ocfs2: Plugs race between the dc thread and an unlock ast message

This patch plugs a race between the downconvert thread and an unlock ast message.
Specifically, after the downconvert worker has done its task, the dc thread needs
to check whether an unlock ast made the downconvert moot.

Reported-by: David Teigland <teigland@redhat.com>
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Acked-by: Mark Fasheh <mfasheh@sus.com>
Signed-off-by: Joel Becker <joel.becker@oracle.com>

diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
index ce8e061..e044019 100644 (file)
--- a/fs/ocfs2/dlmglue.c
+++ b/fs/ocfs2/dlmglue.c
@@ -3384,6 +3384,7 @@ static int ocfs2_unblock_lock(struct ocfs2_super *osb,
        unsigned long flags;
        int blocking;
        int new_level;
+       int level;
        int ret = 0;
        int set_lvb = 0;
        unsigned int gen;
@@ -3503,6 +3504,7 @@ recheck:
         * may sleep, so we save off a copy of what we're blocking as
         * it may change while we're not holding the spin lock. */
        blocking = lockres->l_blocking;
+       level = lockres->l_level;
        spin_unlock_irqrestore(&lockres->l_lock, flags);
 
        ctl->unblock_action = lockres->l_ops->downconvert_worker(lockres, blocking);
@@ -3511,7 +3513,7 @@ recheck:
                goto leave;
 
        spin_lock_irqsave(&lockres->l_lock, flags);
-       if (blocking != lockres->l_blocking) {
+       if ((blocking != lockres->l_blocking) || (level != lockres->l_level)) {
                /* If this changed underneath us, then we can't drop
                 * it just yet. */
                goto recheck;