netfilter: two xtables matches

Add an iptables match based on skb fields, such as mark, priority, input interface and rxhash. The match supports range based matching on one field, with optional inversion and masking.

v2: switches from xt_priority to xt_skbuff. Pablo, is this what you had in mind? It doesn't perfectly duplicate the values from nftables xt_meta. Needs more testing to cover the field-specific codepaths.

Tested by inserting

iptables -t mangle -A PREROUTING -s $SRC -j MARK --set-mark 10 iptables -A INPUT -m skbuff --min 10 --max 10 -j TRACE

The userspace tool needs work, too. For one, I just hardcoded the field_id to be skb_field_mark for this test. That's why it's missing from the command line.

Signed-off-by Willem de Bruijn <[email protected]>

---
 include/linux/netfilter/xt_skbuff.h |   33 ++++++++
 net/netfilter/Kconfig               |    9 ++
 net/netfilter/Makefile              |    1 +
 net/netfilter/xt_skbuff.c           |  141 +++++++++++++++++++++++++++++++++++
 4 files changed, 184 insertions(+), 0 deletions(-)
 create mode 100644 include/linux/netfilter/xt_skbuff.h
 create mode 100644 net/netfilter/xt_skbuff.c

include/linux/netfilter/xt_skbuff.h new file +0 -0

new file mode 100644
index 0000000..10eb8d8

include/linux/netfilter/xt_skbuff.h new file +6 -0

+	skb_field_csum = 0,
+	skb_field_hatype,
+	skb_field_iif,
+	skb_field_len,
+	skb_field_mark,
+	skb_field_pkt_type,

include/linux/netfilter/xt_skbuff.h new file +11 -0

+};
+
+struct xt_skbuff_info {
+	__u16 field_id;		/* an xt_skbuff_field_selector value */
+	__u8  invert;
+	__u64 min;
+	__u64 max;
+	__u64 mask;
+};
+
+#endif /*_XT_SKBUFF_H */

net/netfilter/Kconfig +9 -0

index fefa514..3a07a86 100644
@@ -1093,6 +1093,15 @@ config NETFILTER_XT_MATCH_PKTTYPE
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
+config NETFILTER_XT_MATCH_SKBUFF
+	tristate '"skbuff" match support'
+	depends on NETFILTER_ADVANCED
+	help
+	  This option adds a match based on the value of a chosen sk_buff
+	  field.
+
+	  To compile it as a module, choose M here.  If unsure, say N.
+
 config NETFILTER_XT_MATCH_QUOTA
 	tristate '"quota" match support'
 	depends on NETFILTER_ADVANCED

net/netfilter/Makefile +1 -0

index 3259697..9bc95e0 100644
@@ -129,6 +129,7 @@ obj-$(CONFIG_NETFILTER_XT_MATCH_RATEEST) += xt_rateest.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_REALM) += xt_realm.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_RECENT) += xt_recent.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_SCTP) += xt_sctp.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_SKBUFF) += xt_skbuff.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_SOCKET) += xt_socket.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_STATE) += xt_state.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_STATISTIC) += xt_statistic.o

net/netfilter/xt_skbuff.c new file +11 -0

new file mode 100644
index 0000000..5ca30eb
@@ -0,0 +1,141 @@
+/* Xtables module to match packets based on sk_buff fields.
+ * Copyright 2012 Google Inc.
+ * Written by Willem de Bruijn <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/skbuff.h>

net/netfilter/xt_skbuff.c new file +82 -0

+static bool skbuff_mt(const struct sk_buff *skb,
+			struct xt_action_param *par)
+{
+	const struct xt_skbuff_info *info = par->matchinfo;
+	u64 value;
+
+	switch (info->field_id) {
+	case skb_field_csum:
+		if (skb->ip_summed != CHECKSUM_COMPLETE)
+			return false;
+		value = skb->csum;
+		break;
+	case skb_field_hatype:
+		if (!skb->dev)
+			return false;
+		value = skb->dev->type;
+		break;
+	case skb_field_iif:
+		value = skb->skb_iif;
+		break;
+	case skb_field_len:
+		value = skb->len;
+		break;
+	case skb_field_mark:
+		value = skb->mark;
+		break;
+	case skb_field_pkt_type:
+		value = skb->pkt_type;
+		break;
+	case skb_field_priority:
+		value = skb->priority;
+		break;
+	case skb_field_protocol:
+		value = skb->protocol;
+		break;
+	case skb_field_queue_mapping:
+		value = skb->queue_mapping;
+		break;
+	case skb_field_rt_classid:
+#ifdef CONFIG_NET_CLS_ROUTE
+		const struct dst_entry *dst;
+
+		rcu_read_lock();
+		dst = skb_dst(skb);
+		if (dst)
+			value = dst->tclassid;
+		rcu_read_unlock();
+		if (!dst)
+			return false;
+		break;
+#else
+		return false;
+#endif
+	case skb_field_rxhash:
+		value = skb->rxhash;
+		break;
+	case skb_field_secmark:
+#ifdef CONFIG_NETWORK_SECMARK
+		value = skb->secmark;
+		break;
+#else
+		return false;
+#endif
+	case skb_field_sk_uid:
+		if (!skb->sk || !skb->sk->sk_socket || !skb->sk->sk_socket->file)
+			return false;
+		value = skb->sk->sk_socket->file->f_cred->fsuid;
+		break;
+	case skb_field_sk_gid:
+		if (!skb->sk || !skb->sk->sk_socket || !skb->sk->sk_socket->file)
+			return false;
+		value = skb->sk->sk_socket->file->f_cred->fsgid;
+		break;
+	case skb_field_tstamp:
+		value = skb->tstamp.tv64;
+		break;
+	case skb_field_vlan_tci:
+		value = skb->vlan_tci;
+		break;
+	default:
+		return false;
+	}

net/netfilter/xt_skbuff.c new file +26 -0

+
+	return 0;
+}
+
+static struct xt_match skbuff_mt_reg __read_mostly = {
+	.name		= "skbuff",
+	.revision	= 0,
+	.family		= NFPROTO_UNSPEC,
+	.match		= skbuff_mt,
+	.matchsize	= sizeof(struct xt_skbuff_info),
+	.checkentry	= skbuff_mt_check,
+	.me		= THIS_MODULE,
+};
+
+static int __init skbuff_mt_init(void)
+{
+	return xt_register_match(&skbuff_mt_reg);
+}
+
+static void __exit skbuff_mt_exit(void)
+{
+	xt_unregister_match(&skbuff_mt_reg);
+}
+
+module_init(skbuff_mt_init);
+module_exit(skbuff_mt_exit);

Signature · 1 line

1.7.7.3

Quoted 8 lines · Willem de Bruijn

Add an iptables match based on skb fields, such as mark, priority, input interface and rxhash. The match supports range based matching on one field, with optional inversion and masking.

v2: switches from xt_priority to xt_skbuff. Pablo, is this what you had in mind? It doesn't perfectly duplicate the values from nftables xt_meta. Needs more testing to cover the field-specific codepaths.

Yes, I think this is the good direction for it.

Quoted 4 lines

Tested by inserting

iptables -t mangle -A PREROUTING -s $SRC -j MARK --set-mark 10 iptables -A INPUT -m skbuff --min 10 --max 10 -j TRACE

Hm, you have to specify the field type in the iptables rule, right?i

Quoted 3 lines

The userspace tool needs work, too. For one, I just hardcoded the field_id to be skb_field_mark for this test. That's why it's missing from the command line.

Please, once you're done with the user-space part, post it to the ML in order to help testing this.

Quoted 10 lines

Signed-off-by Willem de Bruijn <[email protected]>

---
 include/linux/netfilter/xt_skbuff.h |   33 ++++++++
 net/netfilter/Kconfig               |    9 ++
 net/netfilter/Makefile              |    1 +
 net/netfilter/xt_skbuff.c           |  141 +++++++++++++++++++++++++++++++++++
 4 files changed, 184 insertions(+), 0 deletions(-)
 create mode 100644 include/linux/netfilter/xt_skbuff.h
 create mode 100644 net/netfilter/xt_skbuff.c

include/linux/netfilter/xt_skbuff.h line 6

Quoted 12 lines

include/linux/netfilter/xt_skbuff.h new file +6 -0

new file mode 100644
index 0000000..10eb8d8
@@ -0,0 +1,33 @@
+#ifndef _XT_SKBUFF_H
+#define _XT_SKBUFF_H
+
+#include <linux/types.h>
+
+enum xt_skbuff_field_selector {

these enums in uppercase.

Quoted 6 lines

include/linux/netfilter/xt_skbuff.h new file +6 -0

+     skb_field_csum = 0,
+     skb_field_hatype,
+     skb_field_iif,
+     skb_field_len,
+     skb_field_mark,
+     skb_field_pkt_type,

include/linux/netfilter/xt_skbuff.h line 22

Quoted 10 lines

include/linux/netfilter/xt_skbuff.h new file +10 -0

+     skb_field_priority,
+     skb_field_protocol,
+     skb_field_queue_mapping,
+     skb_field_rt_classid,
+     skb_field_rxhash,
+     skb_field_secmark,
+     skb_field_sk_uid,
+     skb_field_sk_gid,
+     skb_field_tstamp,
+     skb_field_vlan_tci,

Please, check if these all are worth to be exposed to user-space via iptables.

Quoted 60 lines

include/linux/netfilter/xt_skbuff.h new file +11 -0

+};
+
+struct xt_skbuff_info {
+     __u16 field_id;         /* an xt_skbuff_field_selector value */
+     __u8  invert;
+     __u64 min;
+     __u64 max;
+     __u64 mask;
+};
+
+#endif /*_XT_SKBUFF_H */

net/netfilter/Kconfig +9 -0

index fefa514..3a07a86 100644
@@ -1093,6 +1093,15 @@ config NETFILTER_XT_MATCH_PKTTYPE
 
        To compile it as a module, choose M here.  If unsure, say N.
 
+config NETFILTER_XT_MATCH_SKBUFF
+     tristate '"skbuff" match support'
+     depends on NETFILTER_ADVANCED
+     help
+       This option adds a match based on the value of a chosen sk_buff
+       field.
+
+       To compile it as a module, choose M here.  If unsure, say N.
+
 config NETFILTER_XT_MATCH_QUOTA
      tristate '"quota" match support'
      depends on NETFILTER_ADVANCED

net/netfilter/Makefile +1 -0

index 3259697..9bc95e0 100644
@@ -129,6 +129,7 @@ obj-$(CONFIG_NETFILTER_XT_MATCH_RATEEST) += xt_rateest.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_REALM) += xt_realm.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_RECENT) += xt_recent.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_SCTP) += xt_sctp.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_SKBUFF) += xt_skbuff.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_SOCKET) += xt_socket.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_STATE) += xt_state.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_STATISTIC) += xt_statistic.o

net/netfilter/xt_skbuff.c new file +11 -0

new file mode 100644
index 0000000..5ca30eb
@@ -0,0 +1,141 @@
+/* Xtables module to match packets based on sk_buff fields.
+ * Copyright 2012 Google Inc.
+ * Written by Willem de Bruijn <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/skbuff.h>

net/netfilter/xt_skbuff.c line 21

Quoted 10 lines

net/netfilter/xt_skbuff.c new file +10 -0

+#include <net/sock.h>
+
+#include <linux/netfilter/xt_skbuff.h>
+#include <linux/netfilter/x_tables.h>
+
+MODULE_AUTHOR("Willem de Bruijn <[email protected]>");
+MODULE_DESCRIPTION("Xtables: skbuff match");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("ipt_priority");
+MODULE_ALIAS("ip6t_priority");

you'll have to remove this aliasing support. There was not previous priority match in mainstream.

