Re: Slab Corruption with ipv6 and tcp6fuzz

[Patrick McManus <mcmanus@ducksong.com>]

On Sat, 2008-04-26 at 20:05 +0400, Evgeniy Polyakov wrote:

> > tcp_v6_do_rcv()->tcp_rcv_established(), the latter goes to step5, where
> > eventually skb can be freed via tcp_data_queue() (drop: label), then if
> > check for tcp_defer_accept_check() returns true and thus
> > tcp_rcv_established() returns -1, which forces tcp_v6_do_rcv() to jump
> > to reset: label, which in turn will pass through discard: label and free
> > the same skb again.
> >  

yep. That buggy test was logic from a different strategy long
discarded.. the skb is always queued/dropped independently of
defer_accept_check() and as such it shouldn't impact the return value of
rcv_established.

I was able to reproduce by forcing the fuzzer to play with DEFER_ACCEPT
100% of the time, and with the patch tests out fine as do the original
defer_accept tests.

Sorry for the bug, and thanks Evgeniy for fixing it and Eric for finding
it.

Acked-By: Patrick McManus <mcmanus@ducksong.com>

> 
> Please test. This bug affects both ipv6 and ipv4 code actually.
> 
> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> index ac9b848..0298f80 100644
> --- a/net/ipv4/tcp_input.c
> +++ b/net/ipv4/tcp_input.c
> @@ -4925,8 +4925,7 @@ step5:
>  	tcp_data_snd_check(sk);
>  	tcp_ack_snd_check(sk);
>  
> -	if (tcp_defer_accept_check(sk))
> -		return -1;
> +	tcp_defer_accept_check(sk);
>  	return 0;
>  
>  csum_error:
> 
>