[PATCH] Revert "raid5: make release_stripe lockless" because it causes test regression

[PATCH] Revert "raid5: make release_stripe lockless" because it causes test regression

[Mikulas Patocka]

Revert 773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6 and
d265d9dc1d25a69affc21ae9fe5004b9d09c10ef.

The lvm testsuite (from 
ftp://sources.redhat.com/pub/lvm2/LVM2.2.02.104.tgz) fails with unkillable 
lvm and dmsetup processes on kernel 3.12. The testsuite passes on 3.11. 
Bisect shows that the failure is caused by patch 
773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6. When I revert this patch on 3.12 
kernel, the testsuite finishes and there are no unkillable processes.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@kernel.org	# 3.12

---
 drivers/md/raid5.c |   70 +++--------------------------------------------------
 drivers/md/raid5.h |    3 --
 2 files changed, 5 insertions(+), 68 deletions(-)

Index: linux-3.12-fast/drivers/md/raid5.c
===================================================================
--- linux-3.12-fast.orig/drivers/md/raid5.c	2013-11-21 22:45:20.000000000 +0100
+++ linux-3.12-fast/drivers/md/raid5.c	2013-11-21 23:39:01.000000000 +0100
@@ -293,62 +293,12 @@ static void __release_stripe(struct r5co
 		do_release_stripe(conf, sh);
 }
 
-static struct llist_node *llist_reverse_order(struct llist_node *head)
-{
-	struct llist_node *new_head = NULL;
-
-	while (head) {
-		struct llist_node *tmp = head;
-		head = head->next;
-		tmp->next = new_head;
-		new_head = tmp;
-	}
-
-	return new_head;
-}
-
-/* should hold conf->device_lock already */
-static int release_stripe_list(struct r5conf *conf)
-{
-	struct stripe_head *sh;
-	int count = 0;
-	struct llist_node *head;
-
-	head = llist_del_all(&conf->released_stripes);
-	head = llist_reverse_order(head);
-	while (head) {
-		sh = llist_entry(head, struct stripe_head, release_list);
-		head = llist_next(head);
-		/* sh could be readded after STRIPE_ON_RELEASE_LIST is cleard */
-		smp_mb();
-		clear_bit(STRIPE_ON_RELEASE_LIST, &sh->state);
-		/*
-		 * Don't worry the bit is set here, because if the bit is set
-		 * again, the count is always > 1. This is true for
-		 * STRIPE_ON_UNPLUG_LIST bit too.
-		 */
-		__release_stripe(conf, sh);
-		count++;
-	}
-
-	return count;
-}
-
 static void release_stripe(struct stripe_head *sh)
 {
 	struct r5conf *conf = sh->raid_conf;
 	unsigned long flags;
-	bool wakeup;
 
-	if (test_and_set_bit(STRIPE_ON_RELEASE_LIST, &sh->state))
-		goto slow_path;
-	wakeup = llist_add(&sh->release_list, &conf->released_stripes);
-	if (wakeup)
-		md_wakeup_thread(conf->mddev->thread);
-	return;
-slow_path:
 	local_irq_save(flags);
-	/* we are ok here if STRIPE_ON_RELEASE_LIST is set or not */
 	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
 		do_release_stripe(conf, sh);
 		spin_unlock(&conf->device_lock);
@@ -596,8 +546,7 @@ get_active_stripe(struct r5conf *conf, s
 			if (atomic_read(&sh->count)) {
 				BUG_ON(!list_empty(&sh->lru)
 				    && !test_bit(STRIPE_EXPANDING, &sh->state)
-				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state)
-				    && !test_bit(STRIPE_ON_RELEASE_LIST, &sh->state));
+				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state));
 			} else {
 				if (!test_bit(STRIPE_HANDLE, &sh->state))
 					atomic_inc(&conf->active_stripes);
@@ -4293,10 +4242,6 @@ static void raid5_unplug(struct blk_plug
 			 */
 			smp_mb__before_clear_bit();
 			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
-			/*
-			 * STRIPE_ON_RELEASE_LIST could be set here. In that
-			 * case, the count is always > 1 here
-			 */
 			__release_stripe(conf, sh);
 			cnt++;
 		}
@@ -5007,13 +4952,11 @@ static void raid5_do_work(struct work_st
 	handled = 0;
 	spin_lock_irq(&conf->device_lock);
 	while (1) {
-		int batch_size, released;
-
-		released = release_stripe_list(conf);
+		int batch_size;
 
