Re: [PATCH netfilter] netfilter: bridge: make sure to pull arp header in br_nf_forward_arp()

[Florian Westphal <fw@strlen.de>]

Eric Dumazet <edumazet@google.com> wrote:
> syzbot is kind enough to remind us we need to call skb_may_pull()

[..]

> Fixes: c4e70a87d975 ("netfilter: bridge: rename br_netfilter.c to br_netfilter_hooks.c")
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Reported-by: syzbot <syzkaller@googlegroups.com>
> ---
> 
> Note: Fixes: tag does not point to real bug origin, but is old enough
>      to cover all stable versions.

Indeed, looks like a day0 bug.  We don't have this problem for ipv4/6
because the prerouting hook does pskb_may_pull() as part of ipv4/6
header checks.  Arp doesn't have anything like it.

>  		nf_bridge_pull_encap_header(skb);
>  	}
>  
> +	if (unlikely(!pskb_may_pull(skb, sizeof(struct arphdr))))
> +		return NF_DROP;
> +
>  	if (arp_hdr(skb)->ar_pln != 4) {

Thats indeed the only location where we call NFPROTO_ARP hooks,
so this looks like the proper fix/location.

Thanks Eric!

Reviewed-by: Florian Westphal <fw@strlen.de>