Quoted 82 lines

net/netfilter/xt_skbuff.c new file +82 -0

+static bool skbuff_mt(const struct sk_buff *skb,
+                     struct xt_action_param *par)
+{
+     const struct xt_skbuff_info *info = par->matchinfo;
+     u64 value;
+
+     switch (info->field_id) {
+     case skb_field_csum:
+             if (skb->ip_summed != CHECKSUM_COMPLETE)
+                     return false;
+             value = skb->csum;
+             break;
+     case skb_field_hatype:
+             if (!skb->dev)
+                     return false;
+             value = skb->dev->type;
+             break;
+     case skb_field_iif:
+             value = skb->skb_iif;
+             break;
+     case skb_field_len:
+             value = skb->len;
+             break;
+     case skb_field_mark:
+             value = skb->mark;
+             break;
+     case skb_field_pkt_type:
+             value = skb->pkt_type;
+             break;
+     case skb_field_priority:
+             value = skb->priority;
+             break;
+     case skb_field_protocol:
+             value = skb->protocol;
+             break;
+     case skb_field_queue_mapping:
+             value = skb->queue_mapping;
+             break;
+     case skb_field_rt_classid:
+#ifdef CONFIG_NET_CLS_ROUTE
+             const struct dst_entry *dst;
+
+             rcu_read_lock();
+             dst = skb_dst(skb);
+             if (dst)
+                     value = dst->tclassid;
+             rcu_read_unlock();
+             if (!dst)
+                     return false;
+             break;
+#else
+             return false;
+#endif
+     case skb_field_rxhash:
+             value = skb->rxhash;
+             break;
+     case skb_field_secmark:
+#ifdef CONFIG_NETWORK_SECMARK
+             value = skb->secmark;
+             break;
+#else
+             return false;
+#endif
+     case skb_field_sk_uid:
+             if (!skb->sk || !skb->sk->sk_socket || !skb->sk->sk_socket->file)
+                     return false;
+             value = skb->sk->sk_socket->file->f_cred->fsuid;
+             break;
+     case skb_field_sk_gid:
+             if (!skb->sk || !skb->sk->sk_socket || !skb->sk->sk_socket->file)
+                     return false;
+             value = skb->sk->sk_socket->file->f_cred->fsgid;
+             break;
+     case skb_field_tstamp:
+             value = skb->tstamp.tv64;
+             break;
+     case skb_field_vlan_tci:
+             value = skb->vlan_tci;
+             break;
+     default:
+             return false;
+     }

net/netfilter/xt_skbuff.c line 114

Quoted 10 lines

net/netfilter/xt_skbuff.c new file +10 -0

+
+     value &= info->mask;
+     return (value >= info->min && value <= info->max) ^ info->invert;
+}
+
+static int skbuff_mt_check(const struct xt_mtchk_param *par)
+{
+     const struct xt_skbuff_info *info = par->matchinfo;
+
+     if (info->field_id > skb_field_vlan_tci)

Better define SKB_FIELD_MAX ?

net/netfilter/xt_skbuff.c line 115

Quoted 1 line

net/netfilter/xt_skbuff.c new file +1 -0

+             return -EINVAL;

probably -EOPNOTSUPP is better in case we add some new skbuff field that we support.

Quoted 29 lines

net/netfilter/xt_skbuff.c new file +26 -1

+
+     return 0;
+}
+
+static struct xt_match skbuff_mt_reg __read_mostly = {
+     .name           = "skbuff",
+     .revision       = 0,
+     .family         = NFPROTO_UNSPEC,
+     .match          = skbuff_mt,
+     .matchsize      = sizeof(struct xt_skbuff_info),
+     .checkentry     = skbuff_mt_check,
+     .me             = THIS_MODULE,
+};
+
+static int __init skbuff_mt_init(void)
+{
+     return xt_register_match(&skbuff_mt_reg);
+}
+
+static void __exit skbuff_mt_exit(void)
+{
+     xt_unregister_match(&skbuff_mt_reg);
+}
+
+module_init(skbuff_mt_init);
+module_exit(skbuff_mt_exit);
--

1.7.7.3

A new match that executes sk_run_filter on every packet. BPF filters can access skbuff fields that are out of scope for existing iptables rules, allow more expressive logic, and on platforms with JIT support can even be faster.

I have a corresponding iptables patch that takes tcpdump -ddd output, as used in the examples below. The two parts communicate using a variable length structure. This is similar to ebt_among, but new for iptables.

Verified functionality by inserting an ip source filter on chain INPUT and an ip dest filter on chain OUTPUT and noting that ping failed while a rule was active:

iptables -v -A INPUT -m bpf --bytecode '4,32 0 0 12,21 0 1 $SADDR,6 0 0 96,6 0 0 0,' -j DROP iptables -v -A OUTPUT -m bpf --bytecode '4,32 0 0 16,21 0 1 $DADDR,6 0 0 96,6 0 0 0,' -j DROP

Evaluated throughput by running netperf TCP_STREAM over loopback on x86_64. I expected the BPF filter to outperform hardcoded iptables filters when replacing multiple matches with a single bpf match, but even a single comparison to u32 appears to do better. Relative to the benchmark with no filter applied, rate with 100 BPF filters dropped to 81%. With 100 U32 filters it dropped to 55%. The difference sounds excessive to me, but was consistent on my hardware. Commands used:

for i in seq 100; do iptables -A OUTPUT -m bpf --bytecode '4,48 0 0 9,21 0 1 20,6 0 0 96,6 0 0 0,' -j DROP; done for i in seq 3; do netperf -t TCP_STREAM -I 99 -H localhost; done

iptables -F OUTPUT

for i in seq 100; do iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP; done for i in seq 3; do netperf -t TCP_STREAM -I 99 -H localhost; done

FYI: perf top

[bpf]

    33.94%  [kernel]          [k] copy_user_generic_string
     8.92%  [kernel]          [k] sk_run_filter
     7.77%  [ip_tables]       [k] ipt_do_table

[u32]

    22.63%  [kernel]          [k] copy_user_generic_string
    14.46%  [kernel]          [k] memcpy
     9.19%  [ip_tables]       [k] ipt_do_table
     8.47%  [xt_u32]          [k] u32_mt
     5.32%  [kernel]          [k] skb_copy_bits

The big difference appears to be in memory copying. I have not looked into u32, so cannot explain this right now. More interestingly, at higher rate, sk_run_filter appears to use as many cycles as u32_mt (both traces have roughly the same number of events).

One caveat: to work independent of device link layer, the filter expects DLT_RAW style BPF programs, i.e., those that expect the packet to start at the IP layer.

---
 include/linux/netfilter/xt_bpf.h |   17 +++++++
 net/netfilter/Kconfig            |    9 ++++
 net/netfilter/Makefile           |    1 +
 net/netfilter/x_tables.c         |    5 +-
 net/netfilter/xt_bpf.c           |   88 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 118 insertions(+), 2 deletions(-)
 create mode 100644 include/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/linux/netfilter/xt_bpf.h new file +17 -0

new file mode 100644
index 0000000..23502c0
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H
+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+struct xt_bpf_info {
+	__u16 bpf_program_num_elem;
+
+	/* only used in kernel */
+	struct sk_filter *filter __attribute__((aligned(8)));
+
+	/* variable size, based on program_num_elem */
+	struct sock_filter bpf_program[0];
+};
+
+#endif /*_XT_BPF_H */

net/netfilter/Kconfig +9 -0

index c9739c6..c7cc0b8 100644
@@ -798,6 +798,15 @@ config NETFILTER_XT_MATCH_ADDRTYPE
 	  If you want to compile it as a module, say M here and read
 	  <file:Documentation/kbuild/modules.txt>.  If unsure, say `N'.
 
+config NETFILTER_XT_MATCH_BPF
+	tristate '"bpf" match support'
+	depends on NETFILTER_ADVANCED
+	help
+	  BPF matching applies a linux socket filter to each packet and
+          accepts those for which the filter returns non-zero.
+
+	  To compile it as a module, choose M here.  If unsure, say N.
+
 config NETFILTER_XT_MATCH_CLUSTER
 	tristate '"cluster" match support'
 	depends on NF_CONNTRACK

net/netfilter/Makefile +1 -0

index 8e5602f..9f12eeb 100644
@@ -98,6 +98,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_IDLETIMER) += xt_IDLETIMER.o
 
 # matches
 obj-$(CONFIG_NETFILTER_XT_MATCH_ADDRTYPE) += xt_addrtype.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_BPF) += xt_bpf.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CLUSTER) += xt_cluster.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_COMMENT) += xt_comment.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CONNBYTES) += xt_connbytes.o

net/netfilter/x_tables.c +3 -2

index 8d987c3..26306be 100644
@@ -379,8 +379,9 @@ int xt_check_match(struct xt_mtchk_param *par,
 	if (XT_ALIGN(par->match->matchsize) != size &&
 	    par->match->matchsize != -1) {
 		/*
-		 * ebt_among is exempt from centralized matchsize checking
-		 * because it uses a dynamic-size data set.
+		 * matches of variable size length, such as ebt_among,
+		 * are exempt from centralized matchsize checking. They
+		 * skip the test by setting xt_match.matchsize to -1.
 		 */
 		pr_err("%s_tables: %s.%u match: invalid size "
 		       "%u (kernel) != (user) %u\n",

net/netfilter/xt_bpf.c new file +88 -0

new file mode 100644
index 0000000..07077c5
@@ -0,0 +1,88 @@
+/* Xtables module to match packets using a BPF filter.
+ * Copyright 2012 Google Inc.
+ * Written by Willem de Bruijn <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/ipv6.h>
+#include <linux/filter.h>
+#include <net/ip.h>
+
+#include <linux/netfilter/xt_bpf.h>
+#include <linux/netfilter/x_tables.h>
+
+MODULE_AUTHOR("Willem de Bruijn <[email protected]>");
+MODULE_DESCRIPTION("Xtables: BPF filter match");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("ipt_bpf");
+MODULE_ALIAS("ip6t_bpf");
+
+static int bpf_mt_check(const struct xt_mtchk_param *par)
+{
+	struct xt_bpf_info *info = par->matchinfo;
+	const struct xt_entry_match *match;
+	struct sock_fprog program;
+	int expected_len;
+
+	match = container_of(par->matchinfo, const struct xt_entry_match, data);
+	expected_len = sizeof(struct xt_entry_match) +
+		       sizeof(struct xt_bpf_info) +
+		       (sizeof(struct sock_filter) *
+			info->bpf_program_num_elem);
+
+	if (match->u.match_size != expected_len) {
+		pr_info("bpf: check failed: incorrect length\n");
+		return -EINVAL;
+	}
+
+	program.len = info->bpf_program_num_elem;
+	program.filter = info->bpf_program;
+	if (sk_unattached_filter_create(&info->filter, &program)) {
+		pr_info("bpf: check failed: parse error\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static bool bpf_mt(const struct sk_buff *skb, struct xt_action_param *par)
+{
+	const struct xt_bpf_info *info = par->matchinfo;
+
+	return SK_RUN_FILTER(info->filter, skb);
+}
+
+static void bpf_mt_destroy(const struct xt_mtdtor_param *par)
+{
+	const struct xt_bpf_info *info = par->matchinfo;
+	sk_unattached_filter_destroy(info->filter);
+}
+
+static struct xt_match bpf_mt_reg __read_mostly = {
+	.name		= "bpf",
+	.revision	= 0,
+	.family		= NFPROTO_UNSPEC,
+	.checkentry	= bpf_mt_check,
+	.match		= bpf_mt,
+	.destroy	= bpf_mt_destroy,
+	.matchsize	= -1, /* skip xt_check_match because of dynamic len */
+	.me		= THIS_MODULE,
+};
+
+static int __init bpf_mt_init(void)
+{
+	return xt_register_match(&bpf_mt_reg);
+}
+
+static void __exit bpf_mt_exit(void)
+{
+	xt_unregister_match(&bpf_mt_reg);
+}
+
+module_init(bpf_mt_init);
+module_exit(bpf_mt_exit);

Signature · 1 line

1.7.7.3

Hi Willem,

Quoted 10 lines · Willem de Bruijn

A new match that executes sk_run_filter on every packet. BPF filters can access skbuff fields that are out of scope for existing iptables rules, allow more expressive logic, and on platforms with JIT support can even be faster.

I have a corresponding iptables patch that takes tcpdump -ddd output, as used in the examples below. The two parts communicate using a variable length structure. This is similar to ebt_among, but new for iptables.

Quoted 6 lines

Verified functionality by inserting an ip source filter on chain INPUT and an ip dest filter on chain OUTPUT and noting that ping failed while a rule was active:

iptables -v -A INPUT -m bpf --bytecode '4,32 0 0 12,21 0 1 $SADDR,6 0 0 96,6 0 0 0,' -j DROP iptables -v -A OUTPUT -m bpf --bytecode '4,32 0 0 16,21 0 1 $DADDR,6 0 0 96,6 0 0 0,' -j DROP

I like this BPF idea for iptables.

I made a similar extension time ago, but it was taking a file as parameter. That file contained in BPF code. I made a simple bison parser that takes BPF code and put it into the bpf array of instructions. It would be a bit more intuitive to define a filter and we can distribute it with iptables.

Let me check on my internal trees, I can put that user-space code somewhere in case you're interested.

Quoted 216 lines

Evaluated throughput by running netperf TCP_STREAM over loopback on x86_64. I expected the BPF filter to outperform hardcoded iptables filters when replacing multiple matches with a single bpf match, but even a single comparison to u32 appears to do better. Relative to the benchmark with no filter applied, rate with 100 BPF filters dropped to 81%. With 100 U32 filters it dropped to 55%. The difference sounds excessive to me, but was consistent on my hardware. Commands used:

for i in seq 100; do iptables -A OUTPUT -m bpf --bytecode '4,48 0 0 9,21 0 1 20,6 0 0 96,6 0 0 0,' -j DROP; done for i in seq 3; do netperf -t TCP_STREAM -I 99 -H localhost; done

iptables -F OUTPUT

for i in seq 100; do iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP; done for i in seq 3; do netperf -t TCP_STREAM -I 99 -H localhost; done

FYI: perf top

[bpf]

    33.94%  [kernel]          [k] copy_user_generic_string
     8.92%  [kernel]          [k] sk_run_filter
     7.77%  [ip_tables]       [k] ipt_do_table

[u32]

    22.63%  [kernel]          [k] copy_user_generic_string
    14.46%  [kernel]          [k] memcpy
     9.19%  [ip_tables]       [k] ipt_do_table
     8.47%  [xt_u32]          [k] u32_mt
     5.32%  [kernel]          [k] skb_copy_bits

The big difference appears to be in memory copying. I have not looked into u32, so cannot explain this right now. More interestingly, at higher rate, sk_run_filter appears to use as many cycles as u32_mt (both traces have roughly the same number of events).

