Grab a lov llog lock so we ensure the llog records for a single transno

[fs/lustre-release.git] / lustre / lov / lov_log.c

diff --git a/lustre/lov/lov_log.c b/lustre/lov/lov_log.c
index c6174d7..9a483bb 100644 (file)
--- a/lustre/lov/lov_log.c
+++ b/lustre/lov/lov_log.c
@@ -77,6 +77,12 @@ static int lov_llog_origin_add(struct llog_ctxt *ctxt, struct llog_rec_hdr *rec,
 
         LASSERT(logcookies && numcookies >= lsm->lsm_stripe_count);
 
+        /* We need this to serialize llog records between parallel unlinks so
+         * we can replay llog records in strict transno and llog order.  If
+         * and when we want to make this more scalable we need to lock and
+         * write records in strictly ost_idx order not lsm order.  Consider
+         * file 1 on ost_idx [1, 2, 3, 4] and file 2 on ost_idx [3, 4, 1, 2] */
+        down(&lov->lov_llog_sem);
         for (i = 0,loi = lsm->lsm_oinfo; i < lsm->lsm_stripe_count; i++,loi++) {
                 struct obd_device *child =
                         lov->tgts[loi->loi_ost_idx].ltd_exp->exp_obd; 
@@ -90,6 +96,7 @@ static int lov_llog_origin_add(struct llog_ctxt *ctxt, struct llog_rec_hdr *rec,
                                numcookies - rc, NULL,
                                lock != NULL ? lock + rc : NULL, lock_count);
         }
+        up(&lov->lov_llog_sem);
         OBD_FREE(lur, sizeof(*lur));
 
         RETURN(rc);