeof (timestr), - (now.tv_sec + timo), gf_timefmt_FT); - (void) gf_tw_mod_timer (priv->timer_wheel, scrub_monitor->timer, timo); + gf_time_fmt(timestr, sizeof(timestr), (now.tv_sec + timo), gf_timefmt_FT); + (void)gf_tw_mod_timer(priv->timer_wheel, scrub_monitor->timer, timo); - _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_PENDING); - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, "Scrubbing is " - "rescheduled to run at %s", timestr); + _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PENDING); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Scrubbing is " + "rescheduled to run at %s", + timestr); - return 0; + return 0; } int32_t -br_fsscan_reschedule (xlator_t *this) +br_fsscan_reschedule(xlator_t *this) { - int32_t ret = 0; - uint32_t timo = 0; - char timestr[1024] = {0,}; - struct timeval now = {0,}; - br_private_t *priv = NULL; - struct br_scrubber *fsscrub = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - fsscrub = &priv->fsscrub; - scrub_monitor = &priv->scrub_monitor; - - if (!fsscrub->frequency_reconf) - return 0; - - (void) gettimeofday (&now, NULL); - timo = br_fsscan_calculate_timeout (fsscrub->frequency); - if (timo == 0) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG, - "BUG: Zero schedule timeout"); - return -1; - } + int32_t ret = 0; + uint32_t timo = 0; + char timestr[1024] = { + 0, + }; + struct timeval now = { + 0, + }; + br_private_t *priv = NULL; + struct br_scrubber *fsscrub = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + fsscrub = &priv->fsscrub; + scrub_monitor = &priv->scrub_monitor; + + if (!fsscrub->frequency_reconf) + return 0; - gf_time_fmt (timestr, sizeof (timestr), - (now.tv_sec + timo), gf_timefmt_FT); + (void)gettimeofday(&now, NULL); + timo = br_fsscan_calculate_timeout(fsscrub->frequency); + if (timo == 0) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG, + "BUG: Zero schedule timeout"); + return -1; + } - pthread_mutex_lock (&scrub_monitor->donelock); - { - scrub_monitor->done = _gf_false; - } - pthread_mutex_unlock (&scrub_monitor->donelock); - - ret = gf_tw_mod_timer_pending (priv->timer_wheel, scrub_monitor->timer, timo); - if (ret == 0) - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, - "Scrubber is currently running and would be " - "rescheduled after completion"); - else { - _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_PENDING); - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, - "Scrubbing rescheduled to run at %s", timestr); - } + gf_time_fmt(timestr, sizeof(timestr), (now.tv_sec + timo), gf_timefmt_FT); - return 0; + pthread_mutex_lock(&scrub_monitor->donelock); + { + scrub_monitor->done = _gf_false; + } + pthread_mutex_unlock(&scrub_monitor->donelock); + + ret = gf_tw_mod_timer_pending(priv->timer_wheel, scrub_monitor->timer, + timo); + if (ret == 0) + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Scrubber is currently running and would be " + "rescheduled after completion"); + else { + _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PENDING); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Scrubbing rescheduled to run at %s", timestr); + } + + return 0; } int32_t -br_fsscan_ondemand (xlator_t *this) +br_fsscan_ondemand(xlator_t *this) { - int32_t ret = 0; - uint32_t timo = 0; - char timestr[1024] = {0,}; - struct timeval now = {0,}; - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; + int32_t ret = 0; + uint32_t timo = 0; + char timestr[1024] = { + 0, + }; + struct timeval now = { + 0, + }; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; - priv = this->private; - scrub_monitor = &priv->scrub_monitor; + priv = this->private; + scrub_monitor = &priv->scrub_monitor; - (void) gettimeofday (&now, NULL); + (void)gettimeofday(&now, NULL); - timo = BR_SCRUB_ONDEMAND; + timo = BR_SCRUB_ONDEMAND; - gf_time_fmt (timestr, sizeof (timestr), - (now.tv_sec + timo), gf_timefmt_FT); - - pthread_mutex_lock (&scrub_monitor->donelock); - { - scrub_monitor->done = _gf_false; - } - pthread_mutex_unlock (&scrub_monitor->donelock); - - ret = gf_tw_mod_timer_pending (priv->timer_wheel, scrub_monitor->timer, - timo); - if (ret == 0) - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, - "Scrubber is currently running and would be " - "rescheduled after completion"); - else { - _br_monitor_set_scrub_state (scrub_monitor, - BR_SCRUB_STATE_PENDING); - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, - "Ondemand Scrubbing scheduled to run at %s", timestr); - } + gf_time_fmt(timestr, sizeof(timestr), (now.tv_sec + timo), gf_timefmt_FT); - return 0; + pthread_mutex_lock(&scrub_monitor->donelock); + { + scrub_monitor->done = _gf_false; + } + pthread_mutex_unlock(&scrub_monitor->donelock); + + ret = gf_tw_mod_timer_pending(priv->timer_wheel, scrub_monitor->timer, + timo); + if (ret == 0) + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Scrubber is currently running and would be " + "rescheduled after completion"); + else { + _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PENDING); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Ondemand Scrubbing scheduled to run at %s", timestr); + } + + return 0; } -#define BR_SCRUB_THREAD_SCALE_LAZY 0 -#define BR_SCRUB_THREAD_SCALE_NORMAL 0.4 +#define BR_SCRUB_THREAD_SCALE_LAZY 0 +#define BR_SCRUB_THREAD_SCALE_NORMAL 0.4 #define BR_SCRUB_THREAD_SCALE_AGGRESSIVE 1.0 #ifndef M_E @@ -1096,111 +1122,105 @@ br_fsscan_ondemand (xlator_t *this) * the scale based on the number of processor cores too. */ static unsigned int -br_scrubber_calc_scale (xlator_t *this, - br_private_t *priv, scrub_throttle_t throttle) +br_scrubber_calc_scale(xlator_t *this, br_private_t *priv, + scrub_throttle_t throttle) { - unsigned int scale = 0; + unsigned int scale = 0; - switch (throttle) { + switch (throttle) { case BR_SCRUB_THROTTLE_VOID: case BR_SCRUB_THROTTLE_STALLED: - scale = 0; - break; + scale = 0; + break; case BR_SCRUB_THROTTLE_LAZY: - scale = priv->child_count * - pow (M_E, BR_SCRUB_THREAD_SCALE_LAZY); - break; + scale = priv->child_count * pow(M_E, BR_SCRUB_THREAD_SCALE_LAZY); + break; case BR_SCRUB_THROTTLE_NORMAL: - scale = priv->child_count * - pow (M_E, BR_SCRUB_THREAD_SCALE_NORMAL); - break; + scale = priv->child_count * pow(M_E, BR_SCRUB_THREAD_SCALE_NORMAL); + break; case BR_SCRUB_THROTTLE_AGGRESSIVE: - scale = priv->child_count * - pow (M_E, BR_SCRUB_THREAD_SCALE_AGGRESSIVE); - break; + scale = priv->child_count * + pow(M_E, BR_SCRUB_THREAD_SCALE_AGGRESSIVE); + break; default: - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_UNKNOWN_THROTTLE, - "Unknown throttle %d", throttle); - } - - return scale; + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_UNKNOWN_THROTTLE, + "Unknown throttle %d", throttle); + } + return scale; } static br_child_t * -_br_scrubber_get_next_child (struct br_scrubber *fsscrub) +_br_scrubber_get_next_child(struct br_scrubber *fsscrub) { - br_child_t *child = NULL; + br_child_t *child = NULL; - child = list_first_entry (&fsscrub->scrublist, br_child_t, list); - list_rotate_left (&fsscrub->scrublist); + child = list_first_entry(&fsscrub->scrublist, br_child_t, list); + list_rotate_left(&fsscrub->scrublist); - return child; + return child; } static void -_br_scrubber_get_entry (br_child_t *child, struct br_fsscan_entry **fsentry) +_br_scrubber_get_entry(br_child_t *child, struct br_fsscan_entry **fsentry) { - struct br_scanfs *fsscan = &child->fsscan; + struct br_scanfs *fsscan = &child->fsscan; - if (list_empty (&fsscan->ready)) - return; - *fsentry = list_first_entry - (&fsscan->ready, struct br_fsscan_entry, list); - list_del_init (&(*fsentry)->list); + if (list_empty(&fsscan->ready)) + return; + *fsentry = list_first_entry(&fsscan->ready, struct br_fsscan_entry, list); + list_del_init(&(*fsentry)->list); } static void -_br_scrubber_find_scrubbable_entry (struct br_scrubber *fsscrub, - struct br_fsscan_entry **fsentry) +_br_scrubber_find_scrubbable_entry(struct br_scrubber *fsscrub, + struct br_fsscan_entry **fsentry) { - br_child_t *child = NULL; - br_child_t *firstchild = NULL; - - while (1) { - while (list_empty (&fsscrub->scrublist)) - pthread_cond_wait (&fsscrub->cond, &fsscrub->mutex); + br_child_t *child = NULL; + br_child_t *firstchild = NULL; - firstchild = NULL; - for (child = _br_scrubber_get_next_child (fsscrub); - child != firstchild; - child = _br_scrubber_get_next_child (fsscrub)) { + while (1) { + while (list_empty(&fsscrub->scrublist)) + pthread_cond_wait(&fsscrub->cond, &fsscrub->mutex); - if (!firstchild) - firstchild = child; + firstchild = NULL; + for (child = _br_scrubber_get_next_child(fsscrub); child != firstchild; + child = _br_scrubber_get_next_child(fsscrub)) { + if (!firstchild) + firstchild = child; - _br_scrubber_get_entry (child, fsentry); - if (*fsentry) - break; - } + _br_scrubber_get_entry(child, fsentry); + if (*fsentry) + break; + } - if (*fsentry) - break; + if (*fsentry) + break; - /* nothing to work on.. wait till available */ - pthread_cond_wait (&fsscrub->cond, &fsscrub->mutex); - } + /* nothing to work on.. wait till available */ + pthread_cond_wait(&fsscrub->cond, &fsscrub->mutex); + } } static void -br_scrubber_pick_entry (struct br_scrubber *fsscrub, - struct br_fsscan_entry **fsentry) +br_scrubber_pick_entry(struct br_scrubber *fsscrub, + struct br_fsscan_entry **fsentry) { - pthread_cleanup_push (_br_lock_cleaner, &fsscrub->mutex); + pthread_cleanup_push(_br_lock_cleaner, &fsscrub->mutex); - pthread_mutex_lock (&fsscrub->mutex); - { - *fsentry = NULL; - _br_scrubber_find_scrubbable_entry (fsscrub, fsentry); - } - pthread_mutex_unlock (&fsscrub->mutex); + pthread_mutex_lock(&fsscrub->mutex); + { + *fsentry = NULL; + _br_scrubber_find_scrubbable_entry(fsscrub, fsentry); + } + pthread_mutex_unlock(&fsscrub->mutex); - pthread_cleanup_pop (0); + pthread_cleanup_pop(0); } struct br_scrub_entry { - gf_boolean_t scrubbed; - struct br_fsscan_entry *fsentry; + gf_boolean_t scrubbed; + struct br_fsscan_entry *fsentry; }; /** @@ -1210,686 +1230,698 @@ struct br_scrub_entry { * in the ->pending queue or when an object is undergoing scrubbing. */ static void -br_scrubber_entry_handle (void *arg) +br_scrubber_entry_handle(void *arg) { - struct br_scanfs *fsscan = NULL; - struct br_scrub_entry *sentry = NULL; - struct br_fsscan_entry *fsentry = NULL; + struct br_scanfs *fsscan = NULL; + struct br_scrub_entry *sentry = NULL; + struct br_fsscan_entry *fsentry = NULL; - sentry = arg; + sentry = arg; - fsentry = sentry->fsentry; - fsscan = fsentry->fsscan; + fsentry = sentry->fsentry; + fsscan = fsentry->fsscan; - LOCK (&fsscan->entrylock); - { - if (sentry->scrubbed) { - _br_fsscan_dec_entry_count (fsscan); - - /* cleanup ->entry */ - fsentry->data = NULL; - fsentry->fsscan = NULL; - loc_wipe (&fsentry->parent); - gf_dirent_entry_free (fsentry->entry); - - GF_FREE (sentry->fsentry); - } else { - /* (re)queue the entry again for scrub */ - _br_fsscan_collect_entry (fsscan, sentry->fsentry); - } + LOCK(&fsscan->entrylock); + { + if (sentry->scrubbed) { + _br_fsscan_dec_entry_count(fsscan); + + /* cleanup ->entry */ + fsentry->data = NULL; + fsentry->fsscan = NULL; + loc_wipe(&fsentry->parent); + gf_dirent_entry_free(fsentry->entry); + + GF_FREE(sentry->fsentry); + } else { + /* (re)queue the entry again for scrub */ + _br_fsscan_collect_entry(fsscan, sentry->fsentry); } - UNLOCK (&fsscan->entrylock); + } + UNLOCK(&fsscan->entrylock); } static void -br_scrubber_scrub_entry (xlator_t *this, struct br_fsscan_entry *fsentry) +br_scrubber_scrub_entry(xlator_t *this, struct br_fsscan_entry *fsentry) { - struct br_scrub_entry sentry = {0, }; - - sentry.scrubbed = 0; - sentry.fsentry = fsentry; - - pthread_cleanup_push (br_scrubber_entry_handle, &sentry); - { - (void) br_scrubber_scrub_begin (this, fsentry); - sentry.scrubbed = 1; - } - pthread_cleanup_pop (1); + struct br_scrub_entry sentry = { + 0, + }; + + sentry.scrubbed = 0; + sentry.fsentry = fsentry; + + pthread_cleanup_push(br_scrubber_entry_handle, &sentry); + { + (void)br_scrubber_scrub_begin(this, fsentry); + sentry.scrubbed = 1; + } + pthread_cleanup_pop(1); } -void *br_scrubber_proc (void *arg) +void * +br_scrubber_proc(void *arg) { - xlator_t *this = NULL; - struct br_scrubber *fsscrub = NULL; - struct br_fsscan_entry *fsentry = NULL; + xlator_t *this = NULL; + struct br_scrubber *fsscrub = NULL; + struct br_fsscan_entry *fsentry = NULL; - fsscrub = arg; - THIS = this = fsscrub->this; + fsscrub = arg; + THIS = this = fsscrub->this; - while (1) { - br_scrubber_pick_entry (fsscrub, &fsentry); - br_scrubber_scrub_entry (this, fsentry); - sleep (1); - } + while (1) { + br_scrubber_pick_entry(fsscrub, &fsentry); + br_scrubber_scrub_entry(this, fsentry); + sleep(1); + } - return NULL; + return NULL; } static int32_t -br_scrubber_scale_up (xlator_t *this, - struct br_scrubber *fsscrub, - unsigned int v1, unsigned int v2) +br_scrubber_scale_up(xlator_t *this, struct br_scrubber *fsscrub, + unsigned int v1, unsigned int v2) { - int i = 0; - int32_t ret = -1; - int diff = 0; - struct br_scrubbers *scrub = NULL; - - diff = (int)(v2 - v1); - - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCALING_UP_SCRUBBER, - "Scaling up scrubbers [%d => %d]", v1, v2); - - for (i = 0; i < diff; i++) { - scrub = GF_CALLOC (diff, sizeof (*scrub), - gf_br_mt_br_scrubber_t); - if (!scrub) - break; - - INIT_LIST_HEAD (&scrub->list); - ret = gf_thread_create (&scrub->scrubthread, - NULL, br_scrubber_proc, fsscrub, - "brsproc"); - if (ret) - break; - - fsscrub->nr_scrubbers++; - list_add_tail (&scrub->list, &fsscrub->scrubbers); - } + int i = 0; + int32_t ret = -1; + int diff = 0; + struct br_scrubbers *scrub = NULL; - if ((i != diff) && !scrub) - goto error_return; + diff = (int)(v2 - v1); - if (i != diff) /* degraded scaling.. */ - gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SCALE_UP_FAILED, - "Could not fully scale up to %d scrubber(s). Spawned " - "%d/%d [total scrubber(s): %d]", v2, i, diff, (v1 + i)); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCALING_UP_SCRUBBER, + "Scaling up scrubbers [%d => %d]", v1, v2); - return 0; + for (i = 0; i < diff; i++) { + scrub = GF_CALLOC(diff, sizeof(*scrub), gf_br_mt_br_scrubber_t); + if (!scrub) + break; - error_return: - return -1; + INIT_LIST_HEAD(&scrub->list); + ret = gf_thread_create(&scrub->scrubthread, NULL, br_scrubber_proc, + fsscrub, "brsproc"); + if (ret) + break; + + fsscrub->nr_scrubbers++; + list_add_tail(&scrub->list, &fsscrub->scrubbers); + } + + if ((i != diff) && !scrub) + goto error_return; + + if (i != diff) /* degraded scaling.. */ + gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SCALE_UP_FAILED, + "Could not fully scale up to %d scrubber(s). Spawned " + "%d/%d [total scrubber(s): %d]", + v2, i, diff, (v1 + i)); + + return 0; + +error_return: + return -1; } static int32_t -br_scrubber_scale_down (xlator_t *this, - struct br_scrubber *fsscrub, - unsigned int v1, unsigned int v2) +br_scrubber_scale_down(xlator_t *this, struct br_scrubber *fsscrub, + unsigned int v1, unsigned int v2) { - int i = 0; - int diff = 0; - int32_t ret = -1; - struct br_scrubbers *scrub = NULL; - - diff = (int)(v1 - v2); + int i = 0; + int diff = 0; + int32_t ret = -1; + struct br_scrubbers *scrub = NULL; - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCALE_DOWN_SCRUBBER, - "Scaling down scrubbers [%d => %d]", v1, v2); + diff = (int)(v1 - v2); - for (i = 0 ; i < diff; i++) { - scrub = list_first_entry - (&fsscrub->scrubbers, struct br_scrubbers, list); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCALE_DOWN_SCRUBBER, + "Scaling down scrubbers [%d => %d]", v1, v2); - list_del_init (&scrub->list); - ret = gf_thread_cleanup_xint (scrub->scrubthread); - if (ret) - break; - GF_FREE (scrub); + for (i = 0; i < diff; i++) { + scrub = list_first_entry(&fsscrub->scrubbers, struct br_scrubbers, + list); - fsscrub->nr_scrubbers--; - } - - if (ret) { - gf_msg (this->name, GF_LOG_WARNING, 0, - BRB_MSG_SCALE_DOWN_FAILED, "Could not fully scale down " - "to %d scrubber(s). Terminated %d/%d [total " - "scrubber(s): %d]", v1, i, diff, (v2 - i)); - ret = 0; - } + list_del_init(&scrub->list); + ret = gf_thread_cleanup_xint(scrub->scrubthread); + if (ret) + break; + GF_FREE(scrub); + + fsscrub->nr_scrubbers--; + } + + if (ret) { + gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SCALE_DOWN_FAILED, + "Could not fully scale down " + "to %d scrubber(s). Terminated %d/%d [total " + "scrubber(s): %d]", + v1, i, diff, (v2 - i)); + ret = 0; + } - return ret; + return ret; } static int32_t -br_scrubber_configure (xlator_t *this, br_private_t *priv, - struct br_scrubber *fsscrub, scrub_throttle_t nthrottle) +br_scrubber_configure(xlator_t *this, br_private_t *priv, + struct br_scrubber *fsscrub, scrub_throttle_t nthrottle) { - int32_t ret = 0; - unsigned int v1 = 0; - unsigned int v2 = 0; + int32_t ret = 0; + unsigned int v1 = 0; + unsigned int v2 = 0; - v1 = fsscrub->nr_scrubbers; - v2 = br_scrubber_calc_scale (this, priv, nthrottle); + v1 = fsscrub->nr_scrubbers; + v2 = br_scrubber_calc_scale(this, priv, nthrottle); - if (v1 == v2) - return 0; + if (v1 == v2) + return 0; - if (v1 > v2) - ret = br_scrubber_scale_down (this, fsscrub, v1, v2); - else - ret = br_scrubber_scale_up (this, fsscrub, v1, v2); + if (v1 > v2) + ret = br_scrubber_scale_down(this, fsscrub, v1, v2); + else + ret = br_scrubber_scale_up(this, fsscrub, v1, v2); - return ret; + return ret; } static int32_t -br_scrubber_fetch_option (xlator_t *this, - char *opt, dict_t *options, char **value) +br_scrubber_fetch_option(xlator_t *this, char *opt, dict_t *options, + char **value) { - if (options) - GF_OPTION_RECONF (opt, *value, options, str, error_return); - else - GF_OPTION_INIT (opt, *value, str, error_return); + if (options) + GF_OPTION_RECONF(opt, *value, options, str, error_return); + else + GF_OPTION_INIT(opt, *value, str, error_return); - return 0; + return 0; - error_return: - return -1; +error_return: + return -1; } /* internal "throttle" override */ -#define BR_SCRUB_STALLED "STALLED" +#define BR_SCRUB_STALLED "STALLED" /* TODO: token buket spec */ static int32_t -br_scrubber_handle_throttle (xlator_t *this, br_private_t *priv, - dict_t *options, gf_boolean_t scrubstall) +br_scrubber_handle_throttle(xlator_t *this, br_private_t *priv, dict_t *options, + gf_boolean_t scrubstall) { - int32_t ret = 0; - char *tmp = NULL; - struct br_scrubber *fsscrub = NULL; - scrub_throttle_t nthrottle = BR_SCRUB_THROTTLE_VOID; - - fsscrub = &priv->fsscrub; - fsscrub->throttle_reconf = _gf_false; - - ret = br_scrubber_fetch_option (this, "scrub-throttle", options, &tmp); - if (ret) - goto error_return; - - if (scrubstall) - tmp = BR_SCRUB_STALLED; - - if (strcasecmp (tmp, "lazy") == 0) - nthrottle = BR_SCRUB_THROTTLE_LAZY; - else