diff options
| author | Xuan Zhuo | 2022-11-10 10:57:39 +0100 |
|---|---|---|
| committer | Michael S. Tsirkin | 2022-11-10 16:18:55 +0100 |
| commit | f47af0af0db5933664c621b8af8067ebbcbe66cd (patch) | |
| tree | 48ae6c6db7708f472eb4f0c57c78e6c838e0ee09 /include | |
| parent | display: include dependencies explicitly (diff) | |
| download | qemu-f47af0af0db5933664c621b8af8067ebbcbe66cd.tar.gz qemu-f47af0af0db5933664c621b8af8067ebbcbe66cd.tar.xz qemu-f47af0af0db5933664c621b8af8067ebbcbe66cd.zip | |
virtio-net: fix for heap-buffer-overflow
Run shell script:
cat << EOF | valgrind qemu-system-i386 -display none -machine accel=qtest, -m \
512M -M q35 -nodefaults -device virtio-net,netdev=net0 -netdev \
user,id=net0 -qtest stdio
outl 0xcf8 0x80000810
outl 0xcfc 0xc000
outl 0xcf8 0x80000804
outl 0xcfc 0x01
outl 0xc00d 0x0200
outl 0xcf8 0x80000890
outb 0xcfc 0x4
outl 0xcf8 0x80000889
outl 0xcfc 0x1c000000
outl 0xcf8 0x80000893
outw 0xcfc 0x100
EOF
Got:
==68666== Invalid read of size 8
==68666== at 0x688536: virtio_net_queue_enable (virtio-net.c:575)
==68666== by 0x6E31AE: memory_region_write_accessor (memory.c:492)
==68666== by 0x6E098D: access_with_adjusted_size (memory.c:554)
==68666== by 0x6E4DB3: memory_region_dispatch_write (memory.c:1521)
==68666== by 0x6E31AE: memory_region_write_accessor (memory.c:492)
==68666== by 0x6E098D: access_with_adjusted_size (memory.c:554)
==68666== by 0x6E4DB3: memory_region_dispatch_write (memory.c:1521)
==68666== by 0x6EBCD3: flatview_write_continue (physmem.c:2820)
==68666== by 0x6EBFBF: flatview_write (physmem.c:2862)
==68666== by 0x6EF5E7: address_space_write (physmem.c:2958)
==68666== by 0x6DFDEC: cpu_outw (ioport.c:70)
==68666== by 0x6F6DF0: qtest_process_command (qtest.c:480)
==68666== Address 0x29087fe8 is 24 bytes after a block of size 416 in arena "client"
That is reported by Alexander Bulekov. https://gitlab.com/qemu-project/qemu/-/issues/1309
Here, the queue_index is the index of the cvq, but in some cases cvq
does not have the corresponding NetClientState, so overflow appears.
I add a check here, ignore illegal queue_index and cvq queue_index.
Note the queue_index is below the VIRTIO_QUEUE_MAX but greater or equal
than cvq index could hit this. Other devices are similar.
Fixes: 7f863302 ("virtio-net: support queue_enable")
Fixes: https://gitlab.com/qemu-project/qemu/-/issues/1309
Reported-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Message-Id: <20221110095739.130393-1-xuanzhuo@linux.alibaba.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Diffstat (limited to 'include')
| -rw-r--r-- | include/hw/virtio/virtio.h | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h index 141a253a2c..a973811cbf 100644 --- a/include/hw/virtio/virtio.h +++ b/include/hw/virtio/virtio.h @@ -148,7 +148,9 @@ struct VirtioDeviceClass { void (*set_config)(VirtIODevice *vdev, const uint8_t *config); void (*reset)(VirtIODevice *vdev); void (*set_status)(VirtIODevice *vdev, uint8_t val); + /* Device must validate queue_index. */ void (*queue_reset)(VirtIODevice *vdev, uint32_t queue_index); + /* Device must validate queue_index. */ void (*queue_enable)(VirtIODevice *vdev, uint32_t queue_index); /* For transitional devices, this is a bitmap of features * that are only exposed on the legacy interface but not |