On Thu, 1 Aug 2019 16:54:33 +0200
Jesper Dangaard Brouer <brouer@xxxxxxxxxx> wrote:
> On Thu, 1 Aug 2019 10:18:49 +0200
> Jesper Dangaard Brouer <brouer@xxxxxxxxxx> wrote:
>
> [...]
> > Cc. Stephen as this might be related to:
> > commit 458bf2f224f0 ("net: core: support XDP generic on stacked devices.") (Author: Stephen Hemminger).
> >
> > On Wed, 31 Jul 2019 21:12:23 +0000 Brandon Cazander <brandon.cazander@xxxxxxxxxxxxxxx> wrote:
> >
> > > I am having an issue with xdpgeneric specifically when using XDP_PASS after
> > > bpf_xdp_adjust_head to pop some headers off. My test environment is qemu using
> > > virtio_net specifically, but it also happens with e1000 in qemu/physical devices.
> > >
> > > On a real NIC (ixgbe), the same program is successfully passing decapsulated
> > > traffic, but fails in the same way when forcing xdpgeneric mode.
> >
> [...]
> > > Kernel is 5.2.2-1-debug, OS is openSUSE Tumbleweed 20190724.
> >
> > Can you test an earlier kernel, specifically before: commit
> > 458bf2f224f0 ("net: core: support XDP generic on stacked devices.")
> > (Author: Stephen Hemminger)
> >
> > $ git describe --contains 458bf2f224f04
> > v5.2-rc3~26^2~11^2
> >
> > I fear that this commit, which moved generic-XDP to a later call point,
> > might cause this. Because it could be that the SKB network_header
> > update, is now done before calling XDP program (... still looking at
> > code details).
>
> I now have a reproducer myself, and I can confirm that this commit
> 458bf2f224f0 ("net: core: support XDP generic on stacked devices.") is
> the issue... reverting it fixes the problem. (I don't have a fix yet)
I do have a fix now... It was a matter of calling skb_reset_network_header()
which will fix the issue with ip_hdr(skb) as I suspect was wrong (as
explained earlier).
I wonder if we also need to call skb_reset_transport_header(skb) ?
--
Best regards,
Jesper Dangaard Brouer
MSc.CS, Principal Kernel Engineer at Red Hat
LinkedIn: http://www.linkedin.com/in/brouer
|