 		batch_size = handle_active_stripes(conf, group_id, worker);
 		worker->working = false;
-		if (!batch_size && !released)
+		if (!batch_size)
 			break;
 		handled += batch_size;
 	}
@@ -5048,9 +4991,7 @@ static void raid5d(struct md_thread *thr
 	spin_lock_irq(&conf->device_lock);
 	while (1) {
 		struct bio *bio;
-		int batch_size, released;
-
-		released = release_stripe_list(conf);
+		int batch_size;
 
 		if (
 		    !list_empty(&conf->bitmap_list)) {
@@ -5075,7 +5016,7 @@ static void raid5d(struct md_thread *thr
 		}
 
 		batch_size = handle_active_stripes(conf, ANY_GROUP, NULL);
-		if (!batch_size && !released)
+		if (!batch_size)
 			break;
 		handled += batch_size;
 
@@ -5503,7 +5444,6 @@ static struct r5conf *setup_conf(struct 
 	INIT_LIST_HEAD(&conf->delayed_list);
 	INIT_LIST_HEAD(&conf->bitmap_list);
 	INIT_LIST_HEAD(&conf->inactive_list);
-	init_llist_head(&conf->released_stripes);
 	atomic_set(&conf->active_stripes, 0);
 	atomic_set(&conf->preread_active_stripes, 0);
 	atomic_set(&conf->active_aligned_reads, 0);
Index: linux-3.12-fast/drivers/md/raid5.h
===================================================================
--- linux-3.12-fast.orig/drivers/md/raid5.h	2013-11-21 23:29:32.000000000 +0100
+++ linux-3.12-fast/drivers/md/raid5.h	2013-11-21 23:29:35.000000000 +0100
@@ -197,7 +197,6 @@ enum reconstruct_states {
 struct stripe_head {
 	struct hlist_node	hash;
 	struct list_head	lru;	      /* inactive_list or handle_list */
-	struct llist_node	release_list;
 	struct r5conf		*raid_conf;
 	short			generation;	/* increments with every
 						 * reshape */
@@ -324,7 +323,6 @@ enum {
 	STRIPE_OPS_REQ_PENDING,
 	STRIPE_ON_UNPLUG_LIST,
 	STRIPE_DISCARD,
-	STRIPE_ON_RELEASE_LIST,
 };
 
 /*
@@ -463,7 +461,6 @@ struct r5conf {
 	 */
 	atomic_t		active_stripes;
 	struct list_head	inactive_list;
-	struct llist_head	released_stripes;
 	wait_queue_head_t	wait_for_stripe;
 	wait_queue_head_t	wait_for_overlap;
 	int			inactive_blocked;	/* release of inactive stripes blocked,

Re: [PATCH] Revert "raid5: make release_stripe lockless" because it causes test regression

[Brassow Jonathan]

I have seen this and been working on a couple patches.

I have fixed 773ca8 by not calling mddev_suspend in raid_ctr and making the quiesce in mddev_resume contingent on whether the device has been suspended.  This has fixed the problem up to 3.12.

I am currently bisecting to find a new problem that has been introduced since then.

