[PATCH v2] ipv6/tcp: Stop processing ICMPv6 redirect messages

[PATCH v2] ipv6/tcp: Stop processing ICMPv6 redirect messages

[Christoph Paasch]

Tetja Rediske found that if the host receives an ICMPv6 redirect message
after sending a SYN+ACK, the connection will be reset.

He bisected it down to 093d04d (ipv6: Change skb->data before using
icmpv6_notify() to propagate redirect), but the origin of the bug comes
from ec18d9a26 (ipv6: Add redirect support to all protocol icmp error
handlers.). The bug simply did not trigger prior to 093d04d, because
skb->data did not point to the inner IP header and thus icmpv6_notify
did not call the correct err_handler.

This patch adds the missing "goto out;" in tcp_v6_err. After receiving
an ICMPv6 Redirect, we should not continue processing the ICMP in
tcp_v6_err, as this may trigger the removal of request-socks or setting
sk_err(_soft).

Reported-by: Tetja Rediske <tetja@tetja.de>
Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
---
v2: Removed code-cleanup from the patch and improved the commit-message upon 
suggestion from Eric Dumazet.

 net/ipv6/tcp_ipv6.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 1033d2b..e51bd1a 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -386,6 +386,7 @@ static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 
 		if (dst)
 			dst->ops->redirect(dst, sk, skb);
+		goto out;
 	}
 
 	if (type == ICMPV6_PKT_TOOBIG) {
-- 
1.8.1.227.g44fe835

Re: [PATCH v2] ipv6/tcp: Stop processing ICMPv6 redirect messages

[Eric Dumazet]

On Sun, 2013-04-07 at 16:53 +0200, Christoph Paasch wrote:
> Tetja Rediske found that if the host receives an ICMPv6 redirect message
> after sending a SYN+ACK, the connection will be reset.
> 
> He bisected it down to 093d04d (ipv6: Change skb->data before using
> icmpv6_notify() to propagate redirect), but the origin of the bug comes
> from ec18d9a26 (ipv6: Add redirect support to all protocol icmp error
> handlers.). The bug simply did not trigger prior to 093d04d, because
> skb->data did not point to the inner IP header and thus icmpv6_notify
> did not call the correct err_handler.
> 
> This patch adds the missing "goto out;" in tcp_v6_err. After receiving
> an ICMPv6 Redirect, we should not continue processing the ICMP in
> tcp_v6_err, as this may trigger the removal of request-socks or setting
> sk_err(_soft).
> 
> Reported-by: Tetja Rediske <tetja@tetja.de>
> Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
> ---
> v2: Removed code-cleanup from the patch and improved the commit-message upon 
> suggestion from Eric Dumazet.
> 
>  net/ipv6/tcp_ipv6.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
> index 1033d2b..e51bd1a 100644
> --- a/net/ipv6/tcp_ipv6.c
> +++ b/net/ipv6/tcp_ipv6.c
> @@ -386,6 +386,7 @@ static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
>  
>  		if (dst)
>  			dst->ops->redirect(dst, sk, skb);
> +		goto out;
>  	}
>  
>  	if (type == ICMPV6_PKT_TOOBIG) {


Thanks for sorting this out.

Acked-by: Eric Dumazet <edumazet@google.com>

Re: [PATCH v2] ipv6/tcp: Stop processing ICMPv6 redirect messages

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Sun, 07 Apr 2013 08:02:22 -0700

> On Sun, 2013-04-07 at 16:53 +0200, Christoph Paasch wrote:
>> Tetja Rediske found that if the host receives an ICMPv6 redirect message
>> after sending a SYN+ACK, the connection will be reset.
>> 
>> He bisected it down to 093d04d (ipv6: Change skb->data before using
>> icmpv6_notify() to propagate redirect), but the origin of the bug comes
>> from ec18d9a26 (ipv6: Add redirect support to all protocol icmp error
>> handlers.). The bug simply did not trigger prior to 093d04d, because
>> skb->data did not point to the inner IP header and thus icmpv6_notify
>> did not call the correct err_handler.
>> 
>> This patch adds the missing "goto out;" in tcp_v6_err. After receiving
>> an ICMPv6 Redirect, we should not continue processing the ICMP in
>> tcp_v6_err, as this may trigger the removal of request-socks or setting
>> sk_err(_soft).
>> 
>> Reported-by: Tetja Rediske <tetja@tetja.de>
>> Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
>> ---
>> v2: Removed code-cleanup from the patch and improved the commit-message upon 
>> suggestion from Eric Dumazet.
>> 
>>  net/ipv6/tcp_ipv6.c | 1 +
>>  1 file changed, 1 insertion(+)
>> 
>> diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
>> index 1033d2b..e51bd1a 100644
>> --- a/net/ipv6/tcp_ipv6.c
>> +++ b/net/ipv6/tcp_ipv6.c
>> @@ -386,6 +386,7 @@ static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
>>  
>>  		if (dst)
>>  			dst->ops->redirect(dst, sk, skb);
>> +		goto out;
>>  	}
>>  
>>  	if (type == ICMPV6_PKT_TOOBIG) {
> 
> 
> Thanks for sorting this out.
> 
> Acked-by: Eric Dumazet <edumazet@google.com>

Applied and queued up for -stable, thanks everyone!