Re: Warning in net/ipv4/af_inet.c:154

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mercredi 26 mai 2010 à 00:56 -0700, David Miller a écrit :
> From: Anton Blanchard <anton@samba.org>
> Date: Wed, 26 May 2010 13:19:43 +1000
> 
> > I notice we update sk_forward_alloc in sk_mem_charge and sk_mem_uncharge.
> > Since it isn't an atomic variable I went looking for a lock somewhere in
> > the call chain (first thought was the socket lock). I couldn't find
> > anything, but I could easily be missing something.
> 
> We take the lock properly for all of the skb_queue_rcv_skb() cases
> but this rule isn't followed properly for skb_queue_err_skb().
> 
> Eric, look at even things like skb_tstamp_tx().  Nothing locks the
> socket in those cases, yet we dip down into sock_queue_err_skb() and
> thus invoke skb_set_owner_r which goes into sk_mem_charge() and does
> the non-atomic update on ->sk_forward_alloc.
> 
> I am sure there are other cases with this problem involving
> sock_queue_err_skb()...  ip_icmp_error() (via __udp4_lib_err()),
> ipv6_icmp_error(), etc.

All these points are indeed problematic, since a loooong time, so this
is a stable material.

You are 100% right David, maybe we should add a test when changing
sk_forward_alloc to test if socket is locked (lockdep only test), but
that's for 2.6.36 :)

RAW path is not impacted (yet)

Thanks

[PATCH] net: fix sk_forward_alloc corruptions

As David found out, sock_queue_err_skb() should be called with socket
lock hold, or we risk sk_forward_alloc corruption, since we use non
atomic operations to update this field.

This patch adds bh_lock_sock()/bh_unlock_sock() pair to three spots.
(BH already disabled)

1) skb_tstamp_tx() 
2) Before calling ip_icmp_error(), in __udp4_lib_err() 
3) Before calling ipv6_icmp_error(), in __udp6_lib_err()

Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
 net/core/skbuff.c |    4 ++++
 net/ipv4/udp.c    |    2 ++
 net/ipv6/udp.c    |    6 ++++--
 3 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index c543dd2..439e3b9 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -2991,7 +2991,11 @@ void skb_tstamp_tx(struct sk_buff *orig_skb,
 	memset(serr, 0, sizeof(*serr));
 	serr->ee.ee_errno = ENOMSG;
 	serr->ee.ee_origin = SO_EE_ORIGIN_TIMESTAMPING;
+
+	bh_lock_sock(sk);
 	err = sock_queue_err_skb(sk, skb);
+	bh_unlock_sock(sk);
+
 	if (err)
 		kfree_skb(skb);
 }
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index 9de6a69..1d70ff0 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -634,7 +634,9 @@ void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 			goto out;
 	} else {
+		bh_lock_sock(sk);
 		ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
+		bh_unlock_sock(sk);
 	}
 	sk->sk_err = err;
 	sk->sk_error_report(sk);
diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
index 3d7a2c0..f441365 100644
--- a/net/ipv6/udp.c
+++ b/net/ipv6/udp.c
@@ -465,9 +465,11 @@ void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 	if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
 		goto out;
 
-	if (np->recverr)
+	if (np->recverr) {
+		bh_lock_sock(sk);
 		ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
-
+		bh_unlock_sock(sk);
+	}
 	sk->sk_err = err;
 	sk->sk_error_report(sk);
 out: