Re: [Bugme-new] [Bug 16529] New: xennet driver crashes when using with pseudowire aka l2tpv3

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

Le jeudi 26 août 2010 à 09:14 +0100, Ian Campbell a écrit :
> On Thu, 2010-08-26 at 09:03 +0100, Eric Dumazet wrote:
> > Here is the patch, could you test it please ?
> > 
> > Thanks !
> > 
> > [PATCH] l2tp: test for malicious frames in l2tp_eth_dev_recv()
> > 
> > close https://bugzilla.kernel.org/show_bug.cgi?id=16529
> > 
> > Before calling dev_forward_skb(), we should make sure skb contains at
> > least an ethernet header, even if length included in upper layer said
> > so.
> 
> Does this imply that there is some problem with xen-netfront setting
> skb->len or skb->data_len or something incorrectly? It's not clear where
> data_len has come from in this context.

data_len is a 16bit field provided in a prior encapsulation header,
provided by user (untrusted source)

Some buggy or malicious software sent an invalid frame,


< encapsulation [len=1000] >  < 'runt' eth frame (len<14) > 

Another fix would be to change l2tp_recv_dequeue_skb(), and check

L2TP_SKB_CB(skb)->length against skb->len, before calling 

(*session->recv_skb)(session, skb, length);


I prefer the one liner patch I sent you, as a minimum fix.