[v1] virtio: refresh vring region cache after updating a virtqueue size

[PATCH] virtio: refresh vring region cache after updating a virtqueue size

Posted by Carlos López 1 year, 2 months ago

When a virtqueue size is changed by the guest via
virtio_queue_set_num(), its region cache is not automatically updated.
If the size was increased, this could lead to accessing the cache out
of bounds. For example, in vring_get_used_event():

    static inline uint16_t vring_get_used_event(VirtQueue *vq)
    {
        return vring_avail_ring(vq, vq->vring.num);
    }

    static inline uint16_t vring_avail_ring(VirtQueue *vq, int i)
    {
        VRingMemoryRegionCaches *caches = vring_get_region_caches(vq);
        hwaddr pa = offsetof(VRingAvail, ring[i]);

        if (!caches) {
            return 0;
        }

        return virtio_lduw_phys_cached(vq->vdev, &caches->avail, pa);
    }

vq->vring.num will be greater than caches->avail.len, which will
trigger a failed assertion down the call path of
virtio_lduw_phys_cached().

Fix this by calling virtio_queue_update_rings() after
virtio_queue_set_num() if we are not already calling
virtio_queue_set_rings().

Signed-off-by: Carlos López <clopez@suse.de>
---
 hw/s390x/virtio-ccw.c   | 1 +
 hw/virtio/virtio-mmio.c | 5 ++---
 hw/virtio/virtio-pci.c  | 1 +
 3 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/hw/s390x/virtio-ccw.c b/hw/s390x/virtio-ccw.c
index e33e5207ab..89891ac58a 100644
--- a/hw/s390x/virtio-ccw.c
+++ b/hw/s390x/virtio-ccw.c
@@ -237,6 +237,7 @@ static int virtio_ccw_set_vqs(SubchDev *sch, VqInfoBlock *info,
                 return -EINVAL;
             }
             virtio_queue_set_num(vdev, index, num);
+            virtio_queue_update_rings(vdev, index);
         } else if (virtio_queue_get_num(vdev, index) > num) {
             /* Fail if we don't have a big enough queue. */
             return -EINVAL;
diff --git a/hw/virtio/virtio-mmio.c b/hw/virtio/virtio-mmio.c
index 23ba625eb6..c74822308f 100644
--- a/hw/virtio/virtio-mmio.c
+++ b/hw/virtio/virtio-mmio.c
@@ -350,10 +350,9 @@ static void virtio_mmio_write(void *opaque, hwaddr offset, uint64_t value,
     case VIRTIO_MMIO_QUEUE_NUM:
         trace_virtio_mmio_queue_write(value, VIRTQUEUE_MAX_SIZE);
         virtio_queue_set_num(vdev, vdev->queue_sel, value);
+        virtio_queue_update_rings(vdev, vdev->queue_sel);
 
-        if (proxy->legacy) {
-            virtio_queue_update_rings(vdev, vdev->queue_sel);
-        } else {
+        if (!proxy->legacy) {
             proxy->vqs[vdev->queue_sel].num = value;
         }
         break;
diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
index 247325c193..a0a2f2c965 100644
--- a/hw/virtio/virtio-pci.c
+++ b/hw/virtio/virtio-pci.c
@@ -1554,6 +1554,7 @@ static void virtio_pci_common_write(void *opaque, hwaddr addr,
         proxy->vqs[vdev->queue_sel].num = val;
         virtio_queue_set_num(vdev, vdev->queue_sel,
                              proxy->vqs[vdev->queue_sel].num);
+        virtio_queue_update_rings(vdev, vdev->queue_sel);
         break;
     case VIRTIO_PCI_COMMON_Q_MSIX:
         vector = virtio_queue_vector(vdev, vdev->queue_sel);
-- 
2.35.3

Re: [PATCH] virtio: refresh vring region cache after updating a virtqueue size

Posted by Cornelia Huck 1 year, 1 month ago

On Thu, Mar 02 2023, Carlos López <clopez@suse.de> wrote:

> When a virtqueue size is changed by the guest via
> virtio_queue_set_num(), its region cache is not automatically updated.
> If the size was increased, this could lead to accessing the cache out
> of bounds. For example, in vring_get_used_event():
>
>     static inline uint16_t vring_get_used_event(VirtQueue *vq)
>     {
>         return vring_avail_ring(vq, vq->vring.num);
>     }
>
>     static inline uint16_t vring_avail_ring(VirtQueue *vq, int i)
>     {
>         VRingMemoryRegionCaches *caches = vring_get_region_caches(vq);
>         hwaddr pa = offsetof(VRingAvail, ring[i]);
>
>         if (!caches) {
>             return 0;
>         }
>
>         return virtio_lduw_phys_cached(vq->vdev, &caches->avail, pa);
>     }
>
> vq->vring.num will be greater than caches->avail.len, which will
> trigger a failed assertion down the call path of
> virtio_lduw_phys_cached().
>
> Fix this by calling virtio_queue_update_rings() after
> virtio_queue_set_num() if we are not already calling
> virtio_queue_set_rings().

Don't we instead need to call virtio_init_region_cache() to update the
caches? virtio_queue_set_rings() will calculate avail and used from
desc, which looks wrong for modern devices.

>
> Signed-off-by: Carlos López <clopez@suse.de>
> ---
>  hw/s390x/virtio-ccw.c   | 1 +
>  hw/virtio/virtio-mmio.c | 5 ++---
>  hw/virtio/virtio-pci.c  | 1 +
>  3 files changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/hw/s390x/virtio-ccw.c b/hw/s390x/virtio-ccw.c
> index e33e5207ab..89891ac58a 100644
> --- a/hw/s390x/virtio-ccw.c
> +++ b/hw/s390x/virtio-ccw.c
> @@ -237,6 +237,7 @@ static int virtio_ccw_set_vqs(SubchDev *sch, VqInfoBlock *info,
>                  return -EINVAL;
>              }
>              virtio_queue_set_num(vdev, index, num);
> +            virtio_queue_update_rings(vdev, index);

Note that this is the non-legacy path.

>          } else if (virtio_queue_get_num(vdev, index) > num) {
>              /* Fail if we don't have a big enough queue. */
>              return -EINVAL;
> diff --git a/hw/virtio/virtio-mmio.c b/hw/virtio/virtio-mmio.c
> index 23ba625eb6..c74822308f 100644
> --- a/hw/virtio/virtio-mmio.c
> +++ b/hw/virtio/virtio-mmio.c
> @@ -350,10 +350,9 @@ static void virtio_mmio_write(void *opaque, hwaddr offset, uint64_t value,
>      case VIRTIO_MMIO_QUEUE_NUM:
>          trace_virtio_mmio_queue_write(value, VIRTQUEUE_MAX_SIZE);
>          virtio_queue_set_num(vdev, vdev->queue_sel, value);
> +        virtio_queue_update_rings(vdev, vdev->queue_sel);
>  
> -        if (proxy->legacy) {
> -            virtio_queue_update_rings(vdev, vdev->queue_sel);
> -        } else {
> +        if (!proxy->legacy) {
>              proxy->vqs[vdev->queue_sel].num = value;
>          }
>          break;
> diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
> index 247325c193..a0a2f2c965 100644
> --- a/hw/virtio/virtio-pci.c
> +++ b/hw/virtio/virtio-pci.c
> @@ -1554,6 +1554,7 @@ static void virtio_pci_common_write(void *opaque, hwaddr addr,
>          proxy->vqs[vdev->queue_sel].num = val;
>          virtio_queue_set_num(vdev, vdev->queue_sel,
>                               proxy->vqs[vdev->queue_sel].num);
> +        virtio_queue_update_rings(vdev, vdev->queue_sel);
>          break;
>      case VIRTIO_PCI_COMMON_Q_MSIX:
>          vector = virtio_queue_vector(vdev, vdev->queue_sel);

