ipv4: route: fix inet_rtm_getroute induced crash
authorFlorian Westphal <fw@strlen.de>
Sun, 13 Aug 2017 22:52:58 +0000 (00:52 +0200)
committerDavid S. Miller <davem@davemloft.net>
Mon, 14 Aug 2017 18:08:29 +0000 (11:08 -0700)
"ip route get $daddr iif eth0 from $saddr" causes:
 BUG: KASAN: use-after-free in ip_route_input_rcu+0x1535/0x1b50
 Call Trace:
  ip_route_input_rcu+0x1535/0x1b50
  ip_route_input_noref+0xf9/0x190
  tcp_v4_early_demux+0x1a4/0x2b0
  ip_rcv+0xbcb/0xc05
  __netif_receive_skb+0x9c/0xd0
  netif_receive_skb_internal+0x5a8/0x890

Problem is that inet_rtm_getroute calls either ip_route_input_rcu (if an
iif was provided) or ip_route_output_key_hash_rcu.

But ip_route_input_rcu, unlike ip_route_output_key_hash_rcu, already
associates the dst_entry with the skb.  This clears the SKB_DST_NOREF
bit (i.e. skb_dst_drop will release/free the entry while it should not).

Thus only set the dst if we called ip_route_output_key_hash_rcu().

I tested this patch by running:
 while true;do ip r get 10.0.1.2;done > /dev/null &
 while true;do ip r get 10.0.1.2 iif eth0  from 10.0.1.1;done > /dev/null &
... and saw no crash or memory leak.

Cc: Roopa Prabhu <roopa@cumulusnetworks.com>
Cc: David Ahern <dsahern@gmail.com>
Fixes: ba52d61e0ff ("ipv4: route: restore skb_dst_set in inet_rtm_getroute")
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/route.c

index 0383e66f59bcef3bd6b8627edae9aa2d34139f5d..7effa62beed3fa9065ab7d6365b1d79143bfae79 100644 (file)
@@ -2750,12 +2750,13 @@ static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
                err = 0;
                if (IS_ERR(rt))
                        err = PTR_ERR(rt);
+               else
+                       skb_dst_set(skb, &rt->dst);
        }
 
        if (err)
                goto errout_free;
 
-       skb_dst_set(skb, &rt->dst);
        if (rtm->rtm_flags & RTM_F_NOTIFY)
                rt->rt_flags |= RTCF_NOTIFY;