From: wang di <di.wang@intel.com>
Date: Thu, 9 Jul 2015 03:11:53 +0000 (-0700)
Subject: LU-6837 update: re-lookup the dtrq in the replay list.
X-Git-Tag: 2.7.57~32
X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=commitdiff_plain;h=f836ef39fba1d6884aac9ff37479cbaac6a400b4

LU-6837 update: re-lookup the dtrq in the replay list.

In insert_update_records_to_replay_list(), if the
dtrq request has been added to the list by another
thread, it should lookup the request in the list
again, otherwise it will cause the following sub req
lookup panic, because the current dtrq was just freed.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I766cddbe5abcd96e97a611985032208585573b43
Reviewed-on: http://review.whamcloud.com/15573
Tested-by: Jenkins
Reviewed-by: Niu Yawei <yawei.niu@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
---

diff --git a/lustre/target/update_recovery.c b/lustre/target/update_recovery.c
index cad0183..3ec884d 100644
--- a/lustre/target/update_recovery.c
+++ b/lustre/target/update_recovery.c
@@ -368,6 +368,7 @@ insert_update_records_to_replay_list(struct target_distribute_txn_data *tdtd,
 	       " mdt_index %u\n", tdtd->tdtd_lut->lut_obd->obd_name,
 	       record->ur_batchid, record->ur_master_transno, mdt_index);
 
+again:
 	/* First try to build the replay update request with the records */
 	spin_lock(&tdtd->tdtd_replay_list_lock);
 	dtrq = dtrq_lookup(tdtd, record->ur_batchid);
@@ -391,7 +392,7 @@ insert_update_records_to_replay_list(struct target_distribute_txn_data *tdtd,
 		if (rc == -EEXIST) {
 			/* Some one else already add the record */
 			dtrq_destroy(dtrq);
-			rc = 0;
+			goto again;
 		}
 	} else {
 		struct update_records *dtrq_rec;