From 74678fcc65207e965b74466c254af4feec002984 Mon Sep 17 00:00:00 2001
From: shadow <shadow>
Date: Thu, 24 Apr 2008 11:14:25 +0000
Subject: [PATCH] don't put request into delay list while invalidate in flight.
restore generation check. b=15069 i=johan i=rread
---
lustre/ChangeLog | 7 +++++++
lustre/mgc/mgc_request.c | 1 -
lustre/ptlrpc/client.c | 33 ++++++++++++++++++++-------------
lustre/ptlrpc/events.c | 2 --
lustre/ptlrpc/niobuf.c | 6 +-----
5 files changed, 28 insertions(+), 21 deletions(-)
diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index e29af998ed..09be0432d7 100644
--- a/lustre/ChangeLog
+++ b/lustre/ChangeLog
@@ -12,6 +12,13 @@ tbd Sun Microsystems, Inc.
* RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
removed cwd "./" (refer to Bugzilla 14399).
+Severity : normal
+Bugzilla : 15069
+Description: don't put request into delay list while invalidate in flight.
+Details : ptlrpc_delay_request sometimes put in delay list while invalidate
+ import in flight. this produce timeout for invalidate and sometimes
+ can cause stale data.
+
Severity : minor
Bugzilla : 14856
Frequency : on ppc only
diff --git a/lustre/mgc/mgc_request.c b/lustre/mgc/mgc_request.c
index ee5aaa9575..dab90012cd 100644
--- a/lustre/mgc/mgc_request.c
+++ b/lustre/mgc/mgc_request.c
@@ -868,7 +868,6 @@ static int mgc_import_event(struct obd_device *obd,
switch (event) {
case IMP_EVENT_DISCON:
/* MGC imports should not wait for recovery */
- ptlrpc_invalidate_import(imp);
break;
case IMP_EVENT_INACTIVE:
break;
diff --git a/lustre/ptlrpc/client.c b/lustre/ptlrpc/client.c
index 6e138034b1..f4b6a80eb6 100644
--- a/lustre/ptlrpc/client.c
+++ b/lustre/ptlrpc/client.c
@@ -663,26 +663,34 @@ static int ptlrpc_import_delay_req(struct obd_import *imp,
} else if (req->rq_send_state == LUSTRE_IMP_CONNECTING &&
imp->imp_state == LUSTRE_IMP_CONNECTING) {
/* allow CONNECT even if import is invalid */ ;
- } else if (imp->imp_invalid && (!imp->imp_recon_bk ||
- imp->imp_obd->obd_no_recov)) {
+ if (atomic_read(&imp->imp_inval_count) != 0) {
+ DEBUG_REQ(D_ERROR, req, "invalidate in flight");
+ *status = -EIO;
+ }
+
+ } else if ((imp->imp_invalid && (!imp->imp_recon_bk)) ||
+ imp->imp_obd->obd_no_recov) {
/* If the import has been invalidated (such as by an OST
- * failure), and if the import(MGC) tried all of its connection
- * list (Bug 13464), the request must fail with -ESHUTDOWN.
+ * failure), and if the import(MGC) tried all of its connection
+ * list (Bug 13464), the request must fail with -ESHUTDOWN.
* This indicates the requests should be discarded; an -EIO
- * may result in a resend of the request. */
+ * may result in a resend of the request. */
if (!imp->imp_deactive)
- DEBUG_REQ(D_ERROR, req, "IMP_INVALID");
+ DEBUG_REQ(D_ERROR, req, "IMP_INVALID");
*status = -ESHUTDOWN; /* bz 12940 */
} else if (req->rq_import_generation != imp->imp_generation) {
DEBUG_REQ(D_ERROR, req, "req wrong generation:");
*status = -EIO;
} else if (req->rq_send_state != imp->imp_state) {
- if (imp->imp_obd->obd_no_recov)
- *status = -ESHUTDOWN;
- else if (imp->imp_dlm_fake || req->rq_no_delay)
+ /* invalidate in progress - any requests should be drop */
+ if (atomic_read(&imp->imp_inval_count) != 0) {
+ DEBUG_REQ(D_ERROR, req, "invalidate in flight");
+ *status = -EIO;
+ } else if (imp->imp_dlm_fake || req->rq_no_delay) {
*status = -EWOULDBLOCK;
- else
+ } else {
delay = 1;
+ }
}
RETURN(delay);
@@ -1168,9 +1176,8 @@ check_ctx:
libcfs_nid2str(imp->imp_connection->c_peer.nid),
lustre_msg_get_opc(req->rq_reqmsg));
- set->set_remaining--;
-
atomic_dec(&imp->imp_inflight);
+ set->set_remaining--;
cfs_waitq_signal(&imp->imp_recovery_waitq);
}
@@ -1758,8 +1765,8 @@ int ptlrpc_queue_wait(struct ptlrpc_request *req)
req->rq_phase = RQ_PHASE_RPC;
spin_lock(&imp->imp_lock);
-restart:
req->rq_import_generation = imp->imp_generation;
+restart:
if (ptlrpc_import_delay_req(imp, req, &rc)) {
list_del(&req->rq_list);
diff --git a/lustre/ptlrpc/events.c b/lustre/ptlrpc/events.c
index affe7dd3c9..d24c92d477 100644
--- a/lustre/ptlrpc/events.c
+++ b/lustre/ptlrpc/events.c
@@ -69,8 +69,6 @@ void request_out_callback(lnet_event_t *ev)
ptlrpc_wake_client_req(req);
}
- /* these balance the references in ptl_send_rpc() */
- atomic_dec(&req->rq_import->imp_inflight);
ptlrpc_req_finished(req);
EXIT;
diff --git a/lustre/ptlrpc/niobuf.c b/lustre/ptlrpc/niobuf.c
index 8c9f09a586..d4c8fe04e9 100644
--- a/lustre/ptlrpc/niobuf.c
+++ b/lustre/ptlrpc/niobuf.c
@@ -505,9 +505,8 @@ int ptl_send_rpc(struct ptlrpc_request *request, int noreply)
request->rq_reply_portal);
}
- /* add references on request and import for request_out_callback */
+ /* add references on request for request_out_callback */
ptlrpc_request_addref(request);
- atomic_inc(&request->rq_import->imp_inflight);
if (obd->obd_svc_stats != NULL)
lprocfs_counter_add(obd->obd_svc_stats, PTLRPC_REQACTIVE_CNTR,
request->rq_import->imp_inflight.counter);
@@ -528,10 +527,7 @@ int ptl_send_rpc(struct ptlrpc_request *request, int noreply)
RETURN(rc);
}
- /* drop request_out_callback refs, we couldn't start the send */
- atomic_dec(&request->rq_import->imp_inflight);
ptlrpc_req_finished(request);
-
if (noreply)
RETURN(rc);
else
--
GitLab