Hi Philippe,
On 7/17/23 19:26, Philippe Mathieu-Daudé wrote:
> On 17/7/23 18:21, Eric Auger wrote:
>> In the virtio_iommu_handle_command() when a PROBE request is handled,
>> output_size takes a value greater than the tail size and on a subsequent
>> iteration we can get a stack out-of-band access. Initialize the
>> output_size on each iteration.
>
> A backtrace would be helpful (or reproducer).
Well the way to hit it in explained above. PROBE request followed by a
different request consumed within the same virtio_iommu_handle_command()
call
>
>> The issue was found with ASAN. Credits to:
>> Yiming Tao(Zhejiang University)
>> Gaoning Pan(Zhejiang University)
>>
>> Fixes: 1733eebb9e7 ("virtio-iommu: Implement RESV_MEM probe request")
>> Signed-off-by: Eric Auger <eric.au...@redhat.com>
>> Reported-by: Mauro Matteo Cascella <mcasc...@redhat.com>
>
> Did you mean:
>
> Reported-by: Yiming Tao <ta...@zju.edu.cn>
>
> ?
If I understand correctly trhe actual people who found the issue using
the tool were those I mentionned in the credits section in the commit
msg. but unfortunately I don't have their full address. Mauro Matteo
forwarded me the report which transited though Yiming Tao. Anyway I will
be glad to put more precise in the R-b tags if I get some additional info.
Thanks
Eric
>
>> Cc: qemu-sta...@nongnu.org
>>
>> ---
>> - added the Cc: qemu-sta...@nongnu.org and copied 2 persons involved
>> in the reporting loop
>> ---
>> hw/virtio/virtio-iommu.c | 4 +++-
>> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> Reviewed-by: Philippe Mathieu-Daudé <phi...@linaro.org>
>
