From 2cc418892732f5750584a908d91795b8d136a7f5 Mon Sep 17 00:00:00 2001
From: shadow <shadow>
Date: Thu, 25 Oct 2007 16:52:02 +0000
Subject: [PATCH] interrupt oig_wait produce painc on resend.
b=13888
i=nikita
i=alex
---
lustre/ChangeLog | 10 ++++++++++
lustre/include/lustre_net.h | 1 -
lustre/osc/osc_request.c | 10 +++++++---
lustre/ptlrpc/ptlrpcd.c | 28 ++++++++++------------------
lustre/tests/sanity.sh | 18 ++++++++++++++++++
5 files changed, 45 insertions(+), 22 deletions(-)
diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index 46bc23b301..0e9e25dac5 100644
--- a/lustre/ChangeLog
+++ b/lustre/ChangeLog
@@ -26,6 +26,16 @@ Bugzilla : 13556
Description: conf-sanity.sh test_33 failed with 1
Details : change mgsnode
+Severity : normal
+Bugzilla : 13888
+Description: interrupt oig_wait produce painc on resend.
+Details : brw_redo_request can be used for resend requests from ptlrpcd and
+ private set, and this produce situation when rq_ptlrpcd_data not
+ copyed to new allocated request and triggered LBUG on assert
+ req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
+ ptlrpcd set if request is changed and can be safety changed to use
+ rq_set directly.
+
--------------------------------------------------------------------------------
2007-10-26 Cluster File Systems, Inc. <info@clusterfs.com>
diff --git a/lustre/include/lustre_net.h b/lustre/include/lustre_net.h
index e2ec068a21..d59777c0af 100644
--- a/lustre/include/lustre_net.h
+++ b/lustre/include/lustre_net.h
@@ -355,7 +355,6 @@ struct ptlrpc_request {
struct ptlrpc_request_set *rq_set;
void *rq_interpret_reply; /* Async completion handler */
union ptlrpc_async_args rq_async_args; /* Async completion context */
- void *rq_ptlrpcd_data;
struct ptlrpc_request_pool *rq_pool; /* Pool if request from
preallocated list */
};
diff --git a/lustre/osc/osc_request.c b/lustre/osc/osc_request.c
index 04df048063..25e265651c 100644
--- a/lustre/osc/osc_request.c
+++ b/lustre/osc/osc_request.c
@@ -1341,12 +1341,16 @@ int osc_brw_redo_request(struct ptlrpc_request *request,
oap->oap_request = ptlrpc_request_addref(new_req);
}
}
- client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
-
- DEBUG_REQ(D_INFO, new_req, "new request");
+ /* use ptlrpc_set_add_req is safe because interpret functions work
+ * in check_set context. only one way exist with access to request
+ * from different thread got -EINTR - this way protected with
+ * cl_loi_list_lock */
ptlrpc_set_add_req(set, new_req);
+ client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
+
+ DEBUG_REQ(D_INFO, new_req, "new request");
RETURN(0);
}
diff --git a/lustre/ptlrpc/ptlrpcd.c b/lustre/ptlrpc/ptlrpcd.c
index 16c448c218..d0abb2eb85 100644
--- a/lustre/ptlrpc/ptlrpcd.c
+++ b/lustre/ptlrpc/ptlrpcd.c
@@ -48,8 +48,6 @@ struct ptlrpcd_ctl {
spinlock_t pc_lock;
struct completion pc_starting;
struct completion pc_finishing;
- struct list_head pc_req_list;
- cfs_waitq_t pc_waitq;
struct ptlrpc_request_set *pc_set;
char pc_name[16];
#ifndef __KERNEL__
@@ -67,11 +65,11 @@ static int ptlrpcd_users = 0;
void ptlrpcd_wake(struct ptlrpc_request *req)