I hope to post the patches soon.

 brassow


On Nov 22, 2013, at 2:07 PM, Mikulas Patocka wrote:

> Revert 773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6 and
> d265d9dc1d25a69affc21ae9fe5004b9d09c10ef.
> 
> The lvm testsuite (from 
> ftp://sources.redhat.com/pub/lvm2/LVM2.2.02.104.tgz) fails with unkillable 
> lvm and dmsetup processes on kernel 3.12. The testsuite passes on 3.11. 
> Bisect shows that the failure is caused by patch 
> 773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6. When I revert this patch on 3.12 
> kernel, the testsuite finishes and there are no unkillable processes.
> 
> Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
> Cc: stable@kernel.org	# 3.12
> 
> ---
> drivers/md/raid5.c |   70 +++--------------------------------------------------
> drivers/md/raid5.h |    3 --
> 2 files changed, 5 insertions(+), 68 deletions(-)
> 
> Index: linux-3.12-fast/drivers/md/raid5.c
> ===================================================================
> --- linux-3.12-fast.orig/drivers/md/raid5.c	2013-11-21 22:45:20.000000000 +0100
> +++ linux-3.12-fast/drivers/md/raid5.c	2013-11-21 23:39:01.000000000 +0100
> @@ -293,62 +293,12 @@ static void __release_stripe(struct r5co
> 		do_release_stripe(conf, sh);
> }
> 
> -static struct llist_node *llist_reverse_order(struct llist_node *head)
> -{
> -	struct llist_node *new_head = NULL;
> -
> -	while (head) {
> -		struct llist_node *tmp = head;
> -		head = head->next;
> -		tmp->next = new_head;
> -		new_head = tmp;
> -	}
> -
> -	return new_head;
> -}
> -
> -/* should hold conf->device_lock already */
> -static int release_stripe_list(struct r5conf *conf)
> -{
> -	struct stripe_head *sh;
> -	int count = 0;
> -	struct llist_node *head;
> -
> -	head = llist_del_all(&conf->released_stripes);
> -	head = llist_reverse_order(head);
> -	while (head) {
> -		sh = llist_entry(head, struct stripe_head, release_list);
> -		head = llist_next(head);
> -		/* sh could be readded after STRIPE_ON_RELEASE_LIST is cleard */
> -		smp_mb();
> -		clear_bit(STRIPE_ON_RELEASE_LIST, &sh->state);
> -		/*
> -		 * Don't worry the bit is set here, because if the bit is set
> -		 * again, the count is always > 1. This is true for
> -		 * STRIPE_ON_UNPLUG_LIST bit too.
> -		 */
> -		__release_stripe(conf, sh);
> -		count++;
> -	}
> -
> -	return count;
> -}
> -
> static void release_stripe(struct stripe_head *sh)
> {
> 	struct r5conf *conf = sh->raid_conf;
> 	unsigned long flags;
> -	bool wakeup;
> 
> -	if (test_and_set_bit(STRIPE_ON_RELEASE_LIST, &sh->state))
> -		goto slow_path;
> -	wakeup = llist_add(&sh->release_list, &conf->released_stripes);
> -	if (wakeup)
> -		md_wakeup_thread(conf->mddev->thread);
> -	return;
> -slow_path:
> 	local_irq_save(flags);
> -	/* we are ok here if STRIPE_ON_RELEASE_LIST is set or not */
> 	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
> 		do_release_stripe(conf, sh);
> 		spin_unlock(&conf->device_lock);
> @@ -596,8 +546,7 @@ get_active_stripe(struct r5conf *conf, s
> 			if (atomic_read(&sh->count)) {
> 				BUG_ON(!list_empty(&sh->lru)
> 				    && !test_bit(STRIPE_EXPANDING, &sh->state)
