LU-8197 ptlrpc: set correct rq_timeout for normal replay reply

Fix from f3ce48ce543791d60326777d95ab2c9c03826965 introduced a
regression which set larger rq_timeout of normal replay reply
mistakenly, that result in incorrect estimated service time
increase during recovery.

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Change-Id: If62c94f6147e6d6d9ae5954e4847725e9cd66a67
Reviewed-on: https://review.whamcloud.com/23833
Tested-by: Jenkins
Reviewed-by: Lai Siyao <lai.siyao@intel.com>
Reviewed-by: Hongchao Zhang <hongchao.zhang@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

diff --git a/lustre/ptlrpc/niobuf.c b/lustre/ptlrpc/niobuf.c
index b0f375e..49b9c64 100644 (file)
--- a/lustre/ptlrpc/niobuf.c
+++ b/lustre/ptlrpc/niobuf.c
@@ -507,6 +507,7 @@ static void ptlrpc_at_set_reply(struct ptlrpc_request *req, int flags)
                __u32 timeout;
 
                if (req->rq_export && req->rq_reqmsg != NULL &&
+                   (flags & PTLRPC_REPLY_EARLY) &&
                    lustre_msg_get_flags(req->rq_reqmsg) &
                    (MSG_REPLAY | MSG_REQ_REPLAY_DONE | MSG_LOCK_REPLAY_DONE))
                        timeout = cfs_time_current_sec() -