Skip to contentSkip to conversation

iptables: Module unload causing NULL pointer reference.

PATCH In review David Wilder 2 participants 6 messages Jun 22, 2020 – Jun 24, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover iptables: Module unload causing NULL pointer reference.

A crash happened on ppc64le when running ltp network tests triggered by "rmmod iptable_mangle".

See previous discussion in this thread: https://lists.openwall.net/netdev/2020/06/03/161 .

In the crash I found in iptable_mangle_hook() that state->net->ipv4.iptable_mangle=NULL causing a NULL pointer dereference. net->ipv4.iptable_mangle is set to NULL in iptable_mangle_net_exit() and called when ip_mangle modules is unloaded. A rmmod task was found running in the crash dump. A 2nd crash showed the same problem when running "rmmod iptable_filter" (net->ipv4.iptable_filter=NULL).

To fix this I added .pre_exit hook in all iptable_foo.c. The pre_exit will un-register the underlying hook and exit would do the table freeing. The netns core does an unconditional synchronize_rcu after the pre_exit hooks insuring no packets are in flight that have picked up the pointer before completing the un-register.

These patches include changes for both iptables and ip6tables.

We tested this fix with ltp running iptables01.sh and iptables01.sh -6 a loop for 72 hours.

David Wilder (4):
netfilter: Split ipt_unregister_table() into pre_exit and exit
helpers.
netfilter: Add a .pre_exit hook in all iptable_foo.c.
netfilter: Split ip6t_unregister_table() into pre_exit and exit
helpers.
netfilter: Add a .pre_exit hook in all ip6table_foo.c.

 include/linux/netfilter_ipv4/ip_tables.h  |  6 ++++++
 include/linux/netfilter_ipv6/ip6_tables.h |  3 +++
 net/ipv4/netfilter/ip_tables.c            | 15 ++++++++++++++-
 net/ipv4/netfilter/iptable_filter.c       | 10 +++++++++-
 net/ipv4/netfilter/iptable_mangle.c       | 10 +++++++++-
 net/ipv4/netfilter/iptable_nat.c          | 10 ++++++++--
 net/ipv4/netfilter/iptable_raw.c          | 10 +++++++++-
 net/ipv4/netfilter/iptable_security.c     | 11 +++++++++--
 net/ipv6/netfilter/ip6_tables.c           | 15 ++++++++++++++-
 net/ipv6/netfilter/ip6table_filter.c      | 10 +++++++++-
 net/ipv6/netfilter/ip6table_mangle.c      | 10 +++++++++-
 net/ipv6/netfilter/ip6table_nat.c         | 10 ++++++++--
 net/ipv6/netfilter/ip6table_raw.c         | 10 +++++++++-
 net/ipv6/netfilter/ip6table_security.c    | 10 +++++++++-
 14 files changed, 125 insertions(+), 15 deletions(-)
Signature · 1 line
1.8.3.1
#1 lore raw

PATCH 1/4 netfilter: Split ipt_unregister_table() into pre_exit and exit helpers.

The pre_exit will un-register the underlying hook and .exit will do the table freeing. The netns core does an unconditional synchronize_rcu after the pre_exit hooks insuring no packets are in flight that have picked up the pointer before completing the un-register.

---
 include/linux/netfilter_ipv4/ip_tables.h |  6 ++++++
 net/ipv4/netfilter/ip_tables.c           | 15 ++++++++++++++-
 2 files changed, 20 insertions(+), 1 deletion(-)
include/linux/netfilter_ipv4/ip_tables.h +6 -0
index b394bd4..c4676d6 100644
@@ -25,6 +25,12 @@
 int ipt_register_table(struct net *net, const struct xt_table *table,
 		       const struct ipt_replace *repl,
 		       const struct nf_hook_ops *ops, struct xt_table **res);
+
+void ipt_unregister_table_pre_exit(struct net *net, struct xt_table *table,
+		       const struct nf_hook_ops *ops);
+
+void ipt_unregister_table_exit(struct net *net, struct xt_table *table);
+
 void ipt_unregister_table(struct net *net, struct xt_table *table,
 			  const struct nf_hook_ops *ops);
 
net/ipv4/netfilter/ip_tables.c +14 -1
index c2670ea..5bf9fa0 100644
@@ -1797,11 +1797,22 @@ int ipt_register_table(struct net *net, const struct xt_table *table,
 	return ret;
 }
 
