Re: [Patch net v2] ip_gre: set dev->hard_header_len and dev->needed_headroom properly

From: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
To: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Network Development <netdev@vger.kernel.org>,
	syzbot <syzbot+4a2c52677a8a1aa283cb@syzkaller.appspotmail.com>,
	Xie He <xie.he.0141@gmail.com>, William Tu <u9012063@gmail.com>,
	Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Subject: Re: [Patch net v2] ip_gre: set dev->hard_header_len and dev->needed_headroom properly
Date: Sun, 11 Oct 2020 17:25:41 -0400	[thread overview]
Message-ID: <CA+FuTSdWd0mG7W3Cb2PaAVceV5C0DkYkmGtqKiP6+cD_f3k-zA@mail.gmail.com> (raw)
In-Reply-To: <CA+FuTSfeTWBpOp+ZCBMBQPqcPUAhZcAv2unwMLqgGt_x_PkrqA@mail.gmail.com>

On Sun, Oct 11, 2020 at 5:00 PM Willem de Bruijn
<willemdebruijn.kernel@gmail.com> wrote:
>
> On Sun, Oct 11, 2020 at 3:11 PM Cong Wang <xiyou.wangcong@gmail.com> wrote:
> >
> > GRE tunnel has its own header_ops, ipgre_header_ops, and sets it
> > conditionally. When it is set, it assumes the outer IP header is
> > already created before ipgre_xmit().
> >
> > This is not true when we send packets through a raw packet socket,
> > where L2 headers are supposed to be constructed by user. Packet
> > socket calls dev_validate_header() to validate the header. But
> > GRE tunnel does not set dev->hard_header_len, so that check can
> > be simply bypassed, therefore uninit memory could be passed down
> > to ipgre_xmit(). Similar for dev->needed_headroom.
> >
> > dev->hard_header_len is supposed to be the length of the header
> > created by dev->header_ops->create(), so it should be used whenever
> > header_ops is set, and dev->needed_headroom should be used when it
> > is not set.
> >
> > Reported-and-tested-by: syzbot+4a2c52677a8a1aa283cb@syzkaller.appspotmail.com
> > Cc: Xie He <xie.he.0141@gmail.com>
> > Cc: William Tu <u9012063@gmail.com>
> > Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
> > Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>

> I agree that then swapping dev->needed_headroom for
> dev->hard_header_len at init is then a good approach.
>
> Underlying functions like ip_tunnel_xmit can modify needed_headroom at
> runtime, not sure how that affects runtime changes in
> ipgre_link_update:

Never mind, the patch only selects between needed_headroom or
hard_header_link at ipgre_tunnel_init.

> > @@ -987,10 +991,14 @@ static int ipgre_tunnel_init(struct net_device *dev)
> >                                 return -EINVAL;
> >                         dev->flags = IFF_BROADCAST;
> >                         dev->header_ops = &ipgre_header_ops;
> > +                       dev->hard_header_len = tunnel->hlen + sizeof(*iph);
> > +                       dev->needed_headroom = 0;
>
> here and below, perhaps more descriptive of what is being done, something like:
>
> /* If device has header_ops.create, then headers are part of hard_header_len */
> swap(dev->needed_headroom, dev->hard_header_len)

Arguably nitpicking. Otherwise, patch looks great to me, thanks, so

Acked-by: Willem de Bruijn <willemb@google.com>