Skip to contentSkip to conversation

netfilter 00/41: Netfilter update for 2.6.30

PATCH Applied Patrick McHardy 5 participants 58 messages Mar 24, 2009 – Mar 25, 2009

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

Hi Dave,

the following patches contain (the major part of) my netfilter updates for 2.6.30. Besides the usual small fixes and cleanups, there is:

- a new LED trigger target by Adam Nielsen

- device name match optimizations for *tables from Eric Dumazet

- Conversion of *tables to use RCU for rulesets and counters from Stephen and Eric

- unification of the TTL/HL targets and ttl/hl matches from Jan Engelhardt

- a new "cluster" match from Pablo to transparently distribute traffic over a cluster of machines without a load-balancer

- a set of patches to make netfilter logging backends selectable per address family from Eric Leblond

Please apply or pull from:

git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6.git

Thanks!

 drivers/leds/Kconfig                           |    3 +
 include/linux/netfilter/Kbuild                 |    8 +
 include/linux/netfilter/nfnetlink.h            |    1 +
 include/linux/netfilter/x_tables.h             |   15 +-
 include/linux/netfilter/xt_LED.h               |   13 ++
 include/linux/netfilter/xt_cluster.h           |   15 ++
 include/linux/netfilter/xt_limit.h             |    9 +-
 include/linux/netfilter/xt_quota.h             |    4 +-
 include/linux/netfilter/xt_statistic.h         |    7 +-
 include/linux/netfilter_ipv6/Kbuild            |    1 +
 include/net/netfilter/nf_conntrack.h           |    2 +-
 include/net/netfilter/nf_conntrack_l4proto.h   |   10 +-
 include/net/netfilter/nf_log.h                 |   11 +-
 net/bridge/netfilter/ebtable_broute.c          |    1 -
 net/bridge/netfilter/ebtable_filter.c          |    1 -
 net/bridge/netfilter/ebtable_nat.c             |    1 -
 net/ipv4/netfilter/Kconfig                     |   30 ++--
 net/ipv4/netfilter/Makefile                    |    2 -
 net/ipv4/netfilter/arp_tables.c                |  159 ++++++++++++++-----
 net/ipv4/netfilter/arptable_filter.c           |    2 -
 net/ipv4/netfilter/ip_queue.c                  |    2 +
 net/ipv4/netfilter/ip_tables.c                 |  153 +++++++++++++------
 net/ipv4/netfilter/ipt_LOG.c                   |    2 +-
 net/ipv4/netfilter/ipt_TTL.c                   |   97 ------------
 net/ipv4/netfilter/ipt_ULOG.c                  |    2 +-
 net/ipv4/netfilter/ipt_ttl.c                   |   63 --------
 net/ipv4/netfilter/iptable_filter.c            |    1 -
 net/ipv4/netfilter/iptable_mangle.c            |    1 -
 net/ipv4/netfilter/iptable_raw.c               |    1 -
 net/ipv4/netfilter/iptable_security.c          |    1 -
 net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c |    4 +-
 net/ipv4/netfilter/nf_nat_rule.c               |    1 -
 net/ipv6/netfilter/Kconfig                     |   38 ++---
 net/ipv6/netfilter/Makefile                    |    2 -
 net/ipv6/netfilter/ip6_queue.c                 |    1 +
 net/ipv6/netfilter/ip6_tables.c                |  152 ++++++++++++------
 net/ipv6/netfilter/ip6t_HL.c                   |   95 -----------
 net/ipv6/netfilter/ip6t_LOG.c                  |    2 +-
 net/ipv6/netfilter/ip6t_hl.c                   |   68 --------
 net/ipv6/netfilter/ip6table_filter.c           |    1 -
 net/ipv6/netfilter/ip6table_mangle.c           |    1 -
 net/ipv6/netfilter/ip6table_raw.c              |    1 -
 net/ipv6/netfilter/ip6table_security.c         |    1 -
 net/ipv6/netfilter/nf_conntrack_l3proto_ipv6.c |    1 +
 net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c |    4 +
 net/netfilter/Kconfig                          |   63 ++++++++
 net/netfilter/Makefile                         |    4 +
 net/netfilter/core.c                           |    2 -
 net/netfilter/nf_conntrack_core.c              |   14 +-
 net/netfilter/nf_conntrack_expect.c            |    3 +-
 net/netfilter/nf_conntrack_netlink.c           |  161 +++++++++-----------
 net/netfilter/nf_conntrack_proto.c             |   21 ---
 net/netfilter/nf_conntrack_proto_dccp.c        |  145 +++++++++++++-----
 net/netfilter/nf_conntrack_proto_generic.c     |    2 +-
 net/netfilter/nf_conntrack_proto_tcp.c         |    2 +
 net/netfilter/nf_conntrack_proto_udp.c         |    2 +
 net/netfilter/nf_log.c                         |  201 ++++++++++++++++++++----
 net/netfilter/nfnetlink.c                      |    6 +
 net/netfilter/nfnetlink_log.c                  |   18 ++-
 net/netfilter/x_tables.c                       |   26 +++-
 net/netfilter/xt_HL.c                          |  171 ++++++++++++++++++++
 net/netfilter/xt_LED.c                         |  161 +++++++++++++++++++
 net/netfilter/xt_cluster.c                     |  164 +++++++++++++++++++
 net/netfilter/xt_hashlimit.c                   |    7 +-
 net/netfilter/xt_hl.c                          |  108 +++++++++++++
 net/netfilter/xt_limit.c                       |   40 ++++--
 net/netfilter/xt_physdev.c                     |   37 +++--
 net/netfilter/xt_quota.c                       |   31 +++-
 net/netfilter/xt_statistic.c                   |   28 +++-
 net/netlink/af_netlink.c                       |    1 +
 net/sysctl_net.c                               |    2 +-
 71 files changed, 1626 insertions(+), 784 deletions(-)
 create mode 100644 include/linux/netfilter/xt_LED.h
 create mode 100644 include/linux/netfilter/xt_cluster.h
 delete mode 100644 net/ipv4/netfilter/ipt_TTL.c
 delete mode 100644 net/ipv4/netfilter/ipt_ttl.c
 delete mode 100644 net/ipv6/netfilter/ip6t_HL.c
 delete mode 100644 net/ipv6/netfilter/ip6t_hl.c
 create mode 100644 net/netfilter/xt_HL.c
 create mode 100644 net/netfilter/xt_LED.c
 create mode 100644 net/netfilter/xt_cluster.c
 create mode 100644 net/netfilter/xt_hl.c

Adam Nielsen (1):
netfilter: x_tables: add LED trigger target

Christoph Paasch (2):
netfilter: change generic l4 protocol number
netfilter: remove IPvX specific parts from nf_conntrack_l4proto.h

Cyrill Gorcunov (2):
net: sysctl_net - use net_eq to compare nets
net: netfilter conntrack - add per-net functionality for DCCP protocol

Eric Dumazet (7):
netfilter: arp_tables: unfold two critical loops in arp_packet_match()
netfilter: xt_physdev fixes
netfilter: xt_physdev: unfold two loops in physdev_mt()
netfilter: ip6_tables: unfold two loops in ip6_packet_match()
netfilter: ip_tables: unfold two critical loops in ip_packet_match()
netfilter: xt_hashlimit fix
netfilter: nf_conntrack: Reduce conntrack count in nf_conntrack_free()

Eric Leblond (4):
netfilter: log invalid new icmpv6 packet with nf_log_packet()
netfilter: use a linked list of loggers
netfilter: print the list of register loggers
netfilter: sysctl support of logger choice

Florian Westphal (2):
netfilter: ctnetlink: remove remaining module refcounting
netfilter: remove nf_ct_l4proto_find_get/nf_ct_l4proto_put

Hagen Paul Pfeifer (2):
netfilter: nf_conntrack: table max size should hold at least table size
netfilter: fix hardcoded size assumptions

Jan Engelhardt (6):
netfilter: remove unneeded goto
netfilter: Combine ipt_TTL and ip6t_HL source
netfilter: Combine ipt_ttl and ip6t_hl source
netfilter: xtables: add backward-compat options
netfilter: install missing headers
netfilter: xtables: avoid pointer to self

Pablo Neira Ayuso (7):
netfilter: nf_conntrack: account packets drop by tcp_packet()
netfilter: conntrack: increase drop stats if sequence adjustment fails
netfilter: ctnetlink: cleanup master conntrack assignation
netfilter: ctnetlink: cleanup conntrack update preliminary checkings
netfilter: ctnetlink: move event reporting for new entries outside the lock
netfilter: xtables: add cluster match
nefilter: nfnetlink: add nfnetlink_set_err and use it in ctnetlink

Patrick McHardy (1):
netfilter: ctnetlink: fix rcu context imbalance

Scott James Remnant (2):
netfilter: auto-load ip6_queue module when socket opened
netfilter: auto-load ip_queue module when socket opened

Stephen Hemminger (5):
netfilter: x_tables: change elements in x_tables
netfilter: x_tables: remove unneeded initializations
netfilter: ebtables: remove unneeded initializations
netfilter: iptables: lock free counters
netfilter: Kconfig spelling fixes (trivial)

#1 lore raw

netfilter 01/41: change generic l4 protocol number

commit fe2a7ce4de07472ace0cdf460a41f462a4621687

Author: Christoph Paasch <[email protected]>
Date:   Wed Feb 18 16:28:35 2009 +0100

netfilter: change generic l4 protocol number

0 is used by Hop-by-hop header and so this may cause confusion. 255 is stated as 'Reserved' by IANA.

net/netfilter/nf_conntrack_proto_generic.c +1 -1
index 4be80d7..829374f 100644
@@ -92,7 +92,7 @@ static struct ctl_table generic_compat_sysctl_table[] = {
 struct nf_conntrack_l4proto nf_conntrack_l4proto_generic __read_mostly =
 {
 	.l3proto		= PF_UNSPEC,
-	.l4proto		= 0,
+	.l4proto		= 255,
 	.name			= "unknown",
 	.pkt_to_tuple		= generic_pkt_to_tuple,
 	.invert_tuple		= generic_invert_tuple,
#2 lore raw

netfilter 02/41: remove unneeded goto

In reply to Patrick McHardy

commit fecea3a389c89de9afae2eda74fad894d5677229

Author: Jan Engelhardt <[email protected]>
Date:   Wed Feb 18 16:29:08 2009 +0100

netfilter: remove unneeded goto

net/netfilter/core.c +0 -2
index a90ac83..5bb3473 100644
@@ -174,7 +174,6 @@ next_hook:
 			     outdev, &elem, okfn, hook_thresh);
 	if (verdict == NF_ACCEPT || verdict == NF_STOP) {
 		ret = 1;
-		goto unlock;
 	} else if (verdict == NF_DROP) {
 		kfree_skb(skb);
 		ret = -EPERM;
@@ -183,7 +182,6 @@ next_hook:
 			      verdict >> NF_VERDICT_BITS))
 			goto next_hook;
 	}
-unlock:
 	rcu_read_unlock();
 	return ret;
 }
#3 lore raw

netfilter 03/41: x_tables: change elements in x_tables

In reply to Patrick McHardy

commit 4a2f965ca5a4e2593744bf75425d85e0e8ff814a

Author: Stephen Hemminger <[email protected]>
Date:   Wed Feb 18 16:29:44 2009 +0100

netfilter: x_tables: change elements in x_tables

Change to proper type on private pointer rather than anonymous void. Keep active elements on same cache line.

include/linux/netfilter/x_tables.h +4 -5
index c7ee874..9fac88f 100644
@@ -349,9 +349,6 @@ struct xt_table
 {
 	struct list_head list;
 
-	/* A unique name... */
-	const char name[XT_TABLE_MAXNAMELEN];
-
 	/* What hooks you will enter on */
 	unsigned int valid_hooks;
 
@@ -359,13 +356,15 @@ struct xt_table
 	rwlock_t lock;
 
