Skip to contentSkip to conversation

iptables: ensure number of counters is >0 in do_replace()

PATCH Applied Dave Jones 4 participants 6 messages May 19, 2015 – May 21, 2015

Subsystems (4)
Networking Network drivers netfilter Ethernet bridge
lore
Share & tools
lore

Dave Jones

author
#0 lore raw

After improving setsockopt() coverage in trinity, I started triggering vmalloc failures pretty reliably from this code path:

warn_alloc_failed+0xe9/0x140
__vmalloc_node_range+0x1be/0x270
vzalloc+0x4b/0x50
__do_replace+0x52/0x260 [ip_tables]
do_ipt_set_ctl+0x15d/0x1d0 [ip_tables]
nf_setsockopt+0x65/0x90
ip_setsockopt+0x61/0xa0
raw_setsockopt+0x16/0x60
sock_common_setsockopt+0x14/0x20
SyS_setsockopt+0x71/0xd0

It turns out we don't validate that the num_counters field in the struct we pass in from userspace is initialized.

Signed-off-by Dave Jones <[email protected]>
net/ipv4/netfilter/ip_tables.c +3 -0
index c69db7fa25ee..12a33d178614 100644
@@ -1262,6 +1262,9 @@ do_replace(struct net *net, const void __user *user, unsigned int len)
 	/* overflow check */
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+
 	tmp.name[sizeof(tmp.name)-1] = 0;
 
 	newinfo = xt_alloc_table_info(tmp.size);
maintainer
#1 lore raw
Quoted 24 lines · Dave Jones

After improving setsockopt() coverage in trinity, I started triggering vmalloc failures pretty reliably from this code path:

warn_alloc_failed+0xe9/0x140
__vmalloc_node_range+0x1be/0x270
vzalloc+0x4b/0x50
__do_replace+0x52/0x260 [ip_tables]
do_ipt_set_ctl+0x15d/0x1d0 [ip_tables]
nf_setsockopt+0x65/0x90
ip_setsockopt+0x61/0xa0
raw_setsockopt+0x16/0x60
sock_common_setsockopt+0x14/0x20
SyS_setsockopt+0x71/0xd0

It turns out we don't validate that the num_counters field in the struct we pass in from userspace is initialized.

Signed-off-by Dave Jones <[email protected]>
net/ipv4/netfilter/ip_tables.c +0 -0
index c69db7fa25ee..12a33d178614 100644

net/ipv4/netfilter/ip_tables.c do_replace(struct net *net, const void __user *user, unsigned int len) · line 1267

Quoted 6 lines
net/ipv4/netfilter/ip_tables.c +3 -0
@@ -1262,6 +1262,9 @@ do_replace(struct net *net, const void __user *user, unsigned int len)
 	/* overflow check */
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+

Good catch. Due to iptables wonderful copy-paste programming model, this also needs fixing in compat_do_replace, ip6tables, arbtables and ebtables (8 different places, ick).

author
#2 lore raw
Quoted 20 lines · Florian Westphal
Quoted 19 lines · Dave Jones

After improving setsockopt() coverage in trinity, I started triggering vmalloc failures pretty reliably from this code path:

warn_alloc_failed+0xe9/0x140
__vmalloc_node_range+0x1be/0x270
vzalloc+0x4b/0x50
__do_replace+0x52/0x260 [ip_tables]
do_ipt_set_ctl+0x15d/0x1d0 [ip_tables]
nf_setsockopt+0x65/0x90
ip_setsockopt+0x61/0xa0
raw_setsockopt+0x16/0x60
sock_common_setsockopt+0x14/0x20
SyS_setsockopt+0x71/0xd0

It turns out we don't validate that the num_counters field in the struct we pass in from userspace is initialized.

Signed-off-by Dave Jones <[email protected]>
Quoted 15 lines

net/ipv4/netfilter/ip_tables.c do_replace(struct net *net, const void __user *user, unsigned int len) · line 1267

Quoted 11 lines
net/ipv4/netfilter/ip_tables.c +3 -0
index c69db7fa25ee..12a33d178614 100644
@@ -1262,6 +1262,9 @@ do_replace(struct net *net, const void __user *user, unsigned int len)
 	/* overflow check */
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+

Good catch. Due to iptables wonderful copy-paste programming model, this also needs fixing in compat_do_replace, ip6tables, arbtables and ebtables (8 different places, ick).

ugh. ok, I'll fix them all up.
What's the preferred format, 8 separate patches, or 1 all-in-one diff for all instances of this bug ?

thanks,

Dave

maintainer
#3 lore raw

From: Dave Jones <[email protected]>
Date: Tue, 19 May 2015 19:42:52 -0400

Quoted 2 lines

What's the preferred format, 8 separate patches, or 1 all-in-one diff for all instances of this bug ?