if (strcasecmp (tmp, "normal") == 0) - nthrottle = BR_SCRUB_THROTTLE_NORMAL; - else if (strcasecmp (tmp, "aggressive") == 0) - nthrottle = BR_SCRUB_THROTTLE_AGGRESSIVE; - else if (strcasecmp (tmp, BR_SCRUB_STALLED) == 0) - nthrottle = BR_SCRUB_THROTTLE_STALLED; - else - goto error_return; - - /* on failure old throttling value is preserved */ - ret = br_scrubber_configure (this, priv, fsscrub, nthrottle); - if (ret) - goto error_return; - - if (fsscrub->throttle != nthrottle) - fsscrub->throttle_reconf = _gf_true; - - fsscrub->throttle = nthrottle; - return 0; - - error_return: - return -1; + int32_t ret = 0; + char *tmp = NULL; + struct br_scrubber *fsscrub = NULL; + scrub_throttle_t nthrottle = BR_SCRUB_THROTTLE_VOID; + + fsscrub = &priv->fsscrub; + fsscrub->throttle_reconf = _gf_false; + + ret = br_scrubber_fetch_option(this, "scrub-throttle", options, &tmp); + if (ret) + goto error_return; + + if (scrubstall) + tmp = BR_SCRUB_STALLED; + + if (strcasecmp(tmp, "lazy") == 0) + nthrottle = BR_SCRUB_THROTTLE_LAZY; + else if (strcasecmp(tmp, "normal") == 0) + nthrottle = BR_SCRUB_THROTTLE_NORMAL; + else if (strcasecmp(tmp, "aggressive") == 0) + nthrottle = BR_SCRUB_THROTTLE_AGGRESSIVE; + else if (strcasecmp(tmp, BR_SCRUB_STALLED) == 0) + nthrottle = BR_SCRUB_THROTTLE_STALLED; + else + goto error_return; + + /* on failure old throttling value is preserved */ + ret = br_scrubber_configure(this, priv, fsscrub, nthrottle); + if (ret) + goto error_return; + + if (fsscrub->throttle != nthrottle) + fsscrub->throttle_reconf = _gf_true; + + fsscrub->throttle = nthrottle; + return 0; + +error_return: + return -1; } static int32_t -br_scrubber_handle_stall (xlator_t *this, br_private_t *priv, - dict_t *options, gf_boolean_t *scrubstall) +br_scrubber_handle_stall(xlator_t *this, br_private_t *priv, dict_t *options, + gf_boolean_t *scrubstall) { - int32_t ret = 0; - char *tmp = NULL; + int32_t ret = 0; + char *tmp = NULL; - ret = br_scrubber_fetch_option (this, "scrub-state", options, &tmp); - if (ret) - goto error_return; + ret = br_scrubber_fetch_option(this, "scrub-state", options, &tmp); + if (ret) + goto error_return; - if (strcasecmp (tmp, "pause") == 0) /* anything else is active */ - *scrubstall = _gf_true; + if (strcasecmp(tmp, "pause") == 0) /* anything else is active */ + *scrubstall = _gf_true; - return 0; + return 0; - error_return: - return -1; +error_return: + return -1; } static int32_t -br_scrubber_handle_freq (xlator_t *this, br_private_t *priv, - dict_t *options, gf_boolean_t scrubstall) +br_scrubber_handle_freq(xlator_t *this, br_private_t *priv, dict_t *options, + gf_boolean_t scrubstall) { - int32_t ret = -1; - char *tmp = NULL; - scrub_freq_t frequency = BR_FSSCRUB_FREQ_HOURLY; - struct br_scrubber *fsscrub = NULL; - - fsscrub = &priv->fsscrub; - fsscrub->frequency_reconf = _gf_true; - - ret = br_scrubber_fetch_option (this, "scrub-freq", options, &tmp); - if (ret) - goto error_return; - - if (scrubstall) - tmp = BR_SCRUB_STALLED; - - if (strcasecmp (tmp, "hourly") == 0) { - frequency = BR_FSSCRUB_FREQ_HOURLY; - } else if (strcasecmp (tmp, "daily") == 0) { - frequency = BR_FSSCRUB_FREQ_DAILY; - } else if (strcasecmp (tmp, "weekly") == 0) { - frequency = BR_FSSCRUB_FREQ_WEEKLY; - } else if (strcasecmp (tmp, "biweekly") == 0) { - frequency = BR_FSSCRUB_FREQ_BIWEEKLY; - } else if (strcasecmp (tmp, "monthly") == 0) { - frequency = BR_FSSCRUB_FREQ_MONTHLY; - } else if (strcasecmp (tmp, "minute") == 0) { - frequency = BR_FSSCRUB_FREQ_MINUTE; - } else if (strcasecmp (tmp, BR_SCRUB_STALLED) == 0) { - frequency = BR_FSSCRUB_FREQ_STALLED; - } else - goto error_return; - - if (fsscrub->frequency == frequency) - fsscrub->frequency_reconf = _gf_false; - else - fsscrub->frequency = frequency; - - return 0; - - error_return: - return -1; + int32_t ret = -1; + char *tmp = NULL; + scrub_freq_t frequency = BR_FSSCRUB_FREQ_HOURLY; + struct br_scrubber *fsscrub = NULL; + + fsscrub = &priv->fsscrub; + fsscrub->frequency_reconf = _gf_true; + + ret = br_scrubber_fetch_option(this, "scrub-freq", options, &tmp); + if (ret) + goto error_return; + + if (scrubstall) + tmp = BR_SCRUB_STALLED; + + if (strcasecmp(tmp, "hourly") == 0) { + frequency = BR_FSSCRUB_FREQ_HOURLY; + } else if (strcasecmp(tmp, "daily") == 0) { + frequency = BR_FSSCRUB_FREQ_DAILY; + } else if (strcasecmp(tmp, "weekly") == 0) { + frequency = BR_FSSCRUB_FREQ_WEEKLY; + } else if (strcasecmp(tmp, "biweekly") == 0) { + frequency = BR_FSSCRUB_FREQ_BIWEEKLY; + } else if (strcasecmp(tmp, "monthly") == 0) { + frequency = BR_FSSCRUB_FREQ_MONTHLY; + } else if (strcasecmp(tmp, "minute") == 0) { + frequency = BR_FSSCRUB_FREQ_MINUTE; + } else if (strcasecmp(tmp, BR_SCRUB_STALLED) == 0) { + frequency = BR_FSSCRUB_FREQ_STALLED; + } else + goto error_return; + + if (fsscrub->frequency == frequency) + fsscrub->frequency_reconf = _gf_false; + else + fsscrub->frequency = frequency; + + return 0; + +error_return: + return -1; } -static void br_scrubber_log_option (xlator_t *this, - br_private_t *priv, gf_boolean_t scrubstall) +static void +br_scrubber_log_option(xlator_t *this, br_private_t *priv, + gf_boolean_t scrubstall) { - struct br_scrubber *fsscrub = &priv->fsscrub; - char *scrub_throttle_str[] = { - [BR_SCRUB_THROTTLE_LAZY] = "lazy", - [BR_SCRUB_THROTTLE_NORMAL] = "normal", - [BR_SCRUB_THROTTLE_AGGRESSIVE] = "aggressive", - }; - - char *scrub_freq_str[] = { - [BR_FSSCRUB_FREQ_HOURLY] = "hourly", - [BR_FSSCRUB_FREQ_DAILY] = "daily", - [BR_FSSCRUB_FREQ_WEEKLY] = "weekly", - [BR_FSSCRUB_FREQ_BIWEEKLY] = "biweekly", - [BR_FSSCRUB_FREQ_MONTHLY] = "monthly (30 days)", - [BR_FSSCRUB_FREQ_MINUTE] = "every minute", - }; - - if (scrubstall) - return; /* logged as pause */ - - if (fsscrub->frequency_reconf || fsscrub->throttle_reconf) { - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_TUNABLE, - "SCRUB TUNABLES:: [Frequency: %s, Throttle: %s]", - scrub_freq_str[fsscrub->frequency], - scrub_throttle_str[fsscrub->throttle]); - } + struct br_scrubber *fsscrub = &priv->fsscrub; + char *scrub_throttle_str[] = { + [BR_SCRUB_THROTTLE_LAZY] = "lazy", + [BR_SCRUB_THROTTLE_NORMAL] = "normal", + [BR_SCRUB_THROTTLE_AGGRESSIVE] = "aggressive", + }; + + char *scrub_freq_str[] = { + [BR_FSSCRUB_FREQ_HOURLY] = "hourly", + [BR_FSSCRUB_FREQ_DAILY] = "daily", + [BR_FSSCRUB_FREQ_WEEKLY] = "weekly", + [BR_FSSCRUB_FREQ_BIWEEKLY] = "biweekly", + [BR_FSSCRUB_FREQ_MONTHLY] = "monthly (30 days)", + [BR_FSSCRUB_FREQ_MINUTE] = "every minute", + }; + + if (scrubstall) + return; /* logged as pause */ + + if (fsscrub->frequency_reconf || fsscrub->throttle_reconf) { + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_TUNABLE, + "SCRUB TUNABLES:: [Frequency: %s, Throttle: %s]", + scrub_freq_str[fsscrub->frequency], + scrub_throttle_str[fsscrub->throttle]); + } } int32_t -br_scrubber_handle_options (xlator_t *this, br_private_t *priv, dict_t *options) +br_scrubber_handle_options(xlator_t *this, br_private_t *priv, dict_t *options) { - int32_t ret = 0; - gf_boolean_t scrubstall = _gf_false; /* not as dangerous as it sounds */ + int32_t ret = 0; + gf_boolean_t scrubstall = _gf_false; /* not as dangerous as it sounds */ - ret = br_scrubber_handle_stall (this, priv, options, &scrubstall); - if (ret) - goto error_return; + ret = br_scrubber_handle_stall(this, priv, options, &scrubstall); + if (ret) + goto error_return; - ret = br_scrubber_handle_throttle (this, priv, options, scrubstall); - if (ret) - goto error_return; + ret = br_scrubber_handle_throttle(this, priv, options, scrubstall); + if (ret) + goto error_return; - ret = br_scrubber_handle_freq (this, priv, options, scrubstall); - if (ret) - goto error_return; + ret = br_scrubber_handle_freq(this, priv, options, scrubstall); + if (ret) + goto error_return; - br_scrubber_log_option (this, priv, scrubstall); + br_scrubber_log_option(this, priv, scrubstall); - return 0; + return 0; - error_return: - return -1; +error_return: + return -1; } inode_t * -br_lookup_bad_obj_dir (xlator_t *this, br_child_t *child, uuid_t gfid) +br_lookup_bad_obj_dir(xlator_t *this, br_child_t *child, uuid_t gfid) { - struct iatt statbuf = {0, }; - inode_table_t *table = NULL; - int32_t ret = -1; - loc_t loc = {0, }; - inode_t *linked_inode = NULL; - int32_t op_errno = 0; - - GF_VALIDATE_OR_GOTO ("bit-rot-scrubber", this, out); - GF_VALIDATE_OR_GOTO (this->name, this->private, out); - GF_VALIDATE_OR_GOTO (this->name, child, out); - - table = child->table; - - loc.inode = inode_new (table); - if (!loc.inode) { - gf_msg (this->name, GF_LOG_ERROR, ENOMEM, - BRB_MSG_NO_MEMORY, "failed to allocate a new inode for" - "bad object directory"); - goto out; - } - - gf_uuid_copy (loc.gfid, gfid); - - ret = syncop_lookup (child->xl, &loc, &statbuf, NULL, NULL, NULL); - if (ret < 0) { - op_errno = -ret; - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_LOOKUP_FAILED, "failed to lookup the bad " - "objects directory (gfid: %s (%s))", uuid_utoa (gfid), - strerror (op_errno)); - goto out; - } - - linked_inode = inode_link (loc.inode, NULL, NULL, &statbuf); - if (linked_inode) - inode_lookup (linked_inode); + struct iatt statbuf = { + 0, + }; + inode_table_t *table = NULL; + int32_t ret = -1; + loc_t loc = { + 0, + }; + inode_t *linked_inode = NULL; + int32_t op_errno = 0; + + GF_VALIDATE_OR_GOTO("bit-rot-scrubber", this, out); + GF_VALIDATE_OR_GOTO(this->name, this->private, out); + GF_VALIDATE_OR_GOTO(this->name, child, out); + + table = child->table; + + loc.inode = inode_new(table); + if (!loc.inode) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, + "failed to allocate a new inode for" + "bad object directory"); + goto out; + } + + gf_uuid_copy(loc.gfid, gfid); + + ret = syncop_lookup(child->xl, &loc, &statbuf, NULL, NULL, NULL); + if (ret < 0) { + op_errno = -ret; + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_LOOKUP_FAILED, + "failed to lookup the bad " + "objects directory (gfid: %s (%s))", + uuid_utoa(gfid), strerror(op_errno)); + goto out; + } + + linked_inode = inode_link(loc.inode, NULL, NULL, &statbuf); + if (linked_inode) + inode_lookup(linked_inode); out: - loc_wipe (&loc); - return linked_inode; + loc_wipe(&loc); + return linked_inode; } int32_t -br_read_bad_object_dir (xlator_t *this, br_child_t *child, fd_t *fd, - dict_t *dict) +br_read_bad_object_dir(xlator_t *this, br_child_t *child, fd_t *fd, + dict_t *dict) { - gf_dirent_t entries; - gf_dirent_t *entry = NULL; - int32_t ret = -1; - off_t offset = 0; - int32_t count = 0; - char key[PATH_MAX] = {0, }; - dict_t *out_dict = NULL; - - INIT_LIST_HEAD (&entries.list); - - while ((ret = syncop_readdir (child->xl, fd, 131072, offset, &entries, - NULL, &out_dict))) { - if (ret < 0) - goto out; - - list_for_each_entry (entry, &entries.list, list) { - offset = entry->d_off; - - snprintf (key, sizeof (key), "quarantine-%d", count); - - /* - * ignore the dict_set errors for now. The intention is - * to get as many bad objects as possible instead of - * erroring out at the first failure. - */ - ret = dict_set_dynstr_with_alloc (dict, key, - entry->d_name); - if (!ret) - count++; - - if (out_dict) { - dict_copy (out_dict, dict); - dict_unref (out_dict); - out_dict = NULL; - } - } - - gf_dirent_free (&entries); - } + gf_dirent_t entries; + gf_dirent_t *entry = NULL; + int32_t ret = -1; + off_t offset = 0; + int32_t count = 0; + char key[PATH_MAX] = { + 0, + }; + dict_t *out_dict = NULL; + + INIT_LIST_HEAD(&entries.list); + + while ((ret = syncop_readdir(child->xl, fd, 131072, offset, &entries, NULL, + &out_dict))) { + if (ret < 0) + goto out; + + list_for_each_entry(entry, &entries.list, list) + { + offset = entry->d_off; - ret = count; - ret = dict_set_int32 (dict, "count", count); + snprintf(key, sizeof(key), "quarantine-%d", count); -out: - return ret; -} - -int32_t -br_get_bad_objects_from_child (xlator_t *this, dict_t *dict, br_child_t *child) -{ - inode_t *inode = NULL; - inode_table_t *table = NULL; - fd_t *fd = NULL; - int32_t ret = -1; - loc_t loc = {0, }; - int32_t op_errno = 0; - - GF_VALIDATE_OR_GOTO ("bit-rot-scrubber", this, out); - GF_VALIDATE_OR_GOTO (this->name, this->private, out); - GF_VALIDATE_OR_GOTO (this->name, child, out); - GF_VALIDATE_OR_GOTO (this->name, dict, out); - - table = child->table; - - inode = inode_find (table, BR_BAD_OBJ_CONTAINER); - if (!inode) { - inode = br_lookup_bad_obj_dir (this, child, - BR_BAD_OBJ_CONTAINER); - if (!inode) - goto out; - } + /* + * ignore the dict_set errors for now. The intention is + * to get as many bad objects as possible instead of + * erroring out at the first failure. + */ + ret = dict_set_dynstr_with_alloc(dict, key, entry->d_name); + if (!ret) + count++; - fd = fd_create (inode, 0); - if (!fd) { - gf_msg (this->name, GF_LOG_ERROR, ENOMEM, - BRB_MSG_FD_CREATE_FAILED, "fd creation for the bad " - "objects directory failed (gfid: %s)", - uuid_utoa (BR_BAD_OBJ_CONTAINER)); - goto out; + if (out_dict) { + dict_copy(out_dict, dict); + dict_unref(out_dict); + out_dict = NULL; + } } - loc.inode = inode; - gf_uuid_copy (loc.gfid, inode->gfid); - - ret = syncop_opendir (child->xl, &loc, fd, NULL, NULL); - if (ret < 0) { - op_errno = -ret; - fd_unref (fd); - fd = NULL; - gf_msg (this->name, GF_LOG_ERROR, op_errno, - BRB_MSG_FD_CREATE_FAILED, "failed to open the bad " - "objects directory %s", - uuid_utoa (BR_BAD_OBJ_CONTAINER)); - goto out; - } + gf_dirent_free(&entries); + } - fd_bind (fd); + ret = count; + ret = dict_set_int32(dict, "count", count); - ret = br_read_bad_object_dir (this, child, fd, dict); - if (ret < 0) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_BAD_OBJ_READDIR_FAIL, "readdir of the bad " - "objects directory (%s) failed ", - uuid_utoa (BR_BAD_OBJ_CONTAINER)); - goto out; - } +out: + return ret; +} - ret = 0; +int32_t +br_get_bad_objects_from_child(xlator_t *this, dict_t *dict, br_child_t *child) +{ + inode_t *inode = NULL; + inode_table_t *table = NULL; + fd_t *fd = NULL; + int32_t ret = -1; + loc_t loc = { + 0, + }; + int32_t op_errno = 0; + + GF_VALIDATE_OR_GOTO("bit-rot-scrubber", this, out); + GF_VALIDATE_OR_GOTO(this->name, this->private, out); + GF_VALIDATE_OR_GOTO(this->name, child, out); + GF_VALIDATE_OR_GOTO(this->name, dict, out); + + table = child->table; + + inode = inode_find(table, BR_BAD_OBJ_CONTAINER); + if (!inode) { + inode = br_lookup_bad_obj_dir(this, child, BR_BAD_OBJ_CONTAINER); + if (!inode) + goto out; + } + + fd = fd_create(inode, 0); + if (!fd) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_FD_CREATE_FAILED, + "fd creation for the bad " + "objects directory failed (gfid: %s)", + uuid_utoa(BR_BAD_OBJ_CONTAINER)); + goto out; + } + + loc.inode = inode; + gf_uuid_copy(loc.gfid, inode->gfid); + + ret = syncop_opendir(child->xl, &loc, fd, NULL, NULL); + if (ret < 0) { + op_errno = -ret; + fd_unref(fd); + fd = NULL; + gf_msg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_FD_CREATE_FAILED, + "failed to open the bad " + "objects directory %s", + uuid_utoa(BR_BAD_OBJ_CONTAINER)); + goto out; + } + + fd_bind(fd); + + ret = br_read_bad_object_dir(this, child, fd, dict); + if (ret < 0) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_BAD_OBJ_READDIR_FAIL, + "readdir of the bad " + "objects directory (%s) failed ", + uuid_utoa(BR_BAD_OBJ_CONTAINER)); + goto out; + } + + ret = 0; out: - loc_wipe (&loc); - if (fd) - fd_unref (fd); - return ret; + loc_wipe(&loc); + if (fd) + fd_unref(fd); + return ret; } int32_t -br_collect_bad_objects_of_child (xlator_t *this, br_child_t *child, - dict_t *dict, dict_t *child_dict, - int32_t total_count) +br_collect_bad_objects_of_child(xlator_t *this, br_child_t *child, dict_t *dict, + dict_t *child_dict, int32_t total_count) { - - int32_t ret = -1; - int32_t count = 0; - char key[PATH_MAX] = {0, }; - char main_key[PATH_MAX] = {0, }; - int32_t j = 0; - int32_t tmp_count = 0; - char *entry = NULL; - char tmp[PATH_MAX] = {0, }; - char *path = NULL; - int32_t len = 0; - - ret = dict_get_int32 (child_dict, "count", &count); + int32_t ret = -1; + int32_t count = 0; + char key[PATH_MAX] = { + 0, + }; + char main_key[PATH_MAX] = { + 0, + }; + int32_t j = 0; + int32_t tmp_count = 0; + char *entry = NULL; + char tmp[PATH_MAX] = { + 0, + }; + char *path = NULL; + int32_t len = 0; + + ret = dict_get_int32(child_dict, "count", &count); + if (ret) + goto out; + + tmp_count = total_count; + + for (j = 0; j < count; j++) { + snprintf(key, PATH_MAX, "quarantine-%d", j); + ret = dict_get_str(child_dict, key, &entry); if (ret) - goto out; - - tmp_count = total_count; - - for (j = 0; j < count; j++) { - snprintf (key, PATH_MAX, "quarantine-%d", j); - ret = dict_get_str (child_dict, key, &entry); - if (ret) - continue; - - ret = dict_get_str (child_dict, entry, &path); - len = snprintf (tmp, PATH_MAX, "%s ==> BRICK: %s\n path: %s", - entry, child->brick_path, path); - if ((len < 0) || (len >= PATH_MAX)) { - continue; - } - snprintf (main_key, PATH_MAX, "quarantine-%d", - tmp_count); - - ret = dict_set_dynstr_with_alloc (dict, main_key, tmp); - if (!ret) - tmp_count++; - path = NULL; + continue; + + ret = dict_get_str(child_dict, entry, &path); + len = snprintf(tmp, PATH_MAX, "%s ==> BRICK: %s\n path: %s", entry, + child->brick_path, path); + if ((len < 0) || (len >= PATH_MAX)) { + continue; } + snprintf(main_key, PATH_MAX, "quarantine-%d", tmp_count); - ret = tmp_count; + ret = dict_set_dynstr_with_alloc(dict, main_key, tmp); + if (!ret) + tmp_count++; + path = NULL; + } + + ret = tmp_count; out: - return ret; + return ret; } int32_t -br_collect_bad_objects_from_children (xlator_t *this, dict_t *dict) +br_collect_bad_objects_from_children(xlator_t *this, dict_t *dict) { - int32_t ret = -1; - dict_t *child_dict = NULL; - int32_t i = 0; - int32_t total_count = 0; - br_child_t *child = NULL; - br_private_t *priv = NULL; - dict_t *tmp_dict = NULL; - - priv = this->private; - tmp_dict = dict; - - for (i = 0; i < priv->child_count; i++) { - child = &priv->children[i]; - GF_ASSERT (child); - if (!_br_is_child_connected (child)) - continue; - - child_dict = dict_new (); - if (!child_dict) { - gf_msg (this->name, GF_LOG_ERROR, ENOMEM, - BRB_MSG_NO_MEMORY, "failed to allocate dict"); - continue; - } - ret = br_get_bad_objects_from_child (this, child_dict, child); - /* - * Continue asking the remaining children for the list of - * bad objects even though getting the list from one of them - * fails. - */ - if (ret) { - dict_unref (child_dict); - continue; - } - - ret = br_collect_bad_objects_of_child (this, child, tmp_dict, - child_dict, total_count); - if (ret < 0) { - dict_unref (child_dict); - continue; - } - - total_count = ret; - dict_unref (child_dict); - child_dict = NULL; + int32_t ret = -1; + dict_t *child_dict = NULL; + int32_t i = 0; + int32_t total_count = 0; + br_child_t *child = NULL; + br_private_t *priv = NULL; + dict_t *tmp_dict = NULL; + + priv = this->private; + tmp_dict = dict; + + for (i = 0; i < priv->child_count; i++) { + child = &priv->children[i]; + GF_ASSERT(child); + if (!