> -				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state)
> -				    && !test_bit(STRIPE_ON_RELEASE_LIST, &sh->state));
> +				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state));
> 			} else {
> 				if (!test_bit(STRIPE_HANDLE, &sh->state))
> 					atomic_inc(&conf->active_stripes);
> @@ -4293,10 +4242,6 @@ static void raid5_unplug(struct blk_plug
> 			 */
> 			smp_mb__before_clear_bit();
> 			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
> -			/*
> -			 * STRIPE_ON_RELEASE_LIST could be set here. In that
> -			 * case, the count is always > 1 here
> -			 */
> 			__release_stripe(conf, sh);
> 			cnt++;
> 		}
> @@ -5007,13 +4952,11 @@ static void raid5_do_work(struct work_st
> 	handled = 0;
> 	spin_lock_irq(&conf->device_lock);
> 	while (1) {
> -		int batch_size, released;
> -
> -		released = release_stripe_list(conf);
> +		int batch_size;
> 
> 		batch_size = handle_active_stripes(conf, group_id, worker);
> 		worker->working = false;
> -		if (!batch_size && !released)
> +		if (!batch_size)
> 			break;
> 		handled += batch_size;
> 	}
> @@ -5048,9 +4991,7 @@ static void raid5d(struct md_thread *thr
> 	spin_lock_irq(&conf->device_lock);
> 	while (1) {
> 		struct bio *bio;
> -		int batch_size, released;
> -
> -		released = release_stripe_list(conf);
> +		int batch_size;
> 
> 		if (
> 		    !list_empty(&conf->bitmap_list)) {
> @@ -5075,7 +5016,7 @@ static void raid5d(struct md_thread *thr
> 		}
> 
> 		batch_size = handle_active_stripes(conf, ANY_GROUP, NULL);
> -		if (!batch_size && !released)
> +		if (!batch_size)
> 			break;
> 		handled += batch_size;
> 
> @@ -5503,7 +5444,6 @@ static struct r5conf *setup_conf(struct 
> 	INIT_LIST_HEAD(&conf->delayed_list);
> 	INIT_LIST_HEAD(&conf->bitmap_list);
> 	INIT_LIST_HEAD(&conf->inactive_list);
> -	init_llist_head(&conf->released_stripes);
> 	atomic_set(&conf->active_stripes, 0);
> 	atomic_set(&conf->preread_active_stripes, 0);
> 	atomic_set(&conf->active_aligned_reads, 0);
> Index: linux-3.12-fast/drivers/md/raid5.h
> ===================================================================
> --- linux-3.12-fast.orig/drivers/md/raid5.h	2013-11-21 23:29:32.000000000 +0100
> +++ linux-3.12-fast/drivers/md/raid5.h	2013-11-21 23:29:35.000000000 +0100
> @@ -197,7 +197,6 @@ enum reconstruct_states {
> struct stripe_head {
> 	struct hlist_node	hash;
> 	struct list_head	lru;	      /* inactive_list or handle_list */
> -	struct llist_node	release_list;
> 	struct r5conf		*raid_conf;
> 	short			generation;	/* increments with every
> 						 * reshape */
> @@ -324,7 +323,6 @@ enum {
> 	STRIPE_OPS_REQ_PENDING,
> 	STRIPE_ON_UNPLUG_LIST,
> 	STRIPE_DISCARD,
> -	STRIPE_ON_RELEASE_LIST,
> };
> 
> /*
> @@ -463,7 +461,6 @@ struct r5conf {
> 	 */
> 	atomic_t		active_stripes;
> 	struct list_head	inactive_list;
> -	struct llist_head	released_stripes;
> 	wait_queue_head_t	wait_for_stripe;
> 	wait_queue_head_t	wait_for_overlap;
> 	int			inactive_blocked;	/* release of inactive stripes blocked,
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] Revert "raid5: make release_stripe lockless" because it causes test regression

