[v2] virtio-scsi: fix hotplug ->reset() vs event race

[Qemu-devel] [PATCH v2 2/2] virtio-scsi: fix hotplug ->reset() vs event race

Posted by Stefan Hajnoczi 7 years, 3 months ago

There is a race condition during hotplug when iothread is used.  It
occurs because virtio-scsi may be processing command queues in the
iothread while the monitor performs SCSI device hotplug.

When a SCSI device is hotplugged the HotplugHandler->plug() callback is
invoked and virtio-scsi emits a rescan event to the guest.

If the guest submits a SCSI command at this point then it may be
cancelled before hotplug completes.  This happens because ->reset() is
called by hw/core/qdev.c:device_set_realized() after
HotplugHandler->plug() has been called and
hw/scsi/scsi-disk.c:scsi_disk_reset() purges all requests.

This patch uses the new HotplugHandler->post_plug() callback to emit the
rescan event after ->reset().  This eliminates the race conditions where
requests could be cancelled.

Reported-by: l00284672 <lizhengui@huawei.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 hw/scsi/virtio-scsi.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/hw/scsi/virtio-scsi.c b/hw/scsi/virtio-scsi.c
index 3aa99717e2..5a3057d1f8 100644
--- a/hw/scsi/virtio-scsi.c
+++ b/hw/scsi/virtio-scsi.c
@@ -797,8 +797,16 @@ static void virtio_scsi_hotplug(HotplugHandler *hotplug_dev, DeviceState *dev,
         virtio_scsi_acquire(s);
         blk_set_aio_context(sd->conf.blk, s->ctx);
         virtio_scsi_release(s);
-
     }
+}
+
+/* Announce the new device after it has been plugged */
+static void virtio_scsi_post_hotplug(HotplugHandler *hotplug_dev,
+                                     DeviceState *dev)
+{
+    VirtIODevice *vdev = VIRTIO_DEVICE(hotplug_dev);
+    VirtIOSCSI *s = VIRTIO_SCSI(vdev);
+    SCSIDevice *sd = SCSI_DEVICE(dev);
 
     if (virtio_vdev_has_feature(vdev, VIRTIO_SCSI_F_HOTPLUG)) {
         virtio_scsi_acquire(s);
@@ -968,6 +976,7 @@ static void virtio_scsi_class_init(ObjectClass *klass, void *data)
     vdc->start_ioeventfd = virtio_scsi_dataplane_start;
     vdc->stop_ioeventfd = virtio_scsi_dataplane_stop;
     hc->plug = virtio_scsi_hotplug;
+    hc->post_plug = virtio_scsi_post_hotplug;
     hc->unplug = virtio_scsi_hotunplug;
 }
 
-- 
2.17.1

Re: [Qemu-devel] [PATCH v2 2/2] virtio-scsi: fix hotplug ->reset() vs event race

Posted by Michael S. Tsirkin 7 years, 3 months ago

On Mon, Jul 16, 2018 at 09:37:32AM +0100, Stefan Hajnoczi wrote:
> There is a race condition during hotplug when iothread is used.  It
> occurs because virtio-scsi may be processing command queues in the
> iothread while the monitor performs SCSI device hotplug.
> 
> When a SCSI device is hotplugged the HotplugHandler->plug() callback is
> invoked and virtio-scsi emits a rescan event to the guest.
> 
> If the guest submits a SCSI command at this point then it may be
> cancelled before hotplug completes.  This happens because ->reset() is
> called by hw/core/qdev.c:device_set_realized() after
> HotplugHandler->plug() has been called and
> hw/scsi/scsi-disk.c:scsi_disk_reset() purges all requests.
> 
> This patch uses the new HotplugHandler->post_plug() callback to emit the
> rescan event after ->reset().  This eliminates the race conditions where
> requests could be cancelled.
> 
> Reported-by: l00284672 <lizhengui@huawei.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Fam Zheng <famz@redhat.com>
> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

Acked-by: Michael S. Tsirkin <mst@redhat.com>

Pls merge through scsi tree.

