Commit a9ee4c8a authored by Kurt Hackel's avatar Kurt Hackel Committed by Mark Fasheh
Browse files

ocfs2: better error handling during assert master message 



handle errors during lock assert master by either killing self or other node

Signed-off-by: default avatarKurt Hackel <kurt.hackel@oracle.com>
Signed-off-by: default avatarMark Fasheh <mark.fasheh@oracle.com>
parent a7f90d83

fs/ocfs2/dlm/dlmmaster.c

+14 −4
Original line number Original line Diff line number Diff line
@@ -1633,6 +1633,8 @@ static int dlm_do_assert_master(struct dlm_ctxt *dlm, const char *lockname, 
	dlm_node_iter_init(nodemap, &iter);

	dlm_node_iter_init(nodemap, &iter);

	while ((to = dlm_node_iter_next(&iter)) >= 0) {

	while ((to = dlm_node_iter_next(&iter)) >= 0) {

		int r = 0;

		int r = 0;

		struct dlm_master_list_entry *mle = NULL;



		mlog(0, "sending assert master to %d (%.*s)\n", to,

		mlog(0, "sending assert master to %d (%.*s)\n", to,

		     namelen, lockname);

		     namelen, lockname);

		memset(&assert, 0, sizeof(assert));

		memset(&assert, 0, sizeof(assert));
@@ -1657,7 +1659,15 @@ static int dlm_do_assert_master(struct dlm_ctxt *dlm, const char *lockname, 
			/* ok, something horribly messed.  kill thyself. */

			/* ok, something horribly messed.  kill thyself. */

			mlog(ML_ERROR,"during assert master of %.*s to %u, "

			mlog(ML_ERROR,"during assert master of %.*s to %u, "

			     "got %d.\n", namelen, lockname, to, r);

			     "got %d.\n", namelen, lockname, to, r);

			dlm_dump_lock_resources(dlm);

			spin_lock(&dlm->spinlock);

			spin_lock(&dlm->master_lock);

			if (dlm_find_mle(dlm, &mle, (char *)lockname,

					 namelen)) {

				dlm_print_one_mle(mle);

				__dlm_put_mle(mle);

			}

			spin_unlock(&dlm->master_lock);

			spin_unlock(&dlm->spinlock);

			BUG();

			BUG();

		} else if (r == EAGAIN) {

		} else if (r == EAGAIN) {

			mlog(0, "%.*s: node %u create mles on other "

			mlog(0, "%.*s: node %u create mles on other "
@@ -1922,12 +1932,12 @@ int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data) 




kill:

kill:

	/* kill the caller! */

	/* kill the caller! */

	mlog(ML_ERROR, "Bad message received from another node.  Dumping state "

	     "and killing the other node now!  This node is OK and can continue.\n");

	__dlm_print_one_lock_resource(res);

	spin_unlock(&res->spinlock);

	spin_unlock(&res->spinlock);

	spin_unlock(&dlm->spinlock);

	spin_unlock(&dlm->spinlock);

	dlm_lockres_put(res);

	dlm_lockres_put(res);

	mlog(ML_ERROR, "Bad message received from another node.  Dumping state "

	     "and killing the other node now!  This node is OK and can continue.\n");

	dlm_dump_lock_resources(dlm);

	dlm_put(dlm);

	dlm_put(dlm);

	return -EINVAL;

	return -EINVAL;

}

}