 	/* Man behind the curtain... */
-	//struct ip6t_table_info *private;
-	void *private;
+	struct xt_table_info *private;
 
 	/* Set this to THIS_MODULE if you are a module, otherwise NULL */
 	struct module *me;
 
 	u_int8_t af;		/* address/protocol family */
+
+	/* A unique name... */
+	const char name[XT_TABLE_MAXNAMELEN];
 };
 
 #include <linux/netfilter_ipv4.h>
#4 lore raw

netfilter 04/41: x_tables: remove unneeded initializations

In reply to Patrick McHardy

commit 9c8222b9e71b690c8388bb0ebe5c3e5a1469e884

Author: Stephen Hemminger <[email protected]>
Date:   Wed Feb 18 16:30:20 2009 +0100

netfilter: x_tables: remove unneeded initializations

Later patches change the locking on xt_table and the initialization of the lock element is not needed since the lock is always initialized in xt_table_register anyway.

net/ipv4/netfilter/arptable_filter.c +0 -2
index e091187..6ecfdae 100644
@@ -48,8 +48,6 @@ static struct
 static struct xt_table packet_filter = {
 	.name		= "filter",
 	.valid_hooks	= FILTER_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(packet_filter.lock),
-	.private	= NULL,
 	.me		= THIS_MODULE,
 	.af		= NFPROTO_ARP,
 };
net/ipv4/netfilter/iptable_filter.c +0 -1
index 52cb693..c30a969 100644
@@ -56,7 +56,6 @@ static struct
 static struct xt_table packet_filter = {
 	.name		= "filter",
 	.valid_hooks	= FILTER_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(packet_filter.lock),
 	.me		= THIS_MODULE,
 	.af		= AF_INET,
 };
net/ipv4/netfilter/iptable_mangle.c +0 -1
index 3929d20..4087614 100644
@@ -67,7 +67,6 @@ static struct
 static struct xt_table packet_mangler = {
 	.name		= "mangle",
 	.valid_hooks	= MANGLE_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(packet_mangler.lock),
 	.me		= THIS_MODULE,
 	.af		= AF_INET,
 };
net/ipv4/netfilter/iptable_raw.c +0 -1
index 7f65d18..e5356da 100644
@@ -39,7 +39,6 @@ static struct
 static struct xt_table packet_raw = {
 	.name = "raw",
 	.valid_hooks =  RAW_VALID_HOOKS,
-	.lock = __RW_LOCK_UNLOCKED(packet_raw.lock),
 	.me = THIS_MODULE,
 	.af = AF_INET,
 };
net/ipv4/netfilter/iptable_security.c +0 -1
index a52a35f..29ab630 100644
@@ -60,7 +60,6 @@ static struct
 static struct xt_table security_table = {
 	.name		= "security",
 	.valid_hooks	= SECURITY_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(security_table.lock),
 	.me		= THIS_MODULE,
 	.af		= AF_INET,
 };
net/ipv4/netfilter/nf_nat_rule.c +0 -1
index a7eb047..6348a79 100644
@@ -61,7 +61,6 @@ static struct
 static struct xt_table nat_table = {
 	.name		= "nat",
 	.valid_hooks	= NAT_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(nat_table.lock),
 	.me		= THIS_MODULE,
 	.af		= AF_INET,
 };
net/ipv6/netfilter/ip6table_filter.c +0 -1
index 40d2e36..ef5a0a3 100644
@@ -54,7 +54,6 @@ static struct
 static struct xt_table packet_filter = {
 	.name		= "filter",
 	.valid_hooks	= FILTER_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(packet_filter.lock),
 	.me		= THIS_MODULE,
 	.af		= AF_INET6,
 };
net/ipv6/netfilter/ip6table_mangle.c +0 -1
index d0b31b2..ab0d398 100644
@@ -60,7 +60,6 @@ static struct
 static struct xt_table packet_mangler = {
 	.name		= "mangle",
 	.valid_hooks	= MANGLE_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(packet_mangler.lock),
 	.me		= THIS_MODULE,
 	.af		= AF_INET6,
 };
net/ipv6/netfilter/ip6table_raw.c +0 -1
index 109fab6..4b792b6 100644
@@ -38,7 +38,6 @@ static struct
 static struct xt_table packet_raw = {
 	.name = "raw",
 	.valid_hooks = RAW_VALID_HOOKS,
-	.lock = __RW_LOCK_UNLOCKED(packet_raw.lock),
 	.me = THIS_MODULE,
 	.af = AF_INET6,
 };
net/ipv6/netfilter/ip6table_security.c +0 -1
index 20bc52f..0ea37ff 100644
@@ -59,7 +59,6 @@ static struct
 static struct xt_table security_table = {
 	.name		= "security",
 	.valid_hooks	= SECURITY_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(security_table.lock),
 	.me		= THIS_MODULE,
 	.af		= AF_INET6,
 };
#5 lore raw

netfilter 05/41: ebtables: remove unneeded initializations

In reply to Patrick McHardy

commit 842bff366b536787b88c07cbf2416e2cb26cae67

Author: Stephen Hemminger <[email protected]>
Date:   Wed Feb 18 16:30:38 2009 +0100

netfilter: ebtables: remove unneeded initializations

The initialization of the lock element is not needed since the lock is always initialized in ebt_register_table.

net/bridge/netfilter/ebtable_broute.c +0 -1
index 8604dfc..c751111 100644
@@ -46,7 +46,6 @@ static struct ebt_table broute_table =
 	.name		= "broute",
 	.table		= &initial_table,
 	.valid_hooks	= 1 << NF_BR_BROUTING,
-	.lock		= __RW_LOCK_UNLOCKED(broute_table.lock),
 	.check		= check,
 	.me		= THIS_MODULE,
 };
net/bridge/netfilter/ebtable_filter.c +0 -1
index 2b2e804..a5eea72 100644
@@ -55,7 +55,6 @@ static struct ebt_table frame_filter =
 	.name		= "filter",
 	.table		= &initial_table,
 	.valid_hooks	= FILTER_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(frame_filter.lock),
 	.check		= check,
 	.me		= THIS_MODULE,
 };
net/bridge/netfilter/ebtable_nat.c +0 -1
index 3fe1ae8..6024c55 100644
@@ -55,7 +55,6 @@ static struct ebt_table frame_nat =
 	.name		= "nat",
 	.table		= &initial_table,
 	.valid_hooks	= NAT_VALID_HOOKS,
-	.lock		= __RW_LOCK_UNLOCKED(frame_nat.lock),
 	.check		= check,
 	.me		= THIS_MODULE,
 };
#6 lore raw

netfilter 06/41: log invalid new icmpv6 packet with nf_log_packet()

In reply to Patrick McHardy

commit 55df4ac0c927c7f1f84e6d75532f0ca45d391e64

Author: Eric Leblond <[email protected]>
Date:   Wed Feb 18 16:30:56 2009 +0100

netfilter: log invalid new icmpv6 packet with nf_log_packet()

This patch adds a logging message for invalid new icmpv6 packet.

net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c +4 -0
index c323643..165b256 100644
@@ -126,6 +126,10 @@ static bool icmpv6_new(struct nf_conn *ct, const struct sk_buff *skb,
 		pr_debug("icmpv6: can't create new conn with type %u\n",
 			 type + 128);
 		nf_ct_dump_tuple_ipv6(&ct->tuplehash[0].tuple);
+		if (LOG_INVALID(nf_ct_net(ct), IPPROTO_ICMPV6))
+			nf_log_packet(PF_INET6, 0, skb, NULL, NULL, NULL,
+				      "nf_ct_icmpv6: invalid new with type %d ",
+				      type + 128);
 		return false;
 	}
 	atomic_set(&ct->proto.icmp.count, 0);
#7 lore raw

netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

In reply to Patrick McHardy

commit ddc214c43a923e89741e04da2f10e3037a64e222

Author: Eric Dumazet <[email protected]>
Date:   Wed Feb 18 17:47:50 2009 +0100

netfilter: arp_tables: unfold two critical loops in arp_packet_match()

x86 and powerpc can perform long word accesses in an efficient maner. We can use this to unroll two loops in arp_packet_match(), to perform arithmetic on long words instead of bytes. This is a win on x86_64 for example.

net/ipv4/netfilter/arp_tables.c +34 -10
index 7ea88b6..b5db463 100644
@@ -73,6 +73,36 @@ static inline int arp_devaddr_compare(const struct arpt_devaddr_info *ap,
 	return (ret != 0);
 }
 
+/*
+ * Unfortunatly, _b and _mask are not aligned to an int (or long int)
+ * Some arches dont care, unrolling the loop is a win on them.
+ */
+static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
+{
+#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;
+	const unsigned long *mask = (const unsigned long *)_mask;
+	unsigned long ret;
+
+	ret = (a[0] ^ b[0]) & mask[0];
+	if (IFNAMSIZ > sizeof(unsigned long))
+		ret |= (a[1] ^ b[1]) & mask[1];
+	if (IFNAMSIZ > 2 * sizeof(unsigned long))
+		ret |= (a[2] ^ b[2]) & mask[2];
+	if (IFNAMSIZ > 3 * sizeof(unsigned long))
+		ret |= (a[3] ^ b[3]) & mask[3];
+	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
+#else
+	unsigned long ret = 0;
+	int i;
+
+	for (i = 0; i < IFNAMSIZ; i++)
+		ret |= (_a[i] ^ _b[i]) & _mask[i];
+#endif
+	return ret;
+}
+
 /* Returns whether packet matches rule or not. */
 static inline int arp_packet_match(const struct arphdr *arphdr,
 				   struct net_device *dev,
@@ -83,7 +113,7 @@ static inline int arp_packet_match(const struct arphdr *arphdr,
 	const char *arpptr = (char *)(arphdr + 1);
 	const char *src_devaddr, *tgt_devaddr;
 	__be32 src_ipaddr, tgt_ipaddr;
-	int i, ret;
+	long ret;
 
 #define FWINV(bool, invflg) ((bool) ^ !!(arpinfo->invflags & (invflg)))
 
@@ -156,10 +186,7 @@ static inline int arp_packet_match(const struct arphdr *arphdr,
 	}
 
 	/* Look for ifname matches.  */
-	for (i = 0, ret = 0; i < IFNAMSIZ; i++) {
-		ret |= (indev[i] ^ arpinfo->iniface[i])
-			& arpinfo->iniface_mask[i];
-	}
+	ret = ifname_compare(indev, arpinfo->iniface, arpinfo->iniface_mask);
 
 	if (FWINV(ret != 0, ARPT_INV_VIA_IN)) {
 		dprintf("VIA in mismatch (%s vs %s).%s\n",
@@ -168,10 +195,7 @@ static inline int arp_packet_match(const struct arphdr *arphdr,
 		return 0;
 	}
 
-	for (i = 0, ret = 0; i < IFNAMSIZ; i++) {
-		ret |= (outdev[i] ^ arpinfo->outiface[i])
-			& arpinfo->outiface_mask[i];
-	}
+	ret = ifname_compare(outdev, arpinfo->outiface, arpinfo->outiface_mask);
 
 	if (FWINV(ret != 0, ARPT_INV_VIA_OUT)) {
 		dprintf("VIA out mismatch (%s vs %s).%s\n",
@@ -221,7 +245,7 @@ unsigned int arpt_do_table(struct sk_buff *skb,
 			   const struct net_device *out,
 			   struct xt_table *table)
 {
-	static const char nulldevname[IFNAMSIZ];
+	static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 	unsigned int verdict = NF_DROP;
 	const struct arphdr *arp;
 	bool hotdrop = false;
maintainer
#8 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

From: Patrick McHardy <[email protected]>
Date: Tue, 24 Mar 2009 15:03:16 +0100 (MET)

net/ipv4/netfilter/arp_tables.c static inline int arp_devaddr_compare(const struct arpt_devaddr_info *ap, · line 84

Quoted 9 lines
net/ipv4/netfilter/arp_tables.c +9 -0
+/*
+ * Unfortunatly, _b and _mask are not aligned to an int (or long int)
+ * Some arches dont care, unrolling the loop is a win on them.
+ */
+static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
+{
+#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;

I think we can at least give some help for the platforms which require alignment.

