glusterd: acquire lock to update volinfo structure

Problem: With commit cb0339f92, we are using a separate syntask
for restart_bricks. There can be a situation where two threads
are accessing the same volinfo structure at the same time and
updating volinfo structure. This can lead volinfo to have
inconsistent values and assertion failures because of unexpected
values.

Solution: While updating the volinfo structure, acquire a
store_volinfo_lock, and release the lock only when the thread
completed its critical section part.

Fixes: bz#1627610
Signed-off-by: Sanju Rakonde <srakonde@redhat.com>

Change-Id: I545e4e2368e3285d8f7aa28081ff4448abb72f5d
This commit is contained in:
Sanju Rakonde 2018-09-11 14:19:42 +05:30 committed by Atin Mukherjee
parent d345ac0954
commit 484f417da9
3 changed files with 43 additions and 35 deletions

View File

@ -1785,44 +1785,47 @@ glusterd_store_volinfo(glusterd_volinfo_t *volinfo,
GF_ASSERT(volinfo);
glusterd_perform_volinfo_version_action(volinfo, ac);
ret = glusterd_store_create_volume_dir(volinfo);
if (ret)
goto out;
pthread_mutex_lock(&volinfo->store_volinfo_lock);
{
glusterd_perform_volinfo_version_action(volinfo, ac);
ret = glusterd_store_create_volume_dir(volinfo);
if (ret)
goto unlock;
ret = glusterd_store_create_volume_run_dir(volinfo);
if (ret)
goto out;
ret = glusterd_store_create_volume_run_dir(volinfo);
if (ret)
goto unlock;
ret = glusterd_store_create_vol_shandle_on_absence(volinfo);
if (ret)
goto out;
ret = glusterd_store_create_vol_shandle_on_absence(volinfo);
if (ret)
goto unlock;
ret = glusterd_store_create_nodestate_sh_on_absence(volinfo);
if (ret)
goto out;
ret = glusterd_store_create_nodestate_sh_on_absence(volinfo);
if (ret)
goto unlock;
ret = glusterd_store_perform_volume_store(volinfo);
if (ret)
goto out;
ret = glusterd_store_perform_volume_store(volinfo);
if (ret)
goto unlock;
ret = glusterd_store_volume_atomic_update(volinfo);
if (ret) {
glusterd_perform_volinfo_version_action(
volinfo, GLUSTERD_VOLINFO_VER_AC_DECREMENT);
goto out;
ret = glusterd_store_volume_atomic_update(volinfo);
if (ret) {
glusterd_perform_volinfo_version_action(
volinfo, GLUSTERD_VOLINFO_VER_AC_DECREMENT);
goto unlock;
}
ret = glusterd_store_perform_node_state_store(volinfo);
if (ret)
goto unlock;
/* checksum should be computed at the end */
ret = glusterd_compute_cksum(volinfo, _gf_false);
if (ret)
goto unlock;
}
ret = glusterd_store_perform_node_state_store(volinfo);
if (ret)
goto out;
/* checksum should be computed at the end */
ret = glusterd_compute_cksum(volinfo, _gf_false);
if (ret)
goto out;
out:
unlock:
pthread_mutex_unlock(&volinfo->store_volinfo_lock);
if (ret)
glusterd_store_volume_cleanup_tmp(volinfo);

View File

@ -2283,6 +2283,8 @@ glusterd_op_create_volume(dict_t *dict, char **op_errstr)
goto out;
}
pthread_mutex_init(&volinfo->store_volinfo_lock, NULL);
ret = dict_get_strn(dict, "volname", SLEN("volname"), &volname);
if (ret) {

View File

@ -494,9 +494,12 @@ struct glusterd_volinfo_ {
glusterd_tierdsvc_t tierd;
glusterd_gfproxydsvc_t gfproxyd;
int32_t quota_xattr_version;
gf_boolean_t stage_deleted; /* volume has passed staging
* for delete operation
*/
gf_boolean_t stage_deleted; /* volume has passed staging
* for delete operation
*/
pthread_mutex_t store_volinfo_lock; /* acquire lock for
* updating the volinfo
*/
};
typedef enum gd_snap_status_ {