net/sched/act_nat.c BUG

net/sched/act_nat.c BUG

[Rodrigo Partearroyo González]

Hi all,

I have been testing Stateless NAT and found that ICMP packets with length less 
than 20 bytes were not correctly NAT'ed. I have found a BUG that makes taking 
into account IP header length twice, so ICMP packets smaller than 20 bytes 
were being dropped.

The proposed fix is:

Index: net/sched/act_nat.c
===================================================================
--- net/sched/act_nat.c
+++ net/sched/act_nat.c
@@ -202,7 +202,7 @@
        {
                struct icmphdr *icmph;
 
-               if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph)))
+               if (!pskb_may_pull(skb, ihl + sizeof(*icmph)))
                        goto drop;
 
                icmph = (void *)(skb_network_header(skb) + ihl);

Please, consider applying it.
-- 
Rodrigo Partearroyo González
R&D Engineer

Albentia Systems S.A.
http://www.albentia.com
+34 914400213

C\Margarita Salas 22
Parque Tecnológico de Leganés
Leganés (28918)
Madrid
Spain

Re: net/sched/act_nat.c BUG

[Eric Dumazet]

Le vendredi 09 juillet 2010 à 16:37 +0200, Rodrigo Partearroyo González
a écrit :
> Hi all,
> 
> I have been testing Stateless NAT and found that ICMP packets with length less 
> than 20 bytes were not correctly NAT'ed. I have found a BUG that makes taking 
> into account IP header length twice, so ICMP packets smaller than 20 bytes 
> were being dropped.
> 

CC netdev

> The proposed fix is:
> 
> Index: net/sched/act_nat.c
> ===================================================================
> --- net/sched/act_nat.c
> +++ net/sched/act_nat.c
> @@ -202,7 +202,7 @@
>         {
>                 struct icmphdr *icmph;
>  
> -               if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph)))
> +               if (!pskb_may_pull(skb, ihl + sizeof(*icmph)))
>                         goto drop;
>  
>                 icmph = (void *)(skb_network_header(skb) + ihl);
> 
> Please, consider applying it.

Nice catch, but take a look at next lines too,
when call to skb_clone_writable() is done, since same error is present.

	skb_clone_writable(skb,
			   ihl + sizeof(*icmph) + sizeof(*iph))

Please submit a formal patch, with your "Signed-off-by: ...", as
documented in Documentation/SubmittingPatches

Thanks

Re: net/sched/act_nat.c BUG

[Herbert Xu]

On Fri, Jul 09, 2010 at 05:13:40PM +0200, Eric Dumazet wrote:
> Le vendredi 09 juillet 2010 à 16:37 +0200, Rodrigo Partearroyo González
> a écrit :
> > Hi all,
> > 
> > I have been testing Stateless NAT and found that ICMP packets with length less 
> > than 20 bytes were not correctly NAT'ed. I have found a BUG that makes taking 
> > into account IP header length twice, so ICMP packets smaller than 20 bytes 
> > were being dropped.
> > 
> 
> CC netdev
> 
> > The proposed fix is:
> > 
> > Index: net/sched/act_nat.c
> > ===================================================================
> > --- net/sched/act_nat.c
> > +++ net/sched/act_nat.c
> > @@ -202,7 +202,7 @@
> >         {
> >                 struct icmphdr *icmph;
> >  
> > -               if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph)))
> > +               if (!pskb_may_pull(skb, ihl + sizeof(*icmph)))
> >                         goto drop;
> >  
> >                 icmph = (void *)(skb_network_header(skb) + ihl);
> > 
> > Please, consider applying it.
> 
> Nice catch, but take a look at next lines too,
> when call to skb_clone_writable() is done, since same error is present.
> 
> 	skb_clone_writable(skb,
> 			   ihl + sizeof(*icmph) + sizeof(*iph))
> 
> Please submit a formal patch, with your "Signed-off-by: ...", as
> documented in Documentation/SubmittingPatches

No we do need the second IP header, think about it...

However, we should only drop it only if it's long enough and
pskb_may_pull fails.

Thanks,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Re: net/sched/act_nat.c BUG

[David Miller]

From: Herbert Xu <herbert@gondor.apana.org.au>
Date: Sat, 10 Jul 2010 07:23:59 +0800

> On Fri, Jul 09, 2010 at 05:13:40PM +0200, Eric Dumazet wrote:
>> Le vendredi 09 juillet 2010 à 16:37 +0200, Rodrigo Partearroyo González
>> a écrit :
>> > Hi all,
>> > 
>> > I have been testing Stateless NAT and found that ICMP packets with length less 
>> > than 20 bytes were not correctly NAT'ed. I have found a BUG that makes taking 
>> > into account IP header length twice, so ICMP packets smaller than 20 bytes 
>> > were being dropped.
>> > 
>> 
>> CC netdev
>> 
>> > The proposed fix is:
>> > 
>> > Index: net/sched/act_nat.c
>> > ===================================================================
>> > --- net/sched/act_nat.c
>> > +++ net/sched/act_nat.c
>> > @@ -202,7 +202,7 @@
>> >         {
>> >                 struct icmphdr *icmph;
>> >  
>> > -               if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph)))
>> > +               if (!pskb_may_pull(skb, ihl + sizeof(*icmph)))
>> >                         goto drop;
>> >  
>> >                 icmph = (void *)(skb_network_header(skb) + ihl);
>> > 
>> > Please, consider applying it.
>> 
>> Nice catch, but take a look at next lines too,
>> when call to skb_clone_writable() is done, since same error is present.
>> 
>> 	skb_clone_writable(skb,
>> 			   ihl + sizeof(*icmph) + sizeof(*iph))
>> 
>> Please submit a formal patch, with your "Signed-off-by: ...", as
>> documented in Documentation/SubmittingPatches
> 
> No we do need the second IP header, think about it...
> 
> However, we should only drop it only if it's long enough and
> pskb_may_pull fails.

Ok, I've reverted until we come up with the proper fix, thanks.

Re: net/sched/act_nat.c BUG

[Eric Dumazet]

Le samedi 10 juillet 2010 à 07:23 +0800, Herbert Xu a écrit :

> No we do need the second IP header, think about it...
> 
> However, we should only drop it only if it's long enough and
> pskb_may_pull fails.

Indeed right you are !

As Changli suggested, we need another pskb_may_pull() call.

Thanks