One caveat: to work independent of device link layer, the filter expects DLT_RAW style BPF programs, i.e., those that expect the packet to start at the IP layer.

---
 include/linux/netfilter/xt_bpf.h |   17 +++++++
 net/netfilter/Kconfig            |    9 ++++
 net/netfilter/Makefile           |    1 +
 net/netfilter/x_tables.c         |    5 +-
 net/netfilter/xt_bpf.c           |   88 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 118 insertions(+), 2 deletions(-)
 create mode 100644 include/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/linux/netfilter/xt_bpf.h new file +17 -0

new file mode 100644
index 0000000..23502c0
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H
+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+struct xt_bpf_info {
+     __u16 bpf_program_num_elem;
+
+     /* only used in kernel */
+     struct sk_filter *filter __attribute__((aligned(8)));
+
+     /* variable size, based on program_num_elem */
+     struct sock_filter bpf_program[0];
+};
+
+#endif /*_XT_BPF_H */

net/netfilter/Kconfig +9 -0

index c9739c6..c7cc0b8 100644
@@ -798,6 +798,15 @@ config NETFILTER_XT_MATCH_ADDRTYPE
        If you want to compile it as a module, say M here and read
        <file:Documentation/kbuild/modules.txt>.  If unsure, say `N'.
 
+config NETFILTER_XT_MATCH_BPF
+     tristate '"bpf" match support'
+     depends on NETFILTER_ADVANCED
+     help
+       BPF matching applies a linux socket filter to each packet and
+          accepts those for which the filter returns non-zero.
+
+       To compile it as a module, choose M here.  If unsure, say N.
+
 config NETFILTER_XT_MATCH_CLUSTER
      tristate '"cluster" match support'
      depends on NF_CONNTRACK

net/netfilter/Makefile +1 -0

index 8e5602f..9f12eeb 100644
@@ -98,6 +98,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_IDLETIMER) += xt_IDLETIMER.o
 
 # matches
 obj-$(CONFIG_NETFILTER_XT_MATCH_ADDRTYPE) += xt_addrtype.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_BPF) += xt_bpf.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CLUSTER) += xt_cluster.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_COMMENT) += xt_comment.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CONNBYTES) += xt_connbytes.o

net/netfilter/x_tables.c +3 -2

index 8d987c3..26306be 100644
@@ -379,8 +379,9 @@ int xt_check_match(struct xt_mtchk_param *par,
      if (XT_ALIGN(par->match->matchsize) != size &&
          par->match->matchsize != -1) {
              /*
-              * ebt_among is exempt from centralized matchsize checking
-              * because it uses a dynamic-size data set.
+              * matches of variable size length, such as ebt_among,
+              * are exempt from centralized matchsize checking. They
+              * skip the test by setting xt_match.matchsize to -1.
               */
              pr_err("%s_tables: %s.%u match: invalid size "
                     "%u (kernel) != (user) %u\n",

net/netfilter/xt_bpf.c new file +88 -1

new file mode 100644
index 0000000..07077c5
@@ -0,0 +1,88 @@
+/* Xtables module to match packets using a BPF filter.
+ * Copyright 2012 Google Inc.
+ * Written by Willem de Bruijn <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/ipv6.h>
+#include <linux/filter.h>
+#include <net/ip.h>
+
+#include <linux/netfilter/xt_bpf.h>
+#include <linux/netfilter/x_tables.h>
+
+MODULE_AUTHOR("Willem de Bruijn <[email protected]>");
+MODULE_DESCRIPTION("Xtables: BPF filter match");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("ipt_bpf");
+MODULE_ALIAS("ip6t_bpf");
+
+static int bpf_mt_check(const struct xt_mtchk_param *par)
+{
+     struct xt_bpf_info *info = par->matchinfo;
+     const struct xt_entry_match *match;
+     struct sock_fprog program;
+     int expected_len;
+
+     match = container_of(par->matchinfo, const struct xt_entry_match, data);
+     expected_len = sizeof(struct xt_entry_match) +
+                    sizeof(struct xt_bpf_info) +
+                    (sizeof(struct sock_filter) *
+                     info->bpf_program_num_elem);
+
+     if (match->u.match_size != expected_len) {
+             pr_info("bpf: check failed: incorrect length\n");
+             return -EINVAL;
+     }
+
+     program.len = info->bpf_program_num_elem;
+     program.filter = info->bpf_program;
+     if (sk_unattached_filter_create(&info->filter, &program)) {
+             pr_info("bpf: check failed: parse error\n");
+             return -EINVAL;
+     }
+
+     return 0;
+}
+
+static bool bpf_mt(const struct sk_buff *skb, struct xt_action_param *par)
+{
+     const struct xt_bpf_info *info = par->matchinfo;
+
+     return SK_RUN_FILTER(info->filter, skb);
+}
+
+static void bpf_mt_destroy(const struct xt_mtdtor_param *par)
+{
+     const struct xt_bpf_info *info = par->matchinfo;
+     sk_unattached_filter_destroy(info->filter);
+}
+
+static struct xt_match bpf_mt_reg __read_mostly = {
+     .name           = "bpf",
+     .revision       = 0,
+     .family         = NFPROTO_UNSPEC,
+     .checkentry     = bpf_mt_check,
+     .match          = bpf_mt,
+     .destroy        = bpf_mt_destroy,
+     .matchsize      = -1, /* skip xt_check_match because of dynamic len */
+     .me             = THIS_MODULE,
+};
+
+static int __init bpf_mt_init(void)
+{
+     return xt_register_match(&bpf_mt_reg);
+}
+
+static void __exit bpf_mt_exit(void)
+{
+     xt_unregister_match(&bpf_mt_reg);
+}
+
+module_init(bpf_mt_init);
+module_exit(bpf_mt_exit);
--

1.7.7.3

Quoted 20 lines · Pablo Neira Ayuso

Hi Willem,

Quoted 10 lines · Willem de Bruijn

A new match that executes sk_run_filter on every packet. BPF filters can access skbuff fields that are out of scope for existing iptables rules, allow more expressive logic, and on platforms with JIT support can even be faster.

I have a corresponding iptables patch that takes tcpdump -ddd output, as used in the examples below. The two parts communicate using a variable length structure. This is similar to ebt_among, but new for iptables.

Quoted 6 lines

Verified functionality by inserting an ip source filter on chain INPUT and an ip dest filter on chain OUTPUT and noting that ping failed while a rule was active:

iptables -v -A INPUT -m bpf --bytecode '4,32 0 0 12,21 0 1 $SADDR,6 0 0 96,6 0 0 0,' -j DROP iptables -v -A OUTPUT -m bpf --bytecode '4,32 0 0 16,21 0 1 $DADDR,6 0 0 96,6 0 0 0,' -j DROP

Quoted 7 lines

I like this BPF idea for iptables.

I made a similar extension time ago, but it was taking a file as parameter. That file contained in BPF code. I made a simple bison parser that takes BPF code and put it into the bpf array of instructions. It would be a bit more intuitive to define a filter and we can distribute it with iptables.

That's cleaner, indeed. I actually like how tcpdump operates as a code generator if you pass -ddd. Unfortunately, it generates code only for link layer types of its supported devices, such as DLT_EN10MB and DLT_LINUX_SLL. The network layer interface of basic iptables (forgetting device dependent mechanisms as used in xt_mac) is DLT_RAW, but that is rarely supported.

Quoted 2 lines

Let me check on my internal trees, I can put that user-space code somewhere in case you're interested.

Absolutely. I'll be happy to revise to get it in. I'm also considering sending a patch to tcpdump to make it generate code independent of the installed hardware when specifying -y.

Quoted 29 lines · Willem de Bruijn

Quoted 20 lines · Pablo Neira Ayuso

Hi Willem,

Quoted 10 lines · Willem de Bruijn

A new match that executes sk_run_filter on every packet. BPF filters can access skbuff fields that are out of scope for existing iptables rules, allow more expressive logic, and on platforms with JIT support can even be faster.

I have a corresponding iptables patch that takes tcpdump -ddd output, as used in the examples below. The two parts communicate using a variable length structure. This is similar to ebt_among, but new for iptables.

Quoted 6 lines

Verified functionality by inserting an ip source filter on chain INPUT and an ip dest filter on chain OUTPUT and noting that ping failed while a rule was active:

iptables -v -A INPUT -m bpf --bytecode '4,32 0 0 12,21 0 1 $SADDR,6 0 0 96,6 0 0 0,' -j DROP iptables -v -A OUTPUT -m bpf --bytecode '4,32 0 0 16,21 0 1 $DADDR,6 0 0 96,6 0 0 0,' -j DROP

Quoted 7 lines

I like this BPF idea for iptables.

I made a similar extension time ago, but it was taking a file as parameter. That file contained in BPF code. I made a simple bison parser that takes BPF code and put it into the bpf array of instructions. It would be a bit more intuitive to define a filter and we can distribute it with iptables.

Quoted 6 lines

That's cleaner, indeed. I actually like how tcpdump operates as a code generator if you pass -ddd. Unfortunately, it generates code only for link layer types of its supported devices, such as DLT_EN10MB and DLT_LINUX_SLL. The network layer interface of basic iptables (forgetting device dependent mechanisms as used in xt_mac) is DLT_RAW, but that is rarely supported.

Indeed, you'll have to hack on tcpdump to select the offset. In iptables the base is the layer 3 header. With that change you could use tcpdump for generate code automagically from their syntax.

Quoted 6 lines

Quoted 2 lines

Let me check on my internal trees, I can put that user-space code somewhere in case you're interested.

Absolutely. I'll be happy to revise to get it in. I'm also considering sending a patch to tcpdump to make it generate code independent of the installed hardware when specifying -y.

I found a version of the old parser code I made:

http://1984.lsi.us.es/git/nfbpf/

It interprets a filter expressed in a similar way to tcpdump -dd but it's using the BPF constants. It's quite preliminary and simple if you look at the code.

Extending it to interpret some syntax similar to tcpdump -d would even make more readable the BPF filter.

Time ago I also thought about taking the kernel code that checks that the filter is correct. Currently you get -EINVAL if you pass a handcrafted filter which is incorrect, so it's hard task to debug what you made wrong.

It could be added to the iptables tree. Or if generic enough for BPF and the effort is worth, just provide some small library that iptables can link with and a small compiler/checker to help people develop BPF filters.

Back to your xt_bpf thing, we can use the file containing the code instead:

iptables -v -A INPUT -m bpf --bytecode-file filter1.bpf -j DROP iptables -v -A OUTPUT -m bpf --bytecode-file filter2.bpf -j DROP

We can still allow the inlined filter via --bytecode if you want.

Quoted 64 lines · Pablo Neira Ayuso

Quoted 29 lines · Willem de Bruijn

Quoted 20 lines · Pablo Neira Ayuso

Hi Willem,

Quoted 10 lines · Willem de Bruijn

A new match that executes sk_run_filter on every packet. BPF filters can access skbuff fields that are out of scope for existing iptables rules, allow more expressive logic, and on platforms with JIT support can even be faster.

I have a corresponding iptables patch that takes tcpdump -ddd output, as used in the examples below. The two parts communicate using a variable length structure. This is similar to ebt_among, but new for iptables.

Quoted 6 lines

Verified functionality by inserting an ip source filter on chain INPUT and an ip dest filter on chain OUTPUT and noting that ping failed while a rule was active:

iptables -v -A INPUT -m bpf --bytecode '4,32 0 0 12,21 0 1 $SADDR,6 0 0 96,6 0 0 0,' -j DROP iptables -v -A OUTPUT -m bpf --bytecode '4,32 0 0 16,21 0 1 $DADDR,6 0 0 96,6 0 0 0,' -j DROP

Quoted 7 lines

I like this BPF idea for iptables.

I made a similar extension time ago, but it was taking a file as parameter. That file contained in BPF code. I made a simple bison parser that takes BPF code and put it into the bpf array of instructions. It would be a bit more intuitive to define a filter and we can distribute it with iptables.

Quoted 6 lines

That's cleaner, indeed. I actually like how tcpdump operates as a code generator if you pass -ddd. Unfortunately, it generates code only for link layer types of its supported devices, such as DLT_EN10MB and DLT_LINUX_SLL. The network layer interface of basic iptables (forgetting device dependent mechanisms as used in xt_mac) is DLT_RAW, but that is rarely supported.

Indeed, you'll have to hack on tcpdump to select the offset. In iptables the base is the layer 3 header. With that change you could use tcpdump for generate code automagically from their syntax.

Quoted 6 lines

Quoted 2 lines

Let me check on my internal trees, I can put that user-space code somewhere in case you're interested.

Absolutely. I'll be happy to revise to get it in. I'm also considering sending a patch to tcpdump to make it generate code independent of the installed hardware when specifying -y.