Re: [PATCH] virtio: refresh vring region cache after updating a virtqueue size

Posted by Carlos López 1 year, 1 month ago

On 9/3/23 11:43, Cornelia Huck wrote:
> On Thu, Mar 02 2023, Carlos López <clopez@suse.de> wrote:
>> Fix this by calling virtio_queue_update_rings() after
>> virtio_queue_set_num() if we are not already calling
>> virtio_queue_set_rings().
> 
> Don't we instead need to call virtio_init_region_cache() to update the
> caches? virtio_queue_set_rings() will calculate avail and used from
> desc, which looks wrong for modern devices.

I take it you meant virtio_queue_update_rings() instead of 
virtio_queue_set_rings()? Otherwise I'm not sure what you mean.

If this is the case sure - there is this same kind of logic in 
virtio_load():

             /*
              * VIRTIO-1 devices migrate desc, used, and avail ring 
addresses so
              * only the region cache needs to be set up.  Legacy 
devices need
              * to calculate used and avail ring addresses based on the desc
              * address.
              */
             if (virtio_vdev_has_feature(vdev, VIRTIO_F_VERSION_1)) {
                 virtio_init_region_cache(vdev, i);
             } else {
                 virtio_queue_update_rings(vdev, i);
             }

This will require making virtio_init_region_cache() non static of course.

>> diff --git a/hw/s390x/virtio-ccw.c b/hw/s390x/virtio-ccw.c
>> index e33e5207ab..89891ac58a 100644
>> --- a/hw/s390x/virtio-ccw.c
>> +++ b/hw/s390x/virtio-ccw.c
>> @@ -237,6 +237,7 @@ static int virtio_ccw_set_vqs(SubchDev *sch, VqInfoBlock *info,
>>                  return -EINVAL;
>>              }
>>              virtio_queue_set_num(vdev, index, num);
>> +            virtio_queue_update_rings(vdev, index);
> 
> Note that this is the non-legacy path.
>
So if I understand correctly, in virtio_mmio_write() we check via 
proxy->legacy, and in virtio_ccw_set_vqs() we are in the non-legacy 
path. What about virtio_pci_common_write()?

-- 
Carlos López
Security Engineer
SUSE Software Solutions

Re: [PATCH] virtio: refresh vring region cache after updating a virtqueue size

Posted by Cornelia Huck 1 year, 1 month ago

On Mon, Mar 13 2023, Carlos López <clopez@suse.de> wrote:

> On 9/3/23 11:43, Cornelia Huck wrote:
>> On Thu, Mar 02 2023, Carlos López <clopez@suse.de> wrote:
>>> Fix this by calling virtio_queue_update_rings() after
>>> virtio_queue_set_num() if we are not already calling
>>> virtio_queue_set_rings().
>> 
>> Don't we instead need to call virtio_init_region_cache() to update the
>> caches? virtio_queue_set_rings() will calculate avail and used from
>> desc, which looks wrong for modern devices.
>
> I take it you meant virtio_queue_update_rings() instead of 
> virtio_queue_set_rings()? Otherwise I'm not sure what you mean.

I think I had been looking at the code for too long :(

>
> If this is the case sure - there is this same kind of logic in 
> virtio_load():
>
>              /*
>               * VIRTIO-1 devices migrate desc, used, and avail ring 
> addresses so
>               * only the region cache needs to be set up.  Legacy 
> devices need
>               * to calculate used and avail ring addresses based on the desc
>               * address.
>               */
>              if (virtio_vdev_has_feature(vdev, VIRTIO_F_VERSION_1)) {
>                  virtio_init_region_cache(vdev, i);
>              } else {
>                  virtio_queue_update_rings(vdev, i);
>              }

Yes, I think we need to follow the same logic.