We can, for example, assume 16-bit alignment and thus loop over u16's

#9 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

Quoted 16 lines · David Miller

From: Patrick McHardy <[email protected]>
Date: Tue, 24 Mar 2009 15:03:16 +0100 (MET)

Review

Quoted 9 lines
(quoted diff) +9 -0
+/*
+ * Unfortunatly, _b and _mask are not aligned to an int (or long int)
+ * Some arches dont care, unrolling the loop is a win on them.
+ */
+static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
+{
+#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;

I think we can at least give some help for the platforms which require alignment.

Quoted 2 lines

We can, for example, assume 16-bit alignment and thus loop over u16's

Right. How about this incremental patch ?

Thanks

[PATCH] arp_tables: ifname_compare() can assume 16bit alignment

Arches without efficient unaligned access can still perform a loop assuming 16bit alignment in ifname_compare()

net/ipv4/netfilter/arp_tables.c +6 -2
index 64a7c6c..84b9c17 100644
@@ -76,6 +76,7 @@ static inline int arp_devaddr_compare(const struct arpt_devaddr_info *ap,
 /*
  * Unfortunatly, _b and _mask are not aligned to an int (or long int)
  * Some arches dont care, unrolling the loop is a win on them.
+ * For other arches, we only have a 16bit alignement.
  */
 static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
 {
@@ -95,10 +96,13 @@ static unsigned long ifname_compare(const char *_a, const char *_b, const char *
 	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
 #else
 	unsigned long ret = 0;
+	const u16 *a = (const u16 *)_a;
+	const u16 *b = (const u16 *)_b;
+	const u16 *mask = (const u16 *)_mask;
 	int i;
 
-	for (i = 0; i < IFNAMSIZ; i++)
-		ret |= (_a[i] ^ _b[i]) & _mask[i];
+	for (i = 0; i < IFNAMSIZ/sizeof(u16); i++)
+		ret |= (a[i] ^ b[i]) & mask[i];
 #endif
 	return ret;
 }

--
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

maintainer
#10 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

From: Eric Dumazet <[email protected]>
Date: Tue, 24 Mar 2009 22:06:50 +0100

Quoted 6 lines

[PATCH] arp_tables: ifname_compare() can assume 16bit alignment

Arches without efficient unaligned access can still perform a loop assuming 16bit alignment in ifname_compare()

Looks great, applied, thanks Eric!

#11 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

In reply to Eric Dumazet
Quoted 16 lines · Eric Dumazet
Quoted 13 lines

Review

Quoted 9 lines
(quoted diff) +9 -0
+/*
+ * Unfortunatly, _b and _mask are not aligned to an int (or long int)
+ * Some arches dont care, unrolling the loop is a win on them.
+ */
+static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
+{
+#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;

I think we can at least give some help for the platforms which require alignment.

Quoted 2 lines

We can, for example, assume 16-bit alignment and thus loop over u16's

Quoted 8 lines

Right. How about this incremental patch ?

Thanks

[PATCH] arp_tables: ifname_compare() can assume 16bit alignment

Arches without efficient unaligned access can still perform a loop assuming 16bit alignment in ifname_compare()

Allow me some skepticism, but the code looks pretty much like a standard memcmp.

Quoted 13 lines
	unsigned long ret = 0;
net/ipv4/netfilter/arp_tables.c +3 -0
+	const u16 *a = (const u16 *)_a;
+	const u16 *b = (const u16 *)_b;
+	const u16 *mask = (const u16 *)_mask;
	int i;
net/ipv4/netfilter/arp_tables.c +2 -2
-	for (i = 0; i < IFNAMSIZ; i++)
-		ret |= (_a[i] ^ _b[i]) & _mask[i];
+	for (i = 0; i < IFNAMSIZ/sizeof(u16); i++)
+		ret |= (a[i] ^ b[i]) & mask[i];
#endif
	return ret;
}
maintainer
#12 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

From: Jan Engelhardt <[email protected]>
Date: Tue, 24 Mar 2009 22:17:17 +0100 (CET)

Quoted 22 lines · Eric Dumazet
Quoted 20 lines
Quoted 13 lines

Review

Quoted 9 lines
(quoted diff) +9 -0
+/*
+ * Unfortunatly, _b and _mask are not aligned to an int (or long int)
+ * Some arches dont care, unrolling the loop is a win on them.
+ */
+static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
+{
+#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;

I think we can at least give some help for the platforms which require alignment.

Quoted 2 lines

We can, for example, assume 16-bit alignment and thus loop over u16's

Right. How about this incremental patch ?

Thanks

Quoted 7 lines
Quoted 4 lines

[PATCH] arp_tables: ifname_compare() can assume 16bit alignment

Arches without efficient unaligned access can still perform a loop assuming 16bit alignment in ifname_compare()

Allow me some skepticism, but the code looks pretty much like a standard memcmp.

memcmp() can't make any assumptions about alignment. Whereas we _know_ this thing is exactly 16-bit aligned.

All of the optimized memcmp() implementations look for 32-bit alignment and punt to byte at a time comparison loops if things are not aligned enough.

#13 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

Quoted 7 lines · David Miller
Quoted 6 lines
Quoted 3 lines

Arches without efficient unaligned access can still perform a loop assuming 16bit alignment in ifname_compare()

Allow me some skepticism, but the code looks pretty much like a standard memcmp.

Quoted 6 lines

memcmp() can't make any assumptions about alignment. Whereas we _know_ this thing is exactly 16-bit aligned.

All of the optimized memcmp() implementations look for 32-bit alignment and punt to byte at a time comparison loops if things are not aligned enough.

Yes, I seem to remember glibc doing something like

 if ((addr & 0x03) != 0) {
     // process single bytes (increment addr as you go)
     // until addr & 0x03 == 0.

}

/* optimized loop here. also increases addr */

if ((addr & 0x03) != 0)

     // still bytes left after loop - process on a per-byte basis

Is the cost of testing for non-4-divisibility expensive enough to warrant not usnig memcmp?

Irrespective of all that, I think putting the interface comparison code should be agglomerated in a function/header so that it is replicated across iptables, ip6tables, ebtables, arptables, etc.

maintainer
#14 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

From: Jan Engelhardt <[email protected]>
Date: Tue, 24 Mar 2009 22:23:23 +0100 (CET)

Quoted 2 lines

Is the cost of testing for non-4-divisibility expensive enough to warrant not usnig memcmp?

I think so. This is the fast path of these things.

Quoted 3 lines

Irrespective of all that, I think putting the interface comparison code should be agglomerated in a function/header so that it is replicated across iptables, ip6tables, ebtables, arptables, etc.

Agreed.

#15 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

In reply to Jan Engelhardt
Quoted 22 lines · Jan Engelhardt
Quoted 10 lines · David Miller
Quoted 4 lines
Quoted 2 lines

Arches without efficient unaligned access can still perform a loop assuming 16bit alignment in ifname_compare()

Allow me some skepticism, but the code looks pretty much like a standard memcmp.

memcmp() can't make any assumptions about alignment. Whereas we _know_ this thing is exactly 16-bit aligned.

All of the optimized memcmp() implementations look for 32-bit alignment and punt to byte at a time comparison loops if things are not aligned enough.

Yes, I seem to remember glibc doing something like

 if ((addr & 0x03) != 0) {
     // process single bytes (increment addr as you go)
     // until addr & 0x03 == 0.

}

/* optimized loop here. also increases addr */

if ((addr & 0x03) != 0)

Quoted 8 lines
     // still bytes left after loop - process on a per-byte basis

Is the cost of testing for non-4-divisibility expensive enough to warrant not usnig memcmp?

Irrespective of all that, I think putting the interface comparison code should be agglomerated in a function/header so that it is replicated across iptables, ip6tables, ebtables, arptables, etc.

memcmp() is fine, but how is it solving the masking problem we have ?

Also in the case of arp_tables, _a is long word aligned, while _b and _mask are not.

memcmp() in this case is slower, (and dont handle mask thing)

If you look various ifname_compare(), we have two different implementations.

So yes, a factorization is possible for three ip_tables.c, ip6_tables.c and xt_physdev.c

--
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

#16 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

Quoted 12 lines · Eric Dumazet
Quoted 12 lines
Quoted 6 lines

memcmp() can't make any assumptions about alignment. Whereas we _know_ this thing is exactly 16-bit aligned.

All of the optimized memcmp() implementations look for 32-bit alignment and punt to byte at a time comparison loops if things are not aligned enough.

Yes, I seem to remember glibc doing something like [...]
Is the cost of testing for non-4-divisibility expensive enough to warrant not usnig memcmp?

Quoted 5 lines
Quoted 3 lines

Irrespective of all that, I think putting the interface comparison code should be agglomerated in a function/header so that it is replicated across iptables, ip6tables, ebtables, arptables, etc.

memcmp() is fine, but how is it solving the masking problem we have ?

You are right; we would have to calculate the prefix length of the mask first. (And I think we can assume that there will not be any 1s in the mask after the first 0.) It would consume CPU indeed.

Quoted 7 lines

Also in the case of arp_tables, _a is long word aligned, while _b and _mask are not.

If you look various ifname_compare(), we have two different implementations.

So yes, a factorization is possible for three ip_tables.c, ip6_tables.c and xt_physdev.c.

Very well.

#17 lore raw

PATCH 1/1 netfilter: factorize ifname_compare()

Quoted 10 lines · Jan Engelhardt
Quoted 1 line · Eric Dumazet

memcmp() is fine, but how is it solving the masking problem we have ?

You are right; we would have to calculate the prefix length of the mask first. (And I think we can assume that there will not be any 1s in the mask after the first 0.) It would consume CPU indeed.

Quoted 3 lines

Also in the case of arp_tables, _a is long word aligned, while _b and _mask are not.

Quoted 6 lines
Quoted 4 lines

If you look various ifname_compare(), we have two different implementations.

So yes, a factorization is possible for three ip_tables.c, ip6_tables.c and xt_physdev.c.

Very well.

Here is a patch to factorize these functions, as suggested by Jan

Thank you

[PATCH] netfilter: factorize ifname_compare()

We use same not trivial helper function in four places. We can factorize it.

