lkml.org 
[lkml]   [2017]   [Jun]   [14]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 1/7] vfio: Fix group release deadlock
From
Date
Hi Alex,

On 09/06/2017 23:59, Alex Williamson wrote:
> If vfio_iommu_group_notifier() acquires a group reference and that
> reference becomes the last reference to the group, then vfio_group_put
> introduces a deadlock code path where we're trying to unregister from
> the iommu notifier chain from within a callout of that chain. Use a
> work_struct to release this reference asynchronously.
>
> Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
> ---
> drivers/vfio/vfio.c | 37 ++++++++++++++++++++++++++++++++++++-
> 1 file changed, 36 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
> index 561084ab387f..f2e24d5699f2 100644
> --- a/drivers/vfio/vfio.c
> +++ b/drivers/vfio/vfio.c
> @@ -423,6 +423,34 @@ static void vfio_group_put(struct vfio_group *group)
> kref_put_mutex(&group->kref, vfio_group_release, &vfio.group_lock);
> }
>
> +struct vfio_group_put_work {
> + struct work_struct work;
> + struct vfio_group *group;
> +};
> +
> +static void vfio_group_put_bg(struct work_struct *work)
> +{
> + struct vfio_group_put_work *do_work;
> +
> + do_work = container_of(work, struct vfio_group_put_work, work);
> +
> + vfio_group_put(do_work->group);
> + kfree(do_work);
> +}
> +
> +static void vfio_group_schedule_put(struct vfio_group *group)
> +{
> + struct vfio_group_put_work *do_work;
> +
> + do_work = kmalloc(sizeof(*do_work), GFP_KERNEL);
> + if (WARN_ON(!do_work))
> + return;
> +
> + INIT_WORK(&do_work->work, vfio_group_put_bg);
> + do_work->group = group;
> + schedule_work(&do_work->work);
> +}
> +
> /* Assume group_lock or group reference is held */
> static void vfio_group_get(struct vfio_group *group)
> {
> @@ -762,7 +790,14 @@ static int vfio_iommu_group_notifier(struct notifier_block *nb,
> break;
> }
>
> - vfio_group_put(group);
> + /*
> + * If we're the last reference to the group, the group will be
> + * released, which includes unregistering the iommu group notifier.
> + * We hold a read-lock on that notifier list, unregistering needs
> + * a write-lock... deadlock. Release our reference asyncronously
nit: s/asyncronously/asynchronously

Reviewed-by: Eric Auger <eric.auger@redhat.com>

Thanks

Eric


> + * to avoid that situation.
> + */
> + vfio_group_schedule_put(group);
> return NOTIFY_OK;
> }
>
>

\
 
 \ /
  Last update: 2017-06-14 14:33    [W:0.210 / U:0.036 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site