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 6489 bytes --]

On Fri, 22 Nov 2013 15:07:08 -0500 (EST) Mikulas Patocka
<mpatocka@redhat.com> wrote:

> Revert 773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6 and
> d265d9dc1d25a69affc21ae9fe5004b9d09c10ef.
> 
> The lvm testsuite (from 
> ftp://sources.redhat.com/pub/lvm2/LVM2.2.02.104.tgz) fails with unkillable 
> lvm and dmsetup processes on kernel 3.12. The testsuite passes on 3.11. 
> Bisect shows that the failure is caused by patch 
> 773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6. When I revert this patch on 3.12 
> kernel, the testsuite finishes and there are no unkillable processes.
> 
> Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
> Cc: stable@kernel.org	# 3.12

I certainly don't plan to revert these patches unless we fail to understand
and fix the problem.

Does commit ad4068de49862b083ac2a15bc50689bb30ce3e44 fix the problem?

Thanks,
NeilBrown

> 
> ---
>  drivers/md/raid5.c |   70 +++--------------------------------------------------
>  drivers/md/raid5.h |    3 --
>  2 files changed, 5 insertions(+), 68 deletions(-)
> 
> Index: linux-3.12-fast/drivers/md/raid5.c
> ===================================================================
> --- linux-3.12-fast.orig/drivers/md/raid5.c	2013-11-21 22:45:20.000000000 +0100
> +++ linux-3.12-fast/drivers/md/raid5.c	2013-11-21 23:39:01.000000000 +0100
> @@ -293,62 +293,12 @@ static void __release_stripe(struct r5co
>  		do_release_stripe(conf, sh);
>  }
>  
> -static struct llist_node *llist_reverse_order(struct llist_node *head)
> -{
> -	struct llist_node *new_head = NULL;
> -
> -	while (head) {
> -		struct llist_node *tmp = head;
> -		head = head->next;
> -		tmp->next = new_head;
> -		new_head = tmp;
> -	}
> -
> -	return new_head;
> -}
> -
> -/* should hold conf->device_lock already */
> -static int release_stripe_list(struct r5conf *conf)
> -{
> -	struct stripe_head *sh;
> -	int count = 0;
> -	struct llist_node *head;
> -
> -	head = llist_del_all(&conf->released_stripes);
> -	head = llist_reverse_order(head);
> -	while (head) {
> -		sh = llist_entry(head, struct stripe_head, release_list);
> -		head = llist_next(head);
> -		/* sh could be readded after STRIPE_ON_RELEASE_LIST is cleard */
> -		smp_mb();
> -		clear_bit(STRIPE_ON_RELEASE_LIST, &sh->state);
> -		/*
> -		 * Don't worry the bit is set here, because if the bit is set
> -		 * again, the count is always > 1. This is true for
> -		 * STRIPE_ON_UNPLUG_LIST bit too.
> -		 */
> -		__release_stripe(conf, sh);
> -		count++;
> -	}
> -
> -	return count;
> -}
> -
>  static void release_stripe(struct stripe_head *sh)
>  {
>  	struct r5conf *conf = sh->raid_conf;
>  	unsigned long flags;
> -	bool wakeup;
>  
> -	if (test_and_set_bit(STRIPE_ON_RELEASE_LIST, &sh->state))
> -		goto slow_path;
> -	wakeup = llist_add(&sh->release_list, &conf->released_stripes);
> -	if (wakeup)
> -		md_wakeup_thread(conf->mddev->thread);
> -	return;
> -slow_path:
>  	local_irq_save(flags);
> -	/* we are ok here if STRIPE_ON_RELEASE_LIST is set or not */
>  	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
>  		do_release_stripe(conf, sh);
>  		spin_unlock(&conf->device_lock);
> @@ -596,8 +546,7 @@ get_active_stripe(struct r5conf *conf, s
>  			if (atomic_read(&sh->count)) {
>  				BUG_ON(!list_empty(&sh->lru)
>  				    && !test_bit(STRIPE_EXPANDING, &sh->state)
> -				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state)
> -				    && !test_bit(STRIPE_ON_RELEASE_LIST, &sh->state));
> +				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state));
>  			} else {
>  				if (!test_bit(STRIPE_HANDLE, &sh->state))
>  					atomic_inc(&conf->active_stripes);
> @@ -4293,10 +4242,6 @@ static void raid5_unplug(struct blk_plug
>  			 */
>  			smp_mb__before_clear_bit();
>  			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
> -			/*
> -			 * STRIPE_ON_RELEASE_LIST could be set here. In that
> -			 * case, the count is always > 1 here
> -			 */
>  			__release_stripe(conf, sh);
>  			cnt++;
>  		}
> @@ -5007,13 +4952,11 @@ static void raid5_do_work(struct work_st
>  	handled = 0;
>  	spin_lock_irq(&conf->device_lock);
>  	while (1) {
> -		int batch_size, released;
> -
> -		released = release_stripe_list(conf);
> +		int batch_size;
>  
>  		batch_size = handle_active_stripes(conf, group_id, worker);
>  		worker->working = false;
> -		if (!batch_size && !released)
> +		if (!batch_size)
>  			break;
>  		handled += batch_size;
>  	}
> @@ -5048,9 +4991,7 @@ static void raid5d(struct md_thread *thr
>  	spin_lock_irq(&conf->device_lock);
>  	while (1) {
>  		struct bio *bio;
> -		int batch_size, released;
> -
> -		released = release_stripe_list(conf);
> +		int batch_size;
>  
>  		if (
>  		    !list_empty(&conf->bitmap_list)) {
> @@ -5075,7 +5016,7 @@ static void raid5d(struct md_thread *thr
>  		}
>  
>  		batch_size = handle_active_stripes(conf, ANY_GROUP, NULL);
> -		if (!batch_size && !released)
> +		if (!batch_size)
>  			break;
>  		handled += batch_size;
>  
> @@ -5503,7 +5444,6 @@ static struct r5conf *setup_conf(struct 
>  	INIT_LIST_HEAD(&conf->delayed_list);
>  	INIT_LIST_HEAD(&conf->bitmap_list);
>  	INIT_LIST_HEAD(&conf->inactive_list);
> -	init_llist_head(&conf->released_stripes);
>  	atomic_set(&conf->active_stripes, 0);
>  	atomic_set(&conf->preread_active_stripes, 0);
>  	atomic_set(&conf->active_aligned_reads, 0);
> Index: linux-3.12-fast/drivers/md/raid5.h
> ===================================================================
> --- linux-3.12-fast.orig/drivers/md/raid5.h	2013-11-21 23:29:32.000000000 +0100
> +++ linux-3.12-fast/drivers/md/raid5.h	2013-11-21 23:29:35.000000000 +0100
> @@ -197,7 +197,6 @@ enum reconstruct_states {
>  struct stripe_head {
>  	struct hlist_node	hash;
>  	struct list_head	lru;	      /* inactive_list or handle_list */
> -	struct llist_node	release_list;
>  	struct r5conf		*raid_conf;
>  	short			generation;	/* increments with every
>  						 * reshape */
> @@ -324,7 +323,6 @@ enum {
>  	STRIPE_OPS_REQ_PENDING,
>  	STRIPE_ON_UNPLUG_LIST,
>  	STRIPE_DISCARD,
> -	STRIPE_ON_RELEASE_LIST,
>  };
>  
>  /*
> @@ -463,7 +461,6 @@ struct r5conf {
>  	 */
>  	atomic_t		active_stripes;
>  	struct list_head	inactive_list;
> -	struct llist_head	released_stripes;
>  	wait_queue_head_t	wait_for_stripe;
>  	wait_queue_head_t	wait_for_overlap;
>  	int			inactive_blocked;	/* release of inactive stripes blocked,


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [PATCH] Revert "raid5: make release_stripe lockless" because it causes test regression

[Mikulas Patocka]

On Mon, 25 Nov 2013, NeilBrown wrote:

> On Fri, 22 Nov 2013 15:07:08 -0500 (EST) Mikulas Patocka
> <mpatocka@redhat.com> wrote:
> 
> > Revert 773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6 and
> > d265d9dc1d25a69affc21ae9fe5004b9d09c10ef.
> > 
> > The lvm testsuite (from 
> > ftp://sources.redhat.com/pub/lvm2/LVM2.2.02.104.tgz) fails with unkillable 
> > lvm and dmsetup processes on kernel 3.12. The testsuite passes on 3.11. 
> > Bisect shows that the failure is caused by patch 
> > 773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6. When I revert this patch on 3.12 
> > kernel, the testsuite finishes and there are no unkillable processes.
> > 
> > Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
> > Cc: stable@kernel.org	# 3.12
> 
> I certainly don't plan to revert these patches unless we fail to understand
> and fix the problem.
> 
> Does commit ad4068de49862b083ac2a15bc50689bb30ce3e44 fix the problem?
> 
> Thanks,
> NeilBrown

Yes. This patch, applied to 3.12, fixes it.

Mikulas