On 2/24/20 11:32 AM, David Marchand wrote:
> On Mon, Feb 24, 2020 at 9:28 AM Maxime Coquelin
> <maxime.coque...@redhat.com> wrote:
>>
>> Hi David & Thomas,
>>
>> On 2/24/20 4:14 PM, Marvin Liu wrote:
>>> Available buffer ID should be stored in the zmbuf in the packed-ring
>>> dequeue path. There's no guarantee that local queue avail index is
>>> equal to buffer ID.
>>>
>>> Fixes: d1eafb532268 ("vhost: add packed ring zcopy batch and single
>>> dequeue")
>>> Cc: sta...@dpdk.org
>>>
>>> Signed-off-by: Marvin Liu <yong....@intel.com>
>>> Reported-by: Yinan Wang <yinan.w...@intel.com>
>>> Reviewed-by: Maxime Coquelin <maxime.coque...@redhat.com>
>>>
>>
>> If it is not too late, I think we should pick this patch for
>> v20.02. It is fixing a regression introduced in DPDK v19.11.
>
> I might have cold feet, but taking this fix right now feels risky.
> If the problem has been there since 19.11, it can wait 20.05 and it
> will go to 19.11 after proper validation.
Ok, I get your point and it's your call. Now, this fix is really
isolated to zero-copy packed ring, so the only risk IMO is to break
something that is already not working.
Thanks,
Maxime
