diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index 749966abe3465e865226af6a88fb0fb08f204689..af97f6a7b69316990e7959eeb2fd5d8072894246 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 : minor
+Bugzilla : 15716
+Description: timeout with invalidate import.
+Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
+ handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
+ ptlrpc_invalidate_import and as result LASSERT.
+
Severity : normal
Frequency : only with broken builds/installations
Bugzilla : 15779
diff --git a/lustre/include/linux/obd_class.h b/lustre/include/linux/obd_class.h
index 75a77cd333d271e23c2c68ecd229a526c456dfde..dcb5ba12eb65b25afe7344be7927b39c311328f9 100644
--- a/lustre/include/linux/obd_class.h
+++ b/lustre/include/linux/obd_class.h
@@ -38,10 +38,6 @@
#include <linux/timer.h>
#endif
-void obd_zombie_impexp_init(void);
-void obd_zombie_impexp_cull(void);
-extern void (*obd_zombie_impexp_notify)(void);
-
/* obdo.c */
#ifdef __KERNEL__
void obdo_from_iattr(struct obdo *oa, struct iattr *attr, unsigned ia_valid);
diff --git a/lustre/include/obd_class.h b/lustre/include/obd_class.h
index 01dad5a1267e09156ea114b6466acae2aed86335..a6d1483fce3ce8abfd0005152e62588a6f8a5232 100644
--- a/lustre/include/obd_class.h
+++ b/lustre/include/obd_class.h
@@ -84,6 +84,10 @@ char *obd_export_nid2str(struct obd_export *exp);
int obd_export_evict_by_nid(struct obd_device *obd, char *nid);
int obd_export_evict_by_uuid(struct obd_device *obd, char *uuid);
+int obd_zombie_impexp_init(void);
+void obd_zombie_impexp_stop(void);
+void obd_zombie_impexp_cull(void);
+
/* obd_config.c */
int class_process_config(struct lustre_cfg *lcfg);
int class_process_proc_param(char *prefix, struct lprocfs_vars *lvars,
diff --git a/lustre/mds/mds_lov.c b/lustre/mds/mds_lov.c
index f377629045709788118bfbe122a3500ed7ffe064..3495f3a5e15ca54f775ceee7d4d90120ea25297e 100644
--- a/lustre/mds/mds_lov.c
+++ b/lustre/mds/mds_lov.c
@@ -140,7 +140,7 @@ void mds_lov_destroy_objids(struct obd_device *obd)
}
if (mds->mds_lov_objid_filp) {
- rc = filp_close((struct file *)mds->mds_lov_objid_filp, 0);
+ rc = filp_close((struct file *)mds->mds_lov_objid_filp, NULL);
mds->mds_lov_objid_filp = NULL;
if (rc)
CERROR("%s file won't close, rc=%d\n", LOV_OBJID, rc);
@@ -235,7 +235,7 @@ out:
int mds_lov_write_objids(struct obd_device *obd)
{
struct mds_obd *mds = &obd->u.mds;
- int i, rc = 0;
+ int i = 0, rc = 0;
ENTRY;
if (cfs_bitmap_check_empty(mds->mds_lov_page_dirty))
@@ -254,6 +254,7 @@ int mds_lov_write_objids(struct obd_device *obd)
if (i == mds->mds_lov_objid_lastpage)
size = (mds->mds_lov_objid_lastidx + 1) * sizeof(obd_id);
+ CDEBUG(D_INFO,"write %lld - %ld\n", off, size);
rc = fsfilt_write_record(obd, mds->mds_lov_objid_filp, data,
size, &off, 0);
if (rc < 0)
diff --git a/lustre/obdclass/class_obd.c b/lustre/obdclass/class_obd.c
index c97ca561ad852536981de691102a1503f55bb74e..ef4cc0ecdb6658c44eac91ebf7e55b253c43b565 100644
