[PATCH v2] virtio: fix reachable assertion due to stale value of cached region size

[PATCH v2] virtio: fix reachable assertion due to stale value of cached region size

[Carlos López]

In virtqueue_{split,packed}_get_avail_bytes() descriptors are read
in a loop via MemoryRegionCache regions and calls to
vring_{split,packed}_desc_read() - these take a region cache and the
index of the descriptor to be read.

For direct descriptors we use a cache provided by the caller, whose
size matches that of the virtqueue vring. We limit the number of
descriptors we can read by the size of that vring:

    max = vq->vring.num;
    ...
    MemoryRegionCache *desc_cache = &caches->desc;

For indirect descriptors, we initialize a new cache and limit the
number of descriptors by the size of the intermediate descriptor:

    len = address_space_cache_init(&indirect_desc_cache,
                                   vdev->dma_as,
                                   desc.addr, desc.len, false);
    desc_cache = &indirect_desc_cache;
    ...
    max = desc.len / sizeof(VRingDesc);

However, the first initialization of `max` is done outside the loop
where we process guest descriptors, while the second one is done
inside. This means that a sequence of an indirect descriptor followed
by a direct one will leave a stale value in `max`. If the second
descriptor's `next` field is smaller than the stale value, but
greater than the size of the virtqueue ring (and thus the cached
region), a failed assertion will be triggered in
address_space_read_cached() down the call chain.

Fix this by initializing `max` inside the loop in both functions.

Fixes: 9796d0ac8fb0 ("virtio: use address_space_map/unmap to access descriptors")
Signed-off-by: Carlos López <clopez@suse.de>
---
v2: simplify patch by removing variable cleanups

 hw/virtio/virtio.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
index f35178f5fc..98c4819fcc 100644
--- a/hw/virtio/virtio.c
+++ b/hw/virtio/virtio.c
@@ -1069,7 +1069,7 @@ static void virtqueue_split_get_avail_bytes(VirtQueue *vq,
                             VRingMemoryRegionCaches *caches)
 {
     VirtIODevice *vdev = vq->vdev;
-    unsigned int max, idx;
+    unsigned int idx;
     unsigned int total_bufs, in_total, out_total;
     MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
     int64_t len = 0;
@@ -1078,13 +1078,12 @@ static void virtqueue_split_get_avail_bytes(VirtQueue *vq,
     idx = vq->last_avail_idx;
     total_bufs = in_total = out_total = 0;
 
-    max = vq->vring.num;
-
     while ((rc = virtqueue_num_heads(vq, idx)) > 0) {
         MemoryRegionCache *desc_cache = &caches->desc;
         unsigned int num_bufs;
         VRingDesc desc;
         unsigned int i;
+        unsigned int max = vq->vring.num;
 
         num_bufs = total_bufs;
 
@@ -1206,7 +1205,7 @@ static void virtqueue_packed_get_avail_bytes(VirtQueue *vq,
                                              VRingMemoryRegionCaches *caches)
 {
     VirtIODevice *vdev = vq->vdev;
-    unsigned int max, idx;
+    unsigned int idx;
     unsigned int total_bufs, in_total, out_total;
     MemoryRegionCache *desc_cache;
     MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
@@ -1218,14 +1217,14 @@ static void virtqueue_packed_get_avail_bytes(VirtQueue *vq,
     wrap_counter = vq->last_avail_wrap_counter;
     total_bufs = in_total = out_total = 0;
 
-    max = vq->vring.num;
-
     for (;;) {
         unsigned int num_bufs = total_bufs;
         unsigned int i = idx;
         int rc;
+        unsigned int max = vq->vring.num;
 
         desc_cache = &caches->desc;
+
         vring_packed_desc_read(vdev, &desc, desc_cache, idx, true);
         if (!is_desc_avail(desc.flags, wrap_counter)) {
             break;
-- 
2.35.3

Re: [PATCH v2] virtio: fix reachable assertion due to stale value of cached region size

[Michael S. Tsirkin]

