[dpdk-dev] [PATCH v9 8/8] net/virtio: enable packed virtqueues by default

[Jens Freimann]

On Fri, Oct 26, 2018 at 02:06:24PM +0800, Tiwei Bie wrote:
>On Wed, Oct 24, 2018 at 04:32:36PM +0200, Jens Freimann wrote:
>> Signed-off-by: Jens Freimann <jfreimann at redhat.com>
>> ---
>>  drivers/net/virtio/virtio_ethdev.h | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/net/virtio/virtio_ethdev.h b/drivers/net/virtio/virtio_ethdev.h
>> index 6c9247639..d9b4feee2 100644
>> --- a/drivers/net/virtio/virtio_ethdev.h
>> +++ b/drivers/net/virtio/virtio_ethdev.h
>> @@ -34,6 +34,7 @@
>>  	 1u << VIRTIO_RING_F_INDIRECT_DESC |    \
>>  	 1ULL << VIRTIO_F_VERSION_1       |	\
>>  	 1ULL << VIRTIO_F_IN_ORDER        |	\
>> +	 1ULL << VIRTIO_F_RING_PACKED	  |	\
>
>Virtio-user will be broken when packed ring and CQ
>are negotiated. We need to fix this.

How would you fix this? Is it acceptable to disable packed virtqueues
for now in virtio-user when both RING_PACKED and CQ is requested? 

Like this:

diff --git a/drivers/net/virtio/virtio_user/virtio_user_dev.c b/drivers/net/virtio/virtio_user/virtio_user
index 20816c9..3105bef 100644
--- a/drivers/net/virtio/virtio_user/virtio_user_dev.c
+++ b/drivers/net/virtio/virtio_user/virtio_user_dev.c
@@ -474,6 +474,7 @@ int virtio_user_stop_device(struct virtio_user_dev *dev)
                 * so if necessary, we just claim to support CQ
                 */
                dev->frontend_features |= (1ull << VIRTIO_NET_F_CTRL_VQ);
+               dev->unsupported_features |= (1ULL << VIRTIO_F_RING_PACKED);
        } else {
                dev->unsupported_features |= (1ull << VIRTIO_NET_F_CTRL_VQ);
                /* Also disable features that depend on VIRTIO_NET_F_CTRL_VQ */


regards,
Jens