{
- struct ptlrpcd_ctl *pc = req->rq_ptlrpcd_data;
+ struct ptlrpc_request_set *rq_set = req->rq_set;
- LASSERT(pc != NULL);
+ LASSERT(rq_set != NULL);
- cfs_waitq_signal(&pc->pc_waitq);
+ cfs_waitq_signal(&rq_set->set_waitq);
}
/* requests that are added to the ptlrpcd queue are sent via
@@ -85,9 +83,8 @@ void ptlrpcd_add_req(struct ptlrpc_request *req)
else
pc = &ptlrpcd_recovery_pc;
- req->rq_ptlrpcd_data = pc;
ptlrpc_set_add_new_req(pc->pc_set, req);
- wake_up(&pc->pc_waitq);
+ cfs_waitq_signal(&pc->pc_set->set_waitq);
}
static int ptlrpcd_check(struct ptlrpcd_ctl *pc)
@@ -161,19 +158,13 @@ static int ptlrpcd(void *arg)
* on the set's new_req_list and ptlrpcd_check moves them into
* the set. */
while (1) {
- cfs_waitlink_t set_wait;
struct l_wait_info lwi;
cfs_duration_t timeout;
timeout = cfs_time_seconds(ptlrpc_set_next_timeout(pc->pc_set));
lwi = LWI_TIMEOUT(timeout, ptlrpc_expired_set, pc->pc_set);
- /* ala the pinger, wait on pc's waitqueue and the set's */
- cfs_waitlink_init(&set_wait);
- cfs_waitq_add(&pc->pc_set->set_waitq, &set_wait);
- cfs_waitq_forward(&set_wait, &pc->pc_waitq);
- l_wait_event(pc->pc_waitq, ptlrpcd_check(pc), &lwi);
- cfs_waitq_del(&pc->pc_set->set_waitq, &set_wait);
+ l_wait_event(pc->pc_set->set_waitq, ptlrpcd_check(pc), &lwi);
if (test_bit(LIOD_STOP, &pc->pc_flags))
break;
@@ -187,8 +178,11 @@ static int ptlrpcd(void *arg)
static void ptlrpcd_zombie_impexp_notify(void)
{
- cfs_waitq_signal(&ptlrpcd_pc.pc_waitq);
+ LASSERT(ptlrpcd_pc.pc_set != NULL); // call before ptlrpcd inited ?
+
+ cfs_waitq_signal(&ptlrpcd_pc.pc_set->set_waitq);
}
+
#else
int ptlrpcd_check_async_rpcs(void *arg)
@@ -230,10 +224,8 @@ static int ptlrpcd_start(char *name, struct ptlrpcd_ctl *pc)
memset(pc, 0, sizeof(*pc));
init_completion(&pc->pc_starting);
init_completion(&pc->pc_finishing);
- cfs_waitq_init(&pc->pc_waitq);
pc->pc_flags = 0;
spin_lock_init(&pc->pc_lock);
- CFS_INIT_LIST_HEAD(&pc->pc_req_list);
snprintf (pc->pc_name, sizeof (pc->pc_name), name);
pc->pc_set = ptlrpc_prep_set();
@@ -266,7 +258,7 @@ static int ptlrpcd_start(char *name, struct ptlrpcd_ctl *pc)
static void ptlrpcd_stop(struct ptlrpcd_ctl *pc)
{
set_bit(LIOD_STOP, &pc->pc_flags);
- cfs_waitq_signal(&pc->pc_waitq);
+ cfs_waitq_signal(&pc->pc_set->set_waitq);
#ifdef __KERNEL__
obd_zombie_impexp_notify = NULL;
wait_for_completion(&pc->pc_finishing);
diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh
index 9525d16994..8932d821a4 100644
--- a/lustre/tests/sanity.sh
+++ b/lustre/tests/sanity.sh
@@ -4282,6 +4282,24 @@ test_118j() {
}
run_test 118j "Simulate unrecoverable OST side error =========="
+test_118k()
+{
+ #define OBD_FAIL_OST_BRW_WRITE_BULK 0x20e
+ do_facet ost sysctl -w lustre.fail_loc=0x20e
+ mkdir -p $DIR/$tdir
+
+ for ((i=0;i<10;i++)); do
+ dd if=/dev/zero of=$DIR/$tdir/$tdir-$i bs=1M count=10 &
+ SLEEPPID=$!
+ sleep 0.500s
+ kill $SLEEPPID
+ wait $SLEEPPID
+ done
+
+ sysctl -w lustre.fail_loc=0
+}
+run_test 118k "bio alloc -ENOMEM and IO TERM handling ========="
+
test_119a() # bug 11737
{
BSIZE=$((512 * 1024))
--
GitLab