+void ipt_unregister_table_pre_exit(struct net *net, struct xt_table *table,
+				   const struct nf_hook_ops *ops)
+{
+	nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
+}
+
+void ipt_unregister_table_exit(struct net *net, struct xt_table *table)
+{
+	__ipt_unregister_table(net, table);
+}
+
 void ipt_unregister_table(struct net *net, struct xt_table *table,
 			  const struct nf_hook_ops *ops)
 {
 	if (ops)
-		nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
+		ipt_unregister_table_pre_exit(net, table, ops);
 	__ipt_unregister_table(net, table);
 }
 
@@ -1958,6 +1969,8 @@ static void __exit ip_tables_fini(void)
 
 EXPORT_SYMBOL(ipt_register_table);
 EXPORT_SYMBOL(ipt_unregister_table);
+EXPORT_SYMBOL(ipt_unregister_table_pre_exit);
+EXPORT_SYMBOL(ipt_unregister_table_exit);
 EXPORT_SYMBOL(ipt_do_table);
 module_init(ip_tables_init);
 module_exit(ip_tables_fini);
Signature · 1 line
1.8.3.1
#2 lore raw

PATCH 2/4 netfilter: Add a .pre_exit hook in all iptable_foo.c.

In reply to David Wilder

Using new helpers ipt_unregister_table_pre_exit() and ipt_unregister_table_exit().

---
 net/ipv4/netfilter/iptable_filter.c   | 10 +++++++++-
 net/ipv4/netfilter/iptable_mangle.c   | 10 +++++++++-
 net/ipv4/netfilter/iptable_nat.c      | 10 ++++++++--
 net/ipv4/netfilter/iptable_raw.c      | 10 +++++++++-
 net/ipv4/netfilter/iptable_security.c | 11 +++++++++--
 5 files changed, 44 insertions(+), 7 deletions(-)
net/ipv4/netfilter/iptable_filter.c +9 -1
index 9d54b40..8f7bc1e 100644
@@ -72,16 +72,24 @@ static int __net_init iptable_filter_net_init(struct net *net)
 	return 0;
 }
 
+static void __net_exit iptable_filter_net_pre_exit(struct net *net)
+{
+	if (net->ipv4.iptable_filter)
+		ipt_unregister_table_pre_exit(net, net->ipv4.iptable_filter,
+					      filter_ops);
+}
+
 static void __net_exit iptable_filter_net_exit(struct net *net)
 {
 	if (!net->ipv4.iptable_filter)
 		return;
-	ipt_unregister_table(net, net->ipv4.iptable_filter, filter_ops);
+	ipt_unregister_table_exit(net, net->ipv4.iptable_filter);
 	net->ipv4.iptable_filter = NULL;
 }
 
 static struct pernet_operations iptable_filter_net_ops = {
 	.init = iptable_filter_net_init,
+	.pre_exit = iptable_filter_net_pre_exit,
 	.exit = iptable_filter_net_exit,
 };
 
net/ipv4/netfilter/iptable_mangle.c +9 -1
index bb9266e..f703a71 100644
@@ -100,15 +100,23 @@ static int __net_init iptable_mangle_table_init(struct net *net)
 	return ret;
 }
 
+static void __net_exit iptable_mangle_net_pre_exit(struct net *net)
+{
+	if (net->ipv4.iptable_mangle)
+		ipt_unregister_table_pre_exit(net, net->ipv4.iptable_mangle,
+					      mangle_ops);
+}
+
 static void __net_exit iptable_mangle_net_exit(struct net *net)
 {
 	if (!net->ipv4.iptable_mangle)
 		return;
-	ipt_unregister_table(net, net->ipv4.iptable_mangle, mangle_ops);
+	ipt_unregister_table_exit(net, net->ipv4.iptable_mangle);
 	net->ipv4.iptable_mangle = NULL;
 }
 
 static struct pernet_operations iptable_mangle_net_ops = {
+	.pre_exit = iptable_mangle_net_pre_exit,
 	.exit = iptable_mangle_net_exit,
 };
 
net/ipv4/netfilter/iptable_nat.c +8 -2
index ad33687..b0143b1 100644
@@ -113,16 +113,22 @@ static int __net_init iptable_nat_table_init(struct net *net)
 	return ret;
 }
 
