Xue jiufei
2015-Dec-07 03:34 UTC
[Ocfs2-devel] [PATCH] ocfs2/dlm: fix a race between purge and migratio
We found a race between purge and migration when doing code review.
Node A put lockres to purgelist before receiving the migrate message
from node B which is the master. dlm_mig_lockres_handler finds the
lockres and releases the dlm spinlock. Then dlm_thread on node A
can purge the lockres. dlm_mig_lockres_handler then gets the lockres
spinlock and sends assert master message to tell other nodes that node
A is the master now even lockres is purged. If node C set the master to
node A, and another node D master the lockres because no node respond
he is the master. That will make node C crash when node D send assert
master to node C. So check if lockres gets unhashed in
dlm_mig_lockres_handler to fix this race.
Signed-off-by: Jiufei Xue <xuejiufei at huawei.com>
Reviewed-by: Joseph Qi <joseph.qi at huawei.com>
Reviewed-by: Yiwen Jiang <jiangyiwen at huawei.com>
---
fs/ocfs2/dlm/dlmrecovery.c | 22 ++++++++++++++++++++++
1 file changed, 22 insertions(+)
diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
index 58eaa5c..84908a0 100644
--- a/fs/ocfs2/dlm/dlmrecovery.c
+++ b/fs/ocfs2/dlm/dlmrecovery.c
@@ -1400,11 +1400,33 @@ int dlm_mig_lockres_handler(struct o2net_msg *msg, u32
len, void *data,
/* lookup the lock to see if we have a secondary queue for this
* already... just add the locks in and this will have its owner
* and RECOVERY flag changed when it completes. */
+way_up_top:
res = dlm_lookup_lockres(dlm, mres->lockname, mres->lockname_len);
if (res) {
/* this will get a ref on res */
/* mark it as recovering/migrating and hash it */
spin_lock(&res->spinlock);
+
+ /*
+ * Right after dlm spinlock was released, dlm_thread could have
+ * purged the lockres. Check if lockres got unhashed. If so
+ * start over.
+ */
+ if (hlist_unhashed(&res->hash_node)) {
+ spin_unlock(&res->spinlock);
+ dlm_lockres_put(res);
+ goto way_up_top;
+ }
+
+ /* Wait on the resource purge to complete before continuing */
+ if (res->state & DLM_LOCK_RES_DROPPING_REF) {
+ __dlm_wait_on_lockres_flags(res,
+ DLM_LOCK_RES_DROPPING_REF);
+ spin_unlock(&res->spinlock);
+ dlm_lockres_put(res);
+ goto way_up_top;
+ }
+
if (mres->flags & DLM_MRES_RECOVERY) {
res->state |= DLM_LOCK_RES_RECOVERING;
} else {
--
1.8.4.3