"Michael S. Tsirkin" <m...@redhat.com> writes:
> On Sun, Feb 03, 2013 at 03:09:39PM -0600, Anthony Liguori wrote:
>> Michael Tokarev <m...@tls.msk.ru> writes:
>>
>> > 03.02.2013 17:23, Yan Vugenfirer wrote:
>> >
>> >>> If it helps, mq changes the config size from 8 to 16 bytes. If the
>> >>> driver was making an assumption about an 8-byte config size, that's
>> >>> likely what the problem is.
>> >>>
>> >> That's exactly the problem.
>> >
>> > So what do we do? It isn't nice to break existing setups.
>> > Maybe mq can be disabled by default (together with Antony's
>> > patch) until fixed drivers will be more widely available?
>>
>> I've got a patch that does exactly like this. It's pretty ugly though
>> because of the relationship between virtio-pci and virtio-net. I'm
>> going to try to see if I can find a cleaner way to do it on Monday.
>>
>> I'm also contemplating just disabling mq by default. Since a special
>> command line is needed to enable it anyway (on the backend), having to
>> specify mq=on for the device doesn't seem so bad.
>
>
> It's after midnight here so didn't test yet, but wouldn't the following
> achieve this in a way that is a bit cleaner?
>
> Signed-off-by: Michael S. Tsirkin <m...@redhat.com>
>
>
> diff --git a/hw/virtio-net.c b/hw/virtio-net.c
> index 5699f5e..3342a76 100644
> --- a/hw/virtio-net.c
> +++ b/hw/virtio-net.c
> @@ -346,6 +346,10 @@ static uint32_t virtio_net_get_features(VirtIODevice
> *vdev, uint32_t features)
>
> features |= (1 << VIRTIO_NET_F_MAC);
>
> + if (n->max_queues <= 1) {
> + features &= ~(0x1 << VIRTIO_NET_F_MQ);
> + }
> +
This is too late because the config size has already been set.
> if (!peer_has_vnet_hdr(n)) {
> features &= ~(0x1 << VIRTIO_NET_F_CSUM);
> features &= ~(0x1 << VIRTIO_NET_F_HOST_TSO4);
> @@ -1284,7 +1288,9 @@ VirtIODevice *virtio_net_init(DeviceState *dev, NICConf
> *conf,
> int i;
>
> n = (VirtIONet *)virtio_common_init("virtio-net", VIRTIO_ID_NET,
> - sizeof(struct virtio_net_config),
> + conf->queues > 1 ?
> + sizeof(struct virtio_net_config)
> :
> + sizeof(struct
> virtio_net_config_compat),
> sizeof(VirtIONet));
This makes me a bit nervous. It relies on setting the config size to
one thing before we've masked a feature.
What do you think about just defaulting mq=off? The more I think about
it, the nicer of an option that is.
I'm not a huge fan of automagically changing the value of a feature bit
so I think defaulting it off is a bit more straight forward.
We can certainly revisit for 1.5+.
Regards,
Anthony Liguori
>
> n->vdev.get_config = virtio_net_get_config;
> diff --git a/hw/virtio-net.h b/hw/virtio-net.h
> index 1d5c721..95ad54f 100644
> --- a/hw/virtio-net.h
> +++ b/hw/virtio-net.h
> @@ -69,6 +69,15 @@ typedef struct virtio_net_conf
> /* Maximum packet size we can receive from tap device: header + 64k */
> #define VIRTIO_NET_MAX_BUFSIZE (sizeof(struct virtio_net_hdr) + (64 << 10))
>
> +
> +struct virtio_net_config_compat
> +{
> + /* The config defining mac address ($ETH_ALEN bytes) */
> + uint8_t mac[ETH_ALEN];
> + /* See VIRTIO_NET_F_STATUS and VIRTIO_NET_S_* above */
> + uint16_t status;
> +} QEMU_PACKED;
> +
> struct virtio_net_config
> {
> /* The config defining mac address ($ETH_ALEN bytes) */
