Skip to contentSkip to conversation

netfilter: nf_defrag: Skip defrag if NOTRACK is set

PATCH Superseded Subash Abhinov Kasiviswanathan 2 participants 3 messages Jan 4, 2018 – Jan 9, 2018

Subsystems (2)
Networking netfilter

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Subash Abhinov Kasiviswanathan

author
#0 lore raw

PATCH v3 1/1 netfilter: nf_defrag: Skip defrag if NOTRACK is set

conntrack defrag is needed only if some module like CONNTRACK or NAT explicitly requests it. For plain forwarding scenarios, defrag is not needed and can be skipped if NOTRACK is set in a rule.

Since conntrack defrag is currently higher priority than raw table, setting NOTRACK is not sufficient. We need to move raw to a higher priority for iptables only.

This is achieved by introducing a module parameter which allows to modify the priority. By default, the priority is NF_IP_PRI_RAW to support legacy behavior.

v1->v2: Instead of modifying NF_IP_PRI_RAW itself, use a module parameter to pass in the priority during module load as suggested by Pablo. Also update commit text.

v2->v3: Implement similar functionality for IPv6 as well

Signed-off-by Subash Abhinov Kasiviswanathan <[email protected]>
---
 net/ipv4/netfilter/iptable_raw.c          | 14 +++++++++++++-
 net/ipv4/netfilter/nf_defrag_ipv4.c       |  2 +-
 net/ipv6/netfilter/ip6table_raw.c         | 14 +++++++++++++-
 net/ipv6/netfilter/nf_defrag_ipv6_hooks.c |  3 +++
 4 files changed, 30 insertions(+), 3 deletions(-)
net/ipv4/netfilter/iptable_raw.c +13 -1
index 2642ecd..607392b 100644
@@ -12,7 +12,11 @@
 
 static int __net_init iptable_raw_table_init(struct net *net);
 
-static const struct xt_table packet_raw = {
+static int priority __read_mostly = NF_IP_PRI_RAW;
+MODULE_PARM_DESC(priority, "Priority of IPv4 raw table (NF_IP_PRI_RAW)");
+module_param(priority, int, 0000);
+
+static struct xt_table packet_raw = {
 	.name = "raw",
 	.valid_hooks =  RAW_VALID_HOOKS,
 	.me = THIS_MODULE,

@@ -70,6 +74,14 @@ static int __init iptable_raw_init(void)
 {
 	int ret;
 
+	if (priority < NF_IP_PRI_CONNTRACK_DEFRAG &&
+	    priority > NF_IP_PRI_FIRST) {
+		packet_raw.priority = priority;
+
+		pr_info("iptable_raw: Using custom rule priority=%d\n",
+			packet_raw.priority);
+	}
+
 	rawtable_ops = xt_hook_ops_alloc(&packet_raw, iptable_raw_hook);
 	if (IS_ERR(rawtable_ops))
 		return PTR_ERR(rawtable_ops);
net/ipv4/netfilter/nf_defrag_ipv4.c +1 -1
index 37fe1616..cbd987f 100644
@@ -80,7 +80,7 @@ static unsigned int ipv4_conntrack_defrag(void *priv,
 #endif
 #endif
 	/* Gather fragments. */
-	if (ip_is_fragment(ip_hdr(skb))) {
+	if (skb->_nfct != IP_CT_UNTRACKED && ip_is_fragment(ip_hdr(skb))) {
 		enum ip_defrag_users user =
 			nf_ct_defrag_user(state->hook, skb);
net/ipv6/netfilter/ip6table_raw.c +13 -1
index d4bc564..4114e7b 100644
@@ -11,7 +11,11 @@
 
 static int __net_init ip6table_raw_table_init(struct net *net);
 
-static const struct xt_table packet_raw = {
+static int priority __read_mostly = NF_IP6_PRI_RAW;
+MODULE_PARM_DESC(priority, "Priority of IPv6 raw table (NF_IP6_PRI_RAW)");
+module_param(priority, int, 0000);
+
+static struct xt_table packet_raw = {
 	.name = "raw",
 	.valid_hooks = RAW_VALID_HOOKS,
 	.me = THIS_MODULE,

@@ -63,6 +67,14 @@ static int __init ip6table_raw_init(void)
 {
 	int ret;
 
+	if (priority < NF_IP6_PRI_CONNTRACK_DEFRAG &&
+	    priority > NF_IP6_PRI_FIRST) {
+		packet_raw.priority = priority;
+
+		pr_info("ip6table_raw: Using custom rule priority=%d\n",
+			packet_raw.priority);
+	}
+
 	/* Register hooks */
 	rawtable_ops = xt_hook_ops_alloc(&packet_raw, ip6table_raw_hook);
 	if (IS_ERR(rawtable_ops))
net/ipv6/netfilter/nf_defrag_ipv6_hooks.c +3 -0
index b326da5..87b503a 100644
@@ -65,6 +65,9 @@ static unsigned int ipv6_defrag(void *priv,
 		return NF_ACCEPT;
 #endif
 
+	if (skb->_nfct == IP_CT_UNTRACKED)
+		return NF_ACCEPT;
+
 	err = nf_ct_frag6_gather(state->net, skb,
 				 nf_ct6_defrag_user(state->hook, skb));
 	/* queued */
Signature · 1 line
1.9.1
maintainer
#1 lore raw

Hi Subash,

One more concern before this gets upstream.