Personally, I think you could do this in one patch.

author
#4 lore raw

netfilter: ensure number of counters is >0 in do_replace()

After improving setsockopt() coverage in trinity, I started triggering vmalloc failures pretty reliably from this code path:

warn_alloc_failed+0xe9/0x140
__vmalloc_node_range+0x1be/0x270
vzalloc+0x4b/0x50
__do_replace+0x52/0x260 [ip_tables]
do_ipt_set_ctl+0x15d/0x1d0 [ip_tables]
nf_setsockopt+0x65/0x90
ip_setsockopt+0x61/0xa0
raw_setsockopt+0x16/0x60
sock_common_setsockopt+0x14/0x20
SyS_setsockopt+0x71/0xd0

It turns out we don't validate that the num_counters field in the struct we pass in from userspace is initialized.

The same problem also exists in ebtables, arptables, ipv6, and the compat variants.

Signed-off-by Dave Jones <[email protected]>
net/bridge/netfilter/ebtables.c +4 -0
index 91180a7fc943..24c7c96bf5f8 100644
@@ -1117,6 +1117,8 @@ static int do_replace(struct net *net, const void __user *user,
 		return -ENOMEM;
 	if (tmp.num_counters >= INT_MAX / sizeof(struct ebt_counter))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
 
 	tmp.name[sizeof(tmp.name) - 1] = 0;
 

@@ -2159,6 +2161,8 @@ static int compat_copy_ebt_replace_from_user(struct ebt_replace *repl,
 		return -ENOMEM;
 	if (tmp.num_counters >= INT_MAX / sizeof(struct ebt_counter))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
 
 	memcpy(repl, &tmp, offsetof(struct ebt_replace, hook_entry));
net/ipv4/netfilter/arp_tables.c +6 -0
index 13bfe84bf3ca..a61200754f4b 100644
@@ -1075,6 +1075,9 @@ static int do_replace(struct net *net, const void __user *user,
 	/* overflow check */
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+
 	tmp.name[sizeof(tmp.name)-1] = 0;
 
 	newinfo = xt_alloc_table_info(tmp.size);

@@ -1499,6 +1502,9 @@ static int compat_do_replace(struct net *net, void __user *user,
 		return -ENOMEM;
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+
 	tmp.name[sizeof(tmp.name)-1] = 0;
 
 	newinfo = xt_alloc_table_info(tmp.size);
net/ipv4/netfilter/ip_tables.c +6 -0
index c69db7fa25ee..2d0e265fef6e 100644
@@ -1262,6 +1262,9 @@ do_replace(struct net *net, const void __user *user, unsigned int len)
 	/* overflow check */
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+
 	tmp.name[sizeof(tmp.name)-1] = 0;
 
 	newinfo = xt_alloc_table_info(tmp.size);

@@ -1809,6 +1812,9 @@ compat_do_replace(struct net *net, void __user *user, unsigned int len)
 		return -ENOMEM;
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+
 	tmp.name[sizeof(tmp.name)-1] = 0;
 
 	newinfo = xt_alloc_table_info(tmp.size);
net/ipv6/netfilter/ip6_tables.c +6 -0
index 1a732a1d3c8e..62f5b0d0bc9b 100644
@@ -1275,6 +1275,9 @@ do_replace(struct net *net, const void __user *user, unsigned int len)
 	/* overflow check */
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+
 	tmp.name[sizeof(tmp.name)-1] = 0;
 
 	newinfo = xt_alloc_table_info(tmp.size);

@@ -1822,6 +1825,9 @@ compat_do_replace(struct net *net, void __user *user, unsigned int len)
 		return -ENOMEM;
 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 		return -ENOMEM;
+	if (tmp.num_counters == 0)
+		return -EINVAL;
+
 	tmp.name[sizeof(tmp.name)-1] = 0;
 
 	newinfo = xt_alloc_table_info(tmp.size);
maintainer Applied
#5 lore raw

Re: netfilter: ensure number of counters is >0 in do_replace()

Quoted 14 lines · Dave Jones

After improving setsockopt() coverage in trinity, I started triggering vmalloc failures pretty reliably from this code path:

warn_alloc_failed+0xe9/0x140
__vmalloc_node_range+0x1be/0x270
vzalloc+0x4b/0x50
__do_replace+0x52/0x260 [ip_tables]
do_ipt_set_ctl+0x15d/0x1d0 [ip_tables]
nf_setsockopt+0x65/0x90
ip_setsockopt+0x61/0xa0
raw_setsockopt+0x16/0x60
sock_common_setsockopt+0x14/0x20
SyS_setsockopt+0x71/0xd0
Quoted 5 lines

It turns out we don't validate that the num_counters field in the struct we pass in from userspace is initialized.

The same problem also exists in ebtables, arptables, ipv6, and the compat variants.

Applied.

This also applies to -stable kernels:

3.2.x
3.4.x
3.10.x
3.12.x
3.14.x
3.18.x
4.0.x

so after some testing and a little while, I'll pass this on.

Thanks.