_br_is_child_connected(child)) + continue; + + child_dict = dict_new(); + if (!child_dict) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, + "failed to allocate dict"); + continue; + } + ret = br_get_bad_objects_from_child(this, child_dict, child); + /* + * Continue asking the remaining children for the list of + * bad objects even though getting the list from one of them + * fails. + */ + if (ret) { + dict_unref(child_dict); + continue; } - ret = dict_set_int32 (tmp_dict, "total-count", total_count); + ret = br_collect_bad_objects_of_child(this, child, tmp_dict, child_dict, + total_count); + if (ret < 0) { + dict_unref(child_dict); + continue; + } + + total_count = ret; + dict_unref(child_dict); + child_dict = NULL; + } - return ret; + ret = dict_set_int32(tmp_dict, "total-count", total_count); + + return ret; } int32_t -br_get_bad_objects_list (xlator_t *this, dict_t **dict) +br_get_bad_objects_list(xlator_t *this, dict_t **dict) { - int32_t ret = -1; - dict_t *tmp_dict = NULL; + int32_t ret = -1; + dict_t *tmp_dict = NULL; - GF_VALIDATE_OR_GOTO ("bir-rot-scrubber", this, out); - GF_VALIDATE_OR_GOTO (this->name, dict, out); + GF_VALIDATE_OR_GOTO("bir-rot-scrubber", this, out); + GF_VALIDATE_OR_GOTO(this->name, dict, out); - tmp_dict = *dict; + tmp_dict = *dict; + if (!tmp_dict) { + tmp_dict = dict_new(); if (!tmp_dict) { - tmp_dict = dict_new (); - if (!tmp_dict) { - gf_msg (this->name, GF_LOG_ERROR, ENOMEM, - BRB_MSG_NO_MEMORY, "failed to allocate dict"); - goto out; - } - *dict = tmp_dict; + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, + "failed to allocate dict"); + goto out; } + *dict = tmp_dict; + } - ret = br_collect_bad_objects_from_children (this, tmp_dict); + ret = br_collect_bad_objects_from_children(this, tmp_dict); out: - return ret; + return ret; } static int -wait_for_scrub_to_finish (xlator_t *this) +wait_for_scrub_to_finish(xlator_t *this) { - int ret = -1; - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - scrub_monitor = &priv->scrub_monitor; - - GF_VALIDATE_OR_GOTO ("bit-rot", scrub_monitor, out); - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, - "Waiting for all children to start and finish scrub"); - - pthread_mutex_lock (&scrub_monitor->donelock); - { - while (!scrub_monitor->done) - pthread_cond_wait (&scrub_monitor->donecond, - &scrub_monitor->donelock); - } - pthread_mutex_unlock (&scrub_monitor->donelock); - ret = 0; + int ret = -1; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + scrub_monitor = &priv->scrub_monitor; + + GF_VALIDATE_OR_GOTO("bit-rot", scrub_monitor, out); + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, + "Waiting for all children to start and finish scrub"); + + pthread_mutex_lock(&scrub_monitor->donelock); + { + while (!scrub_monitor->done) + pthread_cond_wait(&scrub_monitor->donecond, + &scrub_monitor->donelock); + } + pthread_mutex_unlock(&scrub_monitor->donelock); + ret = 0; out: - return ret; + return ret; } /** @@ -1897,153 +1929,150 @@ out: * thread that takes care of state machine. */ void * -br_monitor_thread (void *arg) +br_monitor_thread(void *arg) { - int32_t ret = 0; - xlator_t *this = NULL; - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; - - this = arg; - priv = this->private; - - /* - * Since, this is the topmost xlator, THIS has to be set by bit-rot - * xlator itself (STACK_WIND won't help in this case). Also it has - * to be done for each thread that gets spawned. Otherwise, a new - * thread will get global_xlator's pointer when it does "THIS". - */ - THIS = this; - - scrub_monitor = &priv->scrub_monitor; - - pthread_mutex_lock (&scrub_monitor->mutex); - { - while (!scrub_monitor->inited) - pthread_cond_wait (&scrub_monitor->cond, - &scrub_monitor->mutex); - } - pthread_mutex_unlock (&scrub_monitor->mutex); - - /* this needs to be serialized with reconfigure() */ - pthread_mutex_lock (&priv->lock); - { - ret = br_scrub_state_machine (this, _gf_false); - } - pthread_mutex_unlock (&priv->lock); + int32_t ret = 0; + xlator_t *this = NULL; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; + + this = arg; + priv = this->private; + + /* + * Since, this is the topmost xlator, THIS has to be set by bit-rot + * xlator itself (STACK_WIND won't help in this case). Also it has + * to be done for each thread that gets spawned. Otherwise, a new + * thread will get global_xlator's pointer when it does "THIS". + */ + THIS = this; + + scrub_monitor = &priv->scrub_monitor; + + pthread_mutex_lock(&scrub_monitor->mutex); + { + while (!scrub_monitor->inited) + pthread_cond_wait(&scrub_monitor->cond, &scrub_monitor->mutex); + } + pthread_mutex_unlock(&scrub_monitor->mutex); + + /* this needs to be serialized with reconfigure() */ + pthread_mutex_lock(&priv->lock); + { + ret = br_scrub_state_machine(this, _gf_false); + } + pthread_mutex_unlock(&priv->lock); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_SSM_FAILED, + "Scrub state machine failed"); + goto out; + } + + while (1) { + /* Wait for all children to finish scrubbing */ + ret = wait_for_scrub_to_finish(this); if (ret) { - gf_msg (this->name, GF_LOG_ERROR, -ret, - BRB_MSG_SSM_FAILED, - "Scrub state machine failed"); - goto out; + gf_msg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_SCRUB_WAIT_FAILED, + "Scrub wait failed"); + goto out; } - while (1) { - /* Wait for all children to finish scrubbing */ - ret = wait_for_scrub_to_finish (this); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, -ret, - BRB_MSG_SCRUB_WAIT_FAILED, - "Scrub wait failed"); - goto out; - } - - /* scrub exit criteria: Move the state to PENDING */ - br_scrubber_exit_control (this); - } + /* scrub exit criteria: Move the state to PENDING */ + br_scrubber_exit_control(this); + } out: - return NULL; + return NULL; } static void -br_set_scrub_state (struct br_monitor *scrub_monitor, br_scrub_state_t state) +br_set_scrub_state(struct br_monitor *scrub_monitor, br_scrub_state_t state) { - LOCK (&scrub_monitor->lock); - { - _br_monitor_set_scrub_state (scrub_monitor, state); - } - UNLOCK (&scrub_monitor->lock); + LOCK(&scrub_monitor->lock); + { + _br_monitor_set_scrub_state(scrub_monitor, state); + } + UNLOCK(&scrub_monitor->lock); } int32_t -br_scrubber_monitor_init (xlator_t *this, br_private_t *priv) +br_scrubber_monitor_init(xlator_t *this, br_private_t *priv) { - struct br_monitor *scrub_monitor = NULL; - int ret = 0; + struct br_monitor *scrub_monitor = NULL; + int ret = 0; - scrub_monitor = &priv->scrub_monitor; + scrub_monitor = &priv->scrub_monitor; - LOCK_INIT (&scrub_monitor->lock); - scrub_monitor->this = this; + LOCK_INIT(&scrub_monitor->lock); + scrub_monitor->this = this; - scrub_monitor->inited = _gf_false; - pthread_mutex_init (&scrub_monitor->mutex, NULL); - pthread_cond_init (&scrub_monitor->cond, NULL); + scrub_monitor->inited = _gf_false; + pthread_mutex_init(&scrub_monitor->mutex, NULL); + pthread_cond_init(&scrub_monitor->cond, NULL); - scrub_monitor->kick = _gf_false; - scrub_monitor->active_child_count = 0; - pthread_mutex_init (&scrub_monitor->wakelock, NULL); - pthread_cond_init (&scrub_monitor->wakecond, NULL); + scrub_monitor->kick = _gf_false; + scrub_monitor->active_child_count = 0; + pthread_mutex_init(&scrub_monitor->wakelock, NULL); + pthread_cond_init(&scrub_monitor->wakecond, NULL); - scrub_monitor->done = _gf_false; - pthread_mutex_init (&scrub_monitor->donelock, NULL); - pthread_cond_init (&scrub_monitor->donecond, NULL); - - /* Set the state to INACTIVE */ - br_set_scrub_state (&priv->scrub_monitor, BR_SCRUB_STATE_INACTIVE); - - /* Start the monitor thread */ - ret = gf_thread_create (&scrub_monitor->thread, NULL, - br_monitor_thread, this, "brmon"); - if (ret != 0) { - gf_msg (this->name, GF_LOG_ERROR, -ret, - BRB_MSG_SPAWN_FAILED, "monitor thread creation failed"); - ret = -1; - goto err; - } + scrub_monitor->done = _gf_false; + pthread_mutex_init(&scrub_monitor->donelock, NULL); + pthread_cond_init(&scrub_monitor->donecond, NULL); - return 0; + /* Set the state to INACTIVE */ + br_set_scrub_state(&priv->scrub_monitor, BR_SCRUB_STATE_INACTIVE); + + /* Start the monitor thread */ + ret = gf_thread_create(&scrub_monitor->thread, NULL, br_monitor_thread, + this, "brmon"); + if (ret != 0) { + gf_msg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_SPAWN_FAILED, + "monitor thread creation failed"); + ret = -1; + goto err; + } + + return 0; err: - pthread_mutex_destroy (&scrub_monitor->mutex); - pthread_cond_destroy (&scrub_monitor->cond); + pthread_mutex_destroy(&scrub_monitor->mutex); + pthread_cond_destroy(&scrub_monitor->cond); - pthread_mutex_destroy (&scrub_monitor->wakelock); - pthread_cond_destroy (&scrub_monitor->wakecond); + pthread_mutex_destroy(&scrub_monitor->wakelock); + pthread_cond_destroy(&scrub_monitor->wakecond); - pthread_mutex_destroy (&scrub_monitor->donelock); - pthread_cond_destroy (&scrub_monitor->donecond); + pthread_mutex_destroy(&scrub_monitor->donelock); + pthread_cond_destroy(&scrub_monitor->donecond); - LOCK_DESTROY (&scrub_monitor->lock); + LOCK_DESTROY(&scrub_monitor->lock); - return ret; + return ret; } int32_t -br_scrubber_init (xlator_t *this, br_private_t *priv) +br_scrubber_init(xlator_t *this, br_private_t *priv) { - struct br_scrubber *fsscrub = NULL; - int ret = 0; + struct br_scrubber *fsscrub = NULL; + int ret = 0; - priv->tbf = tbf_init (NULL, 0); - if (!priv->tbf) - return -1; + priv->tbf = tbf_init(NULL, 0); + if (!priv->tbf) + return -1; - ret = br_scrubber_monitor_init (this, priv); - if (ret) - return -1; + ret = br_scrubber_monitor_init(this, priv); + if (ret) + return -1; - fsscrub = &priv->fsscrub; + fsscrub = &priv->fsscrub; - fsscrub->this = this; - fsscrub->throttle = BR_SCRUB_THROTTLE_VOID; + fsscrub->this = this; + fsscrub->throttle = BR_SCRUB_THROTTLE_VOID; - pthread_mutex_init (&fsscrub->mutex, NULL); - pthread_cond_init (&fsscrub->cond, NULL); + pthread_mutex_init(&fsscrub->mutex, NULL); + pthread_cond_init(&fsscrub->cond, NULL); - fsscrub->nr_scrubbers = 0; - INIT_LIST_HEAD (&fsscrub->scrubbers); - INIT_LIST_HEAD (&fsscrub->scrublist); + fsscrub->nr_scrubbers = 0; + INIT_LIST_HEAD(&fsscrub->scrubbers); + INIT_LIST_HEAD(&fsscrub->scrublist); - return 0; + return 0; } diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-ssm.c b/xlators/features/bit-rot/src/bitd/bit-rot-ssm.c index af887a1ff8b..753e31a3b23 100644 --- a/xlators/features/bit-rot/src/bitd/bit-rot-ssm.c +++ b/xlators/features/bit-rot/src/bitd/bit-rot-ssm.c @@ -12,78 +12,79 @@ #include "bit-rot-scrub.h" #include "bit-rot-bitd-messages.h" -int br_scrub_ssm_noop (xlator_t *this) +int +br_scrub_ssm_noop(xlator_t *this) { - return 0; + return 0; } int -br_scrub_ssm_state_pause (xlator_t *this) +br_scrub_ssm_state_pause(xlator_t *this) { - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; - priv = this->private; - scrub_monitor = &priv->scrub_monitor; + priv = this->private; + scrub_monitor = &priv->scrub_monitor; - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO, - "Scrubber paused"); - _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_PAUSED); - return 0; + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO, + "Scrubber paused"); + _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PAUSED); + return 0; } int -br_scrub_ssm_state_ipause (xlator_t *this) +br_scrub_ssm_state_ipause(xlator_t *this) { - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; - priv = this->private; - scrub_monitor = &priv->scrub_monitor; + priv = this->private; + scrub_monitor = &priv->scrub_monitor; - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO, - "Scrubber paused"); - _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_IPAUSED); - return 0; + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO, + "Scrubber paused"); + _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_IPAUSED); + return 0; } int -br_scrub_ssm_state_active (xlator_t *this) +br_scrub_ssm_state_active(xlator_t *this) { - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; - priv = this->private; - scrub_monitor = &priv->scrub_monitor; + priv = this->private; + scrub_monitor = &priv->scrub_monitor; - if (scrub_monitor->done) { - (void) br_fsscan_activate (this); - } else { - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO, - "Scrubbing resumed"); - _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_ACTIVE); - } + if (scrub_monitor->done) { + (void)br_fsscan_activate(this); + } else { + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO, + "Scrubbing resumed"); + _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_ACTIVE); + } - return 0; + return 0; } int -br_scrub_ssm_state_stall (xlator_t *this) +br_scrub_ssm_state_stall(xlator_t *this) { - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; - priv = this->private; - scrub_monitor = &priv->scrub_monitor; + priv = this->private; + scrub_monitor = &priv->scrub_monitor; - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO, - "Volume is under active scrubbing. Pausing scrub.."); - _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_STALLED); - return 0; + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO, + "Volume is under active scrubbing. Pausing scrub.."); + _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_STALLED); + return 0; } -static br_scrub_ssm_call * -br_scrub_ssm[BR_SCRUB_MAXSTATES][BR_SCRUB_MAXEVENTS] = { +static br_scrub_ssm_call *br_scrub_ssm[BR_SCRUB_MAXSTATES][BR_SCRUB_MAXEVENTS] = + { /* INACTIVE */ {br_fsscan_schedule, br_scrub_ssm_state_ipause, br_scrub_ssm_noop}, /* PENDING */ @@ -99,25 +100,25 @@ br_scrub_ssm[BR_SCRUB_MAXSTATES][BR_SCRUB_MAXEVENTS] = { }; int32_t -br_scrub_state_machine (xlator_t *this, gf_boolean_t scrub_ondemand) +br_scrub_state_machine(xlator_t *this, gf_boolean_t scrub_ondemand) { - br_private_t *priv = NULL; - br_scrub_ssm_call *call = NULL; - struct br_scrubber *fsscrub = NULL; - br_scrub_state_t currstate = 0; - br_scrub_event_t event = 0; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - fsscrub = &priv->fsscrub; - scrub_monitor = &priv->scrub_monitor; - - currstate = scrub_monitor->state; - if (scrub_ondemand) - event = BR_SCRUB_EVENT_ONDEMAND; - else - event = _br_child_get_scrub_event (fsscrub); - - call = br_scrub_ssm[currstate][event]; - return call (this); + br_private_t *priv = NULL; + br_scrub_ssm_call *call = NULL; + struct br_scrubber *fsscrub = NULL; + br_scrub_state_t currstate = 0; + br_scrub_event_t event = 0; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + fsscrub = &priv->fsscrub; + scrub_monitor = &priv->scrub_monitor; + + currstate = scrub_monitor->state; + if (scrub_ondemand) + event = BR_SCRUB_EVENT_ONDEMAND; + else + event = _br_child_get_scrub_event(fsscrub); + + call = br_scrub_ssm[currstate][event]; + return call(this); } diff --git a/xlators/features/bit-rot/src/bitd/bit-rot.c b/xlators/features/bit-rot/src/bitd/bit-rot.c index cf59232263b..75fa0490c90 100644 --- a/xlators/features/bit-rot/src/bitd/bit-rot.c +++ b/xlators/features/bit-rot/src/bitd/bit-rot.c @@ -21,72 +21,72 @@ #include <pthread.h> #include "bit-rot-bitd-messages.h" -#define BR_HASH_CALC_READ_SIZE (128 * 1024) +#define BR_HASH_CALC_READ_SIZE (128 * 1024) -typedef int32_t (br_child_handler)(xlator_t *, br_child_t *); +typedef int32_t(br_child_handler)(xlator_t *, br_child_t *); struct br_child_event { - xlator_t *this; + xlator_t *this; - br_child_t *child; + br_child_t *child; - br_child_handler *call; + br_child_handler *call; - struct list_head list; + struct list_head list; }; static int -br_find_child_index (xlator_t *this, xlator_t *child) +br_find_child_index(xlator_t *this, xlator_t *child) { - br_private_t *priv = NULL; - int i = -1; - int index = -1; + br_private_t *priv = NULL; + int i = -1; + int index = -1; - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, this->private, out); - GF_VALIDATE_OR_GOTO (this->name, child, out); + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, this->private, out); + GF_VALIDATE_OR_GOTO(this->name, child, out); - priv = this->private; + priv = this->private; - for (i = 0; i < priv->child_count; i++) { - if (child == priv->children[i].xl) { - index = i; - break; - } + for (i = 0; i < priv->child_count; i++) { + if (child == priv->children[i].xl) { + index = i; + break; } + } out: - return index; + return index; } br_child_t * -br_get_child_from_brick_path (xlator_t *this, char *brick_path) +br_get_child_from_brick_path(xlator_t *this, char *brick_path) { - br_private_t *priv = NULL; - br_child_t *child = NULL; - br_child_t *tmp = NULL; - int i = 0; + br_private_t *priv = NULL; + br_child_t *child = NULL; + br_child_t *tmp = NULL; + int i = 0; - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, this->private, out); - GF_VALIDATE_OR_GOTO (this->name, brick_path, out); + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, this->private, out); + GF_VALIDATE_OR_GOTO(this->name, brick_path, out); - priv = this->private; + priv = this->private; - pthread_mutex_lock (&priv->lock); - { - for (i = 0; i < priv->child_count; i++) { - tmp = &priv->children[i]; - if (!strcmp (tmp->brick_path, brick_path)) { - child = tmp; - break; - } - } + pthread_mutex_lock(&priv->lock); + { + for (i = 0; i < priv->child_count; i++) { + tmp = &priv->children[i]; + if (!strcmp(tmp->brick_path, brick_path)) { + child = tmp; + break; + } } - pthread_mutex_unlock (&priv->lock); + } + pthread_mutex_unlock(&priv->lock); out: - return child; + return child; } /** @@ -94,18 +94,18 @@ out: * needed -- later. */ void * -br_brick_init (void *xl, struct gf_brick_spec *brick) +br_brick_init(void *xl, struct gf_brick_spec *brick) { - return brick; + return brick; } /** * and cleanup things here when allocated br_brick_init(). */ void -br_brick_fini (void *xl, char *brick, void *data) +br_brick_fini(void *xl, char *brick, void *data) { - return; + return; } /** @@ -117,109 +117,109 @@ br_brick_fini (void *xl, char *brick, void *data) * change stub to handle this change. */ static br_isignature_t * -br_prepare_signature (const unsigned char *sign, - unsigned long hashlen, - int8_t hashtype, br_object_t *object) +br_prepare_signature(const unsigned char *sign, unsigned long hashlen, + int8_t hashtype, br_object_t *object) { - br_isignature_t *signature = NULL; + br_isignature_t *signature = NULL; - /* TODO: use mem-pool */ - signature = GF_CALLOC (1, signature_size (hashlen + 1), - gf_br_stub_mt_signature_t); - if (!signature) - return NULL; + /* TODO: use mem-pool */ + signature = GF_CALLOC(1, signature_size(hashlen + 1), + gf_br_stub_mt_signature_t); + if (!signature) + return NULL; - /* object version */ - signature->signedversion = object->signedversion; + /* object version */ + signature->signedversion = object->signedversion; - /* signature length & type */ - signature->signaturelen = hashlen; - signature->signaturetype = hashtype; + /* signature length & type */ + signature->signaturelen = hashlen; + signature->signaturetype = hashtype; - /* signature itself */ - memcpy (signature->signature, (char *)sign, hashlen); - signature->signature[hashlen+1] = '\0'; + /* signature itself */ + memcpy(signature->signature, (char *)sign, hashlen); + signature->signature[hashlen + 1] = '\0'; - return signature; + return signature; } gf_boolean_t -bitd_is_bad_file (xlator_t *this, br_child_t *child, loc_t *loc, fd_t *fd) +bitd_is_bad_file(xlator_t *this, br_child_t *child, loc_t *loc, fd_t *fd) { - int32_t ret = -1; - dict_t *xattr = NULL; - inode_t *inode = NULL; - gf_boolean_t bad_file = _gf_false; + int32_t ret = -1; + dict_t *xattr = NULL; + inode_t *inode = NULL; + gf_boolean_t bad_file = _gf_false; - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); + GF_VALIDATE_OR_GOTO("bit-rot", this, out); - inode = (loc) ? loc->inode : fd->inode; + inode = (loc) ? loc->inode : fd->inode; - if (fd) - ret = syncop_fgetxattr (child->xl, fd, &xattr, - BITROT_OBJECT_BAD_KEY, NULL, NULL); - else if (loc) - ret = syncop_getxattr (child->xl, loc, - &xattr, BITROT_OBJECT_BAD_KEY, NULL, - NULL); + if (fd) + ret = syncop_fgetxattr(child->xl, fd, &xattr, BITROT_OBJECT_BAD_KEY, + NULL, NULL); + else if (loc) + ret = syncop_getxattr(child->xl, loc, &xattr, BITROT_OBJECT_BAD_KEY, + NULL, NULL); - if (!ret) { - gf_msg_debug (this->name, 0, "[GFID: %s] is marked corrupted", - uuid_utoa (inode->gfid)); - bad_file = _gf_true; - } + if (!ret) { + gf_msg_debug(this->name, 0, "[GFID: %s] is marked corrupted", + uuid_utoa(inode->gfid)); + bad_file = _gf_true; + } - if (xattr) - dict_unref (xattr); + if (xattr) + dict_unref(xattr); out: - return bad_file; + return bad_file; } /** * Do a lookup on the gfid present within the object. */ static int32_t -br_object_lookup (xlator_t *this, br_object_t *object, - struct iatt *iatt, inode_t **linked_inode) +br_object_lookup(xlator_t *this, br_object_t *object, struct iatt *iatt, + inode_t **linked_inode) { - int ret = -EINVAL; - loc_t loc = {0, }; - inode_t *inode = NULL; - - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, object, out); - - inode = inode_find (object->child->table, object->gfid); - - if (inode) - loc.inode = inode; - else - loc.inode = inode_new (object->child->table); - - if (!loc.inode) { - ret = -ENOMEM; - goto out; - } - - gf_uuid_copy (loc.gfid, object->gfid); - - ret = syncop_lookup (object->child->xl, &loc, iatt, NULL, NULL, NULL); - if (ret < 0) - goto out; - - /* - * The file might have been deleted by the application - * after getting the event, but before doing a lookup. - * So use linked_inode after inode_link is done. - */ - *linked_inode = inode_link (loc.inode, NULL, NULL, iatt); - if (*linked_inode) - inode_lookup (*linked_inode); + int ret = -EINVAL; + loc_t loc = { + 0, + }; + inode_t *inode = NULL; + + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, object, out); + + inode = inode_find(object->child->table, object->gfid); + + if (inode) + loc.inode = inode; + else + loc.inode = inode_new(object->child->table); + + if (!loc.inode) { + ret = -ENOMEM; + goto out; + } + + gf_uuid_copy(loc.gfid, object->gfid); + + ret = syncop_lookup(object->child->xl, &loc, iatt, NULL, NULL, NULL); + if (ret < 0) + goto out; + + /* + * The file might have been deleted by the application + * after getting the event, but before doing a lookup. + * So use linked_inode after inode_link is done. + */ + *linked_inode = inode_link(loc.inode, NULL, NULL, iatt); + if (*linked_inode) + inode_lookup(*linked_inode); out: - loc_wipe (&loc); - return ret; + loc_wipe(&loc); + return ret; } /** @@ -228,43 +228,44 @@ out: * passing xdata -- may be use frame->root->pid itself. */ static int32_t -br_object_open (xlator_t *this, - br_object_t *object, inode_t *inode, fd_t **openfd) +br_object_open(xlator_t *this, br_object_t *object, inode_t *inode, + fd_t **openfd) { - int32_t ret = -1; - fd_t *fd = NULL; - loc_t loc = {0, }; - - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, object, out); - GF_VALIDATE_OR_GOTO (this->name, inode, out); - - ret = -EINVAL; - fd = fd_create (inode, 0); - if (!fd) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED, - "failed to create fd for the inode %s", - uuid_utoa (inode->gfid)); - goto out; - } - - loc.inode = inode_ref (inode); - gf_uuid_copy (loc.gfid, inode->gfid); - - ret = syncop_open (object->child->xl, &loc, O_RDONLY, fd, NULL, NULL); - if (ret) { - br_log_object (this, "open", inode->gfid, -ret); - fd_unref (fd); - fd = NULL; - } else { - fd_bind (fd); - *openfd = fd; - } - - loc_wipe (&loc); + int32_t ret = -1; + fd_t *fd = NULL; + loc_t loc = { + 0, + }; + + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, object, out); + GF_VALIDATE_OR_GOTO(this->name, inode, out); + + ret = -EINVAL; + fd = fd_create(inode, 0); + if (!fd) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED, + "failed to create fd for the inode %s", uuid_utoa(inode->gfid)); + goto out; + } + + loc.inode = inode_ref(inode); + gf_uuid_copy(loc.gfid, inode->gfid); + + ret = syncop_open(object->child->xl, &loc, O_RDONLY, fd, NULL, NULL); + if (ret) { + br_log_object(this, "open", inode->gfid, -ret); + fd_unref(fd); + fd = NULL; + } else { + fd_bind(fd); + *openfd = fd; + } + + loc_wipe(&loc); out: - return ret; + return ret; } /** @@ -272,287 +273,289 @@ out: * and return the buffer. */ static int32_t -br_object_read_block_and_sign (xlator_t *this, fd_t *fd, br_child_t *child, - off_t offset, size_t size, SHA256_CTX *sha256) +br_object_read_block_and_sign(xlator_t *this, fd_t *fd, br_child_t *child, + off_t offset, size_t size, SHA256_CTX *sha256) { - int32_t ret = -1; - tbf_t *tbf = NULL; - struct iovec *iovec = NULL; - struct iobref *iobref = NULL; - br_private_t *priv = NULL; - int count = 0; - int i = 0; - - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, fd, out); - GF_VALIDATE_OR_GOTO (this->name, fd->inode, out); - GF_VALIDATE_OR_GOTO (this->name, child, out); - GF_VALIDATE_OR_GOTO (this->name, this->private, out); - - priv = this->private; - - GF_VALIDATE_OR_GOTO (this->name, priv->tbf, out); - tbf = priv->tbf; - - ret = syncop_readv (child->xl, fd, - size, offset, 0, &iovec, &count, &iobref, NULL, - NULL, NULL); - - if (ret < 0) { - gf_msg (this->name, GF_LOG_ERROR, errno, BRB_MSG_READV_FAILED, - "readv on %s failed", uuid_utoa (fd->inode->gfid)); - ret = -1; - goto out; - } + int32_t ret = -1; + tbf_t *tbf = NULL; + struct iovec *iovec = NULL; + struct iobref *iobref = NULL; + br_private_t *priv = NULL; + int count = 0; + int i = 0; + + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, fd, out); + GF_VALIDATE_OR_GOTO(this->name, fd->inode, out); + GF_VALIDATE_OR_GOTO(this->name, child, out); + GF_VALIDATE_OR_GOTO(this->name, this->private, out); + + priv = this->private; + + GF_VALIDATE_OR_GOTO(this->name, priv->tbf, out); + tbf = priv->tbf; + + ret = syncop_readv(child->xl, fd, size, offset, 0, &iovec, &count, &iobref, + NULL, NULL, NULL); + + if (ret < 0) { + gf_msg(this->name, GF_LOG_ERROR, errno, BRB_MSG_READV_FAILED, + "readv on %s failed", uuid_utoa(fd->inode->gfid)); + ret = -1; + goto out; + } - if (ret == 0) - goto out; + if (ret == 0) + goto out; - for (i = 0; i < count; i++) { - TBF_THROTTLE_BEGIN (tbf, TBF_OP_HASH, iovec[i].iov_len); - { - SHA256_Update (sha256, (const unsigned char *) - (iovec[i].iov_base), iovec[i].iov_len); - } - TBF_THROTTLE_BEGIN (tbf, TBF_OP_HASH, iovec[i].iov_len); + for (i = 0; i < count; i++) { + TBF_THROTTLE_BEGIN(tbf, TBF_OP_HASH, iovec[i].iov_len); + { + SHA256_Update(sha256, (const unsigned char *)(iovec[i].iov_base), + iovec[i].iov_len); } + TBF_THROTTLE_BEGIN(tbf, TBF_OP_HASH, iovec[i].iov_len); + } - out: - if (iovec) - GF_FREE (iovec); +out: + if (iovec) + GF_FREE(iovec); - if (iobref) - iobref_unref (iobref); + if (iobref) + iobref_unref(iobref); - return ret; + return ret; } int32_t -br_calculate_obj_checksum (unsigned char *md, - br_child_t *child, fd_t *fd, struct iatt *iatt) +br_calculate_obj_checksum(unsigned char *md, br_child_t *child, fd_t *fd, + struct iatt *iatt) { - int32_t ret = -1; - off_t offset = 0; - size_t block = BR_HASH_CALC_READ_SIZE; - xlator_t *this = NULL; - - SHA256_CTX sha256; + int32_t ret = -1; + off_t offset = 0; + size_t block = BR_HASH_CALC_READ_SIZE; + xlator_t *this = NULL; - GF_VALIDATE_OR_GOTO ("bit-rot", child, out); - GF_VALIDATE_OR_GOTO ("bit-rot", iatt, out); - GF_VALIDATE_OR_GOTO ("bit-rot", fd, out); + SHA256_CTX sha256; - this = child->this; + GF_VALIDATE_OR_GOTO("bit-rot", child, out); + GF_VALIDATE_OR_GOTO("bit-rot", iatt, out); + GF_VALIDATE_OR_GOTO("bit-rot", fd, out); - SHA256_Init (&sha256); + this = child->this; - while (1) { - ret = br_object_read_block_and_sign (this, fd, child, - offset, block, &sha256); - if (ret < 0) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_BLOCK_READ_FAILED, "reading block with " - "offset %lu of object %s failed", offset, - uuid_utoa (fd->inode->gfid)); - break; - } + SHA256_Init(&sha256); - if (ret == 0) - break; - - offset += ret; + while (1) { + ret = br_object_read_block_and_sign(this, fd, child, offset, block, + &sha256); + if (ret < 0) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_BLOCK_READ_FAILED, + "reading block with " + "offset %lu of object %s failed", + offset, uuid_utoa(fd->inode->gfid)); + break; } if (ret == 0) - SHA256_Final (md, &sha256); + break; - out: - return ret; + offset += ret; + } + + if (ret == 0) + SHA256_Final(md, &sha256); + +out: + return ret; } static int32_t -br_object_checksum (unsigned char *md, - br_object_t *object, fd_t *fd, struct iatt *iatt) +br_object_checksum(unsigned char *md, br_object_t *object, fd_t *fd, + struct iatt *iatt) { - return br_calculate_obj_checksum (md, object->child, fd, iatt); + return br_calculate_obj_checksum(md, object->child, fd, iatt); } static int32_t -br_object_read_sign (inode_t *linked_inode, fd_t *fd, br_object_t *object, - struct iatt *iatt) +br_object_read_sign(inode_t *linked_inode, fd_t *fd, br_object_t *object, + struct iatt *iatt) { - int32_t ret = -1; - xlator_t *this = NULL; - dict_t *xattr = NULL; - unsigned char *md = NULL; - br_isignature_t *sign = NULL; - - GF_VALIDATE_OR_GOTO ("bit-rot", object, out); - GF_VALIDATE_OR_GOTO ("bit-rot", linked_inode, out); - GF_VALIDATE_OR_GOTO ("bit-rot", fd, out); - - this = object->this; - - md = GF_MALLOC (SHA256_DIGEST_LENGTH, gf_common_mt_char); - if (!md) { - gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, - "failed to allocate memory for saving hash of the " - "object %s", uuid_utoa (fd->inode->gfid)); - goto out; - } - - ret = br_object_checksum (md, object, fd, iatt); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_CALC_CHECKSUM_FAILED, "calculating checksum " - "for the object %s failed", - uuid_utoa (linked_inode->gfid)); - goto free_signature; - } - - sign = br_prepare_signature (md, SHA256_DIGEST_LENGTH, - BR_SIGNATURE_TYPE_SHA256, object); - if (!sign) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED, - "failed to get the signature for the object %s", - uuid_utoa (fd->inode->gfid)); - goto free_signature; - } - - xattr = dict_for_key_value - (GLUSTERFS_SET_OBJECT_SIGNATURE, - (void *)sign, signature_size (SHA256_DIGEST_LENGTH), _gf_true); - - if (!xattr) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED, - "dict allocation for signing failed for the object %s", - uuid_utoa (fd->inode->gfid)); - goto free_isign; - } - - ret = syncop_fsetxattr (object->child->xl, fd, xattr, 0, NULL, NULL); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED, - "fsetxattr of signature to the object %s failed", - uuid_utoa (fd->inode->gfid)); - goto unref_dict; - } - - ret = 0; - - unref_dict: - dict_unref (xattr); - free_isign: - GF_FREE (sign); - free_signature: - GF_FREE (md); - out: - return ret; + int32_t ret = -1; + xlator_t *this = NULL; + dict_t *xattr = NULL; + unsigned char *md = NULL; + br_isignature_t *sign = NULL; + + GF_VALIDATE_OR_GOTO("bit-rot", object, out); + GF_VALIDATE_OR_GOTO("bit-rot", linked_inode, out); + GF_VALIDATE_OR_GOTO("bit-rot", fd, out); + + this = object->this; + + md = GF_MALLOC(SHA256_DIGEST_LENGTH, gf_common_mt_char); + if (!md) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, + "failed to allocate memory for saving hash of the " + "object %s", + uuid_utoa(fd->inode->gfid)); + goto out; + } + + ret = br_object_checksum(md, object, fd, iatt); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_CALC_CHECKSUM_FAILED, + "calculating checksum " + "for the object %s failed", + uuid_utoa(linked_inode->gfid)); + goto free_signature; + } + + sign = br_prepare_signature(md, SHA256_DIGEST_LENGTH, + BR_SIGNATURE_TYPE_SHA256, object); + if (!sign) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED, + "failed to get the signature for the object %s", + uuid_utoa(fd->inode->gfid)); + goto free_signature; + } + + xattr = dict_for_key_value(GLUSTERFS_SET_OBJECT_SIGNATURE, (void *)sign, + signature_size(SHA256_DIGEST_LENGTH), _gf_true); + + if (!xattr) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED, + "dict allocation for signing failed for the object %s", + uuid_utoa(fd->inode->gfid)); + goto free_isign; + } + + ret = syncop_fsetxattr(object->child->xl, fd, xattr, 0, NULL, NULL); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED, + "fsetxattr of signature to the object %s failed", + uuid_utoa(fd->inode->gfid)); + goto unref_dict; + } + + ret = 0; + +unref_dict: + dict_unref(xattr); +free_isign: + GF_FREE(sign); +free_signature: + GF_FREE(md); +out: + return ret; } -static int br_object_sign_softerror (int32_t op_errno) +static int +br_object_sign_softerror(int32_t op_errno) { - return ((op_errno == ENOENT) || (op_errno == ESTALE) - || (op_errno == ENODATA)); + return ((op_errno == ENOENT) || (op_errno == ESTALE) || + (op_errno == ENODATA)); } void -br_log_object (xlator_t *this, char *op, uuid_t gfid, int32_t op_errno) +br_log_object(xlator_t *this, char *op, uuid_t gfid, int32_t op_errno) { - int softerror = br_object_sign_softerror (op_errno); - if (softerror) { - gf_msg_debug (this->name, 0, "%s() failed on object %s " - "[reason: %s]", op, uuid_utoa (gfid), - strerror (op_errno)); - } else { - gf_msg (this->name, GF_LOG_ERROR, op_errno, BRB_MSG_OP_FAILED, - "%s() failed on object %s", op, uuid_utoa (gfid)); - } + int softerror = br_object_sign_softerror(op_errno); + if (softerror) { + gf_msg_debug(this->name, 0, + "%s() failed on object %s " + "[reason: %s]", + op, uuid_utoa(gfid), strerror(op_errno)); + } else { + gf_msg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_OP_FAILED, + "%s() failed on object %s", op, uuid_utoa(gfid)); + } } void -br_log_object_path (xlator_t *this, char *op, - const char *path, int32_t op_errno) +br_log_object_path(xlator_t *this, char *op, const char *path, int32_t op_errno) { - int softerror = br_object_sign_softerror (op_errno); - if (softerror) { - gf_msg_debug (this->name, 0, "%s() failed on object %s " - "[reason: %s]", op, path, strerror (op_errno)); - } else { - gf_msg (this->name, GF_LOG_ERROR, op_errno, BRB_MSG_OP_FAILED, - "%s() failed on object %s", op, path); - } + int softerror = br_object_sign_softerror(op_errno); + if (softerror) { + gf_msg_debug(this->name, 0, + "%s() failed on object %s " + "[reason: %s]", + op, path, strerror(op_errno)); + } else { + gf_msg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_OP_FAILED, + "%s() failed on object %s", op, path); + } } static void -br_trigger_sign (xlator_t *this, br_child_t *child, - inode_t *linked_inode, loc_t *loc, gf_boolean_t need_reopen) +br_trigger_sign(xlator_t *this, br_child_t *child, inode_t *linked_inode, + loc_t *loc, gf_boolean_t need_reopen) { - fd_t *fd = NULL; - int32_t ret = -1; - uint32_t val = 0; - dict_t *dict = NULL; - pid_t pid = GF_CLIENT_PID_BITD; - - syncopctx_setfspid (&pid); - - val = (need_reopen == _gf_true) ? BR_OBJECT_REOPEN : BR_OBJECT_RESIGN; - - dict = dict_new (); - if (!dict) - goto out; - - ret = dict_set_uint32 (dict, BR_REOPEN_SIGN_HINT_KEY, val); - if (ret) - goto cleanup_dict; - - ret = -1; - fd = fd_create (linked_inode, 0); - if (!fd) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED, - "Failed to create fd [GFID %s]", - uuid_utoa (linked_inode->gfid)); - goto cleanup_dict; - } - - ret = syncop_open (child->xl, loc, O_RDWR, fd, NULL, NULL); - if (ret) { - br_log_object (this, "open", linked_inode->gfid, -ret); - goto unref_fd; - } - - fd_bind (fd); - - ret = syncop_fsetxattr (child->xl, fd, dict, 0, NULL, NULL); - if (ret) - br_log_object (this, "fsetxattr", linked_inode->gfid, -ret); - - /* passthough: fd_unref() */ - - unref_fd: - fd_unref (fd); - cleanup_dict: - dict_unref (dict); - out: - if (ret) { - gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_TRIGGER_SIGN, - "Could not trigger signingd for %s (reopen hint: %d)", - uuid_utoa (linked_inode->gfid), val); - } + fd_t *fd = NULL; + int32_t ret = -1; + uint32_t val = 0; + dict_t *dict = NULL; + pid_t pid = GF_CLIENT_PID_BITD; + + syncopctx_setfspid(&pid); + + val = (need_reopen == _gf_true) ? BR_OBJECT_REOPEN : BR_OBJECT_RESIGN; + + dict = dict_new(); + if (!dict) + goto out; + + ret = dict_set_uint32(dict, BR_REOPEN_SIGN_HINT_KEY, val); + if (ret) + goto cleanup_dict; + + ret = -1; + fd = fd_create(linked_inode, 0); + if (!fd) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED, + "Failed to create fd [GFID %s]", uuid_utoa(linked_inode->gfid)); + goto cleanup_dict; + } + + ret = syncop_open(child->xl, loc, O_RDWR, fd, NULL, NULL); + if (ret) { + br_log_object(this, "open", linked_inode->gfid, -ret); + goto unref_fd; + } + + fd_bind(fd); + + ret = syncop_fsetxattr(child->xl, fd, dict, 0, NULL, NULL); + if (ret) + br_log_object(this, "fsetxattr", linked_inode->gfid, -ret); + + /* passthough: fd_unref() */ + +unref_fd: + fd_unref(fd); +cleanup_dict: + dict_unref(dict); +out: + if (ret) { + gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_TRIGGER_SIGN, + "Could not trigger signingd for %s (reopen hint: %d)", + uuid_utoa(linked_inode->gfid), val); + } } static void -br_object_resign (xlator_t *this, - br_object_t *object, inode_t *linked_inode) +br_object_resign(xlator_t *this, br_object_t *object, inode_t *linked_inode) { - loc_t loc = {0, }; + loc_t loc = { + 0, + }; - loc.inode = inode_ref (linked_inode); - gf_uuid_copy (loc.gfid, linked_inode->gfid); + loc.inode = inode_ref(linked_inode); + gf_uuid_copy(loc.gfid, linked_inode->gfid); - br_trigger_sign (this, object->child, linked_inode, &loc, _gf_false); + br_trigger_sign(this, object->child, linked_inode, &loc, _gf_false); - loc_wipe (&loc); + loc_wipe(&loc); } /** @@ -560,122 +563,125 @@ br_object_resign (xlator_t *this, * some form of priority scheduling and/or read burstness to avoid starving * (or kicking) client I/O's. */ -static int32_t br_sign_object (br_object_t *object) +static int32_t +br_sign_object(br_object_t *object) { - int32_t ret = -1; - inode_t *linked_inode = NULL; - xlator_t *this = NULL; - fd_t *fd = NULL; - struct iatt iatt = {0, }; - pid_t pid = GF_CLIENT_PID_BITD; - br_sign_state_t sign_info = BR_SIGN_NORMAL; - - GF_VALIDATE_OR_GOTO ("bit-rot", object, out); - - this = object->this; - - /** - * FIXME: This is required as signing an object is restricted to - * clients with special frame->root->pid. Change the way client - * pid is set. - */ - syncopctx_setfspid (&pid); - - ret = br_object_lookup (this, object, &iatt, &linked_inode); - if (ret) { - br_log_object (this, "lookup", object->gfid, -ret); - goto out; - } - - /** - * For fd's that have notified for reopening, we send an explicit - * open() followed by a dummy write() call. This triggers the - * actual signing of the object. - */ - sign_info = ntohl (object->sign_info); - if (sign_info == BR_SIGN_REOPEN_WAIT) { - br_object_resign (this, object, linked_inode); - goto unref_inode; - } - - ret = br_object_open (this, object, linked_inode, &fd); - if (!fd) { - br_log_object (this, "open", object->gfid, -ret); - goto unref_inode; - } - - /** - * we have an open file descriptor on the object. from here on, - * do not be generous to file operation errors. - */ - gf_msg_debug (this->name, 0, "Signing object [%s]", - uuid_utoa (linked_inode->gfid)); - - ret = br_object_read_sign (linked_inode, fd, object, &iatt); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_READ_AND_SIGN_FAILED, "reading and signing of " - "the object %s failed", uuid_utoa (linked_inode->gfid)); - goto unref_fd; - } - - ret = 0; - - unref_fd: - fd_unref (fd); - unref_inode: - inode_unref (linked_inode); - out: - return ret; + int32_t ret = -1; + inode_t *linked_inode = NULL; + xlator_t *this = NULL; + fd_t *fd = NULL; + struct iatt iatt = { + 0, + }; + pid_t pid = GF_CLIENT_PID_BITD; + br_sign_state_t sign_info = BR_SIGN_NORMAL; + + GF_VALIDATE_OR_GOTO("bit-rot", object, out); + + this = object->this; + + /** + * FIXME: This is required as signing an object is restricted to + * clients with special frame->root->pid. Change the way client + * pid is set. + */ + syncopctx_setfspid(&pid); + + ret = br_object_lookup(this, object, &iatt, &linked_inode); + if (ret) { + br_log_object(this, "lookup", object->gfid, -ret); + goto out; + } + + /** + * For fd's that have notified for reopening, we send an explicit + * open() followed by a dummy write() call. This triggers the + * actual signing of the object. + */ + sign_info = ntohl(object->sign_info); + if (sign_info == BR_SIGN_REOPEN_WAIT) { + br_object_resign(this, object, linked_inode); + goto unref_inode; + } + + ret = br_object_open(this, object, linked_inode, &fd); + if (!fd) { + br_log_object(this, "open", object->gfid, -ret); + goto unref_inode; + } + + /** + * we have an open file descriptor on the object. from here on, + * do not be generous to file operation errors. + */ + gf_msg_debug(this->name, 0, "Signing object [%s]", + uuid_utoa(linked_inode->gfid)); + + ret = br_object_read_sign(linked_inode, fd, object, &iatt); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_READ_AND_SIGN_FAILED, + "reading and signing of " + "the object %s failed", + uuid_utoa(linked_inode->gfid)); + goto unref_fd; + } + + ret = 0; + +unref_fd: + fd_unref(fd); +unref_inode: + inode_unref(linked_inode); +out: + return ret; } -static br_object_t *__br_pick_object (br_private_t *priv) +static br_object_t * +__br_pick_object(br_private_t *priv) { - br_object_t *object = NULL; + br_object_t *object = NULL; - while (list_empty (&priv->obj_queue->objects)) { - pthread_cond_wait (&priv->object_cond, &priv->lock); - } + while (list_empty(&priv->obj_queue->objects)) { + pthread_cond_wait(&priv->object_cond, &priv->lock); + } - object = list_first_entry - (&priv->obj_queue->objects, br_object_t, list); - list_del_init (&object->list); + object = list_first_entry(&priv->obj_queue->objects, br_object_t, list); + list_del_init(&object->list); - return object; + return object; } /** * This is the place where the signing of the objects is triggered. */ void * -br_process_object (void *arg) +br_process_object(void *arg) { - xlator_t *this = NULL; - br_object_t *object = NULL; - br_private_t *priv = NULL; - int32_t ret = -1; - - this = arg; - priv = this->private; - - THIS = this; - - for (;;) { - pthread_mutex_lock (&priv->lock); - { - object = __br_pick_object (priv); - } - pthread_mutex_unlock (&priv->lock); - - ret = br_sign_object (object); - if (ret && !br_object_sign_softerror (-ret)) - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_SIGN_FAILED, "SIGNING FAILURE [%s]", - uuid_utoa (object->gfid)); - GF_FREE (object); + xlator_t *this = NULL; + br_object_t *object = NULL; + br_private_t *priv = NULL; + int32_t ret = -1; + + this = arg; + priv = this->private; + + THIS = this; + + for (;;) { + pthread_mutex_lock(&priv->lock); + { + object = __br_pick_object(priv); } + pthread_mutex_unlock(&priv->lock); - return NULL; + ret = br_sign_object(object); + if (ret && !br_object_sign_softerror(-ret)) + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SIGN_FAILED, + "SIGNING FAILURE [%s]", uuid_utoa(object->gfid)); + GF_FREE(object); + } + + return NULL; } /** @@ -692,98 +698,98 @@ br_process_object (void *arg) * NOTE: use call_time to instrument signing time in br_sign_object(). */ void -br_add_object_to_queue (struct gf_tw_timer_list *timer, - void *data, unsigned long call_time) +br_add_object_to_queue(struct gf_tw_timer_list *timer, void *data, + unsigned long call_time) { - br_object_t *object = NULL; - xlator_t *this = NULL; - br_private_t *priv = NULL; - - object = data; - this = object->this; - priv = this->private; - - THIS = this; - - pthread_mutex_lock (&priv->lock); - { - list_add_tail (&object->list, &priv->obj_queue->objects); - pthread_cond_broadcast (&priv->object_cond); - } - pthread_mutex_unlock (&priv->lock); - - if (timer) - mem_put (timer); - return; + br_object_t *object = NULL; + xlator_t *this = NULL; + br_private_t *priv = NULL; + + object = data; + this = object->this; + priv = this->private; + + THIS = this; + + pthread_mutex_lock(&priv->lock); + { + list_add_tail(&object->list, &priv->obj_queue->objects); + pthread_cond_broadcast(&priv->object_cond); + } + pthread_mutex_unlock(&priv->lock); + + if (timer) + mem_put(timer); + return; } static br_object_t * -br_initialize_object (xlator_t *this, br_child_t *child, changelog_event_t *ev) +br_initialize_object(xlator_t *this, br_child_t *child, changelog_event_t *ev) { - br_object_t *object = NULL; + br_object_t *object = NULL; - object = GF_CALLOC (1, sizeof (*object), gf_br_mt_br_object_t); - if (!object) - goto out; - INIT_LIST_HEAD (&object->list); + object = GF_CALLOC(1, sizeof(*object), gf_br_mt_br_object_t); + if (!object) + goto out; + INIT_LIST_HEAD(&object->list); - object->this = this; - object->child = child; - gf_uuid_copy (object->gfid, ev->u.releasebr.gfid); + object->this = this; + object->child = child; + gf_uuid_copy(object->gfid, ev->u.releasebr.gfid); - /* NOTE: it's BE, but no worry */ - object->signedversion = ev->u.releasebr.version; - object->sign_info = ev->u.releasebr.sign_info; + /* NOTE: it's BE, but no worry */ + object->signedversion = ev->u.releasebr.version; + object->sign_info = ev->u.releasebr.sign_info; out: - return object; + return object; } static struct gf_tw_timer_list * -br_initialize_timer (xlator_t *this, br_object_t *object, br_child_t *child, - changelog_event_t *ev) +br_initialize_timer(xlator_t *this, br_object_t *object, br_child_t *child, + changelog_event_t *ev) { - br_private_t *priv = NULL; - struct gf_tw_timer_list *timer = NULL; + br_private_t *priv = NULL; + struct gf_tw_timer_list *timer = NULL; - priv = this->private; + priv = this->private; - timer = mem_get0 (child->timer_pool); - if (!timer) - goto out; - INIT_LIST_HEAD (&timer->entry); + timer = mem_get0(child->timer_pool); + if (!timer) + goto out; + INIT_LIST_HEAD(&timer->entry); - timer->expires = priv->expiry_time; - if (!timer->expires) - timer->expires = 1; + timer->expires = priv->expiry_time; + if (!timer->expires) + timer->expires = 1; - timer->data = object; - timer->function = br_add_object_to_queue; - gf_tw_add_timer (priv->timer_wheel, timer); + timer->data = object; + timer->function = br_add_object_to_queue; + gf_tw_add_timer(priv->timer_wheel, timer); out: - return timer; + return timer; } static int32_t -br_schedule_object_reopen (xlator_t *this, br_object_t *object, - br_child_t *child, changelog_event_t *ev) +br_schedule_object_reopen(xlator_t *this, br_object_t *object, + br_child_t *child, changelog_event_t *ev) { - struct gf_tw_timer_list *timer = NULL; - - timer = br_initialize_timer (this, object, child, ev); - if (!timer) - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_TIMER_FAILED, - "Failed to allocate object expiry timer [GFID: %s]", - uuid_utoa (object->gfid)); - return timer ? 0 : -1; + struct gf_tw_timer_list *timer = NULL; + + timer = br_initialize_timer(this, object, child, ev); + if (!timer) + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_TIMER_FAILED, + "Failed to allocate object expiry timer [GFID: %s]", + uuid_utoa(object->gfid)); + return timer ? 0 : -1; } static int32_t -br_object_quicksign (xlator_t *this, br_object_t *object) +br_object_quicksign(xlator_t *this, br_object_t *object) { - br_add_object_to_queue (NULL, object, 0ULL); - return 0; + br_add_object_to_queue(NULL, object, 0ULL); + return 0; } /** @@ -796,148 +802,146 @@ br_object_quicksign (xlator_t *this, br_object_t *object) * object as a single alloc and bifurcate their respective pointers. */ void -br_brick_callback (void *xl, char *brick, - void *data, changelog_event_t *ev) +br_brick_callback(void *xl, char *brick, void *data, changelog_event_t *ev) { - int32_t ret = 0; - uuid_t gfid = {0,}; - xlator_t *this = NULL; - br_object_t *object = NULL; - br_child_t *child = NULL; - br_sign_state_t sign_info = BR_SIGN_INVALID; - - this = xl; - - GF_VALIDATE_OR_GOTO (this->name, ev, out); - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, this->private, out); - - GF_ASSERT (ev->ev_type == CHANGELOG_OP_TYPE_BR_RELEASE); - GF_ASSERT (!gf_uuid_is_null (ev->u.releasebr.gfid)); - - gf_uuid_copy (gfid, ev->u.releasebr.gfid); - - gf_msg_debug (this->name, 0, "RELEASE EVENT [GFID %s]", - uuid_utoa (gfid)); - - child = br_get_child_from_brick_path (this, brick); - if (!child) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SUBVOL_FAILED, - "failed to get the subvolume for the brick %s", brick); - goto out; - } - - object = br_initialize_object (this, child, ev); - if (!object) { - gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, - "failed to allocate object memory [GFID: %s]", - uuid_utoa (gfid)); - goto out; - } - - /* sanity check */ - sign_info = ntohl (object->sign_info); - GF_ASSERT (sign_info != BR_SIGN_NORMAL); - - if (sign_info == BR_SIGN_REOPEN_WAIT) - ret = br_schedule_object_reopen (this, object, child, ev); - else - ret = br_object_quicksign (this, object); - - if (ret) - goto free_object; - - gf_msg_debug (this->name, 0, "->callback: brick [%s], type [%d]\n", - brick, ev->ev_type); - return; - - free_object: - GF_FREE (object); - out: - return; + int32_t ret = 0; + uuid_t gfid = { + 0, + }; + xlator_t *this = NULL; + br_object_t *object = NULL; + br_child_t *child = NULL; + br_sign_state_t sign_info = BR_SIGN_INVALID; + + this = xl; + + GF_VALIDATE_OR_GOTO(this->name, ev, out); + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, this->private, out); + + GF_ASSERT(ev->ev_type == CHANGELOG_OP_TYPE_BR_RELEASE); + GF_ASSERT(!gf_uuid_is_null(ev->u.releasebr.gfid)); + + gf_uuid_copy(gfid, ev->u.releasebr.gfid); + + gf_msg_debug(this->name, 0, "RELEASE EVENT [GFID %s]", uuid_utoa(gfid)); + + child = br_get_child_from_brick_path(this, brick); + if (!child) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SUBVOL_FAILED, + "failed to get the subvolume for the brick %s", brick); + goto out; + } + + object = br_initialize_object(this, child, ev); + if (!object) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, + "failed to allocate object memory [GFID: %s]", uuid_utoa(gfid)); + goto out; + } + + /* sanity check */ + sign_info = ntohl(object->sign_info); + GF_ASSERT(sign_info != BR_SIGN_NORMAL); + + if (sign_info == BR_SIGN_REOPEN_WAIT) + ret = br_schedule_object_reopen(this, object, child, ev); + else + ret = br_object_quicksign(this, object); + + if (ret) + goto free_object; + + gf_msg_debug(this->name, 0, "->callback: brick [%s], type [%d]\n", brick, + ev->ev_type); + return; + +free_object: + GF_FREE(object); +out: + return; } void -br_fill_brick_spec (struct gf_brick_spec *brick, char *path) +br_fill_brick_spec(struct gf_brick_spec *brick, char *path) { - brick->brick_path = gf_strdup (path); - brick->filter = CHANGELOG_OP_TYPE_BR_RELEASE; - - brick->init = br_brick_init; - brick->fini = br_brick_fini; - brick->callback = br_brick_callback; - brick->connected = NULL; - brick->disconnected = NULL; + brick->brick_path = gf_strdup(path); + brick->filter = CHANGELOG_OP_TYPE_BR_RELEASE; + + brick->init = br_brick_init; + brick->fini = br_brick_fini; + brick->callback = br_brick_callback; + brick->connected = NULL; + brick->disconnected = NULL; } static gf_boolean_t -br_check_object_need_sign (xlator_t *this, dict_t *xattr, br_child_t *child) +br_check_object_need_sign(xlator_t *this, dict_t *xattr, br_child_t *child) { - int32_t ret = -1; - gf_boolean_t need_sign = _gf_false; - br_isignature_out_t *sign = NULL; - - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, xattr, out); - GF_VALIDATE_OR_GOTO (this->name, child, out); - - ret = dict_get_ptr (xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, - (void **)&sign); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED, - "failed to get object signature info"); - goto out; - } + int32_t ret = -1; + gf_boolean_t need_sign = _gf_false; + br_isignature_out_t *sign = NULL; - /* Object has been opened and hence dirty. Do not sign it */ - if (sign->stale) - need_sign = _gf_true; + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, xattr, out); + GF_VALIDATE_OR_GOTO(this->name, child, out); -out: - return need_sign; -} + ret = dict_get_ptr(xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, (void **)&sign); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED, + "failed to get object signature info"); + goto out; + } + /* Object has been opened and hence dirty. Do not sign it */ + if (sign->stale) + need_sign = _gf_true; +out: + return need_sign; +} int32_t -br_prepare_loc (xlator_t *this, br_child_t *child, loc_t *parent, - gf_dirent_t *entry, loc_t *loc) +br_prepare_loc(xlator_t *this, br_child_t *child, loc_t *parent, + gf_dirent_t *entry, loc_t *loc) { - int32_t ret = -1; - inode_t *inode = NULL; - - inode = inode_grep (child->table, parent->inode, entry->d_name); - if (!inode) - loc->inode = inode_new (child->table); - else { - loc->inode = inode; - if (loc->inode->ia_type != IA_IFREG) { - gf_msg_debug (this->name, 0, "%s is not a regular " - "file", entry->d_name); - ret = 0; - goto out; - } - } - - loc->parent = inode_ref (parent->inode); - gf_uuid_copy (loc->pargfid, parent->inode->gfid); - - ret = inode_path (parent->inode, entry->d_name, (char **)&loc->path); - if (ret < 0 || !loc->path) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_PATH_FAILED, - "inode_path on %s (parent: %s) failed", entry->d_name, - uuid_utoa (parent->inode->gfid)); - goto out; - } - - loc->name = strrchr (loc->path, '/'); - if (loc->name) - loc->name++; - - ret = 1; + int32_t ret = -1; + inode_t *inode = NULL; + + inode = inode_grep(child->table, parent->inode, entry->d_name); + if (!inode) + loc->inode = inode_new(child->table); + else { + loc->inode = inode; + if (loc->inode->ia_type != IA_IFREG) { + gf_msg_debug(this->name, 0, + "%s is not a regular " + "file", + entry->d_name); + ret = 0; + goto out; + } + } + + loc->parent = inode_ref(parent->inode); + gf_uuid_copy(loc->pargfid, parent->inode->gfid); + + ret = inode_path(parent->inode, entry->d_name, (char **)&loc->path); + if (ret < 0 || !loc->path) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_PATH_FAILED, + "inode_path on %s (parent: %s) failed", entry->d_name, + uuid_utoa(parent->inode->gfid)); + goto out; + } + + loc->name = strrchr(loc->path, '/'); + if (loc->name) + loc->name++; + + ret = 1; out: - return ret; + return ret; } /** @@ -951,150 +955,161 @@ out: * (though not back to itself :)) */ int -bitd_oneshot_crawl (xlator_t *subvol, - gf_dirent_t *entry, loc_t *parent, void *data) +bitd_oneshot_crawl(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent, + void *data) { - int op_errno = 0; - br_child_t *child = NULL; - xlator_t *this = NULL; - loc_t loc = {0, }; - struct iatt iatt = {0, }; - struct iatt parent_buf = {0, }; - dict_t *xattr = NULL; - int32_t ret = -1; - inode_t *linked_inode = NULL; - gf_boolean_t need_signing = _gf_false; - - GF_VALIDATE_OR_GOTO ("bit-rot", subvol, out); - GF_VALIDATE_OR_GOTO ("bit-rot", data, out); - - child = data; - this = child->this; - - ret = br_prepare_loc (this, child, parent, entry, &loc); - if (!ret) - goto out; - - ret = syncop_lookup (child->xl, &loc, &iatt, &parent_buf, NULL, NULL); - if (ret) { - br_log_object_path (this, "lookup", loc.path, -ret); - goto out; - } - - linked_inode = inode_link (loc.inode, parent->inode, loc.name, &iatt); - if (linked_inode) - inode_lookup (linked_inode); - - if (iatt.ia_type != IA_IFREG) { - gf_msg_debug (this->name, 0, "%s is not a regular file, " - "skipping..", entry->d_name); - ret = 0; - goto unref_inode; - } + int op_errno = 0; + br_child_t *child = NULL; + xlator_t *this = NULL; + loc_t loc = { + 0, + }; + struct iatt iatt = { + 0, + }; + struct iatt parent_buf = { + 0, + }; + dict_t *xattr = NULL; + int32_t ret = -1; + inode_t *linked_inode = NULL; + gf_boolean_t need_signing = _gf_false; + + GF_VALIDATE_OR_GOTO("bit-rot", subvol, out); + GF_VALIDATE_OR_GOTO("bit-rot", data, out); + + child = data; + this = child->this; + + ret = br_prepare_loc(this, child, parent, entry, &loc); + if (!ret) + goto out; + + ret = syncop_lookup(child->xl, &loc, &iatt, &parent_buf, NULL, NULL); + if (ret) { + br_log_object_path(this, "lookup", loc.path, -ret); + goto out; + } + + linked_inode = inode_link(loc.inode, parent->inode, loc.name, &iatt); + if (linked_inode) + inode_lookup(linked_inode); + + if (iatt.ia_type != IA_IFREG) { + gf_msg_debug(this->name, 0, + "%s is not a regular file, " + "skipping..", + entry->d_name); + ret = 0; + goto unref_inode; + } + + /** + * As of now, 2 cases are possible and handled. + * 1) GlusterFS is upgraded from a previous version which does not + * have any idea about bit-rot and have data in the filesystem. + * In this case syncop_getxattr fails with ENODATA and the object + * is signed. (In real, when crawler sends lookup, bit-rot-stub + * creates the xattrs before returning lookup reply) + * 2) Bit-rot was not enabled or BitD was does for some reasons, during + * which some files were created, but since BitD was down, were not + * signed. + * If the file was just created and was being written some data when + * the down BitD came up, then bit-rot stub should be intelligent to + * identify this case (by comparing the ongoing version or by checking + * if there are any fds present for that inode) and handle properly. + */ + + if (bitd_is_bad_file(this, child, &loc, NULL)) { + gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SKIP_OBJECT, + "Entry [%s] is marked corrupted.. skipping.", loc.path); + goto unref_inode; + } + + ret = syncop_getxattr(child->xl, &loc, &xattr, + GLUSTERFS_GET_OBJECT_SIGNATURE, NULL, NULL); + if (ret < 0) { + op_errno = -ret; + br_log_object(this, "getxattr", linked_inode->gfid, op_errno); /** - * As of now, 2 cases are possible and handled. - * 1) GlusterFS is upgraded from a previous version which does not - * have any idea about bit-rot and have data in the filesystem. - * In this case syncop_getxattr fails with ENODATA and the object - * is signed. (In real, when crawler sends lookup, bit-rot-stub - * creates the xattrs before returning lookup reply) - * 2) Bit-rot was not enabled or BitD was does for some reasons, during - * which some files were created, but since BitD was down, were not - * signed. - * If the file was just created and was being written some data when - * the down BitD came up, then bit-rot stub should be intelligent to - * identify this case (by comparing the ongoing version or by checking - * if there are any fds present for that inode) and handle properly. + * No need to sign the zero byte objects as the signing + * happens upon first modification of the object. */ + if (op_errno == ENODATA && (iatt.ia_size != 0)) + need_signing = _gf_true; + if (op_errno == EINVAL) + gf_msg(this->name, GF_LOG_WARNING, 0, + BRB_MSG_PARTIAL_VERSION_PRESENCE, + "Partial " + "version xattr presence detected, ignoring " + "[GFID: %s]", + uuid_utoa(linked_inode->gfid)); + } else { + need_signing = br_check_object_need_sign(this, xattr, child); + } + + if (!need_signing) + goto unref_dict; + + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_TRIGGER_SIGN, + "Triggering signing for %s [GFID: %s | Brick: %s]", loc.path, + uuid_utoa(linked_inode->gfid), child->brick_path); + br_trigger_sign(this, child, linked_inode, &loc, _gf_true); + + ret = 0; + +unref_dict: + if (xattr) + dict_unref(xattr); +unref_inode: + inode_unref(linked_inode); +out: + loc_wipe(&loc); - if (bitd_is_bad_file (this, child, &loc, NULL)) { - gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SKIP_OBJECT, - "Entry [%s] is marked corrupted.. skipping.", loc.path); - goto unref_inode; - } - - ret = syncop_getxattr (child->xl, &loc, &xattr, - GLUSTERFS_GET_OBJECT_SIGNATURE, NULL, NULL); - if (ret < 0) { - op_errno = -ret; - br_log_object (this, "getxattr", linked_inode->gfid, op_errno); - - /** - * No need to sign the zero byte objects as the signing - * happens upon first modification of the object. - */ - if (op_errno == ENODATA && (iatt.ia_size != 0)) - need_signing = _gf_true; - if (op_errno == EINVAL) - gf_msg (this->name, GF_LOG_WARNING, 0, - BRB_MSG_PARTIAL_VERSION_PRESENCE, "Partial " - "version xattr presence detected, ignoring " - "[GFID: %s]", uuid_utoa (linked_inode->gfid)); - } else { - need_signing = br_check_object_need_sign (this, xattr, child); - } - - if (!need_signing) - goto unref_dict; - - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_TRIGGER_SIGN, - "Triggering signing for %s [GFID: %s | Brick: %s]", - loc.path, uuid_utoa (linked_inode->gfid), child->brick_path); - br_trigger_sign (this, child, linked_inode, &loc, _gf_true); - - ret = 0; - - unref_dict: - if (xattr) - dict_unref (xattr); - unref_inode: - inode_unref (linked_inode); - out: - loc_wipe (&loc); - - return ret; + return ret; } #define BR_CRAWL_THROTTLE_COUNT 50 -#define BR_CRAWL_THROTTLE_ZZZ 5 +#define BR_CRAWL_THROTTLE_ZZZ 5 void * -br_oneshot_signer (void *arg) +br_oneshot_signer(void *arg) { - loc_t loc = {0,}; - xlator_t *this = NULL; - br_child_t *child = NULL; + loc_t loc = { + 0, + }; + xlator_t *this = NULL; + br_child_t *child = NULL; - child = arg; - this = child->this; + child = arg; + this = child->this; - THIS = this; + THIS = this; - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_CRAWLING_START, - "Crawling brick [%s], scanning for unsigned objects", - child->brick_path); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_CRAWLING_START, + "Crawling brick [%s], scanning for unsigned objects", + child->brick_path); - loc.inode = child->table->root; - (void) syncop_ftw_throttle - (child->xl, &loc, - GF_CLIENT_PID_BITD, child, bitd_oneshot_crawl, - BR_CRAWL_THROTTLE_COUNT, BR_CRAWL_THROTTLE_ZZZ); + loc.inode = child->table->root; + (void)syncop_ftw_throttle(child->xl, &loc, GF_CLIENT_PID_BITD, child, + bitd_oneshot_crawl, BR_CRAWL_THROTTLE_COUNT, + BR_CRAWL_THROTTLE_ZZZ); - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_CRAWLING_FINISH, - "Completed crawling brick [%s]", child->brick_path); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_CRAWLING_FINISH, + "Completed crawling brick [%s]", child->brick_path); - return NULL; + return NULL; } static void -br_set_child_state (br_child_t *child, br_child_state_t state) +br_set_child_state(br_child_t *child, br_child_state_t state) { - pthread_mutex_lock (&child->lock); - { - _br_set_child_state (child, state); - } - pthread_mutex_unlock (&child->lock); + pthread_mutex_lock(&child->lock); + { + _br_set_child_state(child, state); + } + pthread_mutex_unlock(&child->lock); } /** @@ -1109,160 +1124,159 @@ br_set_child_state (br_child_t *child, br_child_state_t state) * notifications. */ static int32_t -br_enact_signer (xlator_t *this, br_child_t *child, br_stub_init_t *stub) +br_enact_signer(xlator_t *this, br_child_t *child, br_stub_init_t *stub) { - int32_t ret = 0; - br_private_t *priv = NULL; - struct gf_brick_spec *brick = NULL; - - priv = this->private; - - brick = GF_CALLOC (1, sizeof (struct gf_brick_spec), - gf_common_mt_gf_brick_spec_t); - if (!brick) - goto error_return; - - br_fill_brick_spec (brick, stub->export); - ret = gf_changelog_register_generic - (brick, 1, 1, this->ctx->cmd_args.log_file, -1, this); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, errno, - BRB_MSG_REGISTER_FAILED, "Register to changelog " - "failed"); - goto dealloc; - } - - child->threadrunning = 0; - ret = gf_thread_create (&child->thread, NULL, br_oneshot_signer, child, - "brosign"); - if (ret) - gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SPAWN_FAILED, - "failed to spawn FS crawler thread"); - else - child->threadrunning = 1; - - /* it's OK to continue, "old" objects would be signed when modified */ - list_add_tail (&child->list, &priv->signing); - return 0; - - dealloc: - GF_FREE (brick); - error_return: - return -1; + int32_t ret = 0; + br_private_t *priv = NULL; + struct gf_brick_spec *brick = NULL; + + priv = this->private; + + brick = GF_CALLOC(1, sizeof(struct gf_brick_spec), + gf_common_mt_gf_brick_spec_t); + if (!brick) + goto error_return; + + br_fill_brick_spec(brick, stub->export); + ret = gf_changelog_register_generic(brick, 1, 1, + this->ctx->cmd_args.log_file, -1, this); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, errno, BRB_MSG_REGISTER_FAILED, + "Register to changelog " + "failed"); + goto dealloc; + } + + child->threadrunning = 0; + ret = gf_thread_create(&child->thread, NULL, br_oneshot_signer, child, + "brosign"); + if (ret) + gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SPAWN_FAILED, + "failed to spawn FS crawler thread"); + else + child->threadrunning = 1; + + /* it's OK to continue, "old" objects would be signed when modified */ + list_add_tail(&child->list, &priv->signing); + return 0; + +dealloc: + GF_FREE(brick); +error_return: + return -1; } static int32_t -br_launch_scrubber (xlator_t *this, br_child_t *child, - struct br_scanfs *fsscan, struct br_scrubber *fsscrub) +br_launch_scrubber(xlator_t *this, br_child_t *child, struct br_scanfs *fsscan, + struct br_scrubber *fsscrub) { - int32_t ret = -1; - br_private_t *priv = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - - scrub_monitor = &priv->scrub_monitor; - ret = gf_thread_create (&child->thread, NULL, br_fsscanner, child, - "brfsscan"); - if (ret != 0) { - gf_msg (this->name, GF_LOG_ALERT, 0, BRB_MSG_SPAWN_FAILED, - "failed to spawn bitrot scrubber daemon [Brick: %s]", - child->brick_path); - goto error_return; - } - - /* Signal monitor to kick off state machine*/ - pthread_mutex_lock (&scrub_monitor->mutex); - { - if (!scrub_monitor->inited) - pthread_cond_signal (&scrub_monitor->cond); - scrub_monitor->inited = _gf_true; - } - pthread_mutex_unlock (&scrub_monitor->mutex); - - /** - * Everything has been setup.. add this subvolume to scrubbers - * list. - */ - pthread_mutex_lock (&fsscrub->mutex); - { - list_add_tail (&child->list, &fsscrub->scrublist); - pthread_cond_broadcast (&fsscrub->cond); - } - pthread_mutex_unlock (&fsscrub->mutex); + int32_t ret = -1; + br_private_t *priv = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + + scrub_monitor = &priv->scrub_monitor; + ret = gf_thread_create(&child->thread, NULL, br_fsscanner, child, + "brfsscan"); + if (ret != 0) { + gf_msg(this->name, GF_LOG_ALERT, 0, BRB_MSG_SPAWN_FAILED, + "failed to spawn bitrot scrubber daemon [Brick: %s]", + child->brick_path); + goto error_return; + } + + /* Signal monitor to kick off state machine*/ + pthread_mutex_lock(&scrub_monitor->mutex); + { + if (!scrub_monitor->inited) + pthread_cond_signal(&scrub_monitor->cond); + scrub_monitor->inited = _gf_true; + } + pthread_mutex_unlock(&scrub_monitor->mutex); + + /** + * Everything has been setup.. add this subvolume to scrubbers + * list. + */ + pthread_mutex_lock(&fsscrub->mutex); + { + list_add_tail(&child->list, &fsscrub->scrublist); + pthread_cond_broadcast(&fsscrub->cond); + } + pthread_mutex_unlock(&fsscrub->mutex); + + return 0; - return 0; - - error_return: - return -1; +error_return: + return -1; } static int32_t -br_enact_scrubber (xlator_t *this, br_child_t *child) +br_enact_scrubber(xlator_t *this, br_child_t *child) { - int32_t ret = 0; - br_private_t *priv = NULL; - struct br_scanfs *fsscan = NULL; - struct br_scrubber *fsscrub = NULL; + int32_t ret = 0; + br_private_t *priv = NULL; + struct br_scanfs *fsscan = NULL; + struct br_scrubber *fsscrub = NULL; - priv = this->private; + priv = this->private; - fsscan = &child->fsscan; - fsscrub = &priv->fsscrub; + fsscan = &child->fsscan; + fsscrub = &priv->fsscrub; - /** - * if this child already witnesses a successful connection earlier - * there's no need to initialize mutexes, condvars, etc.. - */ - if (_br_child_witnessed_connection (child)) - return br_launch_scrubber (this, child, fsscan, fsscrub); + /** + * if this child already witnesses a successful connection earlier + * there's no need to initialize mutexes, condvars, etc.. + */ + if (_br_child_witnessed_connection(child)) + return br_launch_scrubber(this, child, fsscan, fsscrub); - LOCK_INIT (&fsscan->entrylock); - pthread_mutex_init (&fsscan->waitlock, NULL); - pthread_cond_init (&fsscan->waitcond, NULL); + LOCK_INIT(&fsscan->entrylock); + pthread_mutex_init(&fsscan->waitlock, NULL); + pthread_cond_init(&fsscan->waitcond, NULL); - fsscan->entries = 0; - INIT_LIST_HEAD (&fsscan->queued); - INIT_LIST_HEAD (&fsscan->ready); + fsscan->entries = 0; + INIT_LIST_HEAD(&fsscan->queued); + INIT_LIST_HEAD(&fsscan->ready); - ret = br_launch_scrubber (this, child, fsscan, fsscrub); - if (ret) - goto error_return; + ret = br_launch_scrubber(this, child, fsscan, fsscrub); + if (ret) + goto error_return; - return 0; + return 0; - error_return: - LOCK_DESTROY (&fsscan->entrylock); - pthread_mutex_destroy (&fsscan->waitlock); - pthread_cond_destroy (&fsscan->waitcond); +error_return: + LOCK_DESTROY(&fsscan->entrylock); + pthread_mutex_destroy(&fsscan->waitlock); + pthread_cond_destroy(&fsscan->waitcond); - return -1; + return -1; } static int32_t -br_child_enaction (xlator_t *this, br_child_t *child, br_stub_init_t *stub) +br_child_enaction(xlator_t *this, br_child_t *child, br_stub_init_t *stub) { - int32_t ret = -1; - br_private_t *priv = this->private; + int32_t ret = -1; + br_private_t *priv = this->private; - pthread_mutex_lock (&child->lock); - { - if (priv->iamscrubber) - ret = br_enact_scrubber (this, child); - else - ret = br_enact_signer (this, child, stub); - - if (!ret) { - child->witnessed = 1; - _br_set_child_state (child, BR_CHILD_STATE_CONNECTED); - gf_msg (this->name, GF_LOG_INFO, - 0, BRB_MSG_CONNECTED_TO_BRICK, - "Connected to brick %s..", child->brick_path); - } + pthread_mutex_lock(&child->lock); + { + if (priv->iamscrubber) + ret = br_enact_scrubber(this, child); + else + ret = br_enact_signer(this, child, stub); + + if (!ret) { + child->witnessed = 1; + _br_set_child_state(child, BR_CHILD_STATE_CONNECTED); + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_CONNECTED_TO_BRICK, + "Connected to brick %s..", child->brick_path); } - pthread_mutex_unlock (&child->lock); + } + pthread_mutex_unlock(&child->lock); - return ret; + return ret; } /** @@ -1273,129 +1287,132 @@ br_child_enaction (xlator_t *this, br_child_t *child, br_stub_init_t *stub) * process either acts as a signer or a scrubber. */ int32_t -br_brick_connect (xlator_t *this, br_child_t *child) +br_brick_connect(xlator_t *this, br_child_t *child) { - int32_t ret = -1; - loc_t loc = {0, }; - struct iatt buf = {0, }; - struct iatt parent = {0, }; - br_stub_init_t *stub = NULL; - dict_t *xattr = NULL; - int op_errno = 0; - - GF_VALIDATE_OR_GOTO ("bit-rot", this, out); - GF_VALIDATE_OR_GOTO (this->name, child, out); - GF_VALIDATE_OR_GOTO (this->name, this->private, out); - - br_child_set_scrub_state (child, _gf_false); - br_set_child_state (child, BR_CHILD_STATE_INITIALIZING); - - loc.inode = inode_ref (child->table->root); - gf_uuid_copy (loc.gfid, loc.inode->gfid); - loc.path = gf_strdup ("/"); - - ret = syncop_lookup (child->xl, &loc, &buf, &parent, NULL, NULL); - if (ret) { - op_errno = -ret; - ret = -1; - gf_msg (this->name, GF_LOG_ERROR, op_errno, - BRB_MSG_LOOKUP_FAILED, "lookup on root failed"); - goto wipeloc; - } - - ret = syncop_getxattr (child->xl, &loc, &xattr, - GLUSTERFS_GET_BR_STUB_INIT_TIME, NULL, NULL); - if (ret) { - op_errno = -ret; - ret = -1; - gf_msg (this->name, GF_LOG_ERROR, op_errno, - BRB_MSG_GET_INFO_FAILED, "failed to get stub info"); - goto wipeloc; - } - - ret = dict_get_ptr (xattr, GLUSTERFS_GET_BR_STUB_INIT_TIME, - (void **)&stub); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_INFO_FAILED, - "failed to extract stub information"); - goto free_dict; - } - - memcpy (child->brick_path, stub->export, strlen (stub->export) + 1); - child->tv.tv_sec = ntohl (stub->timebuf[0]); - child->tv.tv_usec = ntohl (stub->timebuf[1]); - - ret = br_child_enaction (this, child, stub); - - free_dict: - dict_unref (xattr); - wipeloc: - loc_wipe (&loc); - out: - if (ret) - br_set_child_state (child, BR_CHILD_STATE_CONNFAILED); - return ret; + int32_t ret = -1; + loc_t loc = { + 0, + }; + struct iatt buf = { + 0, + }; + struct iatt parent = { + 0, + }; + br_stub_init_t *stub = NULL; + dict_t *xattr = NULL; + int op_errno = 0; + + GF_VALIDATE_OR_GOTO("bit-rot", this, out); + GF_VALIDATE_OR_GOTO(this->name, child, out); + GF_VALIDATE_OR_GOTO(this->name, this->private, out); + + br_child_set_scrub_state(child, _gf_false); + br_set_child_state(child, BR_CHILD_STATE_INITIALIZING); + + loc.inode = inode_ref(child->table->root); + gf_uuid_copy(loc.gfid, loc.inode->gfid); + loc.path = gf_strdup("/"); + + ret = syncop_lookup(child->xl, &loc, &buf, &parent, NULL, NULL); + if (ret) { + op_errno = -ret; + ret = -1; + gf_msg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_LOOKUP_FAILED, + "lookup on root failed"); + goto wipeloc; + } + + ret = syncop_getxattr(child->xl, &loc, &xattr, + GLUSTERFS_GET_BR_STUB_INIT_TIME, NULL, NULL); + if (ret) { + op_errno = -ret; + ret = -1; + gf_msg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_GET_INFO_FAILED, + "failed to get stub info"); + goto wipeloc; + } + + ret = dict_get_ptr(xattr, GLUSTERFS_GET_BR_STUB_INIT_TIME, (void **)&stub); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_INFO_FAILED, + "failed to extract stub information"); + goto free_dict; + } + + memcpy(child->brick_path, stub->export, strlen(stub->export) + 1); + child->tv.tv_sec = ntohl(stub->timebuf[0]); + child->tv.tv_usec = ntohl(stub->timebuf[1]); + + ret = br_child_enaction(this, child, stub); + +free_dict: + dict_unref(xattr); +wipeloc: + loc_wipe(&loc); +out: + if (ret) + br_set_child_state(child, BR_CHILD_STATE_CONNFAILED); + return ret; } /* TODO: cleanup signer */ static int32_t -br_cleanup_signer (xlator_t *this, br_child_t *child) +br_cleanup_signer(xlator_t *this, br_child_t *child) { - return 0; + return 0; } static int32_t -br_cleanup_scrubber (xlator_t *this, br_child_t *child) +br_cleanup_scrubber(xlator_t *this, br_child_t *child) { - int32_t ret = 0; - br_private_t *priv = NULL; - struct br_scrubber *fsscrub = NULL; - struct br_monitor *scrub_monitor = NULL; - - priv = this->private; - fsscrub = &priv->fsscrub; - scrub_monitor = &priv->scrub_monitor; - - if (_br_is_child_scrub_active (child)) { - scrub_monitor->active_child_count--; - br_child_set_scrub_state (child, _gf_false); - } - - /** - * 0x0: child (brick) goes out of rotation - * - * This is fully safe w.r.t. entries for this child being actively - * scrubbed. Each of the scrubber thread(s) would finish scrubbing - * the entry (probably failing due to disconnection) and either - * putting the entry back into the queue or continuing further. - * Either way, pending entries for this child's queue need not be - * drained; entries just sit there in the queued/ready list to be - * consumed later upon re-connection. - */ - pthread_mutex_lock (&fsscrub->mutex); - { - list_del_init (&child->list); - } - pthread_mutex_unlock (&fsscrub->mutex); - - /** - * 0x1: cleanup scanner thread - * - * The pending timer needs to be removed _after_ cleaning up the - * filesystem scanner (scheduling the next scrub time is not a - * cancellation point). - */ - ret = gf_thread_cleanup_xint (child->thread); - if (ret) - gf_msg (this->name, GF_LOG_INFO, - 0, BRB_MSG_SCRUB_THREAD_CLEANUP, - "Error cleaning up scanner thread"); - - gf_msg (this->name, GF_LOG_INFO, - 0, BRB_MSG_SCRUBBER_CLEANED, - "Cleaned up scrubber for brick [%s]", child->brick_path); - - return 0; + int32_t ret = 0; + br_private_t *priv = NULL; + struct br_scrubber *fsscrub = NULL; + struct br_monitor *scrub_monitor = NULL; + + priv = this->private; + fsscrub = &priv->fsscrub; + scrub_monitor = &priv->scrub_monitor; + + if (_br_is_child_scrub_active(child)) { + scrub_monitor->active_child_count--; + br_child_set_scrub_state(child, _gf_false); + } + + /** + * 0x0: child (brick) goes out of rotation + * + * This is fully safe w.r.t. entries for this child being actively + * scrubbed. Each of the scrubber thread(s) would finish scrubbing + * the entry (probably failing due to disconnection) and either + * putting the entry back into the queue or continuing further. + * Either way, pending entries for this child's queue need not be + * drained; entries just sit there in the queued/ready list to be + * consumed later upon re-connection. + */ + pthread_mutex_lock(&fsscrub->mutex); + { + list_del_init(&child->list); + } + pthread_mutex_unlock(&fsscrub->mutex); + + /** + * 0x1: cleanup scanner thread + * + * The pending timer needs to be removed _after_ cleaning up the + * filesystem scanner (scheduling the next scrub time is not a + * cancellation point). + */ + ret = gf_thread_cleanup_xint(child->thread); + if (ret) + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_THREAD_CLEANUP, + "Error cleaning up scanner thread"); + + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUBBER_CLEANED, + "Cleaned up scrubber for brick [%s]", child->brick_path); + + return 0; } /** @@ -1404,38 +1421,38 @@ br_cleanup_scrubber (xlator_t *this, br_child_t *child) * the inode table, it's just reused taking care of stale inodes) */ int32_t -br_brick_disconnect (xlator_t *this, br_child_t *child) +br_brick_disconnect(xlator_t *this, br_child_t *child) { - int32_t ret = 0; - struct br_monitor *scrub_monitor = NULL; - br_private_t *priv = this->private; + int32_t ret = 0; + struct br_monitor *scrub_monitor = NULL; + br_private_t *priv = this->private; + + scrub_monitor = &priv->scrub_monitor; + + /* Lock order should be wakelock and then child lock to + * dead locks. + */ + pthread_mutex_lock(&scrub_monitor->wakelock); + { + pthread_mutex_lock(&child->lock); + { + if (!