I found a version of the old parser code I made:

http://1984.lsi.us.es/git/nfbpf/

It interprets a filter expressed in a similar way to tcpdump -dd but it's using the BPF constants. It's quite preliminary and simple if you look at the code.

Extending it to interpret some syntax similar to tcpdump -d would even make more readable the BPF filter.

Time ago I also thought about taking the kernel code that checks that the filter is correct. Currently you get -EINVAL if you pass a handcrafted filter which is incorrect, so it's hard task to debug what you made wrong.

Quoted 4 lines

It could be added to the iptables tree. Or if generic enough for BPF and the effort is worth, just provide some small library that iptables can link with and a small compiler/checker to help people develop BPF filters.

Or use pcap_compile? I went with the tcpdump output to avoid introducing a direct dependency on pcap to iptables. One possible downside I see to pcap_compile vs. developing from scratch is that it might lag in supporting the LSF ancillary data fields.

Quoted 7 lines

Back to your xt_bpf thing, we can use the file containing the code instead:

iptables -v -A INPUT -m bpf --bytecode-file filter1.bpf -j DROP iptables -v -A OUTPUT -m bpf --bytecode-file filter2.bpf -j DROP

We can still allow the inlined filter via --bytecode if you want.

I'll add that. I'd like to keep --bytecode to able to generate the code inline using backticks.

Quoted 70 lines · Willem de Bruijn

Quoted 64 lines · Pablo Neira Ayuso

Quoted 29 lines · Willem de Bruijn

Quoted 20 lines · Pablo Neira Ayuso

Hi Willem,

Quoted 10 lines · Willem de Bruijn

A new match that executes sk_run_filter on every packet. BPF filters can access skbuff fields that are out of scope for existing iptables rules, allow more expressive logic, and on platforms with JIT support can even be faster.

I have a corresponding iptables patch that takes tcpdump -ddd output, as used in the examples below. The two parts communicate using a variable length structure. This is similar to ebt_among, but new for iptables.

Quoted 6 lines

Verified functionality by inserting an ip source filter on chain INPUT and an ip dest filter on chain OUTPUT and noting that ping failed while a rule was active:

iptables -v -A INPUT -m bpf --bytecode '4,32 0 0 12,21 0 1 $SADDR,6 0 0 96,6 0 0 0,' -j DROP iptables -v -A OUTPUT -m bpf --bytecode '4,32 0 0 16,21 0 1 $DADDR,6 0 0 96,6 0 0 0,' -j DROP

Quoted 7 lines

I like this BPF idea for iptables.

I made a similar extension time ago, but it was taking a file as parameter. That file contained in BPF code. I made a simple bison parser that takes BPF code and put it into the bpf array of instructions. It would be a bit more intuitive to define a filter and we can distribute it with iptables.

Quoted 6 lines

That's cleaner, indeed. I actually like how tcpdump operates as a code generator if you pass -ddd. Unfortunately, it generates code only for link layer types of its supported devices, such as DLT_EN10MB and DLT_LINUX_SLL. The network layer interface of basic iptables (forgetting device dependent mechanisms as used in xt_mac) is DLT_RAW, but that is rarely supported.

Indeed, you'll have to hack on tcpdump to select the offset. In iptables the base is the layer 3 header. With that change you could use tcpdump for generate code automagically from their syntax.

Quoted 6 lines

Quoted 2 lines

Let me check on my internal trees, I can put that user-space code somewhere in case you're interested.

Absolutely. I'll be happy to revise to get it in. I'm also considering sending a patch to tcpdump to make it generate code independent of the installed hardware when specifying -y.

I found a version of the old parser code I made:

http://1984.lsi.us.es/git/nfbpf/

It interprets a filter expressed in a similar way to tcpdump -dd but it's using the BPF constants. It's quite preliminary and simple if you look at the code.

Extending it to interpret some syntax similar to tcpdump -d would even make more readable the BPF filter.

Time ago I also thought about taking the kernel code that checks that the filter is correct. Currently you get -EINVAL if you pass a handcrafted filter which is incorrect, so it's hard task to debug what you made wrong.

Quoted 4 lines

It could be added to the iptables tree. Or if generic enough for BPF and the effort is worth, just provide some small library that iptables can link with and a small compiler/checker to help people develop BPF filters.

Quoted 4 lines

Or use pcap_compile? I went with the tcpdump output to avoid introducing a direct dependency on pcap to iptables. One possible downside I see to pcap_compile vs. developing from scratch is that it might lag in supporting the LSF ancillary data fields.

I suggest to put the code of that preliminary nfbpf utility into iptables to allow to read the BPF filters from a file and put them into the BPF array of instructions. I can help with that.

Quoted 6 lines

Quoted 6 lines

Back to your xt_bpf thing, we can use the file containing the code instead:

iptables -v -A INPUT -m bpf --bytecode-file filter1.bpf -j DROP iptables -v -A OUTPUT -m bpf --bytecode-file filter2.bpf -j DROP

Quoted 4 lines

Quoted 1 line

We can still allow the inlined filter via --bytecode if you want.

I'll add that. I'd like to keep --bytecode to able to generate the code inline using backticks.

As said, I'm fine with that, but I'll be really happy if we can provide some utility to generate that code using backticks for the masses (in case they want to pass it inlined in that format).

Support arbitrary linux socket filter (BPF) programs as iptables match rules. This allows for very expressive filters, and on platforms with BPF JIT appears competitive with traditional hardcoded iptables rules.

At least, on an x86_64 that achieves 40K netperf TCP_STREAM without any iptables rules (40 GBps),

inserting 100x this bpf rule gives 28K

./iptables -A OUTPUT -m bpf --bytecode '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0,' -j

(as generated by tcpdump -i any -ddd ip proto 20 | tr '\n' ',')

inserting 100x this u32 rule gives 21K

    ./iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP

The two are logically equivalent, as far as I can tell. Let me know if my test methodology is flawed in some way. Even in cases where slower, the filter adds functionality currently lacking in iptables, such as access to sk_buff fields like rxhash and queue_mapping.

Signed-off-by Willem de Bruijn <[email protected]>

---
 include/linux/netfilter/xt_bpf.h |   17 +++++++
 net/netfilter/Kconfig            |    9 ++++
 net/netfilter/Makefile           |    1 +
 net/netfilter/x_tables.c         |    5 +-
 net/netfilter/xt_bpf.c           |   86 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 116 insertions(+), 2 deletions(-)
 create mode 100644 include/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/linux/netfilter/xt_bpf.h new file +1 -0

new file mode 100644
index 0000000..23502c0
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H

Hi Willem,

Quoted 41 lines · Willem de Bruijn

Support arbitrary linux socket filter (BPF) programs as iptables match rules. This allows for very expressive filters, and on platforms with BPF JIT appears competitive with traditional hardcoded iptables rules.

At least, on an x86_64 that achieves 40K netperf TCP_STREAM without any iptables rules (40 GBps),

inserting 100x this bpf rule gives 28K

./iptables -A OUTPUT -m bpf --bytecode '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0,' -j

(as generated by tcpdump -i any -ddd ip proto 20 | tr '\n' ',')

inserting 100x this u32 rule gives 21K

    ./iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP

The two are logically equivalent, as far as I can tell. Let me know if my test methodology is flawed in some way. Even in cases where slower, the filter adds functionality currently lacking in iptables, such as access to sk_buff fields like rxhash and queue_mapping.

Signed-off-by Willem de Bruijn <[email protected]>

---
 include/linux/netfilter/xt_bpf.h |   17 +++++++
 net/netfilter/Kconfig            |    9 ++++
 net/netfilter/Makefile           |    1 +
 net/netfilter/x_tables.c         |    5 +-
 net/netfilter/xt_bpf.c           |   86 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 116 insertions(+), 2 deletions(-)
 create mode 100644 include/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/linux/netfilter/xt_bpf.h new file +1 -0

new file mode 100644
index 0000000..23502c0
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H

include/linux/netfilter/xt_bpf.h line 11

Quoted 10 lines

include/linux/netfilter/xt_bpf.h new file +10 -0

+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+struct xt_bpf_info {
+     __u16 bpf_program_num_elem;
+
+     /* only used in kernel */
+     struct sk_filter *filter __attribute__((aligned(8)));

I see. You set match->userspacesize to zero in libxt_bpf to skip the comparison of that internal struct sk_filter *filter.

include/linux/netfilter/xt_bpf.h line 14

Quoted 3 lines

include/linux/netfilter/xt_bpf.h new file +3 -0

+
+     /* variable size, based on program_num_elem */
+     struct sock_filter bpf_program[0];

While testing this I noticed:

iptables -I OUTPUT -m bpf --bytecode   \
        '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0' -j ACCEPT

Note that this works but it should not.

iptables -D OUTPUT -m bpf --bytecode   \
        '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,1 0 0 0' -j ACCEPT
                                                               ^

Mind that 1, it's a different filter, but it deletes the previous filter without problems here.

A quick look at make_delete_mask() in iptables tells me that the changes you made to userspace to allow variable size matches are not enough to generate a sane mask (which is fundamental while looking for a matching rule during the deletion).

Quoted 72 lines · Pablo Neira Ayuso

Hi Willem,

Quoted 41 lines · Willem de Bruijn

Support arbitrary linux socket filter (BPF) programs as iptables match rules. This allows for very expressive filters, and on platforms with BPF JIT appears competitive with traditional hardcoded iptables rules.

At least, on an x86_64 that achieves 40K netperf TCP_STREAM without any iptables rules (40 GBps),

inserting 100x this bpf rule gives 28K

./iptables -A OUTPUT -m bpf --bytecode '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0,' -j

(as generated by tcpdump -i any -ddd ip proto 20 | tr '\n' ',')

inserting 100x this u32 rule gives 21K

    ./iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP

The two are logically equivalent, as far as I can tell. Let me know if my test methodology is flawed in some way. Even in cases where slower, the filter adds functionality currently lacking in iptables, such as access to sk_buff fields like rxhash and queue_mapping.

Signed-off-by Willem de Bruijn <[email protected]>

---
 include/linux/netfilter/xt_bpf.h |   17 +++++++
 net/netfilter/Kconfig            |    9 ++++
 net/netfilter/Makefile           |    1 +
 net/netfilter/x_tables.c         |    5 +-
 net/netfilter/xt_bpf.c           |   86 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 116 insertions(+), 2 deletions(-)
 create mode 100644 include/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/linux/netfilter/xt_bpf.h new file +1 -0

new file mode 100644
index 0000000..23502c0
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H

include/linux/netfilter/xt_bpf.h line 11

Quoted 10 lines

include/linux/netfilter/xt_bpf.h new file +10 -0

+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+struct xt_bpf_info {
+     __u16 bpf_program_num_elem;
+
+     /* only used in kernel */
+     struct sk_filter *filter __attribute__((aligned(8)));

I see. You set match->userspacesize to zero in libxt_bpf to skip the comparison of that internal struct sk_filter *filter.

include/linux/netfilter/xt_bpf.h line 14

Quoted 3 lines

include/linux/netfilter/xt_bpf.h new file +3 -0

+
+     /* variable size, based on program_num_elem */
+     struct sock_filter bpf_program[0];

While testing this I noticed:

iptables -I OUTPUT -m bpf --bytecode   \
        '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0' -j ACCEPT

Note that this works but it should not.

iptables -D OUTPUT -m bpf --bytecode   \
        '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,1 0 0 0' -j ACCEPT
                                                               ^

Quoted 7 lines

Mind that 1, it's a different filter, but it deletes the previous filter without problems here.

A quick look at make_delete_mask() in iptables tells me that the changes you made to userspace to allow variable size matches are not enough to generate a sane mask (which is fundamental while looking for a matching rule during the deletion).

Thanks for finding this, Pablo. I completely forgot to check that.

I've never looked at that deletion code before. Will read it and hopefully propose a simple fix in a few days. An earlier version of the patch used a statically sized struct, by the way, like xt_string does (XT_STRING_MAX_PATTERN_SIZE). If it is easier to incorporate, we can always revert to that.

Hi Willem,

Quoted 83 lines · Willem de Bruijn

Quoted 72 lines · Pablo Neira Ayuso

Hi Willem,

Quoted 41 lines · Willem de Bruijn

Support arbitrary linux socket filter (BPF) programs as iptables match rules. This allows for very expressive filters, and on platforms with BPF JIT appears competitive with traditional hardcoded iptables rules.

At least, on an x86_64 that achieves 40K netperf TCP_STREAM without any iptables rules (40 GBps),

inserting 100x this bpf rule gives 28K

./iptables -A OUTPUT -m bpf --bytecode '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0,' -j

(as generated by tcpdump -i any -ddd ip proto 20 | tr '\n' ',')

inserting 100x this u32 rule gives 21K

    ./iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP

The two are logically equivalent, as far as I can tell. Let me know if my test methodology is flawed in some way. Even in cases where slower, the filter adds functionality currently lacking in iptables, such as access to sk_buff fields like rxhash and queue_mapping.

Signed-off-by Willem de Bruijn <[email protected]>

---
 include/linux/netfilter/xt_bpf.h |   17 +++++++
 net/netfilter/Kconfig            |    9 ++++
 net/netfilter/Makefile           |    1 +
 net/netfilter/x_tables.c         |    5 +-
 net/netfilter/xt_bpf.c           |   86 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 116 insertions(+), 2 deletions(-)
 create mode 100644 include/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/linux/netfilter/xt_bpf.h new file +1 -0

new file mode 100644
index 0000000..23502c0
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H

include/linux/netfilter/xt_bpf.h line 11

Quoted 10 lines

include/linux/netfilter/xt_bpf.h new file +10 -0

+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+struct xt_bpf_info {
+     __u16 bpf_program_num_elem;
+
+     /* only used in kernel */
+     struct sk_filter *filter __attribute__((aligned(8)));

I see. You set match->userspacesize to zero in libxt_bpf to skip the comparison of that internal struct sk_filter *filter.

include/linux/netfilter/xt_bpf.h line 14

Quoted 3 lines

include/linux/netfilter/xt_bpf.h new file +3 -0

+
+     /* variable size, based on program_num_elem */
+     struct sock_filter bpf_program[0];

While testing this I noticed:

iptables -I OUTPUT -m bpf --bytecode   \
        '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0' -j ACCEPT

Note that this works but it should not.

iptables -D OUTPUT -m bpf --bytecode   \
        '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,1 0 0 0' -j ACCEPT
                                                               ^

Quoted 7 lines

Mind that 1, it's a different filter, but it deletes the previous filter without problems here.

A quick look at make_delete_mask() in iptables tells me that the changes you made to userspace to allow variable size matches are not enough to generate a sane mask (which is fundamental while looking for a matching rule during the deletion).

Thanks for finding this, Pablo. I completely forgot to check that.

Quoted 5 lines

I've never looked at that deletion code before. Will read it and hopefully propose a simple fix in a few days. An earlier version of the patch used a statically sized struct, by the way, like xt_string does (XT_STRING_MAX_PATTERN_SIZE). If it is easier to incorporate, we can always revert to that.

I prefer if this sticks to static size by now.

The problem is that
BPF_MAXINSNS is probably too much to allocate per rule. So you'll have to limit this to some reasonable amount of lines in the filter. Please, also check that iptables-save and iptables-restore work fine, there is also some problem with the existing code.

Changes:
- v3: reverted no longer needed changes to x_tables.c
- v2: use a fixed size match structure to communicate between kernel and userspace.

Support arbitrary linux socket filter (BPF) programs as iptables match rules. This allows for very expressive filters, and on platforms with BPF JIT appears competitive with traditional hardcoded iptables rules.

At least, on an x86_64 that achieves 40K netperf TCP_STREAM without any iptables rules (40 GBps),

inserting 100x this bpf rule gives 28K

./iptables -A OUTPUT -m bpf --bytecode '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0,' -j

(as generated by tcpdump -i any -ddd ip proto 20 | tr '\n' ',')

inserting 100x this u32 rule gives 21K

    ./iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP

The two are logically equivalent, as far as I can tell. Let me know if my test methodology is flawed in some way. Even in cases where slower, the filter adds functionality currently lacking in iptables, such as access to sk_buff fields like rxhash and queue_mapping.

---
 include/uapi/linux/netfilter/xt_bpf.h |   17 ++++++++
 net/netfilter/Kconfig                 |    9 ++++
 net/netfilter/Makefile                |    1 +
 net/netfilter/xt_bpf.c                |   73 +++++++++++++++++++++++++++++++++
 4 files changed, 100 insertions(+), 0 deletions(-)
 create mode 100644 include/uapi/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/uapi/linux/netfilter/xt_bpf.h new file +17 -0

new file mode 100644
index 0000000..5dda450
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H
+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+#define XT_BPF_MAX_NUM_INSTR	64
+
+struct xt_bpf_info {
+	__u16 bpf_program_num_elem;
+	struct sock_filter bpf_program[XT_BPF_MAX_NUM_INSTR];
+
+	/* only used in the kernel */
+	struct sk_filter *filter __attribute__((aligned(8)));
+};
+
+#endif /*_XT_BPF_H */

net/netfilter/Kconfig +9 -0

index fefa514..d45720f 100644
@@ -798,6 +798,15 @@ config NETFILTER_XT_MATCH_ADDRTYPE
 	  If you want to compile it as a module, say M here and read
 	  <file:Documentation/kbuild/modules.txt>.  If unsure, say `N'.
 
+config NETFILTER_XT_MATCH_BPF
+	tristate '"bpf" match support'
+	depends on NETFILTER_ADVANCED
+	help
+	  BPF matching applies a linux socket filter to each packet and
+          accepts those for which the filter returns non-zero.
+
+	  To compile it as a module, choose M here.  If unsure, say N.
+
 config NETFILTER_XT_MATCH_CLUSTER
 	tristate '"cluster" match support'
 	depends on NF_CONNTRACK

net/netfilter/Makefile +1 -0

index 3259697..6d6194525 100644
@@ -98,6 +98,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_IDLETIMER) += xt_IDLETIMER.o
 
 # matches
 obj-$(CONFIG_NETFILTER_XT_MATCH_ADDRTYPE) += xt_addrtype.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_BPF) += xt_bpf.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CLUSTER) += xt_cluster.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_COMMENT) += xt_comment.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CONNBYTES) += xt_connbytes.o

net/netfilter/xt_bpf.c new file +11 -0

new file mode 100644
index 0000000..1bdfab8
@@ -0,0 +1,73 @@
+/* Xtables module to match packets using a BPF filter.
+ * Copyright 2013 Google Inc.
+ * Written by Willem de Bruijn <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/skbuff.h>

net/netfilter/xt_bpf.c new file +44 -0

+	if (sk_unattached_filter_create(&info->filter, &program)) {
+		pr_info("bpf: check failed: parse error\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static bool bpf_mt(const struct sk_buff *skb, struct xt_action_param *par)
+{
+	const struct xt_bpf_info *info = par->matchinfo;
+
+	return SK_RUN_FILTER(info->filter, skb);
+}
+
+static void bpf_mt_destroy(const struct xt_mtdtor_param *par)
+{
+	const struct xt_bpf_info *info = par->matchinfo;
+	sk_unattached_filter_destroy(info->filter);
+}
+
+static struct xt_match bpf_mt_reg __read_mostly = {
+	.name		= "bpf",
+	.revision	= 0,
+	.family		= NFPROTO_UNSPEC,
+	.checkentry	= bpf_mt_check,
+	.match		= bpf_mt,
+	.destroy	= bpf_mt_destroy,
+	.matchsize	= sizeof(struct xt_bpf_info),
+	.me		= THIS_MODULE,
+};
+
+static int __init bpf_mt_init(void)
+{
+	return xt_register_match(&bpf_mt_reg);
+}
+
+static void __exit bpf_mt_exit(void)
+{
+	xt_unregister_match(&bpf_mt_reg);
+}
+
+module_init(bpf_mt_init);
+module_exit(bpf_mt_exit);

Signature · 1 line

1.7.7.3

Hi Willem,

Quoted 15 lines · Willem de Bruijn

Changes:
- v3: reverted no longer needed changes to x_tables.c
- v2: use a fixed size match structure to communicate between kernel and userspace.

Support arbitrary linux socket filter (BPF) programs as iptables match rules. This allows for very expressive filters, and on platforms with BPF JIT appears competitive with traditional hardcoded iptables rules.

At least, on an x86_64 that achieves 40K netperf TCP_STREAM without any iptables rules (40 GBps),

inserting 100x this bpf rule gives 28K

Quoted 3 lines

./iptables -A OUTPUT -m bpf --bytecode '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0,' -j

(as generated by tcpdump -i any -ddd ip proto 20 | tr '\n' ',')

That code generated by tcpdump will not work.

tcpdump generates BPF code assuming that offset 0 is the link layer header, while iptables considers that offset 0 is the network layer.

More comments below:

Quoted 89 lines

inserting 100x this u32 rule gives 21K

    ./iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP

The two are logically equivalent, as far as I can tell. Let me know if my test methodology is flawed in some way. Even in cases where slower, the filter adds functionality currently lacking in iptables, such as access to sk_buff fields like rxhash and queue_mapping.

---
 include/uapi/linux/netfilter/xt_bpf.h |   17 ++++++++
 net/netfilter/Kconfig                 |    9 ++++
 net/netfilter/Makefile                |    1 +
 net/netfilter/xt_bpf.c                |   73 +++++++++++++++++++++++++++++++++
 4 files changed, 100 insertions(+), 0 deletions(-)
 create mode 100644 include/uapi/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/uapi/linux/netfilter/xt_bpf.h new file +17 -0

new file mode 100644
index 0000000..5dda450
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H
+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+#define XT_BPF_MAX_NUM_INSTR 64
+
+struct xt_bpf_info {
+     __u16 bpf_program_num_elem;
+     struct sock_filter bpf_program[XT_BPF_MAX_NUM_INSTR];
+
+     /* only used in the kernel */
+     struct sk_filter *filter __attribute__((aligned(8)));
+};
+
+#endif /*_XT_BPF_H */

net/netfilter/Kconfig +9 -0

index fefa514..d45720f 100644
@@ -798,6 +798,15 @@ config NETFILTER_XT_MATCH_ADDRTYPE
        If you want to compile it as a module, say M here and read
        <file:Documentation/kbuild/modules.txt>.  If unsure, say `N'.
 
+config NETFILTER_XT_MATCH_BPF
+     tristate '"bpf" match support'
+     depends on NETFILTER_ADVANCED
+     help
+       BPF matching applies a linux socket filter to each packet and
+          accepts those for which the filter returns non-zero.
+
+       To compile it as a module, choose M here.  If unsure, say N.
+
 config NETFILTER_XT_MATCH_CLUSTER
      tristate '"cluster" match support'
      depends on NF_CONNTRACK

net/netfilter/Makefile +1 -0

index 3259697..6d6194525 100644
@@ -98,6 +98,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_IDLETIMER) += xt_IDLETIMER.o
 
 # matches
 obj-$(CONFIG_NETFILTER_XT_MATCH_ADDRTYPE) += xt_addrtype.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_BPF) += xt_bpf.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CLUSTER) += xt_cluster.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_COMMENT) += xt_comment.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CONNBYTES) += xt_connbytes.o

net/netfilter/xt_bpf.c new file +11 -0

new file mode 100644
index 0000000..1bdfab8
@@ -0,0 +1,73 @@
+/* Xtables module to match packets using a BPF filter.
+ * Copyright 2013 Google Inc.
+ * Written by Willem de Bruijn <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/skbuff.h>

net/netfilter/xt_bpf.c line 21

Quoted 10 lines

net/netfilter/xt_bpf.c new file +10 -0

+#include <linux/ipv6.h>
+#include <linux/filter.h>
+#include <net/ip.h>
+
+#include <linux/netfilter/xt_bpf.h>
+#include <linux/netfilter/x_tables.h>
+
+MODULE_AUTHOR("Willem de Bruijn <[email protected]>");
+MODULE_DESCRIPTION("Xtables: BPF filter match");
+MODULE_LICENSE("GPL");

Please, add

MODULE_ALIAS("ipt_bpf");
MODULE_ALIAS("ip6t_bpf");

otherwise module auto-loading will not work.

net/netfilter/xt_bpf.c line 29

Quoted 8 lines

net/netfilter/xt_bpf.c new file +8 -0

+
+static int bpf_mt_check(const struct xt_mtchk_param *par)
+{
+     struct xt_bpf_info *info = par->matchinfo;
+     struct sock_fprog program;
+
+     program.len = info->bpf_program_num_elem;
+     program.filter = info->bpf_program;

sparse reports a warning here above. I've been trying to find a quick solution, please, have a look at it.

Quoted 47 lines

net/netfilter/xt_bpf.c new file +44 -1

+     if (sk_unattached_filter_create(&info->filter, &program)) {
+             pr_info("bpf: check failed: parse error\n");
+             return -EINVAL;
+     }
+
+     return 0;
+}
+
+static bool bpf_mt(const struct sk_buff *skb, struct xt_action_param *par)
+{
+     const struct xt_bpf_info *info = par->matchinfo;
+
+     return SK_RUN_FILTER(info->filter, skb);
+}
+
+static void bpf_mt_destroy(const struct xt_mtdtor_param *par)
+{
+     const struct xt_bpf_info *info = par->matchinfo;
+     sk_unattached_filter_destroy(info->filter);
+}
+
+static struct xt_match bpf_mt_reg __read_mostly = {
+     .name           = "bpf",
+     .revision       = 0,
+     .family         = NFPROTO_UNSPEC,
+     .checkentry     = bpf_mt_check,
+     .match          = bpf_mt,
+     .destroy        = bpf_mt_destroy,
+     .matchsize      = sizeof(struct xt_bpf_info),
+     .me             = THIS_MODULE,
+};
+
+static int __init bpf_mt_init(void)
+{
+     return xt_register_match(&bpf_mt_reg);
+}
+
+static void __exit bpf_mt_exit(void)
+{
+     xt_unregister_match(&bpf_mt_reg);
+}
+
+module_init(bpf_mt_init);
+module_exit(bpf_mt_exit);
--

1.7.7.3

Changes:
- v4: fixed sparse warning and module autoloading
- v3: reverted no longer needed changes to x_tables.c
- v2: use a fixed size match structure to communicate between kernel and userspace.

Support arbitrary linux socket filter (BPF) programs as iptables match rules. This allows for very expressive filters, and on platforms with BPF JIT appears competitive with traditional hardcoded iptables rules.

At least, on an x86_64 that achieves 40K netperf TCP_STREAM without any iptables rules (40 GBps),

inserting 100x this bpf rule gives 28K

    ./iptables -A OUTPUT -m bpf --bytecode '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0,' -j

inserting 100x this u32 rule gives 21K

    ./iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP

The two are logically equivalent, as far as I can tell. Let me know if my test methodology is flawed in some way. Even in cases where slower, the filter adds functionality currently lacking in iptables, such as access to sk_buff fields like rxhash and queue_mapping.

---
 include/uapi/linux/netfilter/xt_bpf.h |   17 +++++++
 net/netfilter/Kconfig                 |    9 ++++
 net/netfilter/Makefile                |    1 +
 net/netfilter/xt_bpf.c                |   75 +++++++++++++++++++++++++++++++++
 4 files changed, 102 insertions(+), 0 deletions(-)
 create mode 100644 include/uapi/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/uapi/linux/netfilter/xt_bpf.h new file +17 -0

new file mode 100644
index 0000000..5dda450
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H
+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+#define XT_BPF_MAX_NUM_INSTR	64
+
+struct xt_bpf_info {
+	__u16 bpf_program_num_elem;
+	struct sock_filter bpf_program[XT_BPF_MAX_NUM_INSTR];
+
+	/* only used in the kernel */
+	struct sk_filter *filter __attribute__((aligned(8)));
+};
+
+#endif /*_XT_BPF_H */

net/netfilter/Kconfig +0 -0

index bb48607..4017d85 100644

net/netfilter/Kconfig +3 -0

+
+	  To compile it as a module, choose M here.  If unsure, say N.
+
 config NETFILTER_XT_MATCH_CLUSTER
 	tristate '"cluster" match support'
 	depends on NF_CONNTRACK

net/netfilter/Makefile +1 -0

index b3bbda6..a1abf87 100644
@@ -99,6 +99,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_IDLETIMER) += xt_IDLETIMER.o
 
 # matches
 obj-$(CONFIG_NETFILTER_XT_MATCH_ADDRTYPE) += xt_addrtype.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_BPF) += xt_bpf.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CLUSTER) += xt_cluster.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_COMMENT) += xt_comment.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CONNBYTES) += xt_connbytes.o

net/netfilter/xt_bpf.c new file +2 -0

new file mode 100644
index 0000000..62d93f8
@@ -0,0 +1,75 @@
+/* Xtables module to match packets using a BPF filter.
+ * Copyright 2013 Google Inc.

net/netfilter/xt_bpf.c new file +61 -0

+
+#include <linux/netfilter/xt_bpf.h>
+#include <linux/netfilter/x_tables.h>
+
+MODULE_AUTHOR("Willem de Bruijn <[email protected]>");
+MODULE_DESCRIPTION("Xtables: BPF filter match");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("ipt_bpf");
+MODULE_ALIAS("ip6t_bpf");
+
+static int bpf_mt_check(const struct xt_mtchk_param *par)
+{
+	struct xt_bpf_info *info = par->matchinfo;
+	struct sock_fprog program;
+
+	program.len = info->bpf_program_num_elem;
+	program.filter = (struct sock_filter __user *) info->bpf_program;
+	if (sk_unattached_filter_create(&info->filter, &program)) {
+		pr_info("bpf: check failed: parse error\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static bool bpf_mt(const struct sk_buff *skb, struct xt_action_param *par)
+{
+	const struct xt_bpf_info *info = par->matchinfo;
+
+	return SK_RUN_FILTER(info->filter, skb);
+}
+
+static void bpf_mt_destroy(const struct xt_mtdtor_param *par)
+{
+	const struct xt_bpf_info *info = par->matchinfo;
+	sk_unattached_filter_destroy(info->filter);
+}
+
+static struct xt_match bpf_mt_reg __read_mostly = {
+	.name		= "bpf",
+	.revision	= 0,
+	.family		= NFPROTO_UNSPEC,
+	.checkentry	= bpf_mt_check,
+	.match		= bpf_mt,
+	.destroy	= bpf_mt_destroy,
+	.matchsize	= sizeof(struct xt_bpf_info),
+	.me		= THIS_MODULE,
+};
+
+static int __init bpf_mt_init(void)
+{
+	return xt_register_match(&bpf_mt_reg);
+}
+
+static void __exit bpf_mt_exit(void)
+{
+	xt_unregister_match(&bpf_mt_reg);
+}
+
+module_init(bpf_mt_init);
+module_exit(bpf_mt_exit);

Signature · 6 lines

1.7.7.3

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Hi Willem,

I have applied this patch to my nf-next tree with minor changes.

From e7db3cbcd6508235d63ba4a31bbd1ce4fdece6e1 Mon Sep 17 00:00:00 2001

Quoted 3 lines · Pablo Neira Ayuso

Hi Willem,

I have applied this patch to my nf-next tree with minor changes.

Quoted 4 lines · Pablo Neira Ayuso

Quoted 3 lines · Pablo Neira Ayuso

Hi Willem,

I have applied this patch to my nf-next tree with minor changes.

[..]

include/uapi/linux/netfilter/Kbuild header-y += xt_TCPOPTSTRIP.h · line 38

Quoted 9 lines

include/uapi/linux/netfilter/Kbuild +1 -0

index 08f555f..8b4bd36 100644
@@ -35,6 +35,7 @@ header-y += xt_TCPOPTSTRIP.h
 header-y += xt_TEE.h
 header-y += xt_TPROXY.h
 header-y += xt_addrtype.h
+header-y += xt_bpf.h

Good catch. I forgot about that too, so xt_connlabel.h is also missing.

From 7a95ad28c40fec445b0e1463fd8ed638701ad94d Mon Sep 17 00:00:00 2001

Hi Willem,

I have applied this patch to my nf-next tree with minor changes.

You'll get also another review asap for the iptables user-space part, we have more time to get that into shape anyway.

Quoted 11 lines · Willem de Bruijn

Changes:
- v4: fixed sparse warning and module autoloading
- v3: reverted no longer needed changes to x_tables.c
- v2: use a fixed size match structure to communicate between kernel and userspace.

Support arbitrary linux socket filter (BPF) programs as iptables match rules. This allows for very expressive filters, and on platforms with BPF JIT appears competitive with traditional hardcoded iptables rules.

Quoted 6 lines

At least, on an x86_64 that achieves 40K netperf TCP_STREAM without any iptables rules (40 GBps),

inserting 100x this bpf rule gives 28K

    ./iptables -A OUTPUT -m bpf --bytecode '6,40 0 0 14, 21 0 3 2048,48 0 0 25,21 0 1 20,6 0 0 96,6 0 0 0,' -j

I have removed part of the changelog, the performance part. This filter above to be possible confusion. That filter is using the link layer as base.

I think the expressivity of BPF is enough to justify its inclusion into mainline.

I also added a line notice on the artificial limitation to 64 instructions per filter.

More minor glitches below.

Quoted 45 lines

inserting 100x this u32 rule gives 21K

    ./iptables -A OUTPUT -m u32 --u32 '6&0xFF=0x20' -j DROP

The two are logically equivalent, as far as I can tell. Let me know if my test methodology is flawed in some way. Even in cases where slower, the filter adds functionality currently lacking in iptables, such as access to sk_buff fields like rxhash and queue_mapping.

---
 include/uapi/linux/netfilter/xt_bpf.h |   17 +++++++
 net/netfilter/Kconfig                 |    9 ++++
 net/netfilter/Makefile                |    1 +
 net/netfilter/xt_bpf.c                |   75 +++++++++++++++++++++++++++++++++
 4 files changed, 102 insertions(+), 0 deletions(-)
 create mode 100644 include/uapi/linux/netfilter/xt_bpf.h
 create mode 100644 net/netfilter/xt_bpf.c

include/uapi/linux/netfilter/xt_bpf.h new file +17 -0

new file mode 100644
index 0000000..5dda450
@@ -0,0 +1,17 @@
+#ifndef _XT_BPF_H
+#define _XT_BPF_H
+
+#include <linux/filter.h>
+#include <linux/types.h>
+
+#define XT_BPF_MAX_NUM_INSTR 64
+
+struct xt_bpf_info {
+     __u16 bpf_program_num_elem;
+     struct sock_filter bpf_program[XT_BPF_MAX_NUM_INSTR];
+
+     /* only used in the kernel */
+     struct sk_filter *filter __attribute__((aligned(8)));
+};
+
+#endif /*_XT_BPF_H */

net/netfilter/Kconfig +0 -0

index bb48607..4017d85 100644

net/netfilter/Kconfig config NETFILTER_XT_MATCH_ADDRTYPE · line 819

Quoted 9 lines

net/netfilter/Kconfig +6 -0

@@ -811,6 +811,15 @@ config NETFILTER_XT_MATCH_ADDRTYPE
        If you want to compile it as a module, say M here and read
        <file:Documentation/kbuild/modules.txt>.  If unsure, say `N'.
 
+config NETFILTER_XT_MATCH_BPF
+     tristate '"bpf" match support'
+     depends on NETFILTER_ADVANCED
+     help
+       BPF matching applies a linux socket filter to each packet and
+          accepts those for which the filter returns non-zero.

          ^^^

Fixed this minor glitch.

Quoted 26 lines

net/netfilter/Kconfig +3 -0

+
+       To compile it as a module, choose M here.  If unsure, say N.
+
 config NETFILTER_XT_MATCH_CLUSTER
      tristate '"cluster" match support'
      depends on NF_CONNTRACK

net/netfilter/Makefile +1 -0

index b3bbda6..a1abf87 100644
@@ -99,6 +99,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_IDLETIMER) += xt_IDLETIMER.o
 
 # matches
 obj-$(CONFIG_NETFILTER_XT_MATCH_ADDRTYPE) += xt_addrtype.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_BPF) += xt_bpf.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CLUSTER) += xt_cluster.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_COMMENT) += xt_comment.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CONNBYTES) += xt_connbytes.o

net/netfilter/xt_bpf.c new file +2 -0

new file mode 100644
index 0000000..62d93f8
@@ -0,0 +1,75 @@
+/* Xtables module to match packets using a BPF filter.
+ * Copyright 2013 Google Inc.

net/netfilter/xt_bpf.c line 12

Quoted 10 lines

net/netfilter/xt_bpf.c new file +10 -0

+ * Written by Willem de Bruijn <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/ipv6.h>

Removed this header above, we don't need it.

net/netfilter/xt_bpf.c line 14

Quoted 2 lines

net/netfilter/xt_bpf.c new file +2 -0

+#include <linux/filter.h>
+#include <net/ip.h>

Same thing.

Quoted 68 lines

net/netfilter/xt_bpf.c new file +61 -1

+
+#include <linux/netfilter/xt_bpf.h>
+#include <linux/netfilter/x_tables.h>
+
+MODULE_AUTHOR("Willem de Bruijn <[email protected]>");
+MODULE_DESCRIPTION("Xtables: BPF filter match");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("ipt_bpf");
+MODULE_ALIAS("ip6t_bpf");
+
+static int bpf_mt_check(const struct xt_mtchk_param *par)
+{
+     struct xt_bpf_info *info = par->matchinfo;
+     struct sock_fprog program;
+
+     program.len = info->bpf_program_num_elem;
+     program.filter = (struct sock_filter __user *) info->bpf_program;
+     if (sk_unattached_filter_create(&info->filter, &program)) {
+             pr_info("bpf: check failed: parse error\n");
+             return -EINVAL;
+     }
+
+     return 0;
+}
+
+static bool bpf_mt(const struct sk_buff *skb, struct xt_action_param *par)
+{
+     const struct xt_bpf_info *info = par->matchinfo;
+
+     return SK_RUN_FILTER(info->filter, skb);
+}
+
+static void bpf_mt_destroy(const struct xt_mtdtor_param *par)
+{
+     const struct xt_bpf_info *info = par->matchinfo;
+     sk_unattached_filter_destroy(info->filter);
+}
+
+static struct xt_match bpf_mt_reg __read_mostly = {
+     .name           = "bpf",
+     .revision       = 0,
+     .family         = NFPROTO_UNSPEC,
+     .checkentry     = bpf_mt_check,
+     .match          = bpf_mt,
+     .destroy        = bpf_mt_destroy,
+     .matchsize      = sizeof(struct xt_bpf_info),
+     .me             = THIS_MODULE,
+};
+
+static int __init bpf_mt_init(void)
+{
+     return xt_register_match(&bpf_mt_reg);
+}
+
+static void __exit bpf_mt_exit(void)
+{
+     xt_unregister_match(&bpf_mt_reg);
+}
+
+module_init(bpf_mt_init);
+module_exit(bpf_mt_exit);
--

1.7.7.3

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

To compile code right now, the little bpf compiler that I emailed before can be downloaded from
http://code.google.com/p/kernel/downloads/detail?name=bpf2decimal.c

I don't think that a compiler has to be shipped with iptables itself, let alone make iptables link against libraries. That said, it is not impossible to detect pcap.h in configure.ac and optionally enable a "-m bpf --string" mode that calls pcap_compile_nopcap from within libxt_bpf, so let me know if you would like me to code that up. I can also try to send a patch to tcpdump that extends compilation (-ddd -y <type>) to arbitrary link layer types.

On Fri, Jan 18, 2013 at 11:48:34AM -0500, Willem de Bruijn wrote:
[...]

Quoted 11 lines

To compile code right now, the little bpf compiler that I emailed before can be downloaded from
http://code.google.com/p/kernel/downloads/detail?name=bpf2decimal.c

I don't think that a compiler has to be shipped with iptables itself, let alone make iptables link against libraries. That said, it is not impossible to detect pcap.h in configure.ac and optionally enable a "-m bpf --string" mode that calls pcap_compile_nopcap from within libxt_bpf, so let me know if you would like me to code that up. I can also try to send a patch to tcpdump that extends compilation (-ddd -y <type>) to arbitrary link layer types.

We have to decide if:

a) we add a new hard library dependency to iptables (libpcap) for just for one single module, that is, the libxt_bpf depends on libpcap.

or

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

I'd like to hear pro and cons arguments from others on this.

Quoted 19 lines · Pablo Neira Ayuso

On Fri, Jan 18, 2013 at 11:48:34AM -0500, Willem de Bruijn wrote:
[...]

Quoted 11 lines

To compile code right now, the little bpf compiler that I emailed before can be downloaded from
http://code.google.com/p/kernel/downloads/detail?name=bpf2decimal.c

I don't think that a compiler has to be shipped with iptables itself, let alone make iptables link against libraries. That said, it is not impossible to detect pcap.h in configure.ac and optionally enable a "-m bpf --string" mode that calls pcap_compile_nopcap from within libxt_bpf, so let me know if you would like me to code that up. I can also try to send a patch to tcpdump that extends compilation (-ddd -y <type>) to arbitrary link layer types.

We have to decide if:

a) we add a new hard library dependency to iptables (libpcap) for just for one single module, that is, the libxt_bpf depends on libpcap.

Quoted 8 lines

or

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

I'd like to hear pro and cons arguments from others on this.

a) is arguably more user friendly, however, I don't think we can retain the 'text representation' for iptables-save so users would still be confronted with the compiled data at some point (i.e., they need to write down the original expression anyway to figure out what the rule they added 6 months back actually does...)

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

Quoted 35 lines · Florian Westphal

Quoted 19 lines · Pablo Neira Ayuso

On Fri, Jan 18, 2013 at 11:48:34AM -0500, Willem de Bruijn wrote:
[...]

Quoted 11 lines

To compile code right now, the little bpf compiler that I emailed before can be downloaded from
http://code.google.com/p/kernel/downloads/detail?name=bpf2decimal.c

I don't think that a compiler has to be shipped with iptables itself, let alone make iptables link against libraries. That said, it is not impossible to detect pcap.h in configure.ac and optionally enable a "-m bpf --string" mode that calls pcap_compile_nopcap from within libxt_bpf, so let me know if you would like me to code that up. I can also try to send a patch to tcpdump that extends compilation (-ddd -y <type>) to arbitrary link layer types.

We have to decide if:

a) we add a new hard library dependency to iptables (libpcap) for just for one single module, that is, the libxt_bpf depends on libpcap.

Quoted 8 lines

or

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

I'd like to hear pro and cons arguments from others on this.

a) is arguably more user friendly, however, I don't think we can retain the 'text representation' for iptables-save so users would still be confronted with the compiled data at some point (i.e., they need to write down the original expression anyway to figure out what the rule they added 6 months back actually does...)

Quoted 2 lines

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

Yes, let's go with b). (But from packaging point of view utils/bpf2decimal.c depending on libpcap is not much different from extensions/libxt_bpf.c depending on libpcap.)

Quoted 19 lines · Pablo Neira Ayuso

On Fri, Jan 18, 2013 at 11:48:34AM -0500, Willem de Bruijn wrote:
[...]

Quoted 11 lines

To compile code right now, the little bpf compiler that I emailed before can be downloaded from
http://code.google.com/p/kernel/downloads/detail?name=bpf2decimal.c

I don't think that a compiler has to be shipped with iptables itself, let alone make iptables link against libraries. That said, it is not impossible to detect pcap.h in configure.ac and optionally enable a "-m bpf --string" mode that calls pcap_compile_nopcap from within libxt_bpf, so let me know if you would like me to code that up. I can also try to send a patch to tcpdump that extends compilation (-ddd -y <type>) to arbitrary link layer types.

We have to decide if:

a) we add a new hard library dependency to iptables (libpcap) for just for one single module, that is, the libxt_bpf depends on libpcap.

Quoted 8 lines

or

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

I'd like to hear pro and cons arguments from others on this.

a) is arguably more user friendly, however, I don't think we can retain the 'text representation' for iptables-save so users would still be confronted with the compiled data at some point (i.e., they need to write down the original expression anyway to figure out what the rule they added 6 months back actually does...)

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

Quoted 4 lines

Quoted 4 lines

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

Quoted 2 lines

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

This sounds like the consensus (for the record, I also prefer this less disruptive approach). In that case, I can submit a revised libxt_bpf with your suggested changes right away, Pablo, and we can leave the separate userspace tool for a later commit.

Quoted 4 lines · Willem de Bruijn

Quoted 4 lines

Quoted 4 lines

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

[...]

Quoted 3 lines

Quoted 2 lines

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

Quoted 4 lines

This sounds like the consensus (for the record, I also prefer this less disruptive approach). In that case, I can submit a revised libxt_bpf with your suggested changes right away, Pablo, and we can leave the separate userspace tool for a later commit.

Either way is fine, but please we should have that utility compiler integrated in the iptables tree by when 3.9-rc1 is released.

Quoted 9 lines · Pablo Neira Ayuso

Quoted 4 lines · Willem de Bruijn

Quoted 4 lines

Quoted 4 lines

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

[...]

Quoted 3 lines

Quoted 2 lines

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

Quoted 7 lines

Quoted 4 lines

This sounds like the consensus (for the record, I also prefer this less disruptive approach). In that case, I can submit a revised libxt_bpf with your suggested changes right away, Pablo, and we can leave the separate userspace tool for a later commit.

Either way is fine, but please we should have that utility compiler integrated in the iptables tree by when 3.9-rc1 is released.

Okay. I'll prepare a separate patch with the pcap-based utility, then.

Since utils is built as part of the root make invocation, I think it's better to test for pcap.h in the root configure.ac and add a test in utils/Makefile.am to build this tool if found, as opposed to creating a separate configure.ac under utils. We can also discuss these details after the first version of the patch, of course.

A BPF compiler to convert tcpudmp expressions to the decimal format accepted by the libxt_bpf.

utils/bpf_compile.c new file +41 -0

@@ -0,0 +1,55 @@
+/*
+ * BPF program compilation tool
+ *
+ * Generates decimal output, similar to `tcpdump -ddd ...`.
+ * Unlike tcpdump, will generate for any given link layer type.
+ *
+ * There is no makefile:
+ * compile with `gcc -Wall -o bpf2decimal bpf2decimal.c -lpcap` or similar.
+ *
+ * Written by Willem de Bruijn ([email protected])
+ * Copyright Google, Inc. 2013
+ * Licensed under the GNU General Public License version 2 (GPLv2)
+*/
+
+#include <pcap.h>
+#include <stdio.h>
+
+int main(int argc, char **argv)
+{
+	struct bpf_program program;
+	struct bpf_insn *ins;
+	int i, dlt = DLT_RAW;
+
+	if (argc < 2 || argc > 3) {
+		fprintf(stderr, "Usage:    %s [link] '<program>'\n\n"
+				"          link is a pcap linklayer type:\n"
+				"          one of EN10MB, RAW, SLIP, ...\n\n"
+				"Examples: %s RAW 'tcp and greater 100'\n"
+				"          %s EN10MB 'ip proto 47'\n'",
+				argv[0], argv[0], argv[0]);
+		return 1;
+	}
+
+	if (argc == 3) {
+		dlt = pcap_datalink_name_to_val(argv[1]);
+		if (dlt == -1) {
+			fprintf(stderr, "Unknown datalinktype: %s\n", argv[1]);
+			return 1;
+		}
+	}
+

utils/bpf_compile.c new file +3 -0

+	return 0;
+}
+

Quoted 4 lines · Willem de Bruijn

A BPF compiler to convert tcpudmp expressions to the decimal format accepted
by the libxt_bpf.

[...]

Quoted 58 lines

utils/bpf_compile.c new file +8 -0

@@ -0,0 +1,55 @@
+/*
+ * BPF program compilation tool
+ *
+ * Generates decimal output, similar to `tcpdump -ddd ...`.
+ * Unlike tcpdump, will generate for any given link layer type.
+ *
+ * There is no makefile:
+ * compile with `gcc -Wall -o bpf2decimal bpf2decimal.c -lpcap` or

similar.
+ *
+ * Written by Willem de Bruijn ([email protected])
+ * Copyright Google, Inc. 2013
+ * Licensed under the GNU General Public License version 2 (GPLv2)

utils/bpf_compile.c new file +14 -0

+*/
+
+#include <pcap.h>
+#include <stdio.h>
+
+int main(int argc, char **argv)
+{
+       struct bpf_program program;
+       struct bpf_insn *ins;
+       int i, dlt = DLT_RAW;
+
+       if (argc < 2 || argc > 3) {
+               fprintf(stderr, "Usage:    %s [link] '<program>'\n\n"
+                               "          link is a pcap linklayer

type:\n"
+ " one of EN10MB, RAW, SLIP,
...\n\n"

utils/bpf_compile.c new file +10 -0

+                               "Examples: %s RAW 'tcp and greater 100'\n"
+                               "          %s EN10MB 'ip proto 47'\n'",
+                               argv[0], argv[0], argv[0]);
+               return 1;
+       }
+
+       if (argc == 3) {
+               dlt = pcap_datalink_name_to_val(argv[1]);
+               if (dlt == -1) {
+                       fprintf(stderr, "Unknown datalinktype: %s\n",

argv[1]);

utils/bpf_compile.c new file +14 -0

+                       return 1;
+               }
+       }
+
+       if (pcap_compile_nopcap(65535, dlt, &program, argv[argc - 1], 1,
+                               PCAP_NETMASK_UNKNOWN)) {
+               fprintf(stderr, "Compilation error\n");
+               return 1;
+       }
+
+       printf("%d\n", program.bf_len);
+       ins = program.bf_insns;
+       for (i = 0; i < program.bf_len; ++ins, ++i)
+               printf("%u %u %u %u\n", ins->code, ins->jt, ins->jf,

Quoted 1 line

ins->k);

Here I think you should release the internally allocated memory by adding a:

        pcap_freecode(&program);

Quoted 5 lines

utils/bpf_compile.c new file +3 -0

+       return 0;
+}
+

Quoted 71 lines · Daniel Borkmann

Quoted 4 lines · Willem de Bruijn

A BPF compiler to convert tcpudmp expressions to the decimal format accepted
by the libxt_bpf.

[...]

Quoted 58 lines

utils/bpf_compile.c new file +8 -0

@@ -0,0 +1,55 @@
+/*
+ * BPF program compilation tool
+ *
+ * Generates decimal output, similar to `tcpdump -ddd ...`.
+ * Unlike tcpdump, will generate for any given link layer type.
+ *
+ * There is no makefile:
+ * compile with `gcc -Wall -o bpf2decimal bpf2decimal.c -lpcap` or

similar.
+ *
+ * Written by Willem de Bruijn ([email protected])
+ * Copyright Google, Inc. 2013
+ * Licensed under the GNU General Public License version 2 (GPLv2)

utils/bpf_compile.c new file +14 -0

+*/
+
+#include <pcap.h>
+#include <stdio.h>
+
+int main(int argc, char **argv)
+{
+       struct bpf_program program;
+       struct bpf_insn *ins;
+       int i, dlt = DLT_RAW;
+
+       if (argc < 2 || argc > 3) {
+               fprintf(stderr, "Usage:    %s [link] '<program>'\n\n"
+                               "          link is a pcap linklayer

type:\n"
+ " one of EN10MB, RAW, SLIP,
...\n\n"

utils/bpf_compile.c new file +10 -0

+                               "Examples: %s RAW 'tcp and greater 100'\n"
+                               "          %s EN10MB 'ip proto 47'\n'",
+                               argv[0], argv[0], argv[0]);
+               return 1;
+       }
+
+       if (argc == 3) {
+               dlt = pcap_datalink_name_to_val(argv[1]);
+               if (dlt == -1) {
+                       fprintf(stderr, "Unknown datalinktype: %s\n",

argv[1]);

utils/bpf_compile.c new file +14 -0

+                       return 1;
+               }
+       }
+
+       if (pcap_compile_nopcap(65535, dlt, &program, argv[argc - 1], 1,
+                               PCAP_NETMASK_UNKNOWN)) {
+               fprintf(stderr, "Compilation error\n");
+               return 1;
+       }
+
+       printf("%d\n", program.bf_len);
+       ins = program.bf_insns;
+       for (i = 0; i < program.bf_len; ++ins, ++i)
+               printf("%u %u %u %u\n", ins->code, ins->jt, ins->jf,

Quoted 1 line

ins->k);

Here I think you should release the internally allocated memory by adding a:

Quoted 1 line

         pcap_freecode(&program);

Thanks for catching that, Daniel. I'll hold off respinning the patch to see if there is other feedback, but will fix this in the next revision.

V2
- update build process to
- succeed without bpf_compiler when --enable-bpf-compiler is not set
- succeed with bpf_compiler when this flag is set and libpcap is found
- fail when this flag is set but libpcap is absent
- release memory allocated by pcap on exit

Quoted 6 lines · Willem de Bruijn

V2
- update build process to
- succeed without bpf_compiler when --enable-bpf-compiler is not set
- succeed with bpf_compiler when this flag is set and libpcap is found
- fail when this flag is set but libpcap is absent
- release memory allocated by pcap on exit

Applied with changes:

* renamed utility to add the `nf' prefix.

* the output of the utility is now exactly what iptables -m bpf --bytecode needs, just to make it easier to everyone.

Still missing a hard fail if --enable-bpf-compiler is set and libpcap is not available in the system,

I'd appreciate if you send me a follow
up patch to address this. You may want to have a look at how ulogd2 does this in its configure fail as reference.

Quoted 2 lines

Still missing a hard fail if --enable-bpf-compiler is set and libpcap is not available in the system,

Compilation and linking will both fail hard:

if gcc -DHAVE_CONFIG_H -I. -I. -I.. -D_LARGEFILE_SOURCE=1

(quoted diff) +0 -6

-D_LARGE_FILES -D_FILE_OFFSET_BITS=64 -D_REENTRANT
-DXTABLES_LIBDIR=\"/usr/local/lib/xtables\" -DXTABLES_INTERNAL
-I../include -I../include   -Wall -Waggregate-return
-Wmissing-declarations -Wmissing-prototypes -Wredundant-decls -Wshadow
-Wstrict-prototypes -Winline -pipe -g -O2 -MT nfbpf_compile.o -MD -MP
-MF ".deps/nfbpf_compile.Tpo" -c -o nfbpf_compile.o nfbpf_compile.c; \

then mv -f ".deps/nfbpf_compile.Tpo" ".deps/nfbpf_compile.Po"; else rm -f ".deps/nfbpf_compile.Tpo"; exit 1; fi nfbpf_compile.c:12:18: fatal error: pcap.h: No such file or directory compilation terminated.
make: *** [nfbpf_compile.o] Error 1

Do you want the configure script itself to fail before that step?

Quoted 3 lines

I'd appreciate if you send me a follow
up patch to address this. You may want to have a look at how ulogd2 does this in its configure fail as reference.

That skips compilation if HAVE_PCAP is false, but it does not fail hard during configure, either. At least, that's what I gather from reading and running

https://git.netfilter.org/ulogd2/tree/configure.ac#n23

Quoted 15 lines · Willem de Bruijn

Quoted 14 lines · Pablo Neira Ayuso

Quoted 4 lines · Willem de Bruijn

Quoted 4 lines

Quoted 4 lines

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

[...]

Quoted 7 lines

Quoted 2 lines

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

This sounds like the consensus (for the record, I also prefer this less disruptive approach). In that case, I can submit a revised libxt_bpf with your suggested changes right away, Pablo, and we can leave the separate userspace tool for a later commit.

Quoted 4 lines

Quoted 2 lines

Either way is fine, but please we should have that utility compiler integrated in the iptables tree by when 3.9-rc1 is released.

Okay. I'll prepare a separate patch with the pcap-based utility, then.

Quoted 5 lines

Since utils is built as part of the root make invocation, I think it's better to test for pcap.h in the root configure.ac and add a test in utils/Makefile.am to build this tool if found, as opposed to creating a separate configure.ac under utils. We can also discuss these details after the first version of the patch, of course.

That's fine by now, and it's way less bloat.

Quoted 16 lines · Pablo Neira Ayuso

Quoted 15 lines · Willem de Bruijn

Quoted 14 lines · Pablo Neira Ayuso

Quoted 4 lines · Willem de Bruijn

Quoted 4 lines

Quoted 4 lines

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

[...]

Quoted 7 lines

Quoted 2 lines

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

This sounds like the consensus (for the record, I also prefer this less disruptive approach). In that case, I can submit a revised libxt_bpf with your suggested changes right away, Pablo, and we can leave the separate userspace tool for a later commit.

Quoted 4 lines

Quoted 4 lines

Quoted 2 lines

Either way is fine, but please we should have that utility compiler integrated in the iptables tree by when 3.9-rc1 is released.

Okay. I'll prepare a separate patch with the pcap-based utility, then.

Just sent the patch. I'm no expert at autoconf and automake, so the build logic can conceivably be shorter, but it works for me and the logic is straightforward. I forgot to mention in the commit message which versions of the tools I used: tested on a ubuntu 12.04 with autoconf 2.68, automake 1.9.6 and libtool 2.4.2.

Quoted 7 lines

Quoted 5 lines

Since utils is built as part of the root make invocation, I think it's better to test for pcap.h in the root configure.ac and add a test in utils/Makefile.am to build this tool if found, as opposed to creating a separate configure.ac under utils. We can also discuss these details after the first version of the patch, of course.

That's fine by now, and it's way less bloat.

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

at a guess, there should be a --with/without option for it, and if --with=tool is specified build/configure should fail if support libraries are missing

Quoted 34 lines · Pablo Neira Ayuso

Quoted 16 lines · Willem de Bruijn

Quoted 15 lines · Pablo Neira Ayuso

Quoted 14 lines · Willem de Bruijn

Quoted 4 lines

Quoted 4 lines

Quoted 4 lines

b) provide a separate utility to generate the BPF filter in text-based format from some utility that accepts tcpdump-like syntax. The utility can be distributed in the utils directory and it would not be mandatory to compile it if libpcap is not present.

[...]

Quoted 7 lines

Quoted 2 lines

I would go with b) for now; we can always move to a) later on, but not the other way around (would kill backwards compatibility).

This sounds like the consensus (for the record, I also prefer this less disruptive approach). In that case, I can submit a revised libxt_bpf with your suggested changes right away, Pablo, and we can leave the separate userspace tool for a later commit.

Quoted 4 lines

Quoted 4 lines

Quoted 2 lines

Either way is fine, but please we should have that utility compiler integrated in the iptables tree by when 3.9-rc1 is released.

Okay. I'll prepare a separate patch with the pcap-based utility, then.

Just sent the patch. I'm no expert at autoconf and automake, so the build logic can conceivably be shorter, but it works for me and the logic is straightforward. I forgot to mention in the commit message which versions of the tools I used: tested on a ubuntu 12.04 with autoconf 2.68, automake 1.9.6 and libtool 2.4.2.

Quoted 6 lines

Quoted 5 lines

Since utils is built as part of the root make invocation, I think it's better to test for pcap.h in the root configure.ac and add a test in utils/Makefile.am to build this tool if found, as opposed to creating a separate configure.ac under utils. We can also discuss these details after the first version of the patch, of course.

Quoted 5 lines

Quoted 1 line

That's fine by now, and it's way less bloat.

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

The second patch is more speculative and aims to be a more general workaround, as well as a performance optimization: support (preferably JIT compiled) BPF programs as iptables match rules.

Potentially, the skb->priority match can be implemented by applying only the second patch and adding a new BPF_S_ANC ancillary field to Linux Socket Filters.

I also wrote corresponding userspace patches to iptables. The process for submitting both kernel and user patches is not 100% clear to me. Sending the kernel bits to both netdev and netfilter-devel for initial feedback. Please correct me if you want it another way.

The patches apply to net-next.

Somehow, the first part of this email went missing. Not critical, but for completeness:

These two patches each add an xtables match.

The xt_priority match is a straighforward addition in the style of xt_mark, adding the option to filter on one more sk_buff field. I have an immediate application for this. The amount of code (in kernel + userspace) to add a single check proved quite large.

Quoted 3 lines · Willem de Bruijn

Somehow, the first part of this email went missing. Not critical, but for completeness:

Quoted 6 lines

These two patches each add an xtables match.

The xt_priority match is a straighforward addition in the style of xt_mark, adding the option to filter on one more sk_buff field. I have an immediate application for this. The amount of code (in kernel + userspace) to add a single check proved quite large.

Hm so yeah, can't we just place this in xt_mark.c?

Quoted 5 lines · Jan Engelhardt

Quoted 3 lines · Willem de Bruijn

Somehow, the first part of this email went missing. Not critical, but for completeness:

Quoted 8 lines

Quoted 6 lines

These two patches each add an xtables match.

The xt_priority match is a straighforward addition in the style of xt_mark, adding the option to filter on one more sk_buff field. I have an immediate application for this. The amount of code (in kernel + userspace) to add a single check proved quite large.

Hm so yeah, can't we just place this in xt_mark.c?

I'm happy to do so, but note that that breaks the custom of having one static struct xt_$NAME for each file xt_$NAME.[ch].

It may be reasonable, as the same issue may keep popping up as additional sk_buff fields are found useful for filtering. For instance, skb->queue_mapping could be used in conjuction with network flow classification (ethtool -N).

All the ancillary data
accessible from BPF likely has some use and could be ported to iptables (rxhash, pkt_type, ...).

To avoid rule explosion, I considered an xt_skbuff match rule that applies the same mask operation, range and inversion tests, and takes a field id to select the sk_buff field to operate on. I think the BPF patch is a better long term solution.

Quoted 6 lines

Quoted 4 lines

The xt_priority match is a straighforward addition in the style of xt_mark, adding the option to filter on one more sk_buff field. I have an immediate application for this. The amount of code (in kernel + userspace) to add a single check proved quite large.

Hm so yeah, can't we just place this in xt_mark.c?

I'm happy to do so, but note that that breaks the custom of having one static struct xt_$NAME for each file xt_$NAME.[ch].

To avoid rule explosion, I considered an xt_skbuff match rule that applies the same mask operation, range and inversion tests, and takes a field id to select the sk_buff field to operate on. I think the BPF patch is a better long term solution.

Quoted 3 lines · Willem de Bruijn

Somehow, the first part of this email went missing. Not critical, but for completeness:

Quoted 6 lines

These two patches each add an xtables match.

The xt_priority match is a straighforward addition in the style of xt_mark, adding the option to filter on one more sk_buff field. I have an immediate application for this. The amount of code (in kernel + userspace) to add a single check proved quite large.

Hm so yeah, can't we just place this in xt_mark.c?

Quoted 7 lines · Jan Engelhardt

Quoted 5 lines · Willem de Bruijn

Somehow, the first part of this email went missing. Not critical, but for completeness:

These two patches each add an xtables match.

Quoted 6 lines

Quoted 4 lines

The xt_priority match is a straighforward addition in the style of xt_mark, adding the option to filter on one more sk_buff field. I have an immediate application for this. The amount of code (in kernel + userspace) to add a single check proved quite large.

Hm so yeah, can't we just place this in xt_mark.c?

I don't feel this belongs to xt_mark at all.

Quoted 15 lines · Pablo Neira Ayuso

Quoted 13 lines · Jan Engelhardt

Quoted 9 lines · Willem de Bruijn

Somehow, the first part of this email went missing. Not critical, but for completeness:

These two patches each add an xtables match.

The xt_priority match is a straighforward addition in the style of xt_mark, adding the option to filter on one more sk_buff field. I have an immediate application for this. The amount of code (in kernel + userspace) to add a single check proved quite large.

Hm so yeah, can't we just place this in xt_mark.c?

Quoted 1 line

I don't feel this belongs to xt_mark at all.

Do you have other concerns, or can I resubmit as is for merging in a few days if no one raises additional issues?

For this and netfilter changes in general: should these patches be against git://1984.lsi.us.es/nf-next instead of net-next? This patch likely applies cleanly there, but I haven't tried yet. Thanks.

Quoted 13 lines · Pablo Neira Ayuso

Quoted 12 lines · Jan Engelhardt

Quoted 9 lines · Willem de Bruijn

Somehow, the first part of this email went missing. Not critical, but for completeness:

These two patches each add an xtables match.

The xt_priority match is a straighforward addition in the style of xt_mark, adding the option to filter on one more sk_buff field. I have an immediate application for this. The amount of code (in kernel + userspace) to add a single check proved quite large.

Quoted 3 lines

Quoted 1 line

Hm so yeah, can't we just place this in xt_mark.c?

I don't feel this belongs to xt_mark at all.

Do you have other concerns, or can I resubmit as is for merging in a few days if no one raises additional issues?