On Mon, 27 Apr 2020 13:50:15 +0800
Jason Wang <jasowang@xxxxxxxxxx> wrote:
> On 2020/4/23 上午12:09, Jesper Dangaard Brouer wrote:
> > In vhost_net_build_xdp() the 'buf' that gets queued via an xdp_buff
> > have embedded a struct tun_xdp_hdr (located at xdp->data_hard_start)
> > which contains the buffer length 'buflen' (with tailroom for
> > skb_shared_info). Also storing this buflen in xdp->frame_sz, does not
> > obsolete struct tun_xdp_hdr, as it also contains a struct
> > virtio_net_hdr with other information.
> >
> > Cc: Jason Wang <jasowang@xxxxxxxxxx>
> > Signed-off-by: Jesper Dangaard Brouer <brouer@xxxxxxxxxx>
> > ---
> > drivers/vhost/net.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
> > index 87469d67ede8..69af007e22f4 100644
> > --- a/drivers/vhost/net.c
> > +++ b/drivers/vhost/net.c
> > @@ -745,6 +745,7 @@ static int vhost_net_build_xdp(struct vhost_net_virtqueue *nvq,
> > xdp->data = buf + pad;
> > xdp->data_end = xdp->data + len;
> > hdr->buflen = buflen;
> > + xdp->frame_sz = buflen;
> >
> > --net->refcnt_bias;
> > alloc_frag->offset += buflen;
>
>
> Tun_xdp_one() will use hdr->buflen as the frame_sz (patch 19), so it
> looks to me there's no need to do this?
I was thinking to go the "other way", meaning let tun_xdp_one() use
xdp->frame_sz, which gets set here. This would allow us to refactor
the code, and drop struct tun_xdp_hdr, as (see pahole below) it only
carries 'buflen' and the remaining part comes from struct
virtio_net_hdr, which could be used directly instead.
As this will be a code refactor, I would prefer we do it after this
patchseries is agreed upon.
$ pahole -C tun_xdp_hdr drivers/net/tap.o
struct tun_xdp_hdr {
int buflen; /* 0 4 */
struct virtio_net_hdr gso; /* 4 10 */
/* size: 16, cachelines: 1, members: 2 */
/* padding: 2 */
/* last cacheline: 16 bytes */
};
--
Best regards,
Jesper Dangaard Brouer
MSc.CS, Principal Kernel Engineer at Red Hat
LinkedIn: http://www.linkedin.com/in/brouer