+static void __net_exit iptable_nat_net_pre_exit(struct net *net)
+{
+	if (net->ipv4.nat_table)
+		ipt_nat_unregister_lookups(net);
+}
+
 static void __net_exit iptable_nat_net_exit(struct net *net)
 {
 	if (!net->ipv4.nat_table)
 		return;
-	ipt_nat_unregister_lookups(net);
-	ipt_unregister_table(net, net->ipv4.nat_table, NULL);
+	ipt_unregister_table_exit(net, net->ipv4.nat_table);
 	net->ipv4.nat_table = NULL;
 }
 
 static struct pernet_operations iptable_nat_net_ops = {
+	.pre_exit = iptable_nat_net_pre_exit,
 	.exit	= iptable_nat_net_exit,
 };
 
net/ipv4/netfilter/iptable_raw.c +9 -1
index 69697eb..9abfe6b 100644
@@ -67,15 +67,23 @@ static int __net_init iptable_raw_table_init(struct net *net)
 	return ret;
 }
 
+static void __net_exit iptable_raw_net_pre_exit(struct net *net)
+{
+	if (net->ipv4.iptable_raw)
+		ipt_unregister_table_pre_exit(net, net->ipv4.iptable_raw,
+					      rawtable_ops);
+}
+
 static void __net_exit iptable_raw_net_exit(struct net *net)
 {
 	if (!net->ipv4.iptable_raw)
 		return;
-	ipt_unregister_table(net, net->ipv4.iptable_raw, rawtable_ops);
+	ipt_unregister_table_exit(net, net->ipv4.iptable_raw);
 	net->ipv4.iptable_raw = NULL;
 }
 
 static struct pernet_operations iptable_raw_net_ops = {
+	.pre_exit = iptable_raw_net_pre_exit,
 	.exit = iptable_raw_net_exit,
 };
 
net/ipv4/netfilter/iptable_security.c +9 -2
index ac633c1..415c197 100644
@@ -62,16 +62,23 @@ static int __net_init iptable_security_table_init(struct net *net)
 	return ret;
 }
 
+static void __net_exit iptable_security_net_pre_exit(struct net *net)
+{
+	if (net->ipv4.iptable_security)
+		ipt_unregister_table_pre_exit(net, net->ipv4.iptable_security,
+					      sectbl_ops);
+}
+
 static void __net_exit iptable_security_net_exit(struct net *net)
 {
 	if (!net->ipv4.iptable_security)
 		return;
-
-	ipt_unregister_table(net, net->ipv4.iptable_security, sectbl_ops);
+	ipt_unregister_table_exit(net, net->ipv4.iptable_security);
 	net->ipv4.iptable_security = NULL;
 }
 
 static struct pernet_operations iptable_security_net_ops = {
+	.pre_exit = iptable_security_net_pre_exit,
 	.exit = iptable_security_net_exit,
 };
 
Signature · 1 line
1.8.3.1
#3 lore raw

PATCH 3/4 netfilter: Split ip6t_unregister_table() into pre_exit and exit helpers.

In reply to David Wilder

The pre_exit will un-register the underlying hook and .exit will do the table freeing. The netns core does an unconditional synchronize_rcu after the pre_exit hooks insuring no packets are in flight that have picked up the pointer before completing the un-register.

---
 include/linux/netfilter_ipv6/ip6_tables.h |  3 +++
 net/ipv6/netfilter/ip6_tables.c           | 15 ++++++++++++++-
 2 files changed, 17 insertions(+), 1 deletion(-)
include/linux/netfilter_ipv6/ip6_tables.h +3 -0
index 8225f78..1547d5f 100644
@@ -29,6 +29,9 @@ int ip6t_register_table(struct net *net, const struct xt_table *table,
 			const struct nf_hook_ops *ops, struct xt_table **res);
 void ip6t_unregister_table(struct net *net, struct xt_table *table,
 			   const struct nf_hook_ops *ops);
+void ip6t_unregister_table_pre_exit(struct net *net, struct xt_table *table,
+				    const struct nf_hook_ops *ops);
+void ip6t_unregister_table_exit(struct net *net, struct xt_table *table);
 extern unsigned int ip6t_do_table(struct sk_buff *skb,
 				  const struct nf_hook_state *state,
 				  struct xt_table *table);
net/ipv6/netfilter/ip6_tables.c +14 -1
index e273934..e96a431 100644
@@ -1807,11 +1807,22 @@ int ip6t_register_table(struct net *net, const struct xt_table *table,
 	return ret;
 }
 
