All Virtuozzo development lists (kernel + QEMU)
 help / color / mirror / Atom feed
* [Devel] [PATCH hci-8.1 v2 0/2] drivers/vhost/blk: add support for worker ioctls
@ 2026-11-15 10:28 Alexander Ivanov
  2026-11-15 10:28 ` [Devel] [PATCH hci-8.1 v2 1/2] drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers Alexander Ivanov
  2026-11-15 10:28 ` [Devel] [PATCH hci-8.1 v2 2/2] drivers/vhost/blk: add support for worker ioctls Alexander Ivanov
  0 siblings, 2 replies; 6+ messages in thread
From: Alexander Ivanov @ 2026-11-15 10:28 UTC (permalink / raw)
  To: andrey.zhadchenko; +Cc: devel

vhost-blk gets the mainstream worker ioctls, so userspace can move from
our VHOST_SET_NWORKERS to them. Until it does, both interfaces live on
the same device, and the workers VHOST_SET_NWORKERS installs are not
counted in attachment_cnt: VHOST_FREE_WORKER can free a worker a vq
still uses.

Patch 1 attaches the VHOST_SET_NWORKERS workers with
__vhost_vq_attach_worker(), so VHOST_FREE_WORKER refuses to free them.
Patch 2 wires the worker ioctls into vhost-blk.

v2:
- patch 1: new, keeps VHOST_SET_NWORKERS safe next to the worker ioctls
- patch 2: unchanged

https://virtuozzo.atlassian.net/browse/VSTOR-107537

Alexander Ivanov (2):
  drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers
  drivers/vhost/blk: add support for worker ioctls

 drivers/vhost/blk.c   | 8 ++++++++
 drivers/vhost/vhost.c | 8 ++++++--
 2 files changed, 14 insertions(+), 2 deletions(-)

-- 
2.53.0

_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [Devel] [PATCH hci-8.1 v2 1/2] drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers
  2026-11-15 10:28 [Devel] [PATCH hci-8.1 v2 0/2] drivers/vhost/blk: add support for worker ioctls Alexander Ivanov
@ 2026-11-15 10:28 ` Alexander Ivanov
  2026-11-15 10:35   ` no-reply
  2026-11-15 10:28 ` [Devel] [PATCH hci-8.1 v2 2/2] drivers/vhost/blk: add support for worker ioctls Alexander Ivanov
  1 sibling, 1 reply; 6+ messages in thread
From: Alexander Ivanov @ 2026-11-15 10:28 UTC (permalink / raw)
  To: andrey.zhadchenko; +Cc: devel

vhost_set_workers() and vhost_propagate_workers() store workers into
vq->worker directly, bypassing __vhost_vq_attach_worker(). The workers
they install keep attachment_cnt at 0, and the default worker keeps
counting every vq. VHOST_FREE_WORKER then destroys a worker a virtqueue
still points at, and the next vhost_vq_work_queue() on that vq is a
use-after-free. VHOST_ATTACH_VRING_WORKER on such a vq drives
attachment_cnt of its worker to -1. vhost-scsi already exposes both
interfaces to the device owner.

Attach the workers with __vhost_vq_attach_worker(), so VHOST_FREE_WORKER
fails with -EBUSY while a vq uses the worker and both interfaces can
coexist until userspace moves to the worker ioctls. Userspace issues
VHOST_SET_NWORKERS before VHOST_SET_VRING_KICK, so the attach takes the
path without synchronize_rcu() and VM startup is not slowed down.

https://virtuozzo.atlassian.net/browse/VSTOR-107537
Feature: vhost-blk: in-kernel accelerator for virtio-blk guests