_br_is_child_connected(child)) + goto unblock; - scrub_monitor = &priv->scrub_monitor; + /* child is on death row.. */ + _br_set_child_state(child, BR_CHILD_STATE_DISCONNECTED); - /* Lock order should be wakelock and then child lock to - * dead locks. - */ - pthread_mutex_lock (&scrub_monitor->wakelock); - { - pthread_mutex_lock (&child->lock); - { - if (!_br_is_child_connected (child)) - goto unblock; - - /* child is on death row.. */ - _br_set_child_state (child, BR_CHILD_STATE_DISCONNECTED); - - if (priv->iamscrubber) - ret = br_cleanup_scrubber (this, child); - else - ret = br_cleanup_signer (this, child); - } - unblock: - pthread_mutex_unlock (&child->lock); + if (priv->iamscrubber) + ret = br_cleanup_scrubber(this, child); + else + ret = br_cleanup_signer(this, child); } - pthread_mutex_unlock (&scrub_monitor->wakelock); + unblock: + pthread_mutex_unlock(&child->lock); + } + pthread_mutex_unlock(&scrub_monitor->wakelock); - return ret; + return ret; } /** @@ -1444,275 +1461,278 @@ br_brick_disconnect (xlator_t *this, br_child_t *child) * information regarding that brick (such as brick path). */ void * -br_handle_events (void *arg) +br_handle_events(void *arg) { - int32_t ret = 0; - xlator_t *this = NULL; - br_private_t *priv = NULL; - br_child_t *child = NULL; - struct br_child_event *childev = NULL; - - this = arg; - priv = this->private; - - /* - * Since, this is the topmost xlator, THIS has to be set by bit-rot - * xlator itself (STACK_WIND won't help in this case). Also it has - * to be done for each thread that gets spawned. Otherwise, a new - * thread will get global_xlator's pointer when it does "THIS". - */ - THIS = this; - - while (1) { - pthread_mutex_lock (&priv->lock); - { - while (list_empty (&priv->bricks)) - pthread_cond_wait (&priv->cond, &priv->lock); - - childev = list_first_entry - (&priv->bricks, struct br_child_event, list); - list_del_init (&childev->list); - } - pthread_mutex_unlock (&priv->lock); - - child = childev->child; - ret = childev->call (this, child); - if (ret) - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_SUBVOL_CONNECT_FAILED, - "callback handler for subvolume [%s] failed", - child->xl->name); - GF_FREE (childev); + int32_t ret = 0; + xlator_t *this = NULL; + br_private_t *priv = NULL; + br_child_t *child = NULL; + struct br_child_event *childev = NULL; + + this = arg; + priv = this->private; + + /* + * Since, this is the topmost xlator, THIS has to be set by bit-rot + * xlator itself (STACK_WIND won't help in this case). Also it has + * to be done for each thread that gets spawned. Otherwise, a new + * thread will get global_xlator's pointer when it does "THIS". + */ + THIS = this; + + while (1) { + pthread_mutex_lock(&priv->lock); + { + while (list_empty(&priv->bricks)) + pthread_cond_wait(&priv->cond, &priv->lock); + + childev = list_first_entry(&priv->bricks, struct br_child_event, + list); + list_del_init(&childev->list); } + pthread_mutex_unlock(&priv->lock); - return NULL; + child = childev->child; + ret = childev->call(this, child); + if (ret) + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SUBVOL_CONNECT_FAILED, + "callback handler for subvolume [%s] failed", + child->xl->name); + GF_FREE(childev); + } + + return NULL; } int32_t -mem_acct_init (xlator_t *this) +mem_acct_init(xlator_t *this) { - int32_t ret = -1; + int32_t ret = -1; - if (!this) - return ret; - - ret = xlator_mem_acct_init (this, gf_br_stub_mt_end + 1); + if (!this) + return ret; - if (ret != 0) { - gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_MEM_ACNT_FAILED, - "Memory accounting init failed"); - return ret; - } + ret = xlator_mem_acct_init(this, gf_br_stub_mt_end + 1); + if (ret != 0) { + gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_MEM_ACNT_FAILED, + "Memory accounting init failed"); return ret; + } + + return ret; } static void -_br_qchild_event (xlator_t *this, br_child_t *child, br_child_handler *call) +_br_qchild_event(xlator_t *this, br_child_t *child, br_child_handler *call) { - br_private_t *priv = NULL; - struct br_child_event *childev = NULL; + br_private_t *priv = NULL; + struct br_child_event *childev = NULL; - priv = this->private; + priv = this->private; - childev = GF_CALLOC (1, sizeof (*childev), gf_br_mt_br_child_event_t); - if (!childev) { - gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, - "Event unhandled for child.. [Brick: %s]", - child->xl->name); - return; - } + childev = GF_CALLOC(1, sizeof(*childev), gf_br_mt_br_child_event_t); + if (!childev) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, + "Event unhandled for child.. [Brick: %s]", child->xl->name); + return; + } - INIT_LIST_HEAD (&childev->list); - childev->this = this; - childev->child = child; - childev->call = call; + INIT_LIST_HEAD(&childev->list); + childev->this = this; + childev->child = child; + childev->call = call; - list_add_tail (&childev->list, &priv->bricks); + list_add_tail(&childev->list, &priv->bricks); } int -br_scrubber_status_get (xlator_t *this, dict_t **dict) +br_scrubber_status_get(xlator_t *this, dict_t **dict) { - int ret = -1; - br_private_t *priv = NULL; - struct br_scrub_stats *scrub_stats = NULL; - - priv = this->private; - - GF_VALIDATE_OR_GOTO ("bit-rot", priv, out); - - scrub_stats = &priv->scrub_stat; - - ret = br_get_bad_objects_list (this, dict); - if (ret) { - gf_msg_debug (this->name, 0, "Failed to collect corrupt " - "files"); - } - - ret = dict_set_int8 (*dict, "scrub-running", - scrub_stats->scrub_running); - if (ret) { - gf_msg_debug (this->name, 0, "Failed setting scrub_running " - "entry to the dictionary"); - } - - ret = dict_set_uint64 (*dict, "scrubbed-files", - scrub_stats->scrubbed_files); - if (ret) { - gf_msg_debug (this->name, 0, "Failed to setting scrubbed file " - "entry to the dictionary"); - } - - ret = dict_set_uint64 (*dict, "unsigned-files", - scrub_stats->unsigned_files); - if (ret) { - gf_msg_debug (this->name, 0, "Failed to set unsigned file count" - " entry to the dictionary"); - } - - ret = dict_set_uint64 (*dict, "scrub-duration", - scrub_stats->scrub_duration); - if (ret) { - gf_msg_debug (this->name, 0, "Failed to set scrub duration" - " entry to the dictionary"); - } - - ret = dict_set_dynstr_with_alloc (*dict, "last-scrub-time", - scrub_stats->last_scrub_time); - if (ret) { - gf_msg_debug (this->name, 0, "Failed to set " - "last scrub time value"); - } + int ret = -1; + br_private_t *priv = NULL; + struct br_scrub_stats *scrub_stats = NULL; + + priv = this->private; + + GF_VALIDATE_OR_GOTO("bit-rot", priv, out); + + scrub_stats = &priv->scrub_stat; + + ret = br_get_bad_objects_list(this, dict); + if (ret) { + gf_msg_debug(this->name, 0, + "Failed to collect corrupt " + "files"); + } + + ret = dict_set_int8(*dict, "scrub-running", scrub_stats->scrub_running); + if (ret) { + gf_msg_debug(this->name, 0, + "Failed setting scrub_running " + "entry to the dictionary"); + } + + ret = dict_set_uint64(*dict, "scrubbed-files", scrub_stats->scrubbed_files); + if (ret) { + gf_msg_debug(this->name, 0, + "Failed to setting scrubbed file " + "entry to the dictionary"); + } + + ret = dict_set_uint64(*dict, "unsigned-files", scrub_stats->unsigned_files); + if (ret) { + gf_msg_debug(this->name, 0, + "Failed to set unsigned file count" + " entry to the dictionary"); + } + + ret = dict_set_uint64(*dict, "scrub-duration", scrub_stats->scrub_duration); + if (ret) { + gf_msg_debug(this->name, 0, + "Failed to set scrub duration" + " entry to the dictionary"); + } + + ret = dict_set_dynstr_with_alloc(*dict, "last-scrub-time", + scrub_stats->last_scrub_time); + if (ret) { + gf_msg_debug(this->name, 0, + "Failed to set " + "last scrub time value"); + } out: - return ret; + return ret; } int -notify (xlator_t *this, int32_t event, void *data, ...) +notify(xlator_t *this, int32_t event, void *data, ...) { - int idx = -1; - int ret = -1; - xlator_t *subvol = NULL; - br_child_t *child = NULL; - br_private_t *priv = NULL; - dict_t *output = NULL; - va_list ap; - struct br_monitor *scrub_monitor = NULL; + int idx = -1; + int ret = -1; + xlator_t *subvol = NULL; + br_child_t *child = NULL; + br_private_t *priv = NULL; + dict_t *output = NULL; + va_list ap; + struct br_monitor *scrub_monitor = NULL; - subvol = (xlator_t *)data; - priv = this->private; - scrub_monitor = &priv->scrub_monitor; + subvol = (xlator_t *)data; + priv = this->private; + scrub_monitor = &priv->scrub_monitor; - gf_msg_trace (this->name, 0, "Notification received: %d", event); + gf_msg_trace(this->name, 0, "Notification received: %d", event); - idx = br_find_child_index (this, subvol); + idx = br_find_child_index(this, subvol); - switch (event) { + switch (event) { case GF_EVENT_CHILD_UP: - if (idx < 0) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_INVALID_SUBVOL, "Got event %d from " - "invalid subvolume", event); - goto out; - } - - pthread_mutex_lock (&priv->lock); - { - child = &priv->children[idx]; - if (child->child_up == 1) - goto unblock_0; - priv->up_children++; - - child->child_up = 1; - child->xl = subvol; - if (!child->table) - child->table = inode_table_new (4096, subvol); - - _br_qchild_event (this, child, br_brick_connect); - pthread_cond_signal (&priv->cond); - } + if (idx < 0) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_INVALID_SUBVOL, + "Got event %d from " + "invalid subvolume", + event); + goto out; + } + + pthread_mutex_lock(&priv->lock); + { + child = &priv->children[idx]; + if (child->child_up == 1) + goto unblock_0; + priv->up_children++; + + child->child_up = 1; + child->xl = subvol; + if (!child->table) + child->table = inode_table_new(4096, subvol); + + _br_qchild_event(this, child, br_brick_connect); + pthread_cond_signal(&priv->cond); + } unblock_0: - pthread_mutex_unlock (&priv->lock); + pthread_mutex_unlock(&priv->lock); - if (priv->up_children == priv->child_count) - default_notify (this, event, data); - break; + if (priv->up_children == priv->child_count) + default_notify(this, event, data); + break; case GF_EVENT_CHILD_DOWN: - if (idx < 0) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_INVALID_SUBVOL_CHILD, - "Got event %d from invalid subvolume", event); - goto out; - } - - pthread_mutex_lock (&priv->lock); - { - child = &priv->children[idx]; - if (child->child_up == 0) - goto unblock_1; - - child->child_up = 0; - priv->up_children--; - - _br_qchild_event (this, child, br_brick_disconnect); - pthread_cond_signal (&priv->cond); - } + if (idx < 0) { + gf_msg(this->name, GF_LOG_ERROR, 0, + BRB_MSG_INVALID_SUBVOL_CHILD, + "Got event %d from invalid subvolume", event); + goto out; + } + + pthread_mutex_lock(&priv->lock); + { + child = &priv->children[idx]; + if (child->child_up == 0) + goto unblock_1; + + child->child_up = 0; + priv->up_children--; + + _br_qchild_event(this, child, br_brick_disconnect); + pthread_cond_signal(&priv->cond); + } unblock_1: - pthread_mutex_unlock (&priv->lock); + pthread_mutex_unlock(&priv->lock); - if (priv->up_children == 0) - default_notify (this, event, data); - break; + if (priv->up_children == 0) + default_notify(this, event, data); + break; case GF_EVENT_SCRUB_STATUS: - gf_msg_debug (this->name, GF_LOG_INFO, "BitRot scrub status " - "called"); - va_start (ap, data); - output = va_arg (ap, dict_t *); - va_end (ap); - - ret = br_scrubber_status_get (this, &output); - gf_msg_debug (this->name, 0, "returning %d", ret); - break; + gf_msg_debug(this->name, GF_LOG_INFO, + "BitRot scrub status " + "called"); + va_start(ap, data); + output = va_arg(ap, dict_t *); + va_end(ap); + + ret = br_scrubber_status_get(this, &output); + gf_msg_debug(this->name, 0, "returning %d", ret); + break; case GF_EVENT_SCRUB_ONDEMAND: - gf_log (this->name, GF_LOG_INFO, "BitRot scrub ondemand " - "called"); - - if (scrub_monitor->state != BR_SCRUB_STATE_PENDING) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_RESCHEDULE_SCRUBBER_FAILED, - "on demand scrub schedule failed. Scrubber is " - "not in pending state. Current state is %d", - scrub_monitor->state); - return -2; - } - - /* Needs synchronization with reconfigure thread */ - pthread_mutex_lock (&priv->lock); - { - ret = br_scrub_state_machine (this, _gf_true); - } - pthread_mutex_unlock (&priv->lock); - - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_RESCHEDULE_SCRUBBER_FAILED, - "Could not schedule ondemand scrubbing. " - "Scrubbing will continue according to " - "old frequency."); - } - gf_msg_debug (this->name, 0, "returning %d", ret); - break; + gf_log(this->name, GF_LOG_INFO, + "BitRot scrub ondemand " + "called"); + + if (scrub_monitor->state != BR_SCRUB_STATE_PENDING) { + gf_msg(this->name, GF_LOG_ERROR, 0, + BRB_MSG_RESCHEDULE_SCRUBBER_FAILED, + "on demand scrub schedule failed. Scrubber is " + "not in pending state. Current state is %d", + scrub_monitor->state); + return -2; + } + + /* Needs synchronization with reconfigure thread */ + pthread_mutex_lock(&priv->lock); + { + ret = br_scrub_state_machine(this, _gf_true); + } + pthread_mutex_unlock(&priv->lock); + + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, + BRB_MSG_RESCHEDULE_SCRUBBER_FAILED, + "Could not schedule ondemand scrubbing. " + "Scrubbing will continue according to " + "old frequency."); + } + gf_msg_debug(this->name, 0, "returning %d", ret); + break; default: - default_notify (this, event, data); - } + default_notify(this, event, data); + } - out: - return 0; +out: + return 0; } /** @@ -1720,62 +1740,62 @@ notify (xlator_t *this, int32_t event, void *data, ...) */ static void -br_fini_signer (xlator_t *this, br_private_t *priv) +br_fini_signer(xlator_t *this, br_private_t *priv) { - int i = 0; + int i = 0; - for (; i < BR_WORKERS; i++) { - (void) gf_thread_cleanup_xint (priv->obj_queue->workers[i]); - } + for (; i < BR_WORKERS; i++) { + (void)gf_thread_cleanup_xint(priv->obj_queue->workers[i]); + } - pthread_cond_destroy (&priv->object_cond); + pthread_cond_destroy(&priv->object_cond); } static int32_t -br_init_signer (xlator_t *this, br_private_t *priv) +br_init_signer(xlator_t *this, br_private_t *priv) { - int i = 0; - int32_t ret = -1; + int i = 0; + int32_t ret = -1; - /* initialize gfchangelog xlator context */ - ret = gf_changelog_init (this); - if (ret) - goto out; + /* initialize gfchangelog xlator context */ + ret = gf_changelog_init(this); + if (ret) + goto out; - pthread_cond_init (&priv->object_cond, NULL); - - priv->obj_queue = GF_CALLOC (1, sizeof (*priv->obj_queue), - gf_br_mt_br_ob_n_wk_t); - if (!priv->obj_queue) - goto cleanup_cond; - INIT_LIST_HEAD (&priv->obj_queue->objects); - - for (i = 0; i < BR_WORKERS; i++) { - ret = gf_thread_create (&priv->obj_queue->workers[i], NULL, - br_process_object, this, "brpobj"); - if (ret != 0) { - gf_msg (this->name, GF_LOG_ERROR, -ret, - BRB_MSG_SPAWN_FAILED, "thread creation" - " failed"); - ret = -1; - goto cleanup_threads; - } - } + pthread_cond_init(&priv->object_cond, NULL); - return 0; + priv->obj_queue = GF_CALLOC(1, sizeof(*priv->obj_queue), + gf_br_mt_br_ob_n_wk_t); + if (!priv->obj_queue) + goto cleanup_cond; + INIT_LIST_HEAD(&priv->obj_queue->objects); - cleanup_threads: - for (i--; i >= 0; i--) { - (void) gf_thread_cleanup_xint (priv->obj_queue->workers[i]); + for (i = 0; i < BR_WORKERS; i++) { + ret = gf_thread_create(&priv->obj_queue->workers[i], NULL, + br_process_object, this, "brpobj"); + if (ret != 0) { + gf_msg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_SPAWN_FAILED, + "thread creation" + " failed"); + ret = -1; + goto cleanup_threads; } + } + + return 0; + +cleanup_threads: + for (i--; i >= 0; i--) { + (void)gf_thread_cleanup_xint(priv->obj_queue->workers[i]); + } - GF_FREE (priv->obj_queue); + GF_FREE(priv->obj_queue); - cleanup_cond: - /* that's explicit */ - pthread_cond_destroy (&priv->object_cond); - out: - return -1; +cleanup_cond: + /* that's explicit */ + pthread_cond_destroy(&priv->object_cond); +out: + return -1; } /** @@ -1784,360 +1804,356 @@ br_init_signer (xlator_t *this, br_private_t *priv) * throttle. */ static int32_t -br_rate_limit_signer (xlator_t *this, int child_count, int numbricks) +br_rate_limit_signer(xlator_t *this, int child_count, int numbricks) { - br_private_t *priv = NULL; - tbf_opspec_t spec = {0,}; - - priv = this->private; - - spec.op = TBF_OP_HASH; - spec.rate = 0; - spec.maxlimit = 0; - -/** - * OK. Most implementations of TBF I've come across generate tokens - * every second (UML, etc..) and some chose sub-second granularity - * (blk-iothrottle cgroups). TBF algorithm itself does not enforce - * any logic for choosing generation interval and it seems pretty - * logical as one could jack up token count per interval w.r.t. - * generation rate. - * - * Value used here is chosen based on a series of test(s) performed - * to balance object signing time and not maxing out on all available - * CPU cores. It's obvious to have seconds granularity and jack up - * token count per interval, thereby achieving close to similar - * results. Let's stick to this as it seems to be working fine for - * the set of ops that are throttled. - **/ - spec.token_gen_interval = 600000; /* In usec */ - + br_private_t *priv = NULL; + tbf_opspec_t spec = { + 0, + }; + + priv = this->private; + + spec.op = TBF_OP_HASH; + spec.rate = 0; + spec.maxlimit = 0; + + /** + * OK. Most implementations of TBF I've come across generate tokens + * every second (UML, etc..) and some chose sub-second granularity + * (blk-iothrottle cgroups). TBF algorithm itself does not enforce + * any logic for choosing generation interval and it seems pretty + * logical as one could jack up token count per interval w.r.t. + * generation rate. + * + * Value used here is chosen based on a series of test(s) performed + * to balance object signing time and not maxing out on all available + * CPU cores. It's obvious to have seconds granularity and jack up + * token count per interval, thereby achieving close to similar + * results. Let's stick to this as it seems to be working fine for + * the set of ops that are throttled. + **/ + spec.token_gen_interval = 600000; /* In usec */ #ifdef BR_RATE_LIMIT_SIGNER - double contribution = 0; - contribution = ((double)1 - ((double)child_count / (double)numbricks)); - if (contribution == 0) - contribution = 1; - spec.rate = BR_HASH_CALC_READ_SIZE * contribution; - spec.maxlimit = BR_WORKERS * BR_HASH_CALC_READ_SIZE; + double contribution = 0; + contribution = ((double)1 - ((double)child_count / (double)numbricks)); + if (contribution == 0) + contribution = 1; + spec.rate = BR_HASH_CALC_READ_SIZE * contribution; + spec.maxlimit = BR_WORKERS * BR_HASH_CALC_READ_SIZE; #endif - if (!spec.rate) - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_RATE_LIMIT_INFO, - "[Rate Limit Info] \"FULL THROTTLE\""); - else - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_RATE_LIMIT_INFO, - "[Rate Limit Info] \"tokens/sec (rate): %lu, " - "maxlimit: %lu\"", spec.rate, spec.maxlimit); - - priv->tbf = tbf_init (&spec, 1); - return priv->tbf ? 0 : -1; + if (!spec.rate) + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_RATE_LIMIT_INFO, + "[Rate Limit Info] \"FULL THROTTLE\""); + else + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_RATE_LIMIT_INFO, + "[Rate Limit Info] \"tokens/sec (rate): %lu, " + "maxlimit: %lu\"", + spec.rate, spec.maxlimit); + + priv->tbf = tbf_init(&spec, 1); + return priv->tbf ? 0 : -1; } static int32_t -br_signer_handle_options (xlator_t *this, br_private_t *priv, dict_t *options) +br_signer_handle_options(xlator_t *this, br_private_t *priv, dict_t *options) { - if (options) - GF_OPTION_RECONF ("expiry-time", priv->expiry_time, - options, uint32, error_return); - else - GF_OPTION_INIT ("expiry-time", priv->expiry_time, - uint32, error_return); + if (options) + GF_OPTION_RECONF("expiry-time", priv->expiry_time, options, uint32, + error_return); + else + GF_OPTION_INIT("expiry-time", priv->expiry_time, uint32, error_return); - return 0; + return 0; error_return: - return -1; + return -1; } static int32_t -br_signer_init (xlator_t *this, br_private_t *priv) +br_signer_init(xlator_t *this, br_private_t *priv) { - int32_t ret = 0; - int numbricks = 0; + int32_t ret = 0; + int numbricks = 0; - GF_OPTION_INIT ("expiry-time", priv->expiry_time, uint32, error_return); - GF_OPTION_INIT ("brick-count", numbricks, int32, error_return); + GF_OPTION_INIT("expiry-time", priv->expiry_time, uint32, error_return); + GF_OPTION_INIT("brick-count", numbricks, int32, error_return); - ret = br_rate_limit_signer (this, priv->child_count, numbricks); - if (ret) - goto error_return; + ret = br_rate_limit_signer(this, priv->child_count, numbricks); + if (ret) + goto error_return; - ret = br_init_signer (this, priv); - if (ret) - goto cleanup_tbf; + ret = br_init_signer(this, priv); + if (ret) + goto cleanup_tbf; - return 0; - - cleanup_tbf: - /* cleanup TBF */ - error_return: - return -1; + return 0; +cleanup_tbf: + /* cleanup TBF */ +error_return: + return -1; } static void -br_free_scrubber_monitor (xlator_t *this, br_private_t *priv) +br_free_scrubber_monitor(xlator_t *this, br_private_t *priv) { - struct br_monitor *scrub_monitor = &priv->scrub_monitor; + struct br_monitor *scrub_monitor = &priv->scrub_monitor; - if (scrub_monitor->timer) { - (void) gf_tw_del_timer (priv->timer_wheel, scrub_monitor->timer); + if (scrub_monitor->timer) { + (void)gf_tw_del_timer(priv->timer_wheel, scrub_monitor->timer); - GF_FREE (scrub_monitor->timer); - scrub_monitor->timer = NULL; - } + GF_FREE(scrub_monitor->timer); + scrub_monitor->timer = NULL; + } - (void) gf_thread_cleanup_xint (scrub_monitor->thread); + (void)gf_thread_cleanup_xint(scrub_monitor->thread); - /* Clean up cond and mutex variables */ - pthread_mutex_destroy (&scrub_monitor->mutex); - pthread_cond_destroy (&scrub_monitor->cond); + /* Clean up cond and mutex variables */ + pthread_mutex_destroy(&scrub_monitor->mutex); + pthread_cond_destroy(&scrub_monitor->cond); - pthread_mutex_destroy (&scrub_monitor->wakelock); - pthread_cond_destroy (&scrub_monitor->wakecond); + pthread_mutex_destroy(&scrub_monitor->wakelock); + pthread_cond_destroy(&scrub_monitor->wakecond); - pthread_mutex_destroy (&scrub_monitor->donelock); - pthread_cond_destroy (&scrub_monitor->donecond); + pthread_mutex_destroy(&scrub_monitor->donelock); + pthread_cond_destroy(&scrub_monitor->donecond); - LOCK_DESTROY (&scrub_monitor->lock); + LOCK_DESTROY(&scrub_monitor->lock); } static void -br_free_children (xlator_t *this, br_private_t *priv, int count) +br_free_children(xlator_t *this, br_private_t *priv, int count) { - br_child_t *child = NULL; + br_child_t *child = NULL; - for (--count; count >= 0; count--) { - child = &priv->children[count]; - mem_pool_destroy (child->timer_pool); - pthread_mutex_destroy (&child->lock); - } + for (--count; count >= 0; count--) { + child = &priv->children[count]; + mem_pool_destroy(child->timer_pool); + pthread_mutex_destroy(&child->lock); + } - GF_FREE (priv->children); - priv->children = NULL; + GF_FREE(priv->children); + priv->children = NULL; } static int -br_init_children (xlator_t *this, br_private_t *priv) +br_init_children(xlator_t *this, br_private_t *priv) { - int i = 0; - br_child_t *child = NULL; - xlator_list_t *trav = NULL; - - priv->child_count = xlator_subvolume_count (this); - priv->children = GF_CALLOC (priv->child_count, sizeof (*priv->children), - gf_br_mt_br_child_t); - if (!priv->children) - goto err; - - trav = this->children; - while (trav) { - child = &priv->children[i]; - - pthread_mutex_init (&child->lock, NULL); - child->witnessed = 0; - - br_set_child_state (child, BR_CHILD_STATE_DISCONNECTED); - - child->this = this; - child->xl = trav->xlator; - - child->timer_pool = mem_pool_new - (struct gf_tw_timer_list, 4096); - if (!child->timer_pool) { - gf_msg (this->name, GF_LOG_ERROR, - ENOMEM, BRB_MSG_NO_MEMORY, - "failed to allocate mem-pool for timer"); - errno = ENOMEM; - goto freechild; - } - - INIT_LIST_HEAD (&child->list); - - i++; - trav = trav->next; + int i = 0; + br_child_t *child = NULL; + xlator_list_t *trav = NULL; + + priv->child_count = xlator_subvolume_count(this); + priv->children = GF_CALLOC(priv->child_count, sizeof(*priv->children), + gf_br_mt_br_child_t); + if (!priv->children) + goto err; + + trav = this->children; + while (trav) { + child = &priv->children[i]; + + pthread_mutex_init(&child->lock, NULL); + child->witnessed = 0; + + br_set_child_state(child, BR_CHILD_STATE_DISCONNECTED); + + child->this = this; + child->xl = trav->xlator; + + child->timer_pool = mem_pool_new(struct gf_tw_timer_list, 4096); + if (!child->timer_pool) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, + "failed to allocate mem-pool for timer"); + errno = ENOMEM; + goto freechild; } - return 0; + INIT_LIST_HEAD(&child->list); + + i++; + trav = trav->next; + } + + return 0; - freechild: - br_free_children (this, priv, i); - err: - return -1; +freechild: + br_free_children(this, priv, i); +err: + return -1; } int32_t -init (xlator_t *this) +init(xlator_t *this) { - int32_t ret = -1; - br_private_t *priv = NULL; - - if (!this->children) { - gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_NO_CHILD, - "FATAL: no children"); - goto out; - } - - priv = GF_CALLOC (1, sizeof (*priv), gf_br_mt_br_private_t); - if (!priv) { - gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, - "failed to allocate memory (->priv)"); - goto out; - } + int32_t ret = -1; + br_private_t *priv = NULL; - GF_OPTION_INIT ("scrubber", priv->iamscrubber, bool, free_priv); + if (!this->children) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_NO_CHILD, + "FATAL: no children"); + goto out; + } - ret = br_init_children (this, priv); - if (ret) - goto free_priv; + priv = GF_CALLOC(1, sizeof(*priv), gf_br_mt_br_private_t); + if (!priv) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, + "failed to allocate memory (->priv)"); + goto out; + } - pthread_mutex_init (&priv->lock, NULL); - pthread_cond_init (&priv->cond, NULL); + GF_OPTION_INIT("scrubber", priv->iamscrubber, bool, free_priv); - INIT_LIST_HEAD (&priv->bricks); - INIT_LIST_HEAD (&priv->signing); + ret = br_init_children(this, priv); + if (ret) + goto free_priv; - priv->timer_wheel = glusterfs_ctx_tw_get (this->ctx); - if (!priv->timer_wheel) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_TIMER_WHEEL_UNAVAILABLE, - "global timer wheel unavailable"); - goto cleanup; - } + pthread_mutex_init(&priv->lock, NULL); + pthread_cond_init(&priv->cond, NULL); - this->private = priv; + INIT_LIST_HEAD(&priv->bricks); + INIT_LIST_HEAD(&priv->signing); - if (!priv->iamscrubber) { - ret = br_signer_init (this, priv); - if (!ret) - ret = br_signer_handle_options (this, priv, NULL); - } else { - ret = br_scrubber_init (this, priv); - if (!ret) - ret = br_scrubber_handle_options (this, priv, NULL); - } + priv->timer_wheel = glusterfs_ctx_tw_get(this->ctx); + if (!priv->timer_wheel) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_TIMER_WHEEL_UNAVAILABLE, + "global timer wheel unavailable"); + goto cleanup; + } - if (ret) - goto cleanup; + this->private = priv; - ret = gf_thread_create (&priv->thread, NULL, br_handle_events, this, - "brhevent"); - if (ret != 0) { - gf_msg (this->name, GF_LOG_ERROR, -ret, - BRB_MSG_SPAWN_FAILED, "thread creation failed"); - ret = -1; - } + if (!priv->iamscrubber) { + ret = br_signer_init(this, priv); + if (!ret) + ret = br_signer_handle_options(this, priv, NULL); + } else { + ret = br_scrubber_init(this, priv); + if (!ret) + ret = br_scrubber_handle_options(this, priv, NULL); + } - if (!ret) { - gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_BITROT_LOADED, - "bit-rot xlator loaded in \"%s\" mode", - (priv->iamscrubber) ? "SCRUBBER" : "SIGNER"); - return 0; - } + if (ret) + goto cleanup; - cleanup: - (void) pthread_cond_destroy (&priv->cond); - (void) pthread_mutex_destroy (&priv->lock); + ret = gf_thread_create(&priv->thread, NULL, br_handle_events, this, + "brhevent"); + if (ret != 0) { + gf_msg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_SPAWN_FAILED, + "thread creation failed"); + ret = -1; + } - br_free_children (this, priv, priv->child_count); + if (!ret) { + gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_BITROT_LOADED, + "bit-rot xlator loaded in \"%s\" mode", + (priv->iamscrubber) ? "SCRUBBER" : "SIGNER"); + return 0; + } - free_priv: - GF_FREE (priv); - out: - this->private = NULL; - return -1; +cleanup: + (void)pthread_cond_destroy(&priv->cond); + (void)pthread_mutex_destroy(&priv->lock); + + br_free_children(this, priv, priv->child_count); + +free_priv: + GF_FREE(priv); +out: + this->private = NULL; + return -1; } void -fini (xlator_t *this) +fini(xlator_t *this) { - br_private_t *priv = this->private; + br_private_t *priv = this->private; - if (!priv) - return; + if (!priv) + return; - if (!priv->iamscrubber) - br_fini_signer (this, priv); - else - (void) br_free_scrubber_monitor (this, priv); + if (!priv->iamscrubber) + br_fini_signer(this, priv); + else + (void)br_free_scrubber_monitor(this, priv); - br_free_children (this, priv, priv->child_count); + br_free_children(this, priv, priv->child_count); - this->private = NULL; - GF_FREE (priv); + this->private = NULL; + GF_FREE(priv); - glusterfs_ctx_tw_put (this->ctx); + glusterfs_ctx_tw_put(this->ctx); - return; + return; } static void -br_reconfigure_monitor (xlator_t *this) +br_reconfigure_monitor(xlator_t *this) { - int32_t ret = 0; - - ret = br_scrub_state_machine (this, _gf_false); - if (ret) { - gf_msg (this->name, GF_LOG_ERROR, 0, - BRB_MSG_RESCHEDULE_SCRUBBER_FAILED, - "Could not reschedule scrubber for the volume. Scrubbing " - "will continue according to old frequency."); - } + int32_t ret = 0; + + ret = br_scrub_state_machine(this, _gf_false); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_RESCHEDULE_SCRUBBER_FAILED, + "Could not reschedule scrubber for the volume. Scrubbing " + "will continue according to old frequency."); + } } static int -br_reconfigure_scrubber (xlator_t *this, dict_t *options) +br_reconfigure_scrubber(xlator_t *this, dict_t *options) { - int32_t ret = -1; - br_private_t *priv = NULL; + int32_t ret = -1; + br_private_t *priv = NULL; - priv = this->private; + priv = this->private; - pthread_mutex_lock (&priv->lock); - { - ret = br_scrubber_handle_options (this, priv, options); - } - pthread_mutex_unlock (&priv->lock); + pthread_mutex_lock(&priv->lock); + { + ret = br_scrubber_handle_options(this, priv, options); + } + pthread_mutex_unlock(&priv->lock); - if (ret) - goto err; + if (ret) + goto err; - /* change state for all _up_ subvolume(s) */ - pthread_mutex_lock (&priv->lock); - { - br_reconfigure_monitor (this); - } - pthread_mutex_unlock (&priv->lock); + /* change state for all _up_ subvolume(s) */ + pthread_mutex_lock(&priv->lock); + { + br_reconfigure_monitor(this); + } + pthread_mutex_unlock(&priv->lock); - err: - return ret; +err: + return ret; } static int -br_reconfigure_signer (xlator_t *this, dict_t *options) +br_reconfigure_signer(xlator_t *this, dict_t *options) { - br_private_t *priv = this->private; + br_private_t *priv = this->private; - return br_signer_handle_options (this, priv, options); + return br_signer_handle_options(this, priv, options); } int -reconfigure (xlator_t *this, dict_t *options) +reconfigure(xlator_t *this, dict_t *options) { - int ret = 0; - br_private_t *priv = NULL; + int ret = 0; + br_private_t *priv = NULL; - priv = this->private; + priv = this->private; - if (priv->iamscrubber) - ret = br_reconfigure_scrubber (this, options); - else - ret = br_reconfigure_signer (this, options); + if (priv->iamscrubber) + ret = br_reconfigure_scrubber(this, options); + else + ret = br_reconfigure_signer(this, options); - return ret; + return ret; } struct xlator_fops fops; @@ -2145,49 +2161,55 @@ struct xlator_fops fops; struct xlator_cbks cbks; struct volume_options options[] = { - { .key = {"expiry-time"}, - .type = GF_OPTION_TYPE_INT, - .default_value = SIGNING_TIMEOUT, - .op_version = {GD_OP_VERSION_3_7_0}, - .flags = OPT_FLAG_SETTABLE, - .description = "Waiting time for an object on which it waits " - "before it is signed", - }, - { .key = {"brick-count"}, - .type = GF_OPTION_TYPE_STR, - .description = "Total number of bricks for the current node for " - "all volumes in the trusted storage pool.", - }, - { .key = {"scrubber", "scrub"}, - .type = GF_OPTION_TYPE_BOOL, - .default_value = "false", - .op_version = {GD_OP_VERSION_3_7_0}, - .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE, - .description = "option to run as a scrubber", - }, - { .key = {"scrub-throttle"}, - .type = GF_OPTION_TYPE_STR, - .default_value = "lazy", - .op_version = {GD_OP_VERSION_3_7_0}, - .flags = OPT_FLAG_SETTABLE, - .description = "Scrub-throttle value is a measure of how fast " - "or slow the scrubber scrubs the filesystem for " - "volume <VOLNAME>", - }, - { .key = {"scrub-freq"}, - .type = GF_OPTION_TYPE_STR, - .default_value = "biweekly", - .op_version = {GD_OP_VERSION_3_7_0}, - .flags = OPT_FLAG_SETTABLE, - .description = "Scrub frequency for volume <VOLNAME>", - }, - { .key = {"scrub-state"}, - .type = GF_OPTION_TYPE_STR, - .default_value = "active", - .op_version = {GD_OP_VERSION_4_0_0}, - .flags = OPT_FLAG_SETTABLE, - .description = "Pause/Resume scrub. Upon resume, scrubber " - "continues from where it left off.", - }, - { .key = {NULL} }, + { + .key = {"expiry-time"}, + .type = GF_OPTION_TYPE_INT, + .default_value = SIGNING_TIMEOUT, + .op_version = {GD_OP_VERSION_3_7_0}, + .flags = OPT_FLAG_SETTABLE, + .description = "Waiting time for an object on which it waits " + "before it is signed", + }, + { + .key = {"brick-count"}, + .type = GF_OPTION_TYPE_STR, + .description = "Total number of bricks for the current node for " + "all volumes in the trusted storage pool.", + }, + { + .key = {"scrubber", "scrub"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "false", + .op_version = {GD_OP_VERSION_3_7_0}, + .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE, + .description = "option to run as a scrubber", + }, + { + .key = {"scrub-throttle"}, + .type = GF_OPTION_TYPE_STR, + .default_value = "lazy", + .op_version = {GD_OP_VERSION_3_7_0}, + .flags = OPT_FLAG_SETTABLE, + .description = "Scrub-throttle value is a measure of how fast " + "or slow the scrubber scrubs the filesystem for " + "volume <VOLNAME>", + }, + { + .key = {"scrub-freq"}, + .type = GF_OPTION_TYPE_STR, + .default_value = "biweekly", + .op_version = {GD_OP_VERSION_3_7_0}, + .flags = OPT_FLAG_SETTABLE, + .description = "Scrub frequency for volume <VOLNAME>", + }, + { + .key = {"scrub-state"}, + .type = GF_OPTION_TYPE_STR, + .default_value = "active", + .op_version = {GD_OP_VERSION_4_0_0}, + .flags = OPT_FLAG_SETTABLE, + .description = "Pause/Resume scrub. Upon resume, scrubber " + "continues from where it left off.", + }, + {.key = {NULL}}, }; |