Re: [PATCH] net: geneve: modify IP header check in geneve6_xmit_skb

From: Sabrina Dubroca <sd@queasysnail.net>
To: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>,
	Phillip Potter <phil@philpotter.co.uk>,
	David Miller <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>, netdev <netdev@vger.kernel.org>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] net: geneve: modify IP header check in geneve6_xmit_skb
Date: Thu, 22 Apr 2021 19:18:49 +0200	[thread overview]
Message-ID: <YIGv+UOHIl8c/JVk@hog> (raw)
In-Reply-To: <CANn89iJSy82k+5b-vgSE-tD7hc8MhM6Niu=eY8sg-b7LbULouQ@mail.gmail.com>

2021-04-22, 18:52:10 +0200, Eric Dumazet wrote:
> On Thu, Apr 22, 2021 at 6:04 PM Sabrina Dubroca <sd@queasysnail.net> wrote:
> >
> > 2021-04-22, 02:39:42 +0200, Florian Westphal wrote:
> > > Phillip Potter <phil@philpotter.co.uk> wrote:
> > > > Modify the check in geneve6_xmit_skb to use the size of a struct iphdr
> > > > rather than struct ipv6hdr. This fixes two kernel selftest failures
> > > > introduced by commit 6628ddfec758
> > > > ("net: geneve: check skb is large enough for IPv4/IPv6 header"), without
> > > > diminishing the fix provided by that commit.
> > >
> > > What errors?
> > >
> > > > Reported-by: kernel test robot <oliver.sang@intel.com>
> > > > Signed-off-by: Phillip Potter <phil@philpotter.co.uk>
> > > > ---
> > > >  drivers/net/geneve.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
> > > > index 42f31c681846..a57a5e6f614f 100644
> > > > --- a/drivers/net/geneve.c
> > > > +++ b/drivers/net/geneve.c
> > > > @@ -988,7 +988,7 @@ static int geneve6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
> > > >     __be16 sport;
> > > >     int err;
> > > >
> > > > -   if (!pskb_network_may_pull(skb, sizeof(struct ipv6hdr)))
> > > > +   if (!pskb_network_may_pull(skb, sizeof(struct iphdr)))
> > > >             return -EINVAL;
> > >
> > > Seems this is papering over some bug, this change makes no sense to
> > > me.  Can you please explain this?
> >
> > I'm not sure the original commit (6628ddfec758 ("net: geneve: check
> > skb is large enough for IPv4/IPv6 header")) is correct either. GENEVE
> > isn't limited to carrying IP, I think an ethernet header with not much
> > else on top should be valid.
> 
> Maybe, but we still attempt to use ip_hdr() in this case, from
> geneve_get_v6_dst()
> 
> So there is something fishy.

In ip_tunnel_get_dsfield()? Only if there's IP in the packet. Other
tunnel types (except vxlan, which probably has the same problem as
geneve) ues pskb_inet_may_pull, that looks like what we need here as
well.

-- 
Sabrina