Signed-off-by: Alexander Ivanov <alexander.ivanov@virtuozzo.com>
---
 drivers/vhost/vhost.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
index 1057ba7e3fc7..5092eed75b8f 100644
--- a/drivers/vhost/vhost.c
+++ b/drivers/vhost/vhost.c
@@ -1094,7 +1094,11 @@ static void vhost_propagate_workers(struct vhost_dev *dev, int nworkers)
 	int i, j = 0;
 
 	for (i = nworkers; i < dev->nvqs; i++) {
-		dev->vqs[i]->worker = dev->vqs[j]->worker;
+		struct vhost_worker *worker;
+
+		worker = rcu_dereference_check(dev->vqs[j]->worker,
+					       lockdep_is_held(&dev->mutex));
+		__vhost_vq_attach_worker(dev->vqs[i], worker);
 		if (++j >= nworkers)
 			j = 0;
 	}
@@ -1117,7 +1121,7 @@ static int vhost_set_workers(struct vhost_dev *dev, int n)
 			ret = -ENOMEM;
 			break;
 		}
-		dev->vqs[i]->worker = worker;
+		__vhost_vq_attach_worker(dev->vqs[i], worker);
 	}
 
 	vhost_propagate_workers(dev, i);
-- 
2.53.0

_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [Devel] [PATCH hci-8.1 v2 2/2] drivers/vhost/blk: add support for worker ioctls
  2026-11-15 10:28 [Devel] [PATCH hci-8.1 v2 0/2] drivers/vhost/blk: add support for worker ioctls Alexander Ivanov
  2026-11-15 10:28 ` [Devel] [PATCH hci-8.1 v2 1/2] drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers Alexander Ivanov
@ 2026-11-15 10:28 ` Alexander Ivanov
  2026-11-15 10:35   ` no-reply
  1 sibling, 1 reply; 6+ messages in thread
From: Alexander Ivanov @ 2026-11-15 10:28 UTC (permalink / raw)
  To: andrey.zhadchenko; +Cc: devel

Let userspace create vhost workers and bind them to vhost-blk
virtqueues with the mainstream ioctls from c1ecd8e95007 ("vhost: allow
userspace to create workers"), the same way d74b55e65502 ("vhost_scsi:
add support for worker ioctls") does it for vhost-scsi.

VHOST_NEW_WORKER needs the device in vhost_task mode, so userspace has
to issue VHOST_SET_FORK_FROM_OWNER with VHOST_FORK_OWNER_TASK before
VHOST_SET_OWNER: our fork_from_owner_default is kthread.

https://virtuozzo.atlassian.net/browse/VSTOR-107537
Feature: vhost-blk: in-kernel accelerator for virtio-blk guests

Suggested-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Alexander Ivanov <alexander.ivanov@virtuozzo.com>
Reviewed-by: Andrey Zhadchenko <andrey.zhadchenko@virtuozzo.com>
---
 drivers/vhost/blk.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/drivers/vhost/blk.c b/drivers/vhost/blk.c
index b7d3755200ba..1d26560dd4f0 100644
--- a/drivers/vhost/blk.c
+++ b/drivers/vhost/blk.c
@@ -978,6 +978,14 @@ static long vhost_blk_ioctl(struct file *f, unsigned int ioctl,
 		return vhost_blk_set_features(blk, features);
 	case VHOST_RESET_OWNER:
 		return vhost_blk_reset_owner(blk);
+	case VHOST_NEW_WORKER:
+	case VHOST_FREE_WORKER:
+	case VHOST_ATTACH_VRING_WORKER:
+	case VHOST_GET_VRING_WORKER:
+		mutex_lock(&blk->dev.mutex);
+		ret = vhost_worker_ioctl(&blk->dev, ioctl, argp);
+		mutex_unlock(&blk->dev.mutex);
+		return ret;
 	default:
 		mutex_lock(&blk->dev.mutex);
 		ret = vhost_dev_ioctl(&blk->dev, ioctl, argp);
-- 
2.53.0

_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [Devel] [PATCH hci-8.1 v2 1/2] drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers
  2026-11-15 10:28 ` [Devel] [PATCH hci-8.1 v2 1/2] drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers Alexander Ivanov
@ 2026-11-15 10:35   ` no-reply
  2026-11-15 11:39     ` Alexander Ivanov
  0 siblings, 1 reply; 6+ messages in thread
From: no-reply @ 2026-11-15 10:35 UTC (permalink / raw)
  To: Alexander Ivanov, andrey.zhadchenko; +Cc: devel

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [Medium] vhost: NULL worker passed to __vhost_vq_attach_worker() in
  vhost_propagate_workers()
- [Low] vhost: missing Fixes: trailer for use-after-free fix of NWORKERS
  workers
--

commit 67a253d12d149aa08cf82eb2515b2911f096d6d2
Author: Alexander Ivanov <alexander.ivanov@virtuozzo.com>

drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers

The commit makes vhost_set_workers() and vhost_propagate_workers() attach workers through __vhost_vq_attach_worker() instead of storing vq->worker directly. This keeps attachment_cnt correct, so VHOST_FREE_WORKER returns -EBUSY while a virtqueue still uses the worker.

> diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> index 1057ba7e3fc7..5092eed75b8f 100644
> --- a/drivers/vhost/vhost.c
> +++ b/drivers/vhost/vhost.c
> @@ -1094,7 +1094,11 @@ static void vhost_propagate_workers(struct vhost_dev *dev, int nworkers)
>  	int i, j = 0;
>  
>  	for (i = nworkers; i < dev->nvqs; i++) {
> -		dev->vqs[i]->worker = dev->vqs[j]->worker;
> +		struct vhost_worker *worker;
> +
> +		worker = rcu_dereference_check(dev->vqs[j]->worker,
> +					       lockdep_is_held(&dev->mutex));
> +		__vhost_vq_attach_worker(dev->vqs[i], worker);

[Severity: Medium]
Can worker be NULL here? __vhost_vq_attach_worker() starts with mutex_lock(&worker->mutex), so a NULL argument is a NULL pointer dereference.

vhost-vdpa calls vhost_dev_init() with use_worker=false, so no default worker is ever attached and every vq->worker stays NULL. VHOST_SET_NWORKERS has no use_worker check, and vhost_set_workers() only attaches new workers to vqs[1..n-1]. vhost_propagate_workers() then reads NULL from vqs[0] and passes it on. Before this patch the NULL was only copied, which was harmless.

A narrower case exists on use_worker devices. A killed worker leaves vq->worker NULL, so a source vq can be NULL at propagate time.

Would returning -EINVAL from vhost_set_workers() when !dev->use_worker, and skipping a NULL source worker here, cover both cases?

[ ... ]

> -		dev->vqs[i]->worker = worker;
> +		__vhost_vq_attach_worker(dev->vqs[i], worker);

[Severity: Low]
This is a bug fix, but the commit message has no Fixes: tag. Should it include:

  Fixes: 705a4b3eefde ("drivers/vhost: add ioctl to increase the number of workers")

-- 
Sashiko AI review · https://sashiko.vzint.dev/#/patchset/20261115102811.2745757-1-alexander.ivanov@virtuozzo.com?part=1

_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [Devel] [PATCH hci-8.1 v2 2/2] drivers/vhost/blk: add support for worker ioctls
  2026-11-15 10:28 ` [Devel] [PATCH hci-8.1 v2 2/2] drivers/vhost/blk: add support for worker ioctls Alexander Ivanov
@ 2026-11-15 10:35   ` no-reply
  0 siblings, 0 replies; 6+ messages in thread
From: no-reply @ 2026-11-15 10:35 UTC (permalink / raw)
  To: Alexander Ivanov, andrey.zhadchenko; +Cc: devel