+void ip6t_unregister_table_pre_exit(struct net *net, struct xt_table *table,
+				    const struct nf_hook_ops *ops)
+{
+	nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
+}
+
+void ip6t_unregister_table_exit(struct net *net, struct xt_table *table)
+{
+	__ip6t_unregister_table(net, table);
+}
+
 void ip6t_unregister_table(struct net *net, struct xt_table *table,
 			   const struct nf_hook_ops *ops)
 {
 	if (ops)
-		nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
+		ip6t_unregister_table_pre_exit(net, table, ops);
 	__ip6t_unregister_table(net, table);
 }
 
@@ -1969,6 +1980,8 @@ static void __exit ip6_tables_fini(void)
 
 EXPORT_SYMBOL(ip6t_register_table);
 EXPORT_SYMBOL(ip6t_unregister_table);
+EXPORT_SYMBOL(ip6t_unregister_table_pre_exit);
+EXPORT_SYMBOL(ip6t_unregister_table_exit);
 EXPORT_SYMBOL(ip6t_do_table);
 
 module_init(ip6_tables_init);
Signature · 1 line
1.8.3.1
#4 lore raw

PATCH 4/4 netfilter: Add a .pre_exit hook in all ip6table_foo.c.

In reply to David Wilder

Using new helpers ip6t_unregister_table_pre_exit() and ip6t_unregister_table_exit().

---
 net/ipv6/netfilter/ip6table_filter.c   | 10 +++++++++-
 net/ipv6/netfilter/ip6table_mangle.c   | 10 +++++++++-
 net/ipv6/netfilter/ip6table_nat.c      | 10 ++++++++--
 net/ipv6/netfilter/ip6table_raw.c      | 10 +++++++++-
 net/ipv6/netfilter/ip6table_security.c | 10 +++++++++-
 5 files changed, 44 insertions(+), 6 deletions(-)
net/ipv6/netfilter/ip6table_filter.c +9 -1
index 32667f5..88337b5 100644
@@ -73,16 +73,24 @@ static int __net_init ip6table_filter_net_init(struct net *net)
 	return 0;
 }
 
+static void __net_exit ip6table_filter_net_pre_exit(struct net *net)
+{
+	if (net->ipv6.ip6table_filter)
+		ip6t_unregister_table_pre_exit(net, net->ipv6.ip6table_filter,
+					       filter_ops);
+}
+
 static void __net_exit ip6table_filter_net_exit(struct net *net)
 {
 	if (!net->ipv6.ip6table_filter)
 		return;
-	ip6t_unregister_table(net, net->ipv6.ip6table_filter, filter_ops);
+	ip6t_unregister_table_exit(net, net->ipv6.ip6table_filter);
 	net->ipv6.ip6table_filter = NULL;
 }
 
 static struct pernet_operations ip6table_filter_net_ops = {
 	.init = ip6table_filter_net_init,
+	.pre_exit = ip6table_filter_net_pre_exit,
 	.exit = ip6table_filter_net_exit,
 };
 
net/ipv6/netfilter/ip6table_mangle.c +9 -1
index 070afb9..1a27486 100644
@@ -93,16 +93,24 @@ static int __net_init ip6table_mangle_table_init(struct net *net)
 	return ret;
 }
 
+static void __net_exit ip6table_mangle_net_pre_exit(struct net *net)
+{
+	if (net->ipv6.ip6table_mangle)
+		ip6t_unregister_table_pre_exit(net, net->ipv6.ip6table_mangle,
+					       mangle_ops);
+}
+
 static void __net_exit ip6table_mangle_net_exit(struct net *net)
 {
 	if (!net->ipv6.ip6table_mangle)
 		return;
 
-	ip6t_unregister_table(net, net->ipv6.ip6table_mangle, mangle_ops);
+	ip6t_unregister_table_exit(net, net->ipv6.ip6table_mangle);
 	net->ipv6.ip6table_mangle = NULL;
 }
 
 static struct pernet_operations ip6table_mangle_net_ops = {
+	.pre_exit = ip6table_mangle_net_pre_exit,
 	.exit = ip6table_mangle_net_exit,
 };
 
net/ipv6/netfilter/ip6table_nat.c +8 -2
index 0f48759..0a23265 100644
@@ -114,16 +114,22 @@ static int __net_init ip6table_nat_table_init(struct net *net)
 	return ret;
 }
 