> This will require making virtio_init_region_cache() non static of course.
>
>>> diff --git a/hw/s390x/virtio-ccw.c b/hw/s390x/virtio-ccw.c
>>> index e33e5207ab..89891ac58a 100644
>>> --- a/hw/s390x/virtio-ccw.c
>>> +++ b/hw/s390x/virtio-ccw.c
>>> @@ -237,6 +237,7 @@ static int virtio_ccw_set_vqs(SubchDev *sch, VqInfoBlock *info,
>>>                  return -EINVAL;
>>>              }
>>>              virtio_queue_set_num(vdev, index, num);
>>> +            virtio_queue_update_rings(vdev, index);
>> 
>> Note that this is the non-legacy path.
>>
> So if I understand correctly, in virtio_mmio_write() we check via 
> proxy->legacy, and in virtio_ccw_set_vqs() we are in the non-legacy 
> path. What about virtio_pci_common_write()?

IIUC, only modern drivers will write to the modern bar.

Re: [PATCH] virtio: refresh vring region cache after updating a virtqueue size

Posted by Michael S. Tsirkin 1 year, 1 month ago

On Thu, Mar 09, 2023 at 11:43:46AM +0100, Cornelia Huck wrote:
> On Thu, Mar 02 2023, Carlos López <clopez@suse.de> wrote:
> 
> > When a virtqueue size is changed by the guest via
> > virtio_queue_set_num(), its region cache is not automatically updated.
> > If the size was increased, this could lead to accessing the cache out
> > of bounds. For example, in vring_get_used_event():
> >
> >     static inline uint16_t vring_get_used_event(VirtQueue *vq)
> >     {
> >         return vring_avail_ring(vq, vq->vring.num);
> >     }
> >
> >     static inline uint16_t vring_avail_ring(VirtQueue *vq, int i)
> >     {
> >         VRingMemoryRegionCaches *caches = vring_get_region_caches(vq);
> >         hwaddr pa = offsetof(VRingAvail, ring[i]);
> >
> >         if (!caches) {
> >             return 0;
> >         }
> >
> >         return virtio_lduw_phys_cached(vq->vdev, &caches->avail, pa);
> >     }
> >
> > vq->vring.num will be greater than caches->avail.len, which will
> > trigger a failed assertion down the call path of
> > virtio_lduw_phys_cached().
> >
> > Fix this by calling virtio_queue_update_rings() after
> > virtio_queue_set_num() if we are not already calling
> > virtio_queue_set_rings().
> 
> Don't we instead need to call virtio_init_region_cache() to update the
> caches? virtio_queue_set_rings() will calculate avail and used from
> desc, which looks wrong for modern devices.


Carlos?

> >
> > Signed-off-by: Carlos López <clopez@suse.de>
> > ---
> >  hw/s390x/virtio-ccw.c   | 1 +
> >  hw/virtio/virtio-mmio.c | 5 ++---
> >  hw/virtio/virtio-pci.c  | 1 +
> >  3 files changed, 4 insertions(+), 3 deletions(-)
> >
> > diff --git a/hw/s390x/virtio-ccw.c b/hw/s390x/virtio-ccw.c
> > index e33e5207ab..89891ac58a 100644
> > --- a/hw/s390x/virtio-ccw.c
> > +++ b/hw/s390x/virtio-ccw.c
> > @@ -237,6 +237,7 @@ static int virtio_ccw_set_vqs(SubchDev *sch, VqInfoBlock *info,
> >                  return -EINVAL;
> >              }
> >              virtio_queue_set_num(vdev, index, num);
> > +            virtio_queue_update_rings(vdev, index);
> 
> Note that this is the non-legacy path.
> 
> >          } else if (virtio_queue_get_num(vdev, index) > num) {
> >              /* Fail if we don't have a big enough queue. */
> >              return -EINVAL;
> > diff --git a/hw/virtio/virtio-mmio.c b/hw/virtio/virtio-mmio.c
> > index 23ba625eb6..c74822308f 100644
> > --- a/hw/virtio/virtio-mmio.c
> > +++ b/hw/virtio/virtio-mmio.c
> > @@ -350,10 +350,9 @@ static void virtio_mmio_write(void *opaque, hwaddr offset, uint64_t value,
> >      case VIRTIO_MMIO_QUEUE_NUM:
> >          trace_virtio_mmio_queue_write(value, VIRTQUEUE_MAX_SIZE);
> >          virtio_queue_set_num(vdev, vdev->queue_sel, value);
> > +        virtio_queue_update_rings(vdev, vdev->queue_sel);
> >  
> > -        if (proxy->legacy) {
> > -            virtio_queue_update_rings(vdev, vdev->queue_sel);
> > -        } else {
> > +        if (!proxy->legacy) {
> >              proxy->vqs[vdev->queue_sel].num = value;
> >          }
> >          break;
> > diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
> > index 247325c193..a0a2f2c965 100644
> > --- a/hw/virtio/virtio-pci.c
> > +++ b/hw/virtio/virtio-pci.c
> > @@ -1554,6 +1554,7 @@ static void virtio_pci_common_write(void *opaque, hwaddr addr,
> >          proxy->vqs[vdev->queue_sel].num = val;
> >          virtio_queue_set_num(vdev, vdev->queue_sel,
> >                               proxy->vqs[vdev->queue_sel].num);
> > +        virtio_queue_update_rings(vdev, vdev->queue_sel);
> >          break;
> >      case VIRTIO_PCI_COMMON_Q_MSIX:
> >          vector = virtio_queue_vector(vdev, vdev->queue_sel);

Re: [PATCH] virtio: refresh vring region cache after updating a virtqueue size

Posted by Thomas Huth 1 year, 1 month ago

On 02/03/2023 11.14, Carlos López wrote:
> When a virtqueue size is changed by the guest via
> virtio_queue_set_num(), its region cache is not automatically updated.
> If the size was increased, this could lead to accessing the cache out
> of bounds. For example, in vring_get_used_event():
> 
>      static inline uint16_t vring_get_used_event(VirtQueue *vq)
>      {
>          return vring_avail_ring(vq, vq->vring.num);
>      }
> 
>      static inline uint16_t vring_avail_ring(VirtQueue *vq, int i)
>      {
>          VRingMemoryRegionCaches *caches = vring_get_region_caches(vq);
>          hwaddr pa = offsetof(VRingAvail, ring[i]);
> 
>          if (!caches) {
>              return 0;
>          }
> 
>          return virtio_lduw_phys_cached(vq->vdev, &caches->avail, pa);
>      }
> 
> vq->vring.num will be greater than caches->avail.len, which will
> trigger a failed assertion down the call path of
> virtio_lduw_phys_cached().
> 
> Fix this by calling virtio_queue_update_rings() after
> virtio_queue_set_num() if we are not already calling
> virtio_queue_set_rings().
> 
> Signed-off-by: Carlos López <clopez@suse.de>
> ---
>   hw/s390x/virtio-ccw.c   | 1 +
>   hw/virtio/virtio-mmio.c | 5 ++---
>   hw/virtio/virtio-pci.c  | 1 +
>   3 files changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/hw/s390x/virtio-ccw.c b/hw/s390x/virtio-ccw.c
> index e33e5207ab..89891ac58a 100644
> --- a/hw/s390x/virtio-ccw.c
> +++ b/hw/s390x/virtio-ccw.c
> @@ -237,6 +237,7 @@ static int virtio_ccw_set_vqs(SubchDev *sch, VqInfoBlock *info,
>                   return -EINVAL;
>               }
>               virtio_queue_set_num(vdev, index, num);
> +            virtio_queue_update_rings(vdev, index);
>           } else if (virtio_queue_get_num(vdev, index) > num) {
>               /* Fail if we don't have a big enough queue. */
>               return -EINVAL;

FWIW, s390 part:
Acked-by: Thomas Huth <thuth@redhat.com>