---
 include/linux/netfilter/x_tables.h |   23 +++++++++++++++++++++++
 net/ipv4/netfilter/arp_tables.c    |   14 +-------------
 net/ipv4/netfilter/ip_tables.c     |   23 ++---------------------
 net/ipv6/netfilter/ip6_tables.c    |   23 ++---------------------
 net/netfilter/xt_physdev.c         |   21 ++-------------------
 5 files changed, 30 insertions(+), 74 deletions(-)
include/linux/netfilter/x_tables.h +23 -0
index e8e08d0..72918b7 100644
@@ -435,6 +435,29 @@ extern void xt_free_table_info(struct xt_table_info *info);
 extern void xt_table_entry_swap_rcu(struct xt_table_info *old,
 				    struct xt_table_info *new);
 
+/*
+ * This helper is performance critical and must be inlined
+ */
+static inline unsigned long ifname_compare_aligned(const char *_a,
+						   const char *_b,
+						   const char *_mask)
+{
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;
+	const unsigned long *mask = (const unsigned long *)_mask;
+	unsigned long ret;
+
+	ret = (a[0] ^ b[0]) & mask[0];
+	if (IFNAMSIZ > sizeof(unsigned long))
+		ret |= (a[1] ^ b[1]) & mask[1];
+	if (IFNAMSIZ > 2 * sizeof(unsigned long))
+		ret |= (a[2] ^ b[2]) & mask[2];
+	if (IFNAMSIZ > 3 * sizeof(unsigned long))
+		ret |= (a[3] ^ b[3]) & mask[3];
+	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
+	return ret;
+}
+
 #ifdef CONFIG_COMPAT
 #include <net/compat.h>
 
net/ipv4/netfilter/arp_tables.c +1 -13
index 84b9c17..ee1133c 100644
@@ -81,19 +81,7 @@ static inline int arp_devaddr_compare(const struct arpt_devaddr_info *ap,
 static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
 {
 #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
-	const unsigned long *a = (const unsigned long *)_a;
-	const unsigned long *b = (const unsigned long *)_b;
-	const unsigned long *mask = (const unsigned long *)_mask;
-	unsigned long ret;
-
-	ret = (a[0] ^ b[0]) & mask[0];
-	if (IFNAMSIZ > sizeof(unsigned long))
-		ret |= (a[1] ^ b[1]) & mask[1];
-	if (IFNAMSIZ > 2 * sizeof(unsigned long))
-		ret |= (a[2] ^ b[2]) & mask[2];
-	if (IFNAMSIZ > 3 * sizeof(unsigned long))
-		ret |= (a[3] ^ b[3]) & mask[3];
-	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
+	unsigned long ret = ifname_compare_aligned(_a, _b, _mask);
 #else
 	unsigned long ret = 0;
 	const u16 *a = (const u16 *)_a;
net/ipv4/netfilter/ip_tables.c +2 -21
index e5294ae..41c59e3 100644
@@ -74,25 +74,6 @@ do {								\
 
    Hence the start of any table is given by get_table() below.  */
 
-static unsigned long ifname_compare(const char *_a, const char *_b,
-				    const unsigned char *_mask)
-{
-	const unsigned long *a = (const unsigned long *)_a;
-	const unsigned long *b = (const unsigned long *)_b;
-	const unsigned long *mask = (const unsigned long *)_mask;
-	unsigned long ret;
-
-	ret = (a[0] ^ b[0]) & mask[0];
-	if (IFNAMSIZ > sizeof(unsigned long))
-		ret |= (a[1] ^ b[1]) & mask[1];
-	if (IFNAMSIZ > 2 * sizeof(unsigned long))
-		ret |= (a[2] ^ b[2]) & mask[2];
-	if (IFNAMSIZ > 3 * sizeof(unsigned long))
-		ret |= (a[3] ^ b[3]) & mask[3];
-	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
-	return ret;
-}
-
 /* Returns whether matches rule or not. */
 /* Performance critical - called for every packet */
 static inline bool
@@ -121,7 +102,7 @@ ip_packet_match(const struct iphdr *ip,
 		return false;
 	}
 
-	ret = ifname_compare(indev, ipinfo->iniface, ipinfo->iniface_mask);
+	ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 
 	if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 		dprintf("VIA in mismatch (%s vs %s).%s\n",
@@ -130,7 +111,7 @@ ip_packet_match(const struct iphdr *ip,
 		return false;
 	}
 
-	ret = ifname_compare(outdev, ipinfo->outiface, ipinfo->outiface_mask);
+	ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 
 	if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 		dprintf("VIA out mismatch (%s vs %s).%s\n",
net/ipv6/netfilter/ip6_tables.c +2 -21
index 34af7bb..e59662b 100644
@@ -89,25 +89,6 @@ ip6t_ext_hdr(u8 nexthdr)
 		 (nexthdr == IPPROTO_DSTOPTS) );
 }
 
-static unsigned long ifname_compare(const char *_a, const char *_b,
-				    const unsigned char *_mask)
-{
-	const unsigned long *a = (const unsigned long *)_a;
-	const unsigned long *b = (const unsigned long *)_b;
-	const unsigned long *mask = (const unsigned long *)_mask;
-	unsigned long ret;
-
-	ret = (a[0] ^ b[0]) & mask[0];
-	if (IFNAMSIZ > sizeof(unsigned long))
-		ret |= (a[1] ^ b[1]) & mask[1];
-	if (IFNAMSIZ > 2 * sizeof(unsigned long))
-		ret |= (a[2] ^ b[2]) & mask[2];
-	if (IFNAMSIZ > 3 * sizeof(unsigned long))
-		ret |= (a[3] ^ b[3]) & mask[3];
-	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
-	return ret;
-}
-
 /* Returns whether matches rule or not. */
 /* Performance critical - called for every packet */
 static inline bool
@@ -138,7 +119,7 @@ ip6_packet_match(const struct sk_buff *skb,
 		return false;
 	}
 
-	ret = ifname_compare(indev, ip6info->iniface, ip6info->iniface_mask);
+	ret = ifname_compare_aligned(indev, ip6info->iniface, ip6info->iniface_mask);
 
 	if (FWINV(ret != 0, IP6T_INV_VIA_IN)) {
 		dprintf("VIA in mismatch (%s vs %s).%s\n",
@@ -147,7 +128,7 @@ ip6_packet_match(const struct sk_buff *skb,
 		return false;
 	}
 
-	ret = ifname_compare(outdev, ip6info->outiface, ip6info->outiface_mask);
+	ret = ifname_compare_aligned(outdev, ip6info->outiface, ip6info->outiface_mask);
 
 	if (FWINV(ret != 0, IP6T_INV_VIA_OUT)) {
 		dprintf("VIA out mismatch (%s vs %s).%s\n",
net/netfilter/xt_physdev.c +2 -19
index 44a234e..8d28ca5 100644
@@ -20,23 +20,6 @@ MODULE_DESCRIPTION("Xtables: Bridge physical device match");
 MODULE_ALIAS("ipt_physdev");
 MODULE_ALIAS("ip6t_physdev");
 
-static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
-{
-	const unsigned long *a = (const unsigned long *)_a;
-	const unsigned long *b = (const unsigned long *)_b;
-	const unsigned long *mask = (const unsigned long *)_mask;
-	unsigned long ret;
-
-	ret = (a[0] ^ b[0]) & mask[0];
-	if (IFNAMSIZ > sizeof(unsigned long))
-		ret |= (a[1] ^ b[1]) & mask[1];
-	if (IFNAMSIZ > 2 * sizeof(unsigned long))
-		ret |= (a[2] ^ b[2]) & mask[2];
-	if (IFNAMSIZ > 3 * sizeof(unsigned long))
-		ret |= (a[3] ^ b[3]) & mask[3];
-	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
-	return ret;
-}
 
 static bool
 physdev_mt(const struct sk_buff *skb, const struct xt_match_param *par)
@@ -85,7 +68,7 @@ physdev_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 	if (!(info->bitmask & XT_PHYSDEV_OP_IN))
 		goto match_outdev;
 	indev = nf_bridge->physindev ? nf_bridge->physindev->name : nulldevname;
-	ret = ifname_compare(indev, info->physindev, info->in_mask);
+	ret = ifname_compare_aligned(indev, info->physindev, info->in_mask);
 
 	if (!ret ^ !(info->invert & XT_PHYSDEV_OP_IN))
 		return false;
@@ -95,7 +78,7 @@ match_outdev:
 		return true;
 	outdev = nf_bridge->physoutdev ?
 		 nf_bridge->physoutdev->name : nulldevname;
-	ret = ifname_compare(outdev, info->physoutdev, info->out_mask);
+	ret = ifname_compare_aligned(outdev, info->physoutdev, info->out_mask);
 
 	return (!!ret ^ !(info->invert & XT_PHYSDEV_OP_OUT));
 }
author Applied
#18 lore raw

Re: [PATCH] netfilter: factorize ifname_compare()

Quoted 3 lines · Eric Dumazet

[PATCH] netfilter: factorize ifname_compare()

We use same not trivial helper function in four places. We can factorize it.

Applied, thanks a lot Eric.

#19 lore raw

Re: netfilter 07/41: arp_tables: unfold two critical loops in arp_packet_match()

In reply to David Miller
Quoted 2 lines · David Miller

memcmp() can't make any assumptions about alignment.

Newer gcc often[1] knows how to generate specialized aligned memcmp() based on the type alignment.

However you need to make sure the input types are right.

So in theories some casts might be enough given a new enough compiler.

[1] not always unfortunately, sometimes it loses this information.

-Andi

Signature · 1 line
[email protected] -- Speaking for myself only.
#20 lore raw

netfilter 08/41: Combine ipt_TTL and ip6t_HL source

In reply to Patrick McHardy

commit 563d36eb3fb22dd04da9aa6f12e1b9ba0ac115f3

Author: Jan Engelhardt <[email protected]>
Date:   Wed Feb 18 18:38:40 2009 +0100

netfilter: Combine ipt_TTL and ip6t_HL source

Suggested by: James King <[email protected]>

Similarly to commit c9fd49680954714473d6cbd2546d6ff120f96840, merge TTL and HL. Since HL does not depend on any IPv6-specific function, no new module dependencies would arise.

With slight adjustments to the Kconfig help text.

net/ipv4/netfilter/Kconfig +0 -15
index 3816e1d..3ad9f43 100644
@@ -322,21 +322,6 @@ config IP_NF_TARGET_ECN
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
-config IP_NF_TARGET_TTL
-	tristate  'TTL target support'
-	depends on IP_NF_MANGLE
-	depends on NETFILTER_ADVANCED
-	help
-	  This option adds a `TTL' target, which enables the user to modify
-	  the TTL value of the IP header.
-
-	  While it is safe to decrement/lower the TTL, this target also enables
-	  functionality to increment and set the TTL value of the IP header to
-	  arbitrary values.  This is EXTREMELY DANGEROUS since you can easily
-	  create immortal packets that loop forever on the network.
-
-	  To compile it as a module, choose M here.  If unsure, say N.
-
 # raw + specific targets
 config IP_NF_RAW
 	tristate  'raw table support (required for NOTRACK/TRACE)'
net/ipv4/netfilter/Makefile +0 -1
index 5f9b650..20b0c37 100644
@@ -61,7 +61,6 @@ obj-$(CONFIG_IP_NF_TARGET_MASQUERADE) += ipt_MASQUERADE.o
 obj-$(CONFIG_IP_NF_TARGET_NETMAP) += ipt_NETMAP.o
 obj-$(CONFIG_IP_NF_TARGET_REDIRECT) += ipt_REDIRECT.o
 obj-$(CONFIG_IP_NF_TARGET_REJECT) += ipt_REJECT.o
-obj-$(CONFIG_IP_NF_TARGET_TTL) += ipt_TTL.o
 obj-$(CONFIG_IP_NF_TARGET_ULOG) += ipt_ULOG.o
 
 # generic ARP tables
net/ipv4/netfilter/ipt_TTL.c deleted +0 -97

97 lines not on this page · raw message

net/ipv6/netfilter/Kconfig +0 -17
index 53ea512..6a42a96 100644
@@ -170,23 +170,6 @@ config IP6_NF_MANGLE
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
-config IP6_NF_TARGET_HL
-	tristate  'HL (hoplimit) target support'
-	depends on IP6_NF_MANGLE
-	depends on NETFILTER_ADVANCED
-	help
-	  This option adds a `HL' target, which enables the user to decrement
-	  the hoplimit value of the IPv6 header or set it to a given (lower)
-	  value.
-
-	  While it is safe to decrement the hoplimit value, this option also
-	  enables functionality to increment and set the hoplimit value of the
-	  IPv6 header to arbitrary values.  This is EXTREMELY DANGEROUS since
-	  you can easily create immortal packets that loop forever on the
-	  network.
-
-	  To compile it as a module, choose M here.  If unsure, say N.
-
 config IP6_NF_RAW
 	tristate  'raw table support (required for TRACE)'
 	depends on NETFILTER_ADVANCED
net/ipv6/netfilter/Makefile +0 -1
index 3f17c94..61a4570 100644
@@ -27,6 +27,5 @@ obj-$(CONFIG_IP6_NF_MATCH_OPTS) += ip6t_hbh.o
 obj-$(CONFIG_IP6_NF_MATCH_RT) += ip6t_rt.o
 
 # targets
-obj-$(CONFIG_IP6_NF_TARGET_HL) += ip6t_HL.o
 obj-$(CONFIG_IP6_NF_TARGET_LOG) += ip6t_LOG.o
 obj-$(CONFIG_IP6_NF_TARGET_REJECT) += ip6t_REJECT.o
net/ipv6/netfilter/ip6t_HL.c deleted +0 -95

