Skip to contentSkip to conversation

xt_recent: Fix false hit_count match

PATCH In review Tim Gardner 3 participants 10 messages Feb 19, 2010 – Mar 22, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xt_recent: Fix false hit_count match

From 146111514a8c126268e848e45b7dd967329b072f Mon Sep 17 00:00:00 2001
From: Tim Gardner <[email protected]>
Date: Thu, 18 Feb 2010 20:33:00 -0700
Subject: [PATCH] xt_recent: Fix false match.

A rule with a zero hit_count will always match.

---
 net/netfilter/xt_recent.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
net/netfilter/xt_recent.c +1 -1
index 1bb0d6c..43e83a4 100644
@@ -260,7 +260,7 @@ recent_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 		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;
 			}
Signature · 1 line
1.6.2.4
#1 lore raw
Quoted 7 lines · Tim Gardner
Quoted 1 line

From 146111514a8c126268e848e45b7dd967329b072f Mon Sep 17 00:00:00 2001

From: Tim Gardner <[email protected]>
Date: Thu, 18 Feb 2010 20:33:00 -0700
Subject: [PATCH] xt_recent: Fix false match.

A rule with a zero hit_count will always match.

Also applied, thanks Tim.

#2 lore raw
Quoted 10 lines · On Tuesday, 23. February 2010 14:59:46 Patrick McHardy wrote:
Quoted 7 lines · Tim Gardner
Quoted 2 lines

From 146111514a8c126268e848e45b7dd967329b072f Mon Sep 17 00:00:00 2001

From: Tim Gardner <[email protected]>
Date: Thu, 18 Feb 2010 20:33:00 -0700
Subject: [PATCH] xt_recent: Fix false match.

A rule with a zero hit_count will always match.

Also applied, thanks Tim.

I just updated from kernel 2.6.32.9 to kernel 2.6.32.10 which contains the xt_recent "zero hit_count will always match" fix.

After that xt_recent stopped working for this scenario:

iptables -A INPUT -m recent --rcheck --rdest --name INET_IP -j LOG echo "+1.2.3.4" >/proc/net/xt_recent/INET_IP

The ip address 1.2.3.4 represents the current ip of my dial up connection.

If I change "--rcheck" to "--update", it works again. Reverting the patch fixes the issue.

Maybe this is related to the xt_recent
proc interface creating the entry
(with a zero hit count)?

Cheers,
Thomas

#3 lore raw
Quoted 23 lines · Thomas Jarosch
Quoted 4 lines · On Tuesday, 23. February 2010 14:59:46 Patrick McHardy wrote:
Quoted 2 lines · Tim Gardner
Quoted 1 line

From 146111514a8c126268e848e45b7dd967329b072f Mon Sep 17 00:00:00 2001

Quoted 8 lines
Quoted 6 lines

From: Tim Gardner <[email protected]>
Date: Thu, 18 Feb 2010 20:33:00 -0700
Subject: [PATCH] xt_recent: Fix false match.

A rule with a zero hit_count will always match.

Also applied, thanks Tim.

I just updated from kernel 2.6.32.9 to kernel 2.6.32.10 which contains the xt_recent "zero hit_count will always match" fix.

After that xt_recent stopped working for this scenario:

iptables -A INPUT -m recent --rcheck --rdest --name INET_IP -j LOG echo "+1.2.3.4" >/proc/net/xt_recent/INET_IP

Quoted 9 lines

The ip address 1.2.3.4 represents the current ip of my dial up connection.

If I change "--rcheck" to "--update", it works again. Reverting the patch fixes the issue.

Maybe this is related to the xt_recent
proc interface creating the entry
(with a zero hit count)?

Mhh, looking at that patch again, I think it should actually do:

if (!info->hit_count || ++hits >= info->hit_count)

    ...

since a hit_count of 0 implies that the user just wants to check for the presence of the entry. Thomas, could you give that a try?

#4 lore raw
Quoted 34 lines · Patrick McHardy
Quoted 23 lines · Thomas Jarosch
Quoted 4 lines · On Tuesday, 23. February 2010 14:59:46 Patrick McHardy wrote:
Quoted 2 lines · Tim Gardner
Quoted 1 line

From 146111514a8c126268e848e45b7dd967329b072f Mon Sep 17 00:00:00 2001

Quoted 8 lines
Quoted 6 lines

From: Tim Gardner<[email protected]>
Date: Thu, 18 Feb 2010 20:33:00 -0700
Subject: [PATCH] xt_recent: Fix false match.

A rule with a zero hit_count will always match.

Also applied, thanks Tim.

I just updated from kernel 2.6.32.9 to kernel 2.6.32.10 which contains the xt_recent "zero hit_count will always match" fix.

After that xt_recent stopped working for this scenario:

iptables -A INPUT -m recent --rcheck --rdest --name INET_IP -j LOG echo "+1.2.3.4">/proc/net/xt_recent/INET_IP

Quoted 9 lines

The ip address 1.2.3.4 represents the current ip of my dial up connection.

If I change "--rcheck" to "--update", it works again. Reverting the patch fixes the issue.

Maybe this is related to the xt_recent
proc interface creating the entry
(with a zero hit count)?

Quoted 8 lines

Mhh, looking at that patch again, I think it should actually do:

if (!info->hit_count || ++hits>= info->hit_count)

     ...

since a hit_count of 0 implies that the user just wants to check for the presence of the entry. Thomas, could you give that a try?