> ---
>  hw/scsi/virtio-scsi.c | 11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)
> 
> diff --git a/hw/scsi/virtio-scsi.c b/hw/scsi/virtio-scsi.c
> index 3aa99717e2..5a3057d1f8 100644
> --- a/hw/scsi/virtio-scsi.c
> +++ b/hw/scsi/virtio-scsi.c
> @@ -797,8 +797,16 @@ static void virtio_scsi_hotplug(HotplugHandler *hotplug_dev, DeviceState *dev,
>          virtio_scsi_acquire(s);
>          blk_set_aio_context(sd->conf.blk, s->ctx);
>          virtio_scsi_release(s);
> -
>      }
> +}
> +
> +/* Announce the new device after it has been plugged */
> +static void virtio_scsi_post_hotplug(HotplugHandler *hotplug_dev,
> +                                     DeviceState *dev)
> +{
> +    VirtIODevice *vdev = VIRTIO_DEVICE(hotplug_dev);
> +    VirtIOSCSI *s = VIRTIO_SCSI(vdev);
> +    SCSIDevice *sd = SCSI_DEVICE(dev);
>  
>      if (virtio_vdev_has_feature(vdev, VIRTIO_SCSI_F_HOTPLUG)) {
>          virtio_scsi_acquire(s);
> @@ -968,6 +976,7 @@ static void virtio_scsi_class_init(ObjectClass *klass, void *data)
>      vdc->start_ioeventfd = virtio_scsi_dataplane_start;
>      vdc->stop_ioeventfd = virtio_scsi_dataplane_stop;
>      hc->plug = virtio_scsi_hotplug;
> +    hc->post_plug = virtio_scsi_post_hotplug;
>      hc->unplug = virtio_scsi_hotunplug;
>  }
>  
> -- 
> 2.17.1

Re: [Qemu-devel] [PATCH v2 2/2] virtio-scsi: fix hotplug ->reset() vs event race

Posted by Paolo Bonzini 7 years, 3 months ago

On 16/07/2018 14:40, Michael S. Tsirkin wrote:
> On Mon, Jul 16, 2018 at 09:37:32AM +0100, Stefan Hajnoczi wrote:
>> There is a race condition during hotplug when iothread is used.  It
>> occurs because virtio-scsi may be processing command queues in the
>> iothread while the monitor performs SCSI device hotplug.
>>
>> When a SCSI device is hotplugged the HotplugHandler->plug() callback is
>> invoked and virtio-scsi emits a rescan event to the guest.
>>
>> If the guest submits a SCSI command at this point then it may be
>> cancelled before hotplug completes.  This happens because ->reset() is
>> called by hw/core/qdev.c:device_set_realized() after
>> HotplugHandler->plug() has been called and
>> hw/scsi/scsi-disk.c:scsi_disk_reset() purges all requests.
>>
>> This patch uses the new HotplugHandler->post_plug() callback to emit the
>> rescan event after ->reset().  This eliminates the race conditions where
>> requests could be cancelled.
>>
>> Reported-by: l00284672 <lizhengui@huawei.com>
>> Cc: Paolo Bonzini <pbonzini@redhat.com>
>> Cc: Fam Zheng <famz@redhat.com>
>> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
> 
> Acked-by: Michael S. Tsirkin <mst@redhat.com>
> 
> Pls merge through scsi tree.

Will do, thanks Stefan.

Paolo

> 
>> ---
>>  hw/scsi/virtio-scsi.c | 11 ++++++++++-
>>  1 file changed, 10 insertions(+), 1 deletion(-)
>>
>> diff --git a/hw/scsi/virtio-scsi.c b/hw/scsi/virtio-scsi.c
>> index 3aa99717e2..5a3057d1f8 100644
>> --- a/hw/scsi/virtio-scsi.c
>> +++ b/hw/scsi/virtio-scsi.c
>> @@ -797,8 +797,16 @@ static void virtio_scsi_hotplug(HotplugHandler *hotplug_dev, DeviceState *dev,
>>          virtio_scsi_acquire(s);
>>          blk_set_aio_context(sd->conf.blk, s->ctx);
>>          virtio_scsi_release(s);
>> -
>>      }
>> +}
>> +
>> +/* Announce the new device after it has been plugged */
>> +static void virtio_scsi_post_hotplug(HotplugHandler *hotplug_dev,
>> +                                     DeviceState *dev)
>> +{
>> +    VirtIODevice *vdev = VIRTIO_DEVICE(hotplug_dev);
>> +    VirtIOSCSI *s = VIRTIO_SCSI(vdev);
>> +    SCSIDevice *sd = SCSI_DEVICE(dev);
>>  
>>      if (virtio_vdev_has_feature(vdev, VIRTIO_SCSI_F_HOTPLUG)) {
>>          virtio_scsi_acquire(s);
>> @@ -968,6 +976,7 @@ static void virtio_scsi_class_init(ObjectClass *klass, void *data)
>>      vdc->start_ioeventfd = virtio_scsi_dataplane_start;
>>      vdc->stop_ioeventfd = virtio_scsi_dataplane_stop;
>>      hc->plug = virtio_scsi_hotplug;
>> +    hc->post_plug = virtio_scsi_post_hotplug;
>>      hc->unplug = virtio_scsi_hotunplug;
>>  }
>>  
>> -- 
>> 2.17.1