95 lines not on this page · raw message

net/netfilter/Kconfig +15 -0
index c2bac9c..d99f29b 100644
@@ -357,6 +357,21 @@ config NETFILTER_XT_TARGET_DSCP
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
+config NETFILTER_XT_TARGET_HL
+	tristate '"HL" hoplimit target support'
+	depends on IP_NF_MANGLE || IP6_NF_MANGLE
+	depends on NETFILTER_ADVANCED
+	---help---
+	This option adds the "HL" (for IPv6) and "TTL" (for IPv4)
+	targets, which enable the user to change the
+	hoplimit/time-to-live value of the IP header.
+
+	While it is safe to decrement the hoplimit/TTL value, the
+	modules also allow to increment and set the hoplimit value of
+	the header to arbitrary values. This is EXTREMELY DANGEROUS
+	since you can easily create immortal packets that loop
+	forever on the network.
+
 config NETFILTER_XT_TARGET_MARK
 	tristate '"MARK" target support'
 	default m if NETFILTER_ADVANCED=n
net/netfilter/Makefile +1 -0
index da3d909..6ebe048 100644
@@ -45,6 +45,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_CLASSIFY) += xt_CLASSIFY.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_CONNMARK) += xt_CONNMARK.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_CONNSECMARK) += xt_CONNSECMARK.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_DSCP) += xt_DSCP.o
+obj-$(CONFIG_NETFILTER_XT_TARGET_HL) += xt_HL.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_MARK) += xt_MARK.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_NFLOG) += xt_NFLOG.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_NFQUEUE) += xt_NFQUEUE.o
net/netfilter/xt_HL.c new file +171 -0

171 lines not on this page · raw message

#21 lore raw

netfilter 09/41: Combine ipt_ttl and ip6t_hl source

In reply to Patrick McHardy

commit cfac5ef7b92a2d504563989ecd0beb563920444b

Author: Jan Engelhardt <[email protected]>
Date:   Wed Feb 18 18:39:31 2009 +0100

netfilter: Combine ipt_ttl and ip6t_hl source

net/ipv4/netfilter/Kconfig +0 -9
index 3ad9f43..40ad41f 100644
@@ -92,15 +92,6 @@ config IP_NF_MATCH_ECN
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
-config IP_NF_MATCH_TTL
-	tristate '"ttl" match support'
-	depends on NETFILTER_ADVANCED
-	help
-	  This adds CONFIG_IP_NF_MATCH_TTL option, which enabled the user
-	  to match packets by their TTL value.
-
-	  To compile it as a module, choose M here.  If unsure, say N.
-
 # `filter', generic and specific targets
 config IP_NF_FILTER
 	tristate "Packet filtering"
net/ipv4/netfilter/Makefile +0 -1
index 20b0c37..4811159 100644
@@ -51,7 +51,6 @@ obj-$(CONFIG_IP_NF_SECURITY) += iptable_security.o
 obj-$(CONFIG_IP_NF_MATCH_ADDRTYPE) += ipt_addrtype.o
 obj-$(CONFIG_IP_NF_MATCH_AH) += ipt_ah.o
 obj-$(CONFIG_IP_NF_MATCH_ECN) += ipt_ecn.o
-obj-$(CONFIG_IP_NF_MATCH_TTL) += ipt_ttl.o
 
 # targets
 obj-$(CONFIG_IP_NF_TARGET_CLUSTERIP) += ipt_CLUSTERIP.o
net/ipv4/netfilter/ipt_ttl.c deleted +0 -63

63 lines not on this page · raw message

net/ipv6/netfilter/Kconfig +0 -9
index 6a42a96..4a8d7ec 100644
@@ -94,15 +94,6 @@ config IP6_NF_MATCH_OPTS
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
-config IP6_NF_MATCH_HL
-	tristate '"hl" match support'
-	depends on NETFILTER_ADVANCED
-	help
-	  HL matching allows you to match packets based on the hop
-	  limit of the packet.
-
-	  To compile it as a module, choose M here.  If unsure, say N.
-
 config IP6_NF_MATCH_IPV6HEADER
 	tristate '"ipv6header" IPv6 Extension Headers Match'
 	default m if NETFILTER_ADVANCED=n
net/ipv6/netfilter/Makefile +0 -1
index 61a4570..aafbba3 100644
@@ -20,7 +20,6 @@ obj-$(CONFIG_NF_CONNTRACK_IPV6) += nf_conntrack_ipv6.o
 obj-$(CONFIG_IP6_NF_MATCH_AH) += ip6t_ah.o
 obj-$(CONFIG_IP6_NF_MATCH_EUI64) += ip6t_eui64.o
 obj-$(CONFIG_IP6_NF_MATCH_FRAG) += ip6t_frag.o
-obj-$(CONFIG_IP6_NF_MATCH_HL) += ip6t_hl.o
 obj-$(CONFIG_IP6_NF_MATCH_IPV6HEADER) += ip6t_ipv6header.o
 obj-$(CONFIG_IP6_NF_MATCH_MH) += ip6t_mh.o
 obj-$(CONFIG_IP6_NF_MATCH_OPTS) += ip6t_hbh.o
net/ipv6/netfilter/ip6t_hl.c deleted +0 -68

68 lines not on this page · raw message

net/netfilter/Kconfig +8 -0
index d99f29b..0eb98b4 100644
@@ -620,6 +620,14 @@ config NETFILTER_XT_MATCH_HELPER
 
 	  To compile it as a module, choose M here.  If unsure, say Y.
 
+config NETFILTER_XT_MATCH_HL
+	tristate '"hl" hoplimit/TTL match support'
+	depends on NETFILTER_ADVANCED
+	---help---
+	HL matching allows you to match packets based on the hoplimit
+	in the IPv6 header, or the time-to-live field in the IPv4
+	header of the packet.
+
 config NETFILTER_XT_MATCH_IPRANGE
 	tristate '"iprange" address range match support'
 	depends on NETFILTER_ADVANCED
net/netfilter/Makefile +1 -0
index 6ebe048..da73ed2 100644
@@ -68,6 +68,7 @@ obj-$(CONFIG_NETFILTER_XT_MATCH_DSCP) += xt_dscp.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_ESP) += xt_esp.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_HASHLIMIT) += xt_hashlimit.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_HELPER) += xt_helper.o
+obj-$(CONFIG_NETFILTER_XT_MATCH_HL) += xt_hl.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_IPRANGE) += xt_iprange.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_LENGTH) += xt_length.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_LIMIT) += xt_limit.o
net/netfilter/xt_hl.c new file +108 -0

108 lines not on this page · raw message

#22 lore raw

netfilter 10/41: xt_physdev fixes

In reply to Patrick McHardy

commit 4f1c3b7e7ee4d841c8af3a074dc361d6a7a77803

Author: Eric Dumazet <[email protected]>
Date:   Wed Feb 18 19:11:39 2009 +0100

netfilter: xt_physdev fixes

1) physdev_mt() incorrectly assumes nulldevname[] is aligned on an int

2) It also uses word comparisons, while it could use long word ones.

net/netfilter/xt_physdev.c +11 -12
index 1bcdfc1..4b13ef7 100644
@@ -24,9 +24,9 @@ static bool
 physdev_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 {
 	int i;
-	static const char nulldevname[IFNAMSIZ];
+	static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 	const struct xt_physdev_info *info = par->matchinfo;
-	bool ret;
+	unsigned long ret;
 	const char *indev, *outdev;
 	const struct nf_bridge_info *nf_bridge;
 
@@ -68,10 +68,10 @@ physdev_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 	if (!(info->bitmask & XT_PHYSDEV_OP_IN))
 		goto match_outdev;
 	indev = nf_bridge->physindev ? nf_bridge->physindev->name : nulldevname;
-	for (i = 0, ret = false; i < IFNAMSIZ/sizeof(unsigned int); i++) {
-		ret |= (((const unsigned int *)indev)[i]
-			^ ((const unsigned int *)info->physindev)[i])
-			& ((const unsigned int *)info->in_mask)[i];
+	for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
+		ret |= (((const unsigned long *)indev)[i]
+			^ ((const unsigned long *)info->physindev)[i])
+			& ((const unsigned long *)info->in_mask)[i];
 	}
 
 	if (!ret ^ !(info->invert & XT_PHYSDEV_OP_IN))
@@ -82,13 +82,12 @@ match_outdev:
 		return true;
 	outdev = nf_bridge->physoutdev ?
 		 nf_bridge->physoutdev->name : nulldevname;
-	for (i = 0, ret = false; i < IFNAMSIZ/sizeof(unsigned int); i++) {
-		ret |= (((const unsigned int *)outdev)[i]
-			^ ((const unsigned int *)info->physoutdev)[i])
-			& ((const unsigned int *)info->out_mask)[i];
+	for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
+		ret |= (((const unsigned long *)outdev)[i]
+			^ ((const unsigned long *)info->physoutdev)[i])
+			& ((const unsigned long *)info->out_mask)[i];
 	}
-
-	return ret ^ !(info->invert & XT_PHYSDEV_OP_OUT);
+	return (!!ret ^ !(info->invert & XT_PHYSDEV_OP_OUT));
 }
 
 static bool physdev_mt_check(const struct xt_mtchk_param *par)
#23 lore raw

netfilter 11/41: xtables: add backward-compat options

In reply to Patrick McHardy

commit 4323362e49bd10b8ff3fe5cf183fdd52662ff4a3

Author: Jan Engelhardt <[email protected]>
Date:   Thu Feb 19 11:16:03 2009 +0100

netfilter: xtables: add backward-compat options

Concern has been expressed about the changing Kconfig options. Provide the old options that forward-select.

net/ipv4/netfilter/Kconfig +18 -0
index 40ad41f..f8d6180 100644
@@ -92,6 +92,15 @@ config IP_NF_MATCH_ECN
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
+config IP_NF_MATCH_TTL
+	tristate '"ttl" match support'
+	depends on NETFILTER_ADVANCED
+	select NETFILTER_XT_MATCH_HL
+	---help---
+	This is a backwards-compat option for the user's convenience
+	(e.g. when running oldconfig). It selects
+	COFNIG_NETFILTER_XT_MATCH_HL.
+
 # `filter', generic and specific targets
 config IP_NF_FILTER
 	tristate "Packet filtering"
@@ -313,6 +322,15 @@ config IP_NF_TARGET_ECN
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
+config IP_NF_TARGET_TTL
+	tristate '"TTL" target support'
+	depends on NETFILTER_ADVANCED
+	select NETFILTER_XT_TARGET_HL
+	---help---
+	This is a backwards-compat option for the user's convenience
+	(e.g. when running oldconfig). It selects
+	COFNIG_NETFILTER_XT_TARGET_HL.
+
 # raw + specific targets
 config IP_NF_RAW
 	tristate  'raw table support (required for NOTRACK/TRACE)'
net/ipv6/netfilter/Kconfig +18 -0
index 4a8d7ec..625353a 100644
@@ -94,6 +94,15 @@ config IP6_NF_MATCH_OPTS
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
+config IP6_NF_MATCH_HL
+	tristate '"hl" hoplimit match support'
+	depends on NETFILTER_ADVANCED
+	select NETFILTER_XT_MATCH_HL
+	---help---
+	This is a backwards-compat option for the user's convenience
+	(e.g. when running oldconfig). It selects
+	COFNIG_NETFILTER_XT_MATCH_HL.
+
 config IP6_NF_MATCH_IPV6HEADER
 	tristate '"ipv6header" IPv6 Extension Headers Match'
 	default m if NETFILTER_ADVANCED=n
@@ -121,6 +130,15 @@ config IP6_NF_MATCH_RT
 	  To compile it as a module, choose M here.  If unsure, say N.
 
