TCP-MD5 checksum failure on x86_64 SMP

Le lundi 17 mai 2010 à 03:49 +0000, Bijay Singh a écrit :

> I am on quite an old kernel 2.6.27 and could not apply your patches.
> 
> Then i moved on to the kernel 2.6.32.11 however since then I have not been able to bring up my card, this is something i need to fix before i can test you fix. Working on that.
> 

Thanks again for the status report.

I see bug is older than what I stated in my previous mail

I could reproduce it in my lab and confirm following patch fixes it

This is a stable candidate (2.6.27 kernels)

Thanks

[PATCH] tcp: tcp_synack_options() fix 

Commit 33ad798c924b4a (tcp: options clean up) introduced a problem
if MD5+SACK+timestamps were used in initial SYN message.

Some stacks (old linux for example) try to negotiate MD5+SACK+TSTAMP
sessions, but since 20 bytes of tcp options space are not enough to
store all the bits needed, we chose to disable timestamps in this case.

We send a SYN-ACK _without_ timestamp option, but socket has timestamps
enabled and all further outgoing messages contain a TS block, all with
the initial timestamp of the remote peer.

Fix is to really disable timestamps option for the whole session.

Reported-by: Bijay Singh <Bijay.Singh@guavus.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
 net/ipv4/tcp_output.c |    5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)


--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On Mon, 17 May 2010 07:03:49 +0200
Eric Dumazet <eric.dumazet@gmail.com> wrote:

> Le lundi 17 mai 2010 à 03:49 +0000, Bijay Singh a écrit :
> 
> > I am on quite an old kernel 2.6.27 and could not apply your patches.
> > 
> > Then i moved on to the kernel 2.6.32.11 however since then I have not been able to bring up my card, this is something i need to fix before i can test you fix. Working on that.
> > 
> 
> Thanks again for the status report.
> 
> I see bug is older than what I stated in my previous mail
> 
> I could reproduce it in my lab and confirm following patch fixes it
> 
> This is a stable candidate (2.6.27 kernels)
> 
> Thanks
> 
> [PATCH] tcp: tcp_synack_options() fix 
> 
> Commit 33ad798c924b4a (tcp: options clean up) introduced a problem
> if MD5+SACK+timestamps were used in initial SYN message.
> 
> Some stacks (old linux for example) try to negotiate MD5+SACK+TSTAMP
> sessions, but since 20 bytes of tcp options space are not enough to
> store all the bits needed, we chose to disable timestamps in this case.
> 
> We send a SYN-ACK _without_ timestamp option, but socket has timestamps
> enabled and all further outgoing messages contain a TS block, all with
> the initial timestamp of the remote peer.
> 
> Fix is to really disable timestamps option for the whole session.
> 
> Reported-by: Bijay Singh <Bijay.Singh@guavus.com>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> ---
>  net/ipv4/tcp_output.c |    5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
> index 0dda86e..b8bb226 100644
> --- a/net/ipv4/tcp_output.c
> +++ b/net/ipv4/tcp_output.c
> @@ -667,7 +667,7 @@ static unsigned tcp_synack_options(struct sock *sk,
>  	u8 cookie_plus = (xvp != NULL && !xvp->cookie_out_never) ?
>  			 xvp->cookie_plus :
>  			 0;
> -	bool doing_ts = ireq->tstamp_ok;
> +	bool doing_ts;
>  
>  #ifdef CONFIG_TCP_MD5SIG
>  	*md5 = tcp_rsk(req)->af_specific->md5_lookup(sk, req);
> @@ -680,11 +680,12 @@ static unsigned tcp_synack_options(struct sock *sk,
>  		 * rather than TS in order to fit in better with old,
>  		 * buggy kernels, but that was deemed to be unnecessary.
>  		 */
> -		doing_ts &= !ireq->sack_ok;
> +		ireq->tstamp_ok &= !ireq->sack_ok;
>  	}
>  #else
>  	*md5 = NULL;
>  #endif
> +	doing_ts = ireq->tstamp_ok;
>  
>  	/* We always send an MSS option. */
>  	opts->mss = mss;
> 
> 

Make this gets back to stable as well.