Re: [PATCH net] virtio-net: restore VIRTIO_HDR_F_DATA_VALID on receiving

[Jason Wang]

On 2017å01æ21æ 00:45, Michael S. Tsirkin wrote:
> On Fri, Jan 20, 2017 at 02:32:42PM +0800, Jason Wang wrote:
> > Commit 501db511397f ("virtio: don't set VIRTIO_NET_HDR_F_DATA_VALID on
> > xmit") in fact disables VIRTIO_HDR_F_DATA_VALID on receiving path too,
> > fixing this by adding a hint (has_data_valid) and set it only on the
> > receiving path.
> >
> > Cc: Rolf Neugebauer<rolf.neugebauer@xxxxxxxxxx>
> > Signed-off-by: Jason Wang<jasowang@xxxxxxxxxx>
> > ---
> >   drivers/net/macvtap.c      | 2 +-
> >   drivers/net/tun.c          | 2 +-
> >   drivers/net/virtio_net.c   | 2 +-
> >   include/linux/virtio_net.h | 6 +++++-
> >   net/packet/af_packet.c     | 4 ++--
> >   5 files changed, 10 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/net/macvtap.c b/drivers/net/macvtap.c
> > index 5c26653..4026185 100644
> > --- a/drivers/net/macvtap.c
> > +++ b/drivers/net/macvtap.c
> > @@ -825,7 +825,7 @@ static ssize_t macvtap_put_user(struct macvtap_queue *q,
> >   			return -EINVAL;
> >   
> >   		if (virtio_net_hdr_from_skb(skb, &vnet_hdr,
> > -					    macvtap_is_little_endian(q)))
> > +					    macvtap_is_little_endian(q), true))
> >   			BUG();
> >   
> >   		if (copy_to_iter(&vnet_hdr, sizeof(vnet_hdr), iter) !=
> > diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> > index cd8e02c..2cd10b2 100644
> > --- a/drivers/net/tun.c
> > +++ b/drivers/net/tun.c
> > @@ -1360,7 +1360,7 @@ static ssize_t tun_put_user(struct tun_struct *tun,
> >   			return -EINVAL;
> >   
> >   		if (virtio_net_hdr_from_skb(skb, &gso,
> > -					    tun_is_little_endian(tun))) {
> > +					    tun_is_little_endian(tun), true)) {
> >   			struct skb_shared_info *sinfo = skb_shinfo(skb);
> >   			pr_err("unexpected GSO type: "
> >   			       "0x%x, gso_size %d, hdr_len %d\n",
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index 4a10500..3474243 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -1104,7 +1104,7 @@ static int xmit_skb(struct send_queue *sq, struct sk_buff *skb)
> >   		hdr = skb_vnet_hdr(skb);
> >   
> >   	if (virtio_net_hdr_from_skb(skb, &hdr->hdr,
> > -				    virtio_is_little_endian(vi->vdev)))
> > +				    virtio_is_little_endian(vi->vdev), false))
> >   		BUG();
> >   
> >   	if (vi->mergeable_rx_bufs)
> > diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
> > index 5643647..5209b5e 100644
> > --- a/include/linux/virtio_net.h
> > +++ b/include/linux/virtio_net.h
> > @@ -56,7 +56,8 @@ static inline int virtio_net_hdr_to_skb(struct sk_buff *skb,
> >   
> >   static inline int virtio_net_hdr_from_skb(const struct sk_buff *skb,
> >   					  struct virtio_net_hdr *hdr,
> > -					  bool little_endian)
> > +					  bool little_endian,
> > +					  bool has_data_valid)
> >   {
> >   	memset(hdr, 0, sizeof(*hdr));   /* no info leak */
> >   
> I would prefer naming it is_rx. Callers should not know about
> internal details like data valid, the issue we are trying to fix
> here is that tx and tx headers are slightly different.

Actually, I've considered something like this, but the problem is:

- tun use this on xmit, so is_rx = true may cause some confusion here
- I believe we may want to support DATA_VALID (like xen-netback) on tx 
(probably with a feature) in the future.

Thanks