 # The targets
+config IP6_NF_TARGET_HL
+	tristate '"HL" hoplimit target support'
+	depends on NETFILTER_ADVANCED
+	select NETFILTER_XT_TARGET_HL
+	---help---
+	This is a backwards-compat option for the user's convenience
+	(e.g. when running oldconfig). It selects
+	COFNIG_NETFILTER_XT_TARGET_HL.
+
 config IP6_NF_TARGET_LOG
 	tristate "LOG target support"
 	default m if NETFILTER_ADVANCED=n
#24 lore raw

netfilter 12/41: xt_physdev: unfold two loops in physdev_mt()

In reply to Patrick McHardy

commit eacc17fb64f03b6c268aaf6cea320100d19d8af5

Author: Eric Dumazet <[email protected]>
Date:   Thu Feb 19 11:17:17 2009 +0100

netfilter: xt_physdev: unfold two loops in physdev_mt()

xt_physdev netfilter module can use an ifname_compare() helper so that two loops are unfolded.

net/netfilter/xt_physdev.c +21 -11
index 4b13ef7..44a234e 100644
@@ -20,10 +20,27 @@ MODULE_DESCRIPTION("Xtables: Bridge physical device match");
 MODULE_ALIAS("ipt_physdev");
 MODULE_ALIAS("ip6t_physdev");
 
+static unsigned long ifname_compare(const char *_a, const char *_b, const char *_mask)
+{
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;
+	const unsigned long *mask = (const unsigned long *)_mask;
+	unsigned long ret;
+
+	ret = (a[0] ^ b[0]) & mask[0];
+	if (IFNAMSIZ > sizeof(unsigned long))
+		ret |= (a[1] ^ b[1]) & mask[1];
+	if (IFNAMSIZ > 2 * sizeof(unsigned long))
+		ret |= (a[2] ^ b[2]) & mask[2];
+	if (IFNAMSIZ > 3 * sizeof(unsigned long))
+		ret |= (a[3] ^ b[3]) & mask[3];
+	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
+	return ret;
+}
+
 static bool
 physdev_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 {
-	int i;
 	static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 	const struct xt_physdev_info *info = par->matchinfo;
 	unsigned long ret;
@@ -68,11 +85,7 @@ physdev_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 	if (!(info->bitmask & XT_PHYSDEV_OP_IN))
 		goto match_outdev;
 	indev = nf_bridge->physindev ? nf_bridge->physindev->name : nulldevname;
-	for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
-		ret |= (((const unsigned long *)indev)[i]
-			^ ((const unsigned long *)info->physindev)[i])
-			& ((const unsigned long *)info->in_mask)[i];
-	}
+	ret = ifname_compare(indev, info->physindev, info->in_mask);
 
 	if (!ret ^ !(info->invert & XT_PHYSDEV_OP_IN))
 		return false;
@@ -82,11 +95,8 @@ match_outdev:
 		return true;
 	outdev = nf_bridge->physoutdev ?
 		 nf_bridge->physoutdev->name : nulldevname;
-	for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
-		ret |= (((const unsigned long *)outdev)[i]
-			^ ((const unsigned long *)info->physoutdev)[i])
-			& ((const unsigned long *)info->out_mask)[i];
-	}
+	ret = ifname_compare(outdev, info->physoutdev, info->out_mask);
+
 	return (!!ret ^ !(info->invert & XT_PHYSDEV_OP_OUT));
 }
 
#25 lore raw

netfilter 13/41: ip6_tables: unfold two loops in ip6_packet_match()

In reply to Patrick McHardy

commit 323dbf96382f057d035afce0237f08e18571ac1d

Author: Eric Dumazet <[email protected]>
Date:   Thu Feb 19 11:18:23 2009 +0100

netfilter: ip6_tables: unfold two loops in ip6_packet_match()

ip6_tables netfilter module can use an ifname_compare() helper so that two loops are unfolded.

net/ipv6/netfilter/ip6_tables.c +21 -12
index a33485d..d64594b 100644
@@ -89,6 +89,25 @@ ip6t_ext_hdr(u8 nexthdr)
 		 (nexthdr == IPPROTO_DSTOPTS) );
 }
 
+static unsigned long ifname_compare(const char *_a, const char *_b,
+				    const unsigned char *_mask)
+{
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;
+	const unsigned long *mask = (const unsigned long *)_mask;
+	unsigned long ret;
+
+	ret = (a[0] ^ b[0]) & mask[0];
+	if (IFNAMSIZ > sizeof(unsigned long))
+		ret |= (a[1] ^ b[1]) & mask[1];
+	if (IFNAMSIZ > 2 * sizeof(unsigned long))
+		ret |= (a[2] ^ b[2]) & mask[2];
+	if (IFNAMSIZ > 3 * sizeof(unsigned long))
+		ret |= (a[3] ^ b[3]) & mask[3];
+	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
+	return ret;
+}
+
 /* Returns whether matches rule or not. */
 /* Performance critical - called for every packet */
 static inline bool
@@ -99,7 +118,6 @@ ip6_packet_match(const struct sk_buff *skb,
 		 unsigned int *protoff,
 		 int *fragoff, bool *hotdrop)
 {
-	size_t i;
 	unsigned long ret;
 	const struct ipv6hdr *ipv6 = ipv6_hdr(skb);
 
@@ -120,12 +138,7 @@ ip6_packet_match(const struct sk_buff *skb,
 		return false;
 	}
 
-	/* Look for ifname matches; this should unroll nicely. */
-	for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
-		ret |= (((const unsigned long *)indev)[i]
-			^ ((const unsigned long *)ip6info->iniface)[i])
-			& ((const unsigned long *)ip6info->iniface_mask)[i];
-	}
+	ret = ifname_compare(indev, ip6info->iniface, ip6info->iniface_mask);
 
 	if (FWINV(ret != 0, IP6T_INV_VIA_IN)) {
 		dprintf("VIA in mismatch (%s vs %s).%s\n",
@@ -134,11 +147,7 @@ ip6_packet_match(const struct sk_buff *skb,
 		return false;
 	}
 
-	for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
-		ret |= (((const unsigned long *)outdev)[i]
-			^ ((const unsigned long *)ip6info->outiface)[i])
-			& ((const unsigned long *)ip6info->outiface_mask)[i];
-	}
+	ret = ifname_compare(outdev, ip6info->outiface, ip6info->outiface_mask);
 
 	if (FWINV(ret != 0, IP6T_INV_VIA_OUT)) {
 		dprintf("VIA out mismatch (%s vs %s).%s\n",
#26 lore raw

netfilter 14/41: iptables: lock free counters

In reply to Patrick McHardy

commit 784544739a25c30637397ace5489eeb6e15d7d49

Author: Stephen Hemminger <[email protected]>
Date:   Fri Feb 20 10:35:32 2009 +0100

netfilter: iptables: lock free counters

The reader/writer lock in ip_tables is acquired in the critical path of processing packets and is one of the reasons just loading iptables can cause a 20% performance loss. The rwlock serves two functions:

1) it prevents changes to table state (xt_replace) while table is in use. This is now handled by doing rcu on the xt_table. When table is replaced, the new table(s) are put in and the old one table(s) are freed after RCU period.

2) it provides synchronization when accesing the counter values. This is now handled by swapping in new table_info entries for each cpu then summing the old values, and putting the result back onto one cpu. On a busy system it may cause sampling to occur at different times on each cpu, but no packet/byte counts are lost in the process.

Sucessfully tested on my dual quad core machine too, but iptables only (no ipv6 here) BTW, my new "tbench 8" result is 2450 MB/s, (it was 2150 MB/s not so long ago)

include/linux/netfilter/x_tables.h +4 -2
index 9fac88f..e8e08d0 100644
@@ -353,7 +353,7 @@ struct xt_table
 	unsigned int valid_hooks;
 
 	/* Lock for the curtain */
-	rwlock_t lock;
+	struct mutex lock;
 
 	/* Man behind the curtain... */
 	struct xt_table_info *private;
@@ -385,7 +385,7 @@ struct xt_table_info
 
 	/* ipt_entry tables: one per CPU */
 	/* Note : this field MUST be the last one, see XT_TABLE_INFO_SZ */
-	char *entries[1];
+	void *entries[1];
 };
 
 #define XT_TABLE_INFO_SZ (offsetof(struct xt_table_info, entries) \
@@ -432,6 +432,8 @@ extern void xt_proto_fini(struct net *net, u_int8_t af);
 
 extern struct xt_table_info *xt_alloc_table_info(unsigned int size);
 extern void xt_free_table_info(struct xt_table_info *info);
+extern void xt_table_entry_swap_rcu(struct xt_table_info *old,
+				    struct xt_table_info *new);
 
 #ifdef CONFIG_COMPAT
 #include <net/compat.h>
net/ipv4/netfilter/arp_tables.c +88 -27

170 lines not on this page · raw message

net/ipv4/netfilter/ip_tables.c +87 -33

179 lines not on this page · raw message

net/ipv6/netfilter/ip6_tables.c +84 -35

175 lines not on this page · raw message

net/netfilter/x_tables.c +21 -5
index bfbf521..bfcac92 100644
@@ -625,6 +625,20 @@ void xt_free_table_info(struct xt_table_info *info)
 }
 EXPORT_SYMBOL(xt_free_table_info);
 
+void xt_table_entry_swap_rcu(struct xt_table_info *oldinfo,
+			     struct xt_table_info *newinfo)
+{
+	unsigned int cpu;
+
+	for_each_possible_cpu(cpu) {
+		void *p = oldinfo->entries[cpu];
+		rcu_assign_pointer(oldinfo->entries[cpu], newinfo->entries[cpu]);
+		newinfo->entries[cpu] = p;
+	}
+
+}
+EXPORT_SYMBOL_GPL(xt_table_entry_swap_rcu);
+
 /* Find table by name, grabs mutex & ref.  Returns ERR_PTR() on error. */
 struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af,
 				    const char *name)
@@ -671,21 +685,22 @@ xt_replace_table(struct xt_table *table,
 	struct xt_table_info *oldinfo, *private;
 
 	/* Do the substitution. */
-	write_lock_bh(&table->lock);
+	mutex_lock(&table->lock);
 	private = table->private;
 	/* Check inside lock: is the old number correct? */
 	if (num_counters != private->number) {
 		duprintf("num_counters != table->private->number (%u/%u)\n",
 			 num_counters, private->number);
-		write_unlock_bh(&table->lock);
+		mutex_unlock(&table->lock);
 		*error = -EAGAIN;
 		return NULL;
 	}
 	oldinfo = private;
-	table->private = newinfo;
+	rcu_assign_pointer(table->private, newinfo);
 	newinfo->initial_entries = oldinfo->initial_entries;
-	write_unlock_bh(&table->lock);
+	mutex_unlock(&table->lock);
 
+	synchronize_net();
 	return oldinfo;
 }
 EXPORT_SYMBOL_GPL(xt_replace_table);
@@ -719,7 +734,8 @@ struct xt_table *xt_register_table(struct net *net, struct xt_table *table,
 
 	/* Simplifies replace_table code. */
 	table->private = bootstrap;
-	rwlock_init(&table->lock);
+	mutex_init(&table->lock);
+
 	if (!xt_replace_table(table, 0, newinfo, &ret))
 		goto unlock;
 
#27 lore raw

netfilter 15/41: nf_conntrack: table max size should hold at least table size

In reply to Patrick McHardy

commit e478075c6f07a383c378fb400edc1a7407a941b0

Author: Hagen Paul Pfeifer <[email protected]>
Date:   Fri Feb 20 10:47:09 2009 +0100

netfilter: nf_conntrack: table max size should hold at least table size

Table size is defined as unsigned, wheres the table maximum size is defined as a signed integer. The calculation of max is 8 or 4, multiplied the table size. Therefore the max value is aligned to unsigned.

include/net/netfilter/nf_conntrack.h +1 -1
index 2e0c536..4dfb793 100644
@@ -287,7 +287,7 @@ static inline int nf_ct_is_untracked(const struct sk_buff *skb)
 
 extern int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp);
 extern unsigned int nf_conntrack_htable_size;
-extern int nf_conntrack_max;
+extern unsigned int nf_conntrack_max;
 
