From 995c1c5e542b639f42ef2fcf58927b3941e0bd5a Mon Sep 17 00:00:00 2001
From: shadow <shadow>
Date: Tue, 20 May 2008 06:45:57 +0000
Subject: [PATCH] fix ASSERTION(client_stat->nid_exp_ref_count == 0) during run
acc-small. b=15139 i=tappro i=komal
---
lustre/ChangeLog | 7 ++++
lustre/include/obd.h | 2 +-
lustre/obdclass/class_hash.c | 2 +-
lustre/obdclass/lprocfs_status.c | 58 ++++++++++++++++++++------------
lustre/obdclass/obd_config.c | 2 +-
5 files changed, 46 insertions(+), 25 deletions(-)
diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index 63a30f0d3c..8b1978375b 100644
--- a/lustre/ChangeLog
+++ b/lustre/ChangeLog
@@ -24,6 +24,13 @@ tbd Sun Microsystems, Inc.
'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
For more information, please refer to bugzilla 13904.
+Severity : normal
+Bugzilla : 15139
+Frequency : rare, on clear nid stats
+Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
+Details : when clean nid stats sometimes try destroy live entry,
+ and this produce panic in free.
+
Severity : major
Bugzilla : 15575
Description: Stack overflow during MDS log replay
diff --git a/lustre/include/obd.h b/lustre/include/obd.h
index 94593cb04d..40e196de8a 100644
--- a/lustre/include/obd.h
+++ b/lustre/include/obd.h
@@ -793,7 +793,7 @@ struct obd_device {
cfs_waitq_t obd_llog_waitq;
struct list_head obd_exports;
int obd_num_exports;
- spinlock_t nid_lock;
+ spinlock_t obd_nid_lock;
struct ldlm_namespace *obd_namespace;
struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
/* a spinlock is OK for what we do now, may need a semaphore later */
diff --git a/lustre/obdclass/class_hash.c b/lustre/obdclass/class_hash.c
index c578564251..880cd2bdec 100644
--- a/lustre/obdclass/class_hash.c
+++ b/lustre/obdclass/class_hash.c
@@ -669,7 +669,7 @@ void nidstats_refcount_put(struct hlist_node * actual_hnode)
data = hlist_entry(actual_hnode, struct nid_stat, nid_hash);
data->nid_exp_ref_count--;
-
+ EXIT;
}
/*******************************************************************************/
diff --git a/lustre/obdclass/lprocfs_status.c b/lustre/obdclass/lprocfs_status.c
index 9615614c6d..ac9990fc74 100644
--- a/lustre/obdclass/lprocfs_status.c
+++ b/lustre/obdclass/lprocfs_status.c
@@ -753,10 +753,8 @@ int lprocfs_obd_cleanup(struct obd_device *obd)
return 0;
}
-void lprocfs_free_client_stats(void *obj, void *data)
+static void lprocfs_free_client_stats(struct nid_stat *client_stat)
{
- struct nid_stat *client_stat = obj;
-
CDEBUG(D_CONFIG, "stat %p - data %p/%p/%p\n", client_stat,
client_stat->nid_proc, client_stat->nid_stats,
client_stat->nid_brw_stats);
@@ -765,7 +763,6 @@ void lprocfs_free_client_stats(void *obj, void *data)
client_stat->nid_exp_ref_count);
hlist_del_init(&client_stat->nid_hash);
- list_del(&client_stat->nid_list);
if (client_stat->nid_proc)
lprocfs_remove(&client_stat->nid_proc);
@@ -787,10 +784,12 @@ void lprocfs_free_per_client_stats(struct obd_device *obd)
ENTRY;
/* we need extra list - because hash_exit called to early */
+ /* not need locking because all clients is died */
while(!list_empty(&obd->obd_nid_stats)) {
stat = list_entry(obd->obd_nid_stats.next,
struct nid_stat, nid_list);
- lprocfs_free_client_stats(stat, NULL);
+ list_del_init(&stat->nid_list);
+ lprocfs_free_client_stats(stat);
}
EXIT;
@@ -1255,24 +1254,29 @@ EXPORT_SYMBOL(lprocfs_nid_stats_clear_read);
void lprocfs_nid_stats_clear_write_cb(void *obj, void *data)
{
- struct nid_stat *client_stat = obj;
+ struct nid_stat *stat = obj;
int i;
- if(client_stat->nid_exp_ref_count == 1) {
- hlist_del_init(&client_stat->nid_hash);
- lprocfs_free_client_stats(client_stat, data);
- OBD_FREE(client_stat, sizeof(struct nid_stat));
+ /* object has only hash + iterate_all references.
