Re: [Stable-review] [104/145] netfilter: xt_recent: fix false match

[Willy Tarreau]

On Sat, Mar 13, 2010 at 07:40:03AM -0700, Tim Gardner wrote:
> On 03/12/2010 11:24 PM, Willy Tarreau wrote:
> >On Fri, Mar 12, 2010 at 04:27:17PM -0800, Greg KH wrote:
> >>2.6.32-stable review patch.  If anyone has any objections, please let me 
> >>know.
> >>
> >>----------------
> >>From: Tim Gardner<tim.gardner@xxxxxxxxxxxxx>
> >>
> >>commit 8ccb92ad41cb311e52ad1b1fe77992c7f47a3b63 upstream.
> >>
> >>A rule with a zero hit_count will always match.
> >>
> >>Signed-off-by: Tim Gardner<tim.gardner@xxxxxxxxxxxxx>
> >>Signed-off-by: Patrick McHardy<kaber@xxxxxxxxx>
> >>Signed-off-by: Greg Kroah-Hartman<gregkh@xxxxxxx>
> >>
> >>---
> >>  net/netfilter/xt_recent.c |    2 +-
> >>  1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >>--- a/net/netfilter/xt_recent.c
> >>+++ b/net/netfilter/xt_recent.c
> >>@@ -260,7 +260,7 @@ recent_mt(const struct sk_buff *skb, con
> >>  		for (i = 0; i<  e->nstamps; i++) {
> >>  			if (info->seconds&&  time_after(time, e->stamps[i]))
> >>  				continue;
> >>-			if (++hits>= info->hit_count) {
> >>+			if (info->hit_count&&  ++hits>= info->hit_count) {
> >>  				ret = !ret;
> >>  				break;
> >>  			}
> >
> >I don't know if this has any undesired side effect or not, but the
> >logic is changed now since "hits" will not be increased anymore when
> >info->hit_count is zero. And the code does not make it obvious to me
> >what the intended purpose was.
> >
> >For this reason I always find it dangerous to change variables in
> >if() conditions because it's where we change operations the most
> >frequently when fixing bugs.
> >
> >Willy
> >
> 
> Willy - I agree with you that changing variables in an if() clause can 
> be dangerous. I did consider the possibility for side effects in this 
> case, but decided to go with the simplest patch since 'hits' is local to 
> the scope of the the surrounding else if() clause and is used in no 
> other place.

indeed, but before the patch, "hits" was increased at every
pass in the loop. Now it's only increased for passes which
have a non-zero hit_count, so in theory it can change the
result of the test for further passes of the loop. I just
don't know if it can have any effect, but I trust you since
you had to understand the code for the change :-)

Regards,
Willy

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/