bpf: restore skb->sk before pskb_trim() call
authorEric Dumazet <edumazet@google.com>
Wed, 26 Apr 2017 16:09:23 +0000 (09:09 -0700)
committerDavid S. Miller <davem@davemloft.net>
Mon, 1 May 2017 02:23:16 +0000 (22:23 -0400)
While testing a fix [1] in ___pskb_trim(), addressing the WARN_ON_ONCE()
in skb_try_coalesce() reported by Andrey, I found that we had an skb
with skb->sk set but no skb->destructor.

This invalidated heuristic found in commit 158f323b9868 ("net: adjust
skb->truesize in pskb_expand_head()") and in cited patch.

Considering the BUG_ON(skb->sk) we have in skb_orphan(), we should
restrain the temporary setting to a minimal section.

[1] https://patchwork.ozlabs.org/patch/755570/
    net: adjust skb->truesize in ___pskb_trim()

Fixes: 8f917bba0042 ("bpf: pass sk to helper functions")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/core/filter.c

index 9a37860a80fc78378705b681ec3b0468824cbcf4..a253a6197e6b37a7ae2fe451c646b01c861a3e22 100644 (file)
@@ -98,8 +98,8 @@ int sk_filter_trim_cap(struct sock *sk, struct sk_buff *skb, unsigned int cap)
 
                skb->sk = sk;
                pkt_len = bpf_prog_run_save_cb(filter->prog, skb);
-               err = pkt_len ? pskb_trim(skb, max(cap, pkt_len)) : -EPERM;
                skb->sk = save_sk;
+               err = pkt_len ? pskb_trim(skb, max(cap, pkt_len)) : -EPERM;
        }
        rcu_read_unlock();