 #define NF_CT_STAT_INC(net, count)	\
 	(per_cpu_ptr((net)->ct.stat, raw_smp_processor_id())->count++)
net/netfilter/nf_conntrack_core.c +1 -1
index 90ce9dd..f3aa4e6 100644
@@ -54,7 +54,7 @@ EXPORT_SYMBOL_GPL(nf_conntrack_lock);
 unsigned int nf_conntrack_htable_size __read_mostly;
 EXPORT_SYMBOL_GPL(nf_conntrack_htable_size);
 
-int nf_conntrack_max __read_mostly;
+unsigned int nf_conntrack_max __read_mostly;
 EXPORT_SYMBOL_GPL(nf_conntrack_max);
 
 struct nf_conn nf_conntrack_untracked __read_mostly;
#28 lore raw

netfilter 16/41: fix hardcoded size assumptions

In reply to Patrick McHardy

commit af07d241dc76f0a52c7ff04df3a3970020fe6157

Author: Hagen Paul Pfeifer <[email protected]>
Date:   Fri Feb 20 10:48:06 2009 +0100

netfilter: fix hardcoded size assumptions

get_random_bytes() is sometimes called with a hard coded size assumption of an integer. This could not be true for next centuries. This patch replace it with a compile time statement.

net/netfilter/nf_conntrack_core.c +3 -2
index f3aa4e6..2235432 100644
@@ -472,7 +472,8 @@ struct nf_conn *nf_conntrack_alloc(struct net *net,
 	struct nf_conn *ct;
 
 	if (unlikely(!nf_conntrack_hash_rnd_initted)) {
-		get_random_bytes(&nf_conntrack_hash_rnd, 4);
+		get_random_bytes(&nf_conntrack_hash_rnd,
+				sizeof(nf_conntrack_hash_rnd));
 		nf_conntrack_hash_rnd_initted = 1;
 	}
 
@@ -1103,7 +1104,7 @@ int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp)
 
 	/* We have to rehahs for the new table anyway, so we also can
 	 * use a newrandom seed */
-	get_random_bytes(&rnd, 4);
+	get_random_bytes(&rnd, sizeof(rnd));
 
 	/* Lookups in the old hash might happen in parallel, which means we
 	 * might get false negatives during connection lookup. New connections
net/netfilter/nf_conntrack_expect.c +2 -1
index 3a8a34a..357ba39 100644
@@ -72,7 +72,8 @@ static unsigned int nf_ct_expect_dst_hash(const struct nf_conntrack_tuple *tuple
 	unsigned int hash;
 
 	if (unlikely(!nf_ct_expect_hash_rnd_initted)) {
-		get_random_bytes(&nf_ct_expect_hash_rnd, 4);
+		get_random_bytes(&nf_ct_expect_hash_rnd,
+				 sizeof(nf_ct_expect_hash_rnd));
 		nf_ct_expect_hash_rnd_initted = 1;
 	}
 
net/netfilter/xt_hashlimit.c +1 -1
index f97fded..2482055 100644
@@ -149,7 +149,7 @@ dsthash_alloc_init(struct xt_hashlimit_htable *ht,
 	/* initialize hash with random val at the time we allocate
 	 * the first hashtable entry */
 	if (!ht->rnd_initialized) {
-		get_random_bytes(&ht->rnd, 4);
+		get_random_bytes(&ht->rnd, sizeof(ht->rnd));
 		ht->rnd_initialized = 1;
 	}
 
#29 lore raw

netfilter 17/41: x_tables: add LED trigger target

In reply to Patrick McHardy

commit 268cb38e1802db560c73167e643f14a3dcb4b07c

Author: Adam Nielsen <[email protected]>
Date:   Fri Feb 20 10:55:14 2009 +0100

netfilter: x_tables: add LED trigger target

Kernel module providing implementation of LED netfilter target. Each instance of the target appears as a led-trigger device, which can be associated with one or more LEDs in /sys/class/leds/

drivers/leds/Kconfig +3 -0
index 7427136..556aeca 100644
@@ -223,4 +223,7 @@ config LEDS_TRIGGER_DEFAULT_ON
 	  This allows LEDs to be initialised in the ON state.
 	  If unsure, say Y.
 
+comment "iptables trigger is under Netfilter config (LED target)"
+	depends on LEDS_TRIGGERS
+
 endif # NEW_LEDS
include/linux/netfilter/Kbuild +1 -0
index 5a8af87..deeaee5 100644
@@ -7,6 +7,7 @@ header-y += xt_CLASSIFY.h
 header-y += xt_CONNMARK.h
 header-y += xt_CONNSECMARK.h
 header-y += xt_DSCP.h
+header-y += xt_LED.h
 header-y += xt_MARK.h
 header-y += xt_NFLOG.h
 header-y += xt_NFQUEUE.h
include/linux/netfilter/xt_LED.h new file +13 -0
new file mode 100644
index 0000000..4c91a0d
@@ -0,0 +1,13 @@
+#ifndef _XT_LED_H
+#define _XT_LED_H
+
+struct xt_led_info {
+	char id[27];        /* Unique ID for this trigger in the LED class */
+	__u8 always_blink;  /* Blink even if the LED is already on */
+	__u32 delay;        /* Delay until LED is switched off after trigger */
+
+	/* Kernel data used in the module */
+	void *internal_data __attribute__((aligned(8)));
+};
+
+#endif /* _XT_LED_H */
net/netfilter/Kconfig +24 -0
index 0eb98b4..cdbaaff 100644
@@ -372,6 +372,30 @@ config NETFILTER_XT_TARGET_HL
 	since you can easily create immortal packets that loop
 	forever on the network.
 
+config NETFILTER_XT_TARGET_LED
+	tristate '"LED" target support'
+	depends on LEDS_CLASS
+	depends on NETFILTER_ADVANCED
+	help
+	  This option adds a `LED' target, which allows you to blink LEDs in
+	  response to particular packets passing through your machine.
+
+	  This can be used to turn a spare LED into a network activity LED,
+	  which only flashes in response to FTP transfers, for example.  Or
+	  you could have an LED which lights up for a minute or two every time
+	  somebody connects to your machine via SSH.
+
+	  You will need support for the "led" class to make this work.
+
+	  To create an LED trigger for incoming SSH traffic:
+	    iptables -A INPUT -p tcp --dport 22 -j LED --led-trigger-id ssh --led-delay 1000
+
+	  Then attach the new trigger to an LED on your system:
+	    echo netfilter-ssh > /sys/class/leds/<ledname>/trigger
+
+	  For more information on the LEDs available on your system, see
+	  Documentation/leds-class.txt
+
 config NETFILTER_XT_TARGET_MARK
 	tristate '"MARK" target support'
 	default m if NETFILTER_ADVANCED=n
net/netfilter/Makefile +1 -0
index da73ed2..7a9b839 100644
@@ -46,6 +46,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_CONNMARK) += xt_CONNMARK.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_CONNSECMARK) += xt_CONNSECMARK.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_DSCP) += xt_DSCP.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_HL) += xt_HL.o
+obj-$(CONFIG_NETFILTER_XT_TARGET_LED) += xt_LED.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_MARK) += xt_MARK.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_NFLOG) += xt_NFLOG.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_NFQUEUE) += xt_NFQUEUE.o
net/netfilter/xt_LED.c new file +161 -0

161 lines not on this page · raw message

#30 lore raw

netfilter 18/41: ip_tables: unfold two critical loops in ip_packet_match()

In reply to Patrick McHardy

commit 08361aa807ae5e5007cd226ca9e34287512de737

Author: Eric Dumazet <[email protected]>
Date:   Fri Feb 20 11:03:33 2009 +0100

netfilter: ip_tables: unfold two critical loops in ip_packet_match()

While doing oprofile tests I noticed two loops are not properly unrolled by gcc

Using a hand coded unrolled loop provides nice speedup : ipt_do_table credited of 2.52 % of cpu instead of 3.29 % in tbench.

net/ipv4/netfilter/ip_tables.c +21 -12
index 08cde5b..e5294ae 100644
@@ -74,6 +74,25 @@ do {								\
 
    Hence the start of any table is given by get_table() below.  */
 
+static unsigned long ifname_compare(const char *_a, const char *_b,
+				    const unsigned char *_mask)
+{
+	const unsigned long *a = (const unsigned long *)_a;
+	const unsigned long *b = (const unsigned long *)_b;
+	const unsigned long *mask = (const unsigned long *)_mask;
+	unsigned long ret;
+
+	ret = (a[0] ^ b[0]) & mask[0];
+	if (IFNAMSIZ > sizeof(unsigned long))
+		ret |= (a[1] ^ b[1]) & mask[1];
+	if (IFNAMSIZ > 2 * sizeof(unsigned long))
+		ret |= (a[2] ^ b[2]) & mask[2];
+	if (IFNAMSIZ > 3 * sizeof(unsigned long))
+		ret |= (a[3] ^ b[3]) & mask[3];
+	BUILD_BUG_ON(IFNAMSIZ > 4 * sizeof(unsigned long));
+	return ret;
+}
+
 /* Returns whether matches rule or not. */
 /* Performance critical - called for every packet */
 static inline bool
@@ -83,7 +102,6 @@ ip_packet_match(const struct iphdr *ip,
 		const struct ipt_ip *ipinfo,
 		int isfrag)
 {
-	size_t i;
 	unsigned long ret;
 
 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
@@ -103,12 +121,7 @@ ip_packet_match(const struct iphdr *ip,
 		return false;
 	}
 
-	/* Look for ifname matches; this should unroll nicely. */
-	for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
-		ret |= (((const unsigned long *)indev)[i]
-			^ ((const unsigned long *)ipinfo->iniface)[i])
-			& ((const unsigned long *)ipinfo->iniface_mask)[i];
-	}
+	ret = ifname_compare(indev, ipinfo->iniface, ipinfo->iniface_mask);
 
 	if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 		dprintf("VIA in mismatch (%s vs %s).%s\n",
@@ -117,11 +130,7 @@ ip_packet_match(const struct iphdr *ip,
 		return false;
 	}
 
-	for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
-		ret |= (((const unsigned long *)outdev)[i]
-			^ ((const unsigned long *)ipinfo->outiface)[i])
-			& ((const unsigned long *)ipinfo->outiface_mask)[i];
-	}
+	ret = ifname_compare(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 
 	if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 		dprintf("VIA out mismatch (%s vs %s).%s\n",
#31 lore raw

netfilter 19/41: nf_conntrack: account packets drop by tcp_packet()

In reply to Patrick McHardy

commit 7d1e04598e5e92527840b6889fb75b4b30fdd33b

Author: Pablo Neira Ayuso <[email protected]>
Date:   Tue Feb 24 14:48:01 2009 +0100

netfilter: nf_conntrack: account packets drop by tcp_packet()

Since tcp_packet() may return -NF_DROP in two situations, the packet-drop stats must be increased.

net/netfilter/nf_conntrack_core.c +2 -0
index 2235432..ebc2756 100644
@@ -734,6 +734,8 @@ nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum,
 		nf_conntrack_put(skb->nfct);
 		skb->nfct = NULL;
 		NF_CT_STAT_INC_ATOMIC(net, invalid);
+		if (ret == -NF_DROP)
+			NF_CT_STAT_INC_ATOMIC(net, drop);
 		return -ret;
 	}
 
#32 lore raw

netfilter 20/41: install missing headers

In reply to Patrick McHardy

commit d060ffc1840e37100628f520e66600c5ae483b44

Author: Jan Engelhardt <[email protected]>
Date:   Tue Feb 24 15:23:58 2009 +0100

netfilter: install missing headers

iptables imports headers from (the unifdefed headers of a) kernel tree, but some headers happened to not be installed.