> Let userspace create vhost workers and bind them to vhost-blk
> virtqueues with the mainstream ioctls from c1ecd8e95007 ("vhost: allow
> userspace to create workers"), the same way d74b55e65502 ("vhost_scsi:
> add support for worker ioctls") does it for vhost-scsi.
> 
> VHOST_NEW_WORKER needs the device in vhost_task mode, so userspace has
> to issue VHOST_SET_FORK_FROM_OWNER with VHOST_FORK_OWNER_TASK before
> VHOST_SET_OWNER: our fork_from_owner_default is kthread.
> 
> https://virtuozzo.atlassian.net/browse/VSTOR-107537
> Feature: vhost-blk: in-kernel accelerator for virtio-blk guests
> 
> Suggested-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
> Signed-off-by: Alexander Ivanov <alexander.ivanov@virtuozzo.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.vzint.dev/#/patchset/20261115102811.2745757-1-alexander.ivanov@virtuozzo.com?part=2


_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [Devel] [PATCH hci-8.1 v2 1/2] drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers
  2026-11-15 10:35   ` no-reply
@ 2026-11-15 11:39     ` Alexander Ivanov
  0 siblings, 0 replies; 6+ messages in thread
From: Alexander Ivanov @ 2026-11-15 11:39 UTC (permalink / raw)
  To: no-reply, andrey.zhadchenko; +Cc: devel

>> diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
>> index 1057ba7e3fc7..5092eed75b8f 100644
>> --- a/drivers/vhost/vhost.c
>> +++ b/drivers/vhost/vhost.c
>> @@ -1094,7 +1094,11 @@ static void vhost_propagate_workers(struct vhost_dev *dev, int nworkers)
>>   	int i, j = 0;
>>   
>>   	for (i = nworkers; i < dev->nvqs; i++) {
>> -		dev->vqs[i]->worker = dev->vqs[j]->worker;
>> +		struct vhost_worker *worker;
>> +
>> +		worker = rcu_dereference_check(dev->vqs[j]->worker,
>> +					       lockdep_is_held(&dev->mutex));
>> +		__vhost_vq_attach_worker(dev->vqs[i], worker);
> [Severity: Medium]
> Can worker be NULL here? __vhost_vq_attach_worker() starts with mutex_lock(&worker->mutex), so a NULL argument is a NULL pointer dereference.
>
> vhost-vdpa calls vhost_dev_init() with use_worker=false, so no default worker is ever attached and every vq->worker stays NULL. VHOST_SET_NWORKERS has no use_worker check, and vhost_set_workers() only attaches new workers to vqs[1..n-1]. vhost_propagate_workers() then reads NULL from vqs[0] and passes it on. Before this patch the NULL was only copied, which was harmless.
>
> A narrower case exists on use_worker devices. A killed worker leaves vq->worker NULL, so a source vq can be NULL at propagate time.
>
> Would returning -EINVAL from vhost_set_workers() when !dev->use_worker, and skipping a NULL source worker here, cover both cases?
Fixed.
>
> [ ... ]
>
>> -		dev->vqs[i]->worker = worker;
>> +		__vhost_vq_attach_worker(dev->vqs[i], worker);
> [Severity: Low]
> This is a bug fix, but the commit message has no Fixes: tag. Should it include:
>
>    Fixes: 705a4b3eefde ("drivers/vhost: add ioctl to increase the number of workers")

Don't have such hash in the kernel tree.

Added - Fixes: 1491588e71da ("drivers/vhost: add ioctl to increase the 
number of workers")

_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2026-11-15 11:41 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-15 10:28 [Devel] [PATCH hci-8.1 v2 0/2] drivers/vhost/blk: add support for worker ioctls Alexander Ivanov
2026-11-15 10:28 ` [Devel] [PATCH hci-8.1 v2 1/2] drivers/vhost: fix a use-after-free of VHOST_SET_NWORKERS workers Alexander Ivanov
2026-11-15 10:35   ` no-reply
2026-11-15 11:39     ` Alexander Ivanov
2026-11-15 10:28 ` [Devel] [PATCH hci-8.1 v2 2/2] drivers/vhost/blk: add support for worker ioctls Alexander Ivanov
2026-11-15 10:35   ` no-reply

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.