+ * add/delete blocked by hash bucket lock */
+ CDEBUG(D_INFO,"refcnt %d\n", stat->nid_exp_ref_count);
+ if(stat->nid_exp_ref_count == 2) {
+ hlist_del_init(&stat->nid_hash);
+ stat->nid_exp_ref_count--;
+ spin_lock(&stat->nid_obd->obd_nid_lock);
+ list_del_init(&stat->nid_list);
+ spin_unlock(&stat->nid_obd->obd_nid_lock);
+ list_add(&stat->nid_list, data);
EXIT;
return;
}
/* we has reference to object - only clear data*/
- if (client_stat->nid_stats) {
- lprocfs_clear_stats(client_stat->nid_stats);
- }
+ if (stat->nid_stats)
+ lprocfs_clear_stats(stat->nid_stats);
- if (client_stat->nid_brw_stats) {
+ if (stat->nid_brw_stats) {
for (i = 0; i < BRW_LAST; i++)
- lprocfs_oh_clear(&client_stat->nid_brw_stats->hist[i]);
+ lprocfs_oh_clear(&stat->nid_brw_stats->hist[i]);
}
EXIT;
return;
@@ -1283,10 +1287,17 @@ int lprocfs_nid_stats_clear_write(struct file *file, const char *buffer,
unsigned long count, void *data)
{
struct obd_device *obd = (struct obd_device *)data;
-
+ struct nid_stat *client_stat;
+ CFS_LIST_HEAD(free_list);
lustre_hash_iterate_all(obd->obd_nid_stats_hash_body,
- lprocfs_free_client_stats, NULL);
+ lprocfs_nid_stats_clear_write_cb, &free_list);
+
+ while (!list_empty(&free_list)) {
+ client_stat = list_entry(free_list.next, struct nid_stat, nid_list);
+ list_del_init(&client_stat->nid_list);
+ lprocfs_free_client_stats(client_stat);
+ }
return count;
}
@@ -1323,6 +1334,11 @@ int lprocfs_exp_setup(struct obd_export *exp, lnet_nid_t *nid, int *newnid)
tmp->nid_obd = exp->exp_obd;
tmp->nid_exp_ref_count = 1; /* need live in hash after destroy export */
+ /* protect competitive add to list, not need locking on destroy */
+ spin_lock(&obd->obd_nid_lock);
+ list_add(&tmp->nid_list, &obd->obd_nid_stats);
+ spin_unlock(&obd->obd_nid_lock);
+
tmp1= lustre_hash_findadd_unique(obd->obd_nid_stats_hash_body, nid,
&tmp->nid_hash);
CDEBUG(D_INFO, "Found stats %p for nid %s - ref %d\n",
@@ -1348,16 +1364,14 @@ int lprocfs_exp_setup(struct obd_export *exp, lnet_nid_t *nid, int *newnid)
if (rc)
CWARN("Error adding the uuid file\n");
- /* protect competitive add to list, not need locking on destroy */
- spin_lock(&obd->nid_lock);
- list_add(&tmp->nid_list, &obd->obd_nid_stats);
- spin_unlock(&obd->nid_lock);
-
exp->exp_nid_stats = tmp;
*newnid = 1;
RETURN(rc);
destroy_new:
+ spin_lock(&obd->obd_nid_lock);
+ list_del(&tmp->nid_list);
+ spin_unlock(&obd->obd_nid_lock);
OBD_FREE(tmp, sizeof(struct nid_stat));
RETURN(rc);
}
diff --git a/lustre/obdclass/obd_config.c b/lustre/obdclass/obd_config.c
index eba45b5e22..0f98e59013 100644
--- a/lustre/obdclass/obd_config.c
+++ b/lustre/obdclass/obd_config.c
@@ -191,7 +191,7 @@ int class_attach(struct lustre_cfg *lcfg)
CFS_INIT_LIST_HEAD(&obd->obd_exports);
CFS_INIT_LIST_HEAD(&obd->obd_exports_timed);
CFS_INIT_LIST_HEAD(&obd->obd_nid_stats);
- spin_lock_init(&obd->nid_lock);
+ spin_lock_init(&obd->obd_nid_lock);
spin_lock_init(&obd->obd_dev_lock);
sema_init(&obd->obd_dev_sem, 1);
spin_lock_init(&obd->obd_osfs_lock);
--
GitLab