[PATCH] netfilter: use skb_to_full_sk in ip6_route_me_harder

From: Eric Dumazet <edumazet@google.com>

For some reason, Florian forgot to apply to ip6_route_me_harder
the fix that went in commit 29e09229d9f2 ("netfilter: use
skb_to_full_sk in ip_route_me_harder")

Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of listener") 
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <redacted>
---
 net/ipv6/netfilter.c |    9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/net/ipv6/netfilter.c b/net/ipv6/netfilter.c

index
d95ceca7ff8f648ff301d91a2e3eb60fc2050f1c..531d6957af36c4af48176f9360e9d
95f78a45d55 100644

--- a/net/ipv6/netfilter.c
+++ b/net/ipv6/netfilter.c

@@ -21,18 +21,19 @@

 int ip6_route_me_harder(struct net *net, struct sk_buff *skb)
 {
 	const struct ipv6hdr *iph = ipv6_hdr(skb);
+	struct sock *sk = sk_to_full_sk(skb->sk);
 	unsigned int hh_len;
 	struct dst_entry *dst;
 	struct flowi6 fl6 = {
-		.flowi6_oif = skb->sk ? skb->sk->sk_bound_dev_if : 0,
+		.flowi6_oif = sk ? sk->sk_bound_dev_if : 0,
 		.flowi6_mark = skb->mark,
-		.flowi6_uid = sock_net_uid(net, skb->sk),
+		.flowi6_uid = sock_net_uid(net, sk),
 		.daddr = iph->daddr,
 		.saddr = iph->saddr,
 	};
 	int err;
 
-	dst = ip6_route_output(net, skb->sk, &fl6);
+	dst = ip6_route_output(net, sk, &fl6);
 	err = dst->error;
 	if (err) {
 		IP6_INC_STATS(net, ip6_dst_idev(dst),
IPSTATS_MIB_OUTNOROUTES);

@@ -50,7 +51,7 @@ int ip6_route_me_harder(struct net *net, struct

sk_buff *skb)
 	if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 	    xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6)
== 0) {
 		skb_dst_set(skb, NULL);
-		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6),
skb->sk, 0);
+		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk,
0);
 		if (IS_ERR(dst))
 			return PTR_ERR(dst);
 		skb_dst_set(skb, dst);

Eric Dumazet [off-list ref] wrote:

From: Eric Dumazet <edumazet@google.com>

For some reason, Florian forgot to apply to ip6_route_me_harder
the fix that went in commit 29e09229d9f2 ("netfilter: use
skb_to_full_sk in ip_route_me_harder")

Indeed, sorry about that, thanks for taking care of this.

On Sun, 2018-02-25 at 11:43 -0800, Eric Dumazet wrote:

No idea what happened, but it looks like this garbage should not affect
   patchwork.

Tell me if a resend is needed, thanks.

On Sun, Feb 25, 2018 at 11:43:39AM -0800, Eric Dumazet wrote:

From: Eric Dumazet <edumazet@google.com>

For some reason, Florian forgot to apply to ip6_route_me_harder
the fix that went in commit 29e09229d9f2 ("netfilter: use
skb_to_full_sk in ip_route_me_harder")

Applied, thanks Eric.

On Sun, 2018-02-25 at 11:47 -0800, Eric Dumazet wrote:

No idea what happened, but it looks like this garbage should not affect
   patchwork.

Tell me if a resend is needed, thanks.

Hmm... I will send a V2, sorry for this mess.

From: Eric Dumazet <edumazet@google.com>

For some reason, Florian forgot to apply to ip6_route_me_harder
the fix that went in commit 29e09229d9f2 ("netfilter: use
skb_to_full_sk in ip_route_me_harder")

Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of listener") 
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <redacted>
---
 net/ipv6/netfilter.c |    9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/net/ipv6/netfilter.c b/net/ipv6/netfilter.c
index d95ceca7ff8f648ff301d91a2e3eb60fc2050f1c..531d6957af36c4af48176f9360e9d95f78a45d55 100644
--- a/net/ipv6/netfilter.c
+++ b/net/ipv6/netfilter.c

@@ -21,18 +21,19 @@
 int ip6_route_me_harder(struct net *net, struct sk_buff *skb)
 {
 	const struct ipv6hdr *iph = ipv6_hdr(skb);
+	struct sock *sk = sk_to_full_sk(skb->sk);
 	unsigned int hh_len;
 	struct dst_entry *dst;
 	struct flowi6 fl6 = {
-		.flowi6_oif = skb->sk ? skb->sk->sk_bound_dev_if : 0,
+		.flowi6_oif = sk ? sk->sk_bound_dev_if : 0,
 		.flowi6_mark = skb->mark,
-		.flowi6_uid = sock_net_uid(net, skb->sk),
+		.flowi6_uid = sock_net_uid(net, sk),
 		.daddr = iph->daddr,
 		.saddr = iph->saddr,
 	};
 	int err;
 
-	dst = ip6_route_output(net, skb->sk, &fl6);
+	dst = ip6_route_output(net, sk, &fl6);
 	err = dst->error;
 	if (err) {
 		IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);

@@ -50,7 +51,7 @@ int ip6_route_me_harder(struct net *net, struct sk_buff *skb)
 	if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 	    xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
 		skb_dst_set(skb, NULL);
-		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), skb->sk, 0);
+		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
 		if (IS_ERR(dst))
 			return PTR_ERR(dst);
 		skb_dst_set(skb, dst);

On Sun, Feb 25, 2018 at 11:49:07AM -0800, Eric Dumazet wrote:

From: Eric Dumazet <edumazet@google.com>

For some reason, Florian forgot to apply to ip6_route_me_harder
the fix that went in commit 29e09229d9f2 ("netfilter: use
skb_to_full_sk in ip_route_me_harder")

Applied this one, thanks Eric.