On Thu, Mar 02, 2023 at 11:03:59AM +0100, Carlos López wrote:
> In virtqueue_{split,packed}_get_avail_bytes() descriptors are read
> in a loop via MemoryRegionCache regions and calls to
> vring_{split,packed}_desc_read() - these take a region cache and the
> index of the descriptor to be read.
> 
> For direct descriptors we use a cache provided by the caller, whose
> size matches that of the virtqueue vring. We limit the number of
> descriptors we can read by the size of that vring:
> 
>     max = vq->vring.num;
>     ...
>     MemoryRegionCache *desc_cache = &caches->desc;
> 
> For indirect descriptors, we initialize a new cache and limit the
> number of descriptors by the size of the intermediate descriptor:
> 
>     len = address_space_cache_init(&indirect_desc_cache,
>                                    vdev->dma_as,
>                                    desc.addr, desc.len, false);
>     desc_cache = &indirect_desc_cache;
>     ...
>     max = desc.len / sizeof(VRingDesc);
> 
> However, the first initialization of `max` is done outside the loop
> where we process guest descriptors, while the second one is done
> inside. This means that a sequence of an indirect descriptor followed
> by a direct one will leave a stale value in `max`. If the second
> descriptor's `next` field is smaller than the stale value, but
> greater than the size of the virtqueue ring (and thus the cached
> region), a failed assertion will be triggered in
> address_space_read_cached() down the call chain.
> 
> Fix this by initializing `max` inside the loop in both functions.
> 
> Fixes: 9796d0ac8fb0 ("virtio: use address_space_map/unmap to access descriptors")
> Signed-off-by: Carlos López <clopez@suse.de>
> ---
> v2: simplify patch by removing variable cleanups
> 
>  hw/virtio/virtio.c | 11 +++++------
>  1 file changed, 5 insertions(+), 6 deletions(-)
> 
> diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> index f35178f5fc..98c4819fcc 100644
> --- a/hw/virtio/virtio.c
> +++ b/hw/virtio/virtio.c
> @@ -1069,7 +1069,7 @@ static void virtqueue_split_get_avail_bytes(VirtQueue *vq,
>                              VRingMemoryRegionCaches *caches)
>  {
>      VirtIODevice *vdev = vq->vdev;
> -    unsigned int max, idx;
> +    unsigned int idx;
>      unsigned int total_bufs, in_total, out_total;
>      MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
>      int64_t len = 0;
> @@ -1078,13 +1078,12 @@ static void virtqueue_split_get_avail_bytes(VirtQueue *vq,
>      idx = vq->last_avail_idx;
>      total_bufs = in_total = out_total = 0;
>  
> -    max = vq->vring.num;
> -
>      while ((rc = virtqueue_num_heads(vq, idx)) > 0) {
>          MemoryRegionCache *desc_cache = &caches->desc;
>          unsigned int num_bufs;
>          VRingDesc desc;
>          unsigned int i;
> +        unsigned int max = vq->vring.num;
>  
>          num_bufs = total_bufs;
>  
> @@ -1206,7 +1205,7 @@ static void virtqueue_packed_get_avail_bytes(VirtQueue *vq,
>                                               VRingMemoryRegionCaches *caches)
>  {
>      VirtIODevice *vdev = vq->vdev;
> -    unsigned int max, idx;
> +    unsigned int idx;
>      unsigned int total_bufs, in_total, out_total;
>      MemoryRegionCache *desc_cache;
>      MemoryRegionCache indirect_desc_cache = MEMORY_REGION_CACHE_INVALID;
> @@ -1218,14 +1217,14 @@ static void virtqueue_packed_get_avail_bytes(VirtQueue *vq,
>      wrap_counter = vq->last_avail_wrap_counter;
>      total_bufs = in_total = out_total = 0;
>  
> -    max = vq->vring.num;
> -
>      for (;;) {
>          unsigned int num_bufs = total_bufs;
>          unsigned int i = idx;
>          int rc;
> +        unsigned int max = vq->vring.num;
>  
>          desc_cache = &caches->desc;
> +

OK and let's not add cosmetics with a bugfix unnecessarily?

>          vring_packed_desc_read(vdev, &desc, desc_cache, idx, true);
>          if (!is_desc_avail(desc.flags, wrap_counter)) {
>              break;
> -- 
> 2.35.3

Re: [PATCH v2] virtio: fix reachable assertion due to stale value of cached region sizey

[Alexander Bulekov]

On 230302 1103, Carlos López wrote:
> In virtqueue_{split,packed}_get_avail_bytes() descriptors are read
> in a loop via MemoryRegionCache regions and calls to
> vring_{split,packed}_desc_read() - these take a region cache and the
> index of the descriptor to be read.
> 
> For direct descriptors we use a cache provided by the caller, whose
> size matches that of the virtqueue vring. We limit the number of
> descriptors we can read by the size of that vring:
> 
>     max = vq->vring.num;
>     ...
>     MemoryRegionCache *desc_cache = &caches->desc;
> 
> For indirect descriptors, we initialize a new cache and limit the
> number of descriptors by the size of the intermediate descriptor:
> 
>     len = address_space_cache_init(&indirect_desc_cache,
>                                    vdev->dma_as,
>                                    desc.addr, desc.len, false);
>     desc_cache = &indirect_desc_cache;
>     ...
>     max = desc.len / sizeof(VRingDesc);
> 
> However, the first initialization of `max` is done outside the loop
> where we process guest descriptors, while the second one is done
> inside. This means that a sequence of an indirect descriptor followed
> by a direct one will leave a stale value in `max`. If the second
> descriptor's `next` field is smaller than the stale value, but
> greater than the size of the virtqueue ring (and thus the cached
> region), a failed assertion will be triggered in
> address_space_read_cached() down the call chain.
> 
> Fix this by initializing `max` inside the loop in both functions.
> 
> Fixes: 9796d0ac8fb0 ("virtio: use address_space_map/unmap to access descriptors")
> Signed-off-by: Carlos López <clopez@suse.de>

OSS-Fuzz confirmed this fixed an issue:
https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=52152

Never made a gitlab report for it because, it  seemed very similar to
some other issues
https://gitlab.com/qemu-project/qemu/-/issues/301
https://gitlab.com/qemu-project/qemu/-/issues/781
https://gitlab.com/qemu-project/qemu/-/issues/300

But apparently it was a different one.
Thank you
-Alex