netdev - Re: [PATCH v2 bpf-net] bpf: Change bpf_fib

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <f33d42e4-ec29-bb5a-967e-dbacbd8ec076@gmail.com>
Date:   Tue, 26 Jun 2018 07:53:41 -0600
From:   David Ahern <dsahern@...il.com>
To:     Daniel Borkmann <daniel@...earbox.net>, dsahern@...nel.org,
        netdev@...r.kernel.org, borkmann@...earbox.net, ast@...nel.org
Cc:     davem@...emloft.net, kafai@...com
Subject: Re: [PATCH v2 bpf-net] bpf: Change bpf_fib_lookup to return lookup
 status

On 6/26/18 3:50 AM, Daniel Borkmann wrote:

> [...]
> You change all the semantics of return code here, but this breaks bpf_skb_fib_lookup().
> I cannot see how this would work in that case. The code does the following with the
> bpf_ipv{4,6}_fib_lookup() return code:
> 
> [...]
>         switch (params->family) {
> #if IS_ENABLED(CONFIG_INET)
>         case AF_INET:
>                 index = bpf_ipv4_fib_lookup(net, params, flags, false);
>                 break;
> #endif
> #if IS_ENABLED(CONFIG_IPV6)
>         case AF_INET6:
>                 index = bpf_ipv6_fib_lookup(net, params, flags, false);
>                 break;
> #endif
>         }
> 
>         if (index > 0) {
>                 struct net_device *dev;
> 
>                 dev = dev_get_by_index_rcu(net, index);
>                 if (!is_skb_forwardable(dev, skb))
>                         index = 0;
>         }

Yes, I forgot to update the skb path. That should be rc now and then the
dev lookup based on params->ifindex. Will fix.

> [...]
> 
> So the BPF_FIB_LKUP_* results become the dev ifindex here and the !is_skb_forwardable()
> case further suggests that the packet *can* be forwarded based on the new semantics
> whereas MTU check is bypassed on success.
> 
> It probably helps to craft a selftest for XDP *and* tc case in future, so we can be sure
> nothing breaks with new changes.

yes, will do.