include/linux/netfilter/Kbuild +6 -0
index deeaee5..947b47d 100644
@@ -14,8 +14,11 @@ header-y += xt_NFQUEUE.h
 header-y += xt_RATEEST.h
 header-y += xt_SECMARK.h
 header-y += xt_TCPMSS.h
+header-y += xt_TCPOPTSTRIP.h
+header-y += xt_TPROXY.h
 header-y += xt_comment.h
 header-y += xt_connbytes.h
+header-y += xt_connlimit.h
 header-y += xt_connmark.h
 header-y += xt_conntrack.h
 header-y += xt_dccp.h
@@ -31,6 +34,7 @@ header-y += xt_mark.h
 header-y += xt_multiport.h
 header-y += xt_owner.h
 header-y += xt_pkttype.h
+header-y += xt_quota.h
 header-y += xt_rateest.h
 header-y += xt_realm.h
 header-y += xt_recent.h
@@ -40,6 +44,8 @@ header-y += xt_statistic.h
 header-y += xt_string.h
 header-y += xt_tcpmss.h
 header-y += xt_tcpudp.h
+header-y += xt_time.h
+header-y += xt_u32.h
 
 unifdef-y += nf_conntrack_common.h
 unifdef-y += nf_conntrack_ftp.h
include/linux/netfilter_ipv6/Kbuild +1 -0
index 8887a5f..aca4bd1 100644
@@ -11,6 +11,7 @@ header-y += ip6t_length.h
 header-y += ip6t_limit.h
 header-y += ip6t_mac.h
 header-y += ip6t_mark.h
+header-y += ip6t_mh.h
 header-y += ip6t_multiport.h
 header-y += ip6t_opts.h
 header-y += ip6t_owner.h
#33 lore raw

netfilter 21/41: xt_hashlimit fix

In reply to Patrick McHardy

commit 28337ff5438a640afa713d874d076e3a8a9150da

Author: Eric Dumazet <[email protected]>
Date:   Tue Feb 24 15:30:29 2009 +0100

netfilter: xt_hashlimit fix

Commit 784544739a25c30637397ace5489eeb6e15d7d49
(netfilter: iptables: lock free counters) broke xt_hashlimit netfilter module :

This module was storing a pointer inside its xt_hashlimit_info, and this pointer is not relocated when we temporarly switch tables (iptables -L).

This hack is not not needed at all (probably a leftover from ancient time), as each cpu should and can access to its own copy.

net/netfilter/xt_hashlimit.c +1 -4
index 2482055..a5b5369 100644
@@ -565,8 +565,7 @@ hashlimit_init_dst(const struct xt_hashlimit_htable *hinfo,
 static bool
 hashlimit_mt_v0(const struct sk_buff *skb, const struct xt_match_param *par)
 {
-	const struct xt_hashlimit_info *r =
-		((const struct xt_hashlimit_info *)par->matchinfo)->u.master;
+	const struct xt_hashlimit_info *r = par->matchinfo;
 	struct xt_hashlimit_htable *hinfo = r->hinfo;
 	unsigned long now = jiffies;
 	struct dsthash_ent *dh;
@@ -702,8 +701,6 @@ static bool hashlimit_mt_check_v0(const struct xt_mtchk_param *par)
 	}
 	mutex_unlock(&hlimit_mutex);
 
-	/* Ugly hack: For SMP, we only want to use one set */
-	r->u.master = r;
 	return true;
 }
 
#34 lore raw

netfilter 22/41: use a linked list of loggers

In reply to Patrick McHardy

commit ca735b3aaa945626ba65a3e51145bfe4ecd9e222

Author: Eric Leblond <[email protected]>
Date:   Mon Mar 16 14:54:21 2009 +0100

netfilter: use a linked list of loggers

This patch modifies nf_log to use a linked list of loggers for each protocol. This list of loggers is read and write protected with a mutex.

This patch separates registration and binding. To be used as logging module, a module has to register calling nf_log_register() and to bind to a protocol it has to call nf_log_bind_pf().
This patch also converts the logging modules to the new API. For nfnetlink_log, it simply switchs call to register functions to call to bind function and adds a call to nf_log_register() during init. For other modules, it just remove a const flag from the logger structure and replace it with a __read_mostly.

include/net/netfilter/nf_log.h +8 -3
index 7182c06..920997f 100644
@@ -1,6 +1,8 @@
 #ifndef _NF_LOG_H
 #define _NF_LOG_H
 
+#include <linux/netfilter.h>
+
 /* those NF_LOG_* defines and struct nf_loginfo are legacy definitios that will
  * disappear once iptables is replaced with pkttables.  Please DO NOT use them
  * for any new code! */
@@ -40,12 +42,15 @@ struct nf_logger {
 	struct module	*me;
 	nf_logfn 	*logfn;
 	char		*name;
+	struct list_head	list[NFPROTO_NUMPROTO];
 };
 
 /* Function to register/unregister log function. */
-int nf_log_register(u_int8_t pf, const struct nf_logger *logger);
-void nf_log_unregister(const struct nf_logger *logger);
-void nf_log_unregister_pf(u_int8_t pf);
+int nf_log_register(u_int8_t pf, struct nf_logger *logger);
+void nf_log_unregister(struct nf_logger *logger);
+
+int nf_log_bind_pf(u_int8_t pf, const struct nf_logger *logger);
+void nf_log_unbind_pf(u_int8_t pf);
 
 /* Calls the registered backend logging function */
 void nf_log_packet(u_int8_t pf,
net/ipv4/netfilter/ipt_LOG.c +1 -1
index 27a78fb..acc44c6 100644
@@ -464,7 +464,7 @@ static struct xt_target log_tg_reg __read_mostly = {
 	.me		= THIS_MODULE,
 };
 
-static const struct nf_logger ipt_log_logger ={
+static struct nf_logger ipt_log_logger __read_mostly = {
 	.name		= "ipt_LOG",
 	.logfn		= &ipt_log_packet,
 	.me		= THIS_MODULE,
net/ipv4/netfilter/ipt_ULOG.c +1 -1
index 18a2826..d32cc4b 100644
@@ -379,7 +379,7 @@ static struct xt_target ulog_tg_reg __read_mostly = {
 	.me		= THIS_MODULE,
 };
 
-static struct nf_logger ipt_ulog_logger = {
+static struct nf_logger ipt_ulog_logger __read_mostly = {
 	.name		= "ipt_ULOG",
 	.logfn		= ipt_logfn,
 	.me		= THIS_MODULE,
net/ipv6/netfilter/ip6t_LOG.c +1 -1
index 37adf5a..7018cac 100644
@@ -477,7 +477,7 @@ static struct xt_target log_tg6_reg __read_mostly = {
 	.me 		= THIS_MODULE,
 };
 
-static const struct nf_logger ip6t_logger = {
+static struct nf_logger ip6t_logger __read_mostly = {
 	.name		= "ip6t_LOG",
 	.logfn		= &ip6t_log_packet,
 	.me		= THIS_MODULE,
net/netfilter/nf_log.c +60 -30

132 lines not on this page · raw message

net/netfilter/nfnetlink_log.c +13 -5
index fa49dc7..3eae3fc 100644
@@ -691,7 +691,7 @@ nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 	return -ENOTSUPP;
 }
 
-static const struct nf_logger nfulnl_logger = {
+static struct nf_logger nfulnl_logger __read_mostly = {
 	.name	= "nfnetlink_log",
 	.logfn	= &nfulnl_log_packet,
 	.me	= THIS_MODULE,
@@ -723,9 +723,9 @@ nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 		/* Commands without queue context */
 		switch (cmd->command) {
 		case NFULNL_CFG_CMD_PF_BIND:
-			return nf_log_register(pf, &nfulnl_logger);
+			return nf_log_bind_pf(pf, &nfulnl_logger);
 		case NFULNL_CFG_CMD_PF_UNBIND:
-			nf_log_unregister_pf(pf);
+			nf_log_unbind_pf(pf);
 			return 0;
 		}
 	}
@@ -950,17 +950,25 @@ static int __init nfnetlink_log_init(void)
 		goto cleanup_netlink_notifier;
 	}
 
+	status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
+	if (status < 0) {
+		printk(KERN_ERR "log: failed to register logger\n");
+		goto cleanup_subsys;
+	}
+
 #ifdef CONFIG_PROC_FS
 	if (!proc_create("nfnetlink_log", 0440,
 			 proc_net_netfilter, &nful_file_ops))
-		goto cleanup_subsys;
+		goto cleanup_logger;
 #endif
 	return status;
 
 #ifdef CONFIG_PROC_FS
+cleanup_logger:
+	nf_log_unregister(&nfulnl_logger);
+#endif
 cleanup_subsys:
 	nfnetlink_subsys_unregister(&nfulnl_subsys);
-#endif
 cleanup_netlink_notifier:
 	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
 	return status;
#35 lore raw

netfilter 23/41: print the list of register loggers

In reply to Patrick McHardy

commit c7a913cd5535554d6f5d5e1f5ef46c4307cf2afc

Author: Eric Leblond <[email protected]>
Date:   Mon Mar 16 14:55:27 2009 +0100

netfilter: print the list of register loggers

This patch modifies the proc output to add display of registered loggers. The content of /proc/net/netfilter/nf_log is modified. Instead of displaying a protocol per line with format:
proto:logger
it now displays:
proto:logger (comma_separated_list_of_loggers)
NONE is used as keyword if no logger is used.

net/netfilter/nf_log.c +26 -2
index a228b5f..4fcbcc7 100644
@@ -154,13 +154,37 @@ static int seq_show(struct seq_file *s, void *v)
 {
 	loff_t *pos = v;
 	const struct nf_logger *logger;
+	struct nf_logger *t;
+	int ret;
 
 	logger = rcu_dereference(nf_loggers[*pos]);
 
 	if (!logger)
-		return seq_printf(s, "%2lld NONE\n", *pos);
+		ret = seq_printf(s, "%2lld NONE (", *pos);
+	else
+		ret = seq_printf(s, "%2lld %s (", *pos, logger->name);
+
+	if (ret < 0)
+		return ret;
+
+	mutex_lock(&nf_log_mutex);
+	list_for_each_entry(t, &nf_loggers_l[*pos], list[*pos]) {
+		ret = seq_printf(s, "%s", t->name);
+		if (ret < 0) {
+			mutex_unlock(&nf_log_mutex);
+			return ret;
+		}
+		if (&t->list[*pos] != nf_loggers_l[*pos].prev) {
+			ret = seq_printf(s, ",");
+			if (ret < 0) {
+				mutex_unlock(&nf_log_mutex);
+				return ret;
+			}
+		}
+	}
+	mutex_unlock(&nf_log_mutex);
 
-	return seq_printf(s, "%2lld %s\n", *pos, logger->name);
+	return seq_printf(s, ")\n");
 }
 
 static const struct seq_operations nflog_seq_ops = {
#36 lore raw

netfilter 24/41: remove IPvX specific parts from nf_conntrack_l4proto.h

In reply to Patrick McHardy

commit 9d2493f88f846b391a15a736efc7f4b97d6c4046

Author: Christoph Paasch <[email protected]>
Date:   Mon Mar 16 15:15:35 2009 +0100

netfilter: remove IPvX specific parts from nf_conntrack_l4proto.h

Moving the structure definitions to the corresponding IPvX specific header files.

include/net/netfilter/nf_conntrack_l4proto.h +1 -4
index debdaf7..16ab604 100644
@@ -90,10 +90,7 @@ struct nf_conntrack_l4proto
 	struct module *me;
 };
 
-/* Existing built-in protocols */
-extern struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp6;
-extern struct nf_conntrack_l4proto nf_conntrack_l4proto_udp4;
-extern struct nf_conntrack_l4proto nf_conntrack_l4proto_udp6;
+/* Existing built-in generic protocol */
 extern struct nf_conntrack_l4proto nf_conntrack_l4proto_generic;
 
 #define MAX_NF_CT_PROTO 256
net/ipv6/netfilter/nf_conntrack_l3proto_ipv6.c +1 -0