Quoted 31 lines · Subash Abhinov Kasiviswanathan

conntrack defrag is needed only if some module like CONNTRACK or NAT explicitly requests it. For plain forwarding scenarios, defrag is not needed and can be skipped if NOTRACK is set in a rule.

Since conntrack defrag is currently higher priority than raw table, setting NOTRACK is not sufficient. We need to move raw to a higher priority for iptables only.

This is achieved by introducing a module parameter which allows to modify the priority. By default, the priority is NF_IP_PRI_RAW to support legacy behavior.

v1->v2: Instead of modifying NF_IP_PRI_RAW itself, use a module parameter to pass in the priority during module load as suggested by Pablo. Also update commit text.

v2->v3: Implement similar functionality for IPv6 as well

Signed-off-by Subash Abhinov Kasiviswanathan <[email protected]>
---
 net/ipv4/netfilter/iptable_raw.c          | 14 +++++++++++++-
 net/ipv4/netfilter/nf_defrag_ipv4.c       |  2 +-
 net/ipv6/netfilter/ip6table_raw.c         | 14 +++++++++++++-
 net/ipv6/netfilter/nf_defrag_ipv6_hooks.c |  3 +++
 4 files changed, 30 insertions(+), 3 deletions(-)
net/ipv4/netfilter/iptable_raw.c +0 -0
index 2642ecd..607392b 100644

net/ipv4/netfilter/iptable_raw.c line 17

Quoted 7 lines
net/ipv4/netfilter/iptable_raw.c +3 -1
@@ -12,7 +12,11 @@
 
 static int __net_init iptable_raw_table_init(struct net *net);
 
-static const struct xt_table packet_raw = {
+static int priority __read_mostly = NF_IP_PRI_RAW;
+MODULE_PARM_DESC(priority, "Priority of IPv4 raw table (NF_IP_PRI_RAW)");
+module_param(priority, int, 0000);

Do you think we can turn this into an on/off knob instead?

I mean, I think it's good if you add a new NF_IP_PRI_RAW_BEFORE_DEFRAG and we place it into uapi.

I'm just worried about follow up patches from people asking to making this flexible in all other existing tables, I would like this does not happen :-).

Quoted 6 lines
net/ipv4/netfilter/iptable_raw.c +2 -0
+
+static struct xt_table packet_raw = {
 	.name = "raw",
 	.valid_hooks =  RAW_VALID_HOOKS,
 	.me = THIS_MODULE,

net/ipv4/netfilter/iptable_raw.c static int __init iptable_raw_init(void) · line 81

Quoted 8 lines
net/ipv4/netfilter/iptable_raw.c +5 -0
@@ -70,6 +74,14 @@ static int __init iptable_raw_init(void)
 {
 	int ret;
 
+	if (priority < NF_IP_PRI_CONNTRACK_DEFRAG &&
+	    priority > NF_IP_PRI_FIRST) {
+		packet_raw.priority = priority;
+
+		pr_info("iptable_raw: Using custom rule priority=%d\n",

                         ^^^^^^^^^^^^

Probably better if you add:

        #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

in this patch while on this.

Thanks for your patience, we're almost there.

author
#2 lore raw
Quoted 29 lines · Pablo Neira Ayuso

Hi Subash,

One more concern before this gets upstream.

Do you think we can turn this into an on/off knob instead?

I mean, I think it's good if you add a new NF_IP_PRI_RAW_BEFORE_DEFRAG and we place it into uapi.

I'm just worried about follow up patches from people asking to making this flexible in all other existing tables, I would like this does not happen :-).

Quoted 6 lines
net/ipv4/netfilter/iptable_raw.c +2 -0
+
+static struct xt_table packet_raw = {
 	.name = "raw",
 	.valid_hooks =  RAW_VALID_HOOKS,
 	.me = THIS_MODULE,

net/ipv4/netfilter/iptable_raw.c static int __init iptable_raw_init(void) · line 81

Quoted 8 lines
net/ipv4/netfilter/iptable_raw.c +5 -0
@@ -70,6 +74,14 @@ static int __init iptable_raw_init(void)
 {
 	int ret;
 
+	if (priority < NF_IP_PRI_CONNTRACK_DEFRAG &&
+	    priority > NF_IP_PRI_FIRST) {
+		packet_raw.priority = priority;
+
+		pr_info("iptable_raw: Using custom rule priority=%d\n",

                         ^^^^^^^^^^^^
Quoted 7 lines

Probably better if you add:

        #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

in this patch while on this.

Thanks for your patience, we're almost there.

Hi Pablo

Sure, I can update these.

Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project