--- a/lustre/obdclass/class_obd.c
+++ b/lustre/obdclass/class_obd.c
@@ -610,6 +610,7 @@ static void cleanup_obdclass(void)
class_handle_cleanup();
class_exit_uuidlist();
+ obd_zombie_impexp_stop();
memory_leaked = obd_memory_sum();
pages_leaked = obd_pages_sum();
diff --git a/lustre/obdclass/genops.c b/lustre/obdclass/genops.c
index 57eaa22990f6c58678be0f2c4bf548e9d10adfc4..83053aeefb010644f3e2ffc48d42d2a48c6d6738 100644
--- a/lustre/obdclass/genops.c
+++ b/lustre/obdclass/genops.c
@@ -45,9 +45,7 @@ cfs_mem_cache_t *import_cachep;
struct list_head obd_zombie_imports;
struct list_head obd_zombie_exports;
spinlock_t obd_zombie_impexp_lock;
-void (*obd_zombie_impexp_notify)(void) = NULL;
-EXPORT_SYMBOL(obd_zombie_impexp_notify);
-
+static void obd_zombie_impexp_notify(void);
int (*ptlrpc_put_connection_superhack)(struct ptlrpc_connection *c);
@@ -624,8 +622,7 @@ void __class_export_put(struct obd_export *exp)
list_add(&exp->exp_obd_chain, &obd_zombie_exports);
spin_unlock(&obd_zombie_impexp_lock);
- if (obd_zombie_impexp_notify != NULL)
- obd_zombie_impexp_notify();
+ obd_zombie_impexp_notify();
}
}
EXPORT_SYMBOL(__class_export_put);
@@ -766,8 +763,7 @@ void class_import_put(struct obd_import *import)
list_add(&import->imp_zombie_chain, &obd_zombie_imports);
spin_unlock(&obd_zombie_impexp_lock);
- if (obd_zombie_impexp_notify != NULL)
- obd_zombie_impexp_notify();
+ obd_zombie_impexp_notify();
}
EXIT;
@@ -1282,11 +1278,11 @@ int obd_export_evict_by_uuid(struct obd_device *obd, char *uuid)
}
EXPORT_SYMBOL(obd_export_evict_by_uuid);
-void obd_zombie_impexp_cull(void)
+void obd_zombie_impexp_cull(void)
{
struct obd_import *import;
struct obd_export *export;
-
+
do {
spin_lock (&obd_zombie_impexp_lock);
@@ -1297,7 +1293,7 @@ void obd_zombie_impexp_cull(void)
imp_zombie_chain);
list_del(&import->imp_zombie_chain);
}
-
+
export = NULL;
if (!list_empty(&obd_zombie_exports)) {
export = list_entry(obd_zombie_exports.next,
@@ -1307,7 +1303,7 @@ void obd_zombie_impexp_cull(void)
}
spin_unlock(&obd_zombie_impexp_lock);
-
+
if (import != NULL)
class_import_destroy(import);
@@ -1316,11 +1312,121 @@ void obd_zombie_impexp_cull(void)
} while (import != NULL || export != NULL);
}
-EXPORT_SYMBOL(obd_zombie_impexp_cull);
-void obd_zombie_impexp_init(void)
+static struct completion obd_zombie_start;
+static struct completion obd_zombie_stop;
+static unsigned long obd_zombie_flags;
+static cfs_waitq_t obd_zombie_waitq;
+
+enum {
+ OBD_ZOMBIE_STOP = 1
+};
+
+int obd_zombi_impexp_check(void *arg)
{
+ int rc;
+
+ spin_lock(&obd_zombie_impexp_lock);
+ rc = list_empty(&obd_zombie_imports) &&
+ list_empty(&obd_zombie_exports) &&
+ !test_bit(OBD_ZOMBIE_STOP, &obd_zombie_flags);
+
+ spin_unlock(&obd_zombie_impexp_lock);
+
+ RETURN(rc);
+}
+
+static void obd_zombie_impexp_notify(void)
+{
+ cfs_waitq_signal(&obd_zombie_waitq);
+}
+
+#ifdef __KERNEL__
+
+static int obd_zombie_impexp_thread(void *unused)
+{
+ int rc;
+
+ if ((rc = cfs_daemonize_ctxt("obd_zombid"))) {
+ complete(&obd_zombie_start);
+ RETURN(rc);
+ }
+
+ complete(&obd_zombie_start);
+
+ while(!test_bit(OBD_ZOMBIE_STOP, &obd_zombie_flags)) {
+ struct l_wait_info lwi = { 0 };
+
+ l_wait_event(obd_zombie_waitq, !obd_zombi_impexp_check(NULL), &lwi);
+
+ obd_zombie_impexp_cull();
+ }
+
+ complete(&obd_zombie_stop);
+
+ RETURN(0);
+}
+
+#else /* ! KERNEL */
+
+static atomic_t zombi_recur = ATOMIC_INIT(0);
+static void *obd_zombi_impexp_work_cb;
+static void *obd_zombi_impexp_idle_cb;
+
+int obd_zombi_impexp_kill(void *arg)
+{
+ int rc = 0;
+
+ if (atomic_inc_return(&zombi_recur) == 1) {
+ obd_zombie_impexp_cull();
+ rc = 1;
+ }
+ atomic_dec(&zombi_recur);
+ return rc;
+}
+
+#endif
+
+int obd_zombie_impexp_init(void)
+{
+ int rc;
+
CFS_INIT_LIST_HEAD(&obd_zombie_imports);
CFS_INIT_LIST_HEAD(&obd_zombie_exports);
spin_lock_init(&obd_zombie_impexp_lock);
+ init_completion(&obd_zombie_start);
+ init_completion(&obd_zombie_stop);
+ cfs_waitq_init(&obd_zombie_waitq);
+
+#ifdef __KERNEL__
+ rc = cfs_kernel_thread(obd_zombie_impexp_thread, NULL, 0);
+ if (rc < 0)
+ RETURN(rc);
+
+ wait_for_completion(&obd_zombie_start);
+#else
+
+ obd_zombi_impexp_work_cb =
+ liblustre_register_wait_callback("obd_zombi_impexp_kill",
+ &obd_zombi_impexp_kill, NULL);
+
+ obd_zombi_impexp_idle_cb =
+ liblustre_register_idle_callback("obd_zombi_impexp_check",
+ &obd_zombi_impexp_check, NULL);
+ rc = 0;
+
+#endif
+ RETURN(rc);
+}
+
+void obd_zombie_impexp_stop(void)
+{
+ set_bit(OBD_ZOMBIE_STOP, &obd_zombie_flags);
+ obd_zombie_impexp_notify();
+#ifdef __KERNEL__
+ wait_for_completion(&obd_zombie_stop);
+#else
+ liblustre_deregister_wait_callback(obd_zombi_impexp_work_cb);
+ liblustre_deregister_idle_callback(obd_zombi_impexp_idle_cb);
+#endif
}
diff --git a/lustre/obdclass/obd_config.c b/lustre/obdclass/obd_config.c
index cf8c393752a09d290fb6d205230500562407f3cc..924c6fb48f5d6ccab7440bdb83c049f1e5425910 100644
--- a/lustre/obdclass/obd_config.c
+++ b/lustre/obdclass/obd_config.c
@@ -359,10 +359,10 @@ int class_detach(struct obd_device *obd, struct lustre_cfg *lcfg)
obd->obd_name, obd->obd_uuid.uuid);
class_decref(obd);
-
+
/* not strictly necessary, but cleans up eagerly */
obd_zombie_impexp_cull();
-
+
RETURN(0);
}
diff --git a/lustre/ptlrpc/import.c b/lustre/ptlrpc/import.c
index 8d60653d6ee0053220e52a018147bd808d099deb..e65d1457a0e2c310ea10089325df34e8d568f88d 100644
--- a/lustre/ptlrpc/import.c
+++ b/lustre/ptlrpc/import.c
@@ -234,6 +234,7 @@ void ptlrpc_invalidate_import(struct obd_import *imp)
DEBUG_REQ(D_ERROR, req, "still on delayed list");
}
spin_unlock(&imp->imp_lock);
+ LASSERT(atomic_read(&imp->imp_inflight) == 0);
}
out:
diff --git a/lustre/ptlrpc/ptlrpcd.c b/lustre/ptlrpc/ptlrpcd.c
index d0abb2eb85e0fc5f94fb5ace7546ad09c5591e82..5b82f570f03ae206127b07deaeae56d45a4e598d 100644
--- a/lustre/ptlrpc/ptlrpcd.c
+++ b/lustre/ptlrpc/ptlrpcd.c
@@ -97,8 +97,6 @@ static int ptlrpcd_check(struct ptlrpcd_ctl *pc)
if (test_bit(LIOD_STOP, &pc->pc_flags))
RETURN(1);
- obd_zombie_impexp_cull();
-
spin_lock(&pc->pc_set->set_new_req_lock);
list_for_each_safe(pos, tmp, &pc->pc_set->set_new_requests) {
req = list_entry(pos, struct ptlrpc_request, rq_set_chain);
@@ -176,13 +174,6 @@ static int ptlrpcd(void *arg)
return 0;
}
-static void ptlrpcd_zombie_impexp_notify(void)
-{
- 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)
@@ -233,9 +224,6 @@ static int ptlrpcd_start(char *name, struct ptlrpcd_ctl *pc)
RETURN(-ENOMEM);
#ifdef __KERNEL__
- /* wake ptlrpcd when zombie imports or exports exist */
- obd_zombie_impexp_notify = ptlrpcd_zombie_impexp_notify;
-
rc = cfs_kernel_thread(ptlrpcd, pc, 0);
if (rc < 0) {
ptlrpc_set_destroy(pc->pc_set);
@@ -260,7 +248,6 @@ static void ptlrpcd_stop(struct ptlrpcd_ctl *pc)
set_bit(LIOD_STOP, &pc->pc_flags);
cfs_waitq_signal(&pc->pc_set->set_waitq);
#ifdef __KERNEL__
- obd_zombie_impexp_notify = NULL;
wait_for_completion(&pc->pc_finishing);
#else
liblustre_deregister_wait_callback(pc->pc_wait_callback);
diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index e3b7f9750a62a8cb414385893b7e4c0c822d8c2c..446852da2511c91de29da28d0baf547ae878d56c 100644
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -800,6 +800,7 @@ test_22() {
echo Client mount before any osts are in the logs
mount_client $MOUNT
check_mount && return 41
+ umount_client $MOUNT
pass
echo Client mount with ost in logs, but none running
diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 0717e187c940b116bd59d230feeeb41aa9ed6151..aa5a1ddd8b92fe89b3638a01333db4f33842b499 100644
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -268,16 +268,16 @@ mount_facet() {
local dev=${facet}_dev
local opt=${facet}_opt
echo "Starting ${facet}: ${!opt} $@ ${!dev} ${MOUNT%/*}/${facet}"
+ do_facet ${facet} "lctl set_param debug=$PTLDEBUG; \
+ lctl set_param subsystem_debug=${SUBSYSTEM# }; \
+ lctl set_param debug_mb=${DEBUG_SIZE}; \
+ sync"
+
do_facet ${facet} mount -t lustre ${!opt} $@ ${!dev} ${MOUNT%/*}/${facet}
RC=${PIPESTATUS[0]}
if [ $RC -ne 0 ]; then
echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
echo "Start of ${device} on ${facet} failed ${RC}"
- else
- do_facet ${facet} "lctl set_param debug=$PTLDEBUG; \
- lctl set_param subsystem_debug=${SUBSYSTEM# }; \
- lctl set_param debug_mb=${DEBUG_SIZE}; \
- sync"
fi
return $RC
}
@@ -337,11 +337,12 @@ zconf_mount() {
echo "Starting client: $client: $OPTIONS $device $mnt"
do_node $client mkdir -p $mnt
- do_node $client mount -t lustre $OPTIONS $device $mnt || return 1
-
do_node $client "lctl set_param debug=$PTLDEBUG;
lctl set_param subsystem_debug=${SUBSYSTEM# };
lctl set_param debug_mb=${DEBUG_SIZE}"
+
+ do_node $client mount -t lustre $OPTIONS $device $mnt || return 1
+
[ -d /r ] && $LCTL modules > /r/tmp/ogdb-$HOSTNAME
return 0
}