[EXTERNAL] [PATCH] net/virtio-user: reset used index counter in dev reset

Maxime Coquelin maxime.coquelin at redhat.com
Thu Sep 19 13:17:56 CEST 2024



On 9/19/24 12:54, Maxime Coquelin wrote:
> Hi,
> 
> On 9/6/24 12:14, Shiva Shankar Kommula wrote:
>> Hello Maxime,
>> could you please review the following change ?
>>
>> Thanks
>>
>>> Subject: [EXTERNAL] [PATCH] net/virtio-user: reset used index counter 
>>> in dev
>>> reset
>>>
>>> When the virtio device is reinitialized during ethdev 
>>> reconfiguration, all the
>>> virtio rings are recreated and repopulated on the device. 
>>> Accordingly, reset the
>>> used index counter value back to zero. Signed-off-by: Kommula Shiva 
>>> Shankar
>>> <kshankar@ marvell. com>
>>> When the virtio device is reinitialized during ethdev 
>>> reconfiguration, all the
>>> virtio rings are recreated and repopulated on the device.
>>> Accordingly, reset the used index counter value back to zero.
> 
> I think it needs to be backported to stable releases.
> What about this?
> 
> Fixes: 48a4464029a7 ("net/virtio-user: support control VQ for packed")
> Cc: stable at dpdk.org
> 
>>>
>>> Signed-off-by: Kommula Shiva Shankar <kshankar at marvell.com>
>>> ---
>>>   drivers/net/virtio/virtio_user_ethdev.c | 1 +
>>>   1 file changed, 1 insertion(+)
>>>
>>> diff --git a/drivers/net/virtio/virtio_user_ethdev.c
>>> b/drivers/net/virtio/virtio_user_ethdev.c
>>> index ae6593ba0b..d60c7e188c 100644
>>> --- a/drivers/net/virtio/virtio_user_ethdev.c
>>> +++ b/drivers/net/virtio/virtio_user_ethdev.c
>>> @@ -204,6 +204,7 @@ virtio_user_setup_queue_packed(struct virtqueue
>>> *vq,
>>>       vring->device = (void *)(uintptr_t)used_addr;
>>>       dev->packed_queues[queue_idx].avail_wrap_counter = true;
>>>       dev->packed_queues[queue_idx].used_wrap_counter = true;
>>> +    dev->packed_queues[queue_idx].used_idx = 0;
>>>
>>>       for (i = 0; i < vring->num; i++)
>>>           vring->desc[i].flags = 0;
>>> -- 
>>> 2.43.0
>>

With added Fixes tag:
Reviewed-by: Maxime Coquelin <maxime.coquelin at redhat.com>

Thanks,
Maxime



More information about the dev mailing list