mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH] netfilter: ip_tables: propagate compat_table_info() error
@ 2026-11-08 22:33 Noah Weinberger
  2026-11-09  2:43 ` Florian Westphal
  0 siblings, 1 reply; 2+ messages in thread
From: Noah Weinberger @ 2026-11-08 22:33 UTC (permalink / raw)
  To: pablo, fw, kadlec; +Cc: netfilter-devel, netdev, linux-kernel

Hello,

In the IPT_SO_GET_INFO compat path, the return value of
compat_table_info() is ignored: on failure (vmalloc ENOMEM in
xt_compat_init_offsets()), private = &tmp is used anyway, and ret is
overwritten with 0 after a successful copy_to_user - so a 32-bit caller
gets stale or partially adjusted 64-bit-layout size/hook_entry/underflow
values instead of an error. The follow-up GET_ENTRIES then fails with
-EAGAIN. tmp is fully initialized, so there is no disclosure; this is a
robustness fix. The same pattern exists in ip6_tables.c; both are
patched.

Affected version/commit: mainline fa4ad798f6f681031775266044ba1b0cf3c670bd (v7.4-rc1, 2026-11-07), net/ipv4/netfilter/ip_tables.c:968-987, net/ipv6/netfilter/ip6_tables.c:989-1008.
Longstanding code.

This was found during an assisted (AI-supported) static source review of
mainline fa4ad798f6f681031775266044ba1b0cf3c670bd (v7.4-rc1, 2026-11-07); every code path cited below was re-verified by the
reporter against that tree. The patch below has NOT been compiled and has
NOT been runtime tested, and no reproducer was produced - the triggering
conditions are described above instead. Please treat this as a report for
verification rather than a battle-tested fix.

Reported-by: Noah Weinberger <noahweinberger2003@proton.me>
---
diff --git a/net/ipv4/netfilter/ip_tables.c b/net/ipv4/netfilter/ip_tables.c
index 809441c..e11a5af 100644
--- a/net/ipv4/netfilter/ip_tables.c
+++ b/net/ipv4/netfilter/ip_tables.c
@@ -968,6 +968,8 @@ static int get_info(struct net *net, void __user *user, const int *len)
 		if (in_compat_syscall()) {
 			ret = compat_table_info(private, &tmp);
 			xt_compat_flush_offsets(AF_INET);
+			if (ret)
+				goto unlock;
 			private = &tmp;
 		}
 #endif
@@ -986,6 +988,7 @@ static int get_info(struct net *net, void __user *user, const int *len)
 		else
 			ret = 0;
 
+unlock:
 		xt_table_unlock(t);
 		module_put(t->me);
 	} else

diff --git a/net/ipv6/netfilter/ip6_tables.c b/net/ipv6/netfilter/ip6_tables.c
index 313c4aa..1d5749e 100644
--- a/net/ipv6/netfilter/ip6_tables.c
+++ b/net/ipv6/netfilter/ip6_tables.c
@@ -989,6 +989,8 @@ static int get_info(struct net *net, void __user *user, const int *len)
 		if (in_compat_syscall()) {
 			ret = compat_table_info(private, &tmp);
 			xt_compat_flush_offsets(AF_INET6);
+			if (ret)
+				goto unlock;
 			private = &tmp;
 		}
 #endif
@@ -1007,6 +1009,7 @@ static int get_info(struct net *net, void __user *user, const int *len)
 		else
 			ret = 0;
 
+unlock:
 		xt_table_unlock(t);
 		module_put(t->me);
 	} else


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] netfilter: ip_tables: propagate compat_table_info() error
  2026-11-08 22:33 [PATCH] netfilter: ip_tables: propagate compat_table_info() error Noah Weinberger
@ 2026-11-09  2:43 ` Florian Westphal
  0 siblings, 0 replies; 2+ messages in thread
From: Florian Westphal @ 2026-11-09  2:43 UTC (permalink / raw)
  To: Noah Weinberger; +Cc: pablo, kadlec, netfilter-devel, netdev, linux-kernel

Noah Weinberger <NoahWeinberger2003@proton.me> wrote:
> In the IPT_SO_GET_INFO compat path, the return value of
> compat_table_info() is ignored: on failure (vmalloc ENOMEM in
> xt_compat_init_offsets()), private = &tmp is used anyway, and ret is
> overwritten with 0 after a successful copy_to_user - so a 32-bit caller
> gets stale or partially adjusted 64-bit-layout size/hook_entry/underflow
> values instead of an error. The follow-up GET_ENTRIES then fails with
> -EAGAIN. tmp is fully initialized, so there is no disclosure; this is a
> robustness fix. The same pattern exists in ip6_tables.c; both are
> patched.
> 
> Affected version/commit: mainline fa4ad798f6f681031775266044ba1b0cf3c670bd (v7.4-rc1, 2026-11-07), net/ipv4/netfilter/ip_tables.c:968-987, net/ipv6/netfilter/ip6_tables.c:989-1008.
> Longstanding code.
> 
> This was found during an assisted (AI-supported) static source review of
> mainline fa4ad798f6f681031775266044ba1b0cf3c670bd (v7.4-rc1, 2026-11-07); every code path cited below was re-verified by the
> reporter against that tree. The patch below has NOT been compiled and has
> NOT been runtime tested, and no reproducer was produced - the triggering
> conditions are described above instead. Please treat this as a report for
> verification rather than a battle-tested fix.

8-(

net/ipv4/netfilter/ip_tables.c: In function 'get_info':
net/ipv4/netfilter/ip_tables.c:992:1: error: label 'unlock' defined but not used [-Werror=unused-label]
  992 | unlock:

8-(

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2026-11-09  2:43 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-08 22:33 [PATCH] netfilter: ip_tables: propagate compat_table_info() error Noah Weinberger
2026-11-09  2:43 ` Florian Westphal

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®