Linux Virtualization - Sep 2014 - [PATCH v2] virtio-rng: fix stuck of hot-unplugging busy device

When we try to hot-remove a busy virtio-rng device from QEMU monitor,
the device can't be hot-removed. Because virtio-rng driver hangs at
wait_for_completion_killable().

This patch exits the waiting by completing have_data completion before
unregistering, resets data_avail to avoid the hwrng core use wrong
buffer bytes. Before real unregistering we should return -ENODEV for
reading.

When we hot-unplug the device, dd process in guest will exit.
  $ dd if=/dev/hwrng of=/dev/null
  dd: error reading ?/dev/hwrng?: No such device

Signed-off-by: Amos Kong <akong at redhat.com>
Cc: stable at vger.kernel.org
---
V2: reset data_avail (Amit)
    adjust unregister order
---
 drivers/char/hw_random/virtio-rng.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/char/hw_random/virtio-rng.c
b/drivers/char/hw_random/virtio-rng.c
index 2e3139e..e76433b 100644
--- a/drivers/char/hw_random/virtio-rng.c
+++ b/drivers/char/hw_random/virtio-rng.c
@@ -68,6 +68,10 @@ static int virtio_read(struct hwrng *rng, void *buf, size_t
size, bool wait)
 	int ret;
 	struct virtrng_info *vi = (struct virtrng_info *)rng->priv;
 
+	if (!vi->hwrng_register_done) {
+		return -ENODEV;
+	}
+
 	if (!vi->busy) {
 		vi->busy = true;
 		init_completion(&vi->have_data);
@@ -137,10 +141,14 @@ static void remove_common(struct virtio_device *vdev)
 {
 	struct virtrng_info *vi = vdev->priv;
 
+	vi->data_avail = 0;
+	complete(&vi->have_data);
 	vdev->config->reset(vdev);
 	vi->busy = false;
-	if (vi->hwrng_register_done)
+	if (vi->hwrng_register_done) {
+		vi->hwrng_register_done = false;
 		hwrng_unregister(&vi->hwrng);
+	}
 	vdev->config->del_vqs(vdev);
 	ida_simple_remove(&rng_index_ida, vi->index);
 	kfree(vi);
-- 
1.9.3

Hi Amos,

On (Tue) 09 Sep 2014 [19:14:02], Amos Kong wrote:
> When we try to hot-remove a busy virtio-rng device from QEMU monitor,
> the device can't be hot-removed. Because virtio-rng driver hangs at
> wait_for_completion_killable().
> 
> This patch exits the waiting by completing have_data completion before
> unregistering, resets data_avail to avoid the hwrng core use wrong
> buffer bytes. Before real unregistering we should return -ENODEV for
> reading.
> 
> When we hot-unplug the device, dd process in guest will exit.
>   $ dd if=/dev/hwrng of=/dev/null
>   dd: error reading ?/dev/hwrng?: No such device
> 
> Signed-off-by: Amos Kong <akong at redhat.com>
> Cc: stable at vger.kernel.org
> ---
> V2: reset data_avail (Amit)
>     adjust unregister order
Thanks, this looks good.

Can you please split into two parts, the complete() in one, and the
hwrng_register_done stuff into another?

Thanks,

		Amit

On Wed, Sep 10, 2014 at 11:04:54AM +0530, Amit Shah
wrote:
> Hi Amos,
> 
> On (Tue) 09 Sep 2014 [19:14:02], Amos Kong wrote:
> > When we try to hot-remove a busy virtio-rng device from QEMU monitor,
> > the device can't be hot-removed. Because virtio-rng driver hangs
at
> > wait_for_completion_killable().
> > 
> > This patch exits the waiting by completing have_data completion before
> > unregistering, resets data_avail to avoid the hwrng core use wrong
> > buffer bytes. Before real unregistering we should return -ENODEV for
> > reading.
> > 
> > When we hot-unplug the device, dd process in guest will exit.
> >   $ dd if=/dev/hwrng of=/dev/null
> >   dd: error reading ?/dev/hwrng?: No such device
> > 
> > Signed-off-by: Amos Kong <akong at redhat.com>
> > Cc: stable at vger.kernel.org
> > ---
> > V2: reset data_avail (Amit)
> >     adjust unregister order
> 
> Thanks, this looks good.
> 
> Can you please split into two parts, the complete() in one, and the
> hwrng_register_done stuff into another?
I just posted the V3, split to two patches, and updated the commitlog
to describe why we need to return ENODEV for reading.
 
> Thanks,
> 
> 		Amit
-- 
			Amos.