+static void __net_exit ip6table_nat_net_pre_exit(struct net *net)
+{
+	if (net->ipv6.ip6table_nat)
+		ip6t_nat_unregister_lookups(net);
+}
+
 static void __net_exit ip6table_nat_net_exit(struct net *net)
 {
 	if (!net->ipv6.ip6table_nat)
 		return;
-	ip6t_nat_unregister_lookups(net);
-	ip6t_unregister_table(net, net->ipv6.ip6table_nat, NULL);
+	ip6t_unregister_table_exit(net, net->ipv6.ip6table_nat);
 	net->ipv6.ip6table_nat = NULL;
 }
 
 static struct pernet_operations ip6table_nat_net_ops = {
+	.pre_exit = ip6table_nat_net_pre_exit,
 	.exit	= ip6table_nat_net_exit,
 };
 
net/ipv6/netfilter/ip6table_raw.c +9 -1
index a22100b..8f9e742 100644
@@ -66,15 +66,23 @@ static int __net_init ip6table_raw_table_init(struct net *net)
 	return ret;
 }
 
+static void __net_exit ip6table_raw_net_pre_exit(struct net *net)
+{
+	if (net->ipv6.ip6table_raw)
+		ip6t_unregister_table_pre_exit(net, net->ipv6.ip6table_raw,
+					       rawtable_ops);
+}
+
 static void __net_exit ip6table_raw_net_exit(struct net *net)
 {
 	if (!net->ipv6.ip6table_raw)
 		return;
-	ip6t_unregister_table(net, net->ipv6.ip6table_raw, rawtable_ops);
+	ip6t_unregister_table_exit(net, net->ipv6.ip6table_raw);
 	net->ipv6.ip6table_raw = NULL;
 }
 
 static struct pernet_operations ip6table_raw_net_ops = {
+	.pre_exit = ip6table_raw_net_pre_exit,
 	.exit = ip6table_raw_net_exit,
 };
 
net/ipv6/netfilter/ip6table_security.c +9 -1
index a74335f..5e8c48f 100644
@@ -61,15 +61,23 @@ static int __net_init ip6table_security_table_init(struct net *net)
 	return ret;
 }
 
+static void __net_exit ip6table_security_net_pre_exit(struct net *net)
+{
+	if (net->ipv6.ip6table_security)
+		ip6t_unregister_table_pre_exit(net, net->ipv6.ip6table_security,
+					       sectbl_ops);
+}
+
 static void __net_exit ip6table_security_net_exit(struct net *net)
 {
 	if (!net->ipv6.ip6table_security)
 		return;
-	ip6t_unregister_table(net, net->ipv6.ip6table_security, sectbl_ops);
+	ip6t_unregister_table_exit(net, net->ipv6.ip6table_security);
 	net->ipv6.ip6table_security = NULL;
 }
 
 static struct pernet_operations ip6table_security_net_ops = {
+	.pre_exit = ip6table_security_net_pre_exit,
 	.exit = ip6table_security_net_exit,
 };
 
Signature · 1 line
1.8.3.1
#5 lore raw
In reply to David Wilder
Quoted 11 lines · David Wilder

A crash happened on ppc64le when running ltp network tests triggered by "rmmod iptable_mangle".

See previous discussion in this thread: https://lists.openwall.net/netdev/2020/06/03/161 .

In the crash I found in iptable_mangle_hook() that state->net->ipv4.iptable_mangle=NULL causing a NULL pointer dereference. net->ipv4.iptable_mangle is set to NULL in iptable_mangle_net_exit() and called when ip_mangle modules is unloaded. A rmmod task was found running in the crash dump. A 2nd crash showed the same problem when running "rmmod iptable_filter" (net->ipv4.iptable_filter=NULL).

To fix this I added .pre_exit hook in all iptable_foo.c. The pre_exit will un-register the underlying hook and exit would do the table freeing. The netns core does an unconditional synchronize_rcu after the pre_exit hooks insuring no packets are in flight that have picked up the pointer before completing the un-register.

These patches include changes for both iptables and ip6tables.

We tested this fix with ltp running iptables01.sh and iptables01.sh -6 a loop for 72 hours.

Series applied, thanks.