On Mon, 12 Sep 2011 08:33:31 +0300 "Moshe Melnikov" <moshe@xxxxxxxxxxxxxxxxx>
wrote:
> I can reproduce it very easily.
>
> I don't know how to apply this patch. I have very limited knowledge in Linux
> kernel.
If you can't build a kernel there isn't much I can do to help.
If you are using a distro kernel, maybe report the bug and the fix to the
distro and they might release a new kernel in due course.
Thanks for confirming that it is easily reproduced. That challenged me to
examine the problem again and I see that I was missing something, and can see
exactly how the set of actions you described would cause exactly that crash.
This patch fixes it properly and will be going upstream shortly.
thanks for your help,
NeilBrown
From 01f96c0a9922cd9919baf9d16febdf7016177a12 Mon Sep 17 00:00:00 2001
From: NeilBrown <neilb@xxxxxxx>
Date: Wed, 21 Sep 2011 15:30:20 +1000
Subject: [PATCH] md: Avoid waking up a thread after it has been freed.
Two related problems:
1/ some error paths call "md_unregister_thread(mddev->thread)"
without subsequently clearing ->thread. A subsequent call
to mddev_unlock will try to wake the thread, and crash.
2/ Most calls to md_wakeup_thread are protected against the thread
disappeared either by:
- holding the ->mutex
- having an active request, so something else must be keeping
the array active.
However mddev_unlock calls md_wakeup_thread after dropping the
mutex and without any certainty of an active request, so the
->thread could theoretically disappear.
So we need a spinlock to provide some protections.
So change md_unregister_thread to take a pointer to the thread
pointer, and ensure that it always does the required locking, and
clears the pointer properly.
Reported-by: "Moshe Melnikov" <moshe@xxxxxxxxxxxxxxxxx>
Signed-off-by: NeilBrown <neilb@xxxxxxx>
cc: stable@xxxxxxxxxx
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 5404b22..5c95ccb 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -61,6 +61,11 @@
static void autostart_arrays(int part);
#endif
+/* pers_list is a list of registered personalities protected
+ * by pers_lock.
+ * pers_lock does extra service to protect accesses to
+ * mddev->thread when the mutex cannot be held.
+ */
static LIST_HEAD(pers_list);
static DEFINE_SPINLOCK(pers_lock);
@@ -739,7 +744,12 @@ static void mddev_unlock(mddev_t * mddev)
} else
mutex_unlock(&mddev->reconfig_mutex);
+ /* was we've dropped the mutex we need a spinlock to
+ * make sur the thread doesn't disappear
+ */
+ spin_lock(&pers_lock);
md_wakeup_thread(mddev->thread);
+ spin_unlock(&pers_lock);
}
static mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr)
@@ -6429,11 +6439,18 @@ mdk_thread_t *md_register_thread(void (*run) (mddev_t *), mddev_t *mddev,
return thread;
}
-void md_unregister_thread(mdk_thread_t *thread)
+void md_unregister_thread(mdk_thread_t **threadp)
{
+ mdk_thread_t *thread = *threadp;
if (!thread)
return;
dprintk("interrupting MD-thread pid %d\n", task_pid_nr(thread->tsk));
+ /* Locking ensures that mddev_unlock does not wake_up a
+ * non-existent thread
+ */
+ spin_lock(&pers_lock);
+ *threadp = NULL;
+ spin_unlock(&pers_lock);
kthread_stop(thread->tsk);
kfree(thread);
@@ -7340,8 +7357,7 @@ static void reap_sync_thread(mddev_t *mddev)
mdk_rdev_t *rdev;
/* resync has finished, collect result */
- md_unregister_thread(mddev->sync_thread);
- mddev->sync_thread = NULL;
+ md_unregister_thread(&mddev->sync_thread);
if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
/* success...*/
diff --git a/drivers/md/md.h b/drivers/md/md.h
index 1e586bb..0a309dc 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -560,7 +560,7 @@ extern int register_md_personality(struct mdk_personality *p);
extern int unregister_md_personality(struct mdk_personality *p);
extern mdk_thread_t * md_register_thread(void (*run) (mddev_t *mddev),
mddev_t *mddev, const char *name);
-extern void md_unregister_thread(mdk_thread_t *thread);
+extern void md_unregister_thread(mdk_thread_t **threadp);
extern void md_wakeup_thread(mdk_thread_t *thread);
extern void md_check_recovery(mddev_t *mddev);
extern void md_write_start(mddev_t *mddev, struct bio *bi);
diff --git a/drivers/md/multipath.c b/drivers/md/multipath.c
index 3535c23..d5b5fb3 100644
--- a/drivers/md/multipath.c
+++ b/drivers/md/multipath.c
@@ -514,8 +514,7 @@ static int multipath_stop (mddev_t *mddev)
{
multipath_conf_t *conf = mddev->private;
- md_unregister_thread(mddev->thread);
- mddev->thread = NULL;
+ md_unregister_thread(&mddev->thread);
blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
mempool_destroy(conf->pool);
kfree(conf->multipaths);
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index f4622dd..d9587df 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -2562,8 +2562,7 @@ static int stop(mddev_t *mddev)
raise_barrier(conf);
lower_barrier(conf);
- md_unregister_thread(mddev->thread);
- mddev->thread = NULL;
+ md_unregister_thread(&mddev->thread);
if (conf->r1bio_pool)
mempool_destroy(conf->r1bio_pool);
kfree(conf->mirrors);
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index d7a8468..0cd9672 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -2955,7 +2955,7 @@ static int run(mddev_t *mddev)
return 0;
out_free_conf:
- md_unregister_thread(mddev->thread);
+ md_unregister_thread(&mddev->thread);
if (conf->r10bio_pool)
mempool_destroy(conf->r10bio_pool);
safe_put_page(conf->tmppage);
@@ -2973,8 +2973,7 @@ static int stop(mddev_t *mddev)
raise_barrier(conf, 0);
lower_barrier(conf);
- md_unregister_thread(mddev->thread);
- mddev->thread = NULL;
+ md_unregister_thread(&mddev->thread);
blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
if (conf->r10bio_pool)
mempool_destroy(conf->r10bio_pool);
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 43709fa..ac5e8b5 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -4941,8 +4941,7 @@ static int run(mddev_t *mddev)
return 0;
abort:
- md_unregister_thread(mddev->thread);
- mddev->thread = NULL;
+ md_unregister_thread(&mddev->thread);
if (conf) {
print_raid5_conf(conf);
free_conf(conf);
@@ -4956,8 +4955,7 @@ static int stop(mddev_t *mddev)
{
raid5_conf_t *conf = mddev->private;
- md_unregister_thread(mddev->thread);
- mddev->thread = NULL;
+ md_unregister_thread(&mddev->thread);
if (mddev->queue)
mddev->queue->backing_dev_info.congested_fn = NULL;
free_conf(conf);
Attachment:
signature.asc
Description: PGP signature