I think you're right. Its kind of a subtle exit condition.

rtg

Signature · 2 lines
Tim Gardner [email protected] www.tpi.com
OR 503-601-0234 x102 MT 406-443-5357
#5 lore raw
In reply to Patrick McHardy
Quoted 11 lines · On Friday, 19. March 2010 16:41:49 you wrote:
Quoted 3 lines

Maybe this is related to the xt_recent
proc interface creating the entry
(with a zero hit count)?

Mhh, looking at that patch again, I think it should actually do:

if (!info->hit_count || ++hits >= info->hit_count)

    ...

since a hit_count of 0 implies that the user just wants to check for the presence of the entry. Thomas, could you give that a try?

The new code works. Isn't that almost the same as reverting the original patch? info->hit_count == 0 will match again.

So we could just go back to

"if (++hits >= info->hit_count)"

Or am I missing something?

Clearly your new version is more readable about the intent.

Cheers,
Thomas

#6 lore raw
Quoted 15 lines · Thomas Jarosch
Quoted 4 lines · On Friday, 19. March 2010 16:41:49 you wrote:
Quoted 4 lines

Maybe this is related to the xt_recent
proc interface creating the entry
(with a zero hit count)?

Quoted 8 lines

Mhh, looking at that patch again, I think it should actually do:

if (!info->hit_count || ++hits >= info->hit_count)

    ...

since a hit_count of 0 implies that the user just wants to check for the presence of the entry. Thomas, could you give that a try?

Quoted 9 lines

The new code works. Isn't that almost the same as reverting the original patch? info->hit_count == 0 will match again.

So we could just go back to

"if (++hits >= info->hit_count)"

Or am I missing something?

I think you're right. Tim, please remind me, why was the match on zero hits considered a false positive?

#7 lore raw
Quoted 21 lines · Patrick McHardy
Quoted 20 lines · Thomas Jarosch
Quoted 4 lines · On Friday, 19. March 2010 16:41:49 you wrote:
Quoted 4 lines

Maybe this is related to the xt_recent
proc interface creating the entry
(with a zero hit count)?

Quoted 8 lines

Mhh, looking at that patch again, I think it should actually do:

if (!info->hit_count || ++hits>= info->hit_count)

     ...

since a hit_count of 0 implies that the user just wants to check for the presence of the entry. Thomas, could you give that a try?

The new code works. Isn't that almost the same as reverting the original patch? info->hit_count == 0 will match again.

So we could just go back to

Quoted 8 lines
Quoted 4 lines

"if (++hits>= info->hit_count)"

Or am I missing something?

I think you're right. Tim, please remind me, why was the match on zero hits considered a false positive?

Because it looked like it? Maybe its just whining after the fact, but 3 of us missed that it was also an exit condition. IMHO it was too subtle. I like your final patch much better because, as Thomas pointed out, it makes it a bit clearer what that clause is doing.

rtg

Signature · 2 lines
Tim Gardner [email protected] www.tpi.com
OR 503-601-0234 x102 MT 406-443-5357
Tested-by Thomas Jarosch
#8 lore raw
In reply to Thomas Jarosch
Quoted 18 lines · Thomas Jarosch
Quoted 4 lines · On Friday, 19. March 2010 16:41:49 you wrote:
Quoted 4 lines

Maybe this is related to the xt_recent
proc interface creating the entry
(with a zero hit count)?

Quoted 8 lines

Mhh, looking at that patch again, I think it should actually do:

if (!info->hit_count || ++hits >= info->hit_count)

    ...

since a hit_count of 0 implies that the user just wants to check for the presence of the entry. Thomas, could you give that a try?

The new code works. Isn't that almost the same as reverting the original patch? info->hit_count == 0 will match again.

Quoted 7 lines

So we could just go back to

"if (++hits >= info->hit_count)"

Or am I missing something?

Clearly your new version is more readable about the intent.

Thomas, before I send this upstream with a Tested-by tag in your name, could you please confirm that this is the change you've actually tested?

Thanks.

commit ef1691504c83ba3eb636c0cfd3ed33f7a6d0b4ee

Author: Patrick McHardy <[email protected]>
Date:   Mon Mar 22 18:25:20 2010 +0100

netfilter: xt_recent: fix regression in rules using a zero hit_count

Commit 8ccb92ad (netfilter: xt_recent: fix false match) fixed supposedly false matches in rules using a zero hit_count. As it turns out there is nothing false about these matches and people are actually using entries with a hit_count of zero to make rules dependant on addresses inserted manually through /proc.

Since this slipped past the eyes of three reviewers, instead of reverting the commit in question, this patch explicitly checks for a hit_count of zero to make the intentions more clear.

net/netfilter/xt_recent.c +1 -1
index 7073dbb..971d172 100644
@@ -267,7 +267,7 @@ recent_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 		for (i = 0; i < e->nstamps; i++) {
 			if (info->seconds && time_after(time, e->stamps[i]))
 				continue;
-			if (info->hit_count && ++hits >= info->hit_count) {
+			if (!info->hit_count || ++hits >= info->hit_count) {
 				ret = !ret;
 				break;
 			}
#9 lore raw
Quoted 4 lines · Patrick McHardy
Quoted 1 line

Clearly your new version is more readable about the intent.

Thomas, before I send this upstream with a Tested-by tag in your name, could you please confirm that this is the change you've actually tested?

Yes and I just did another test run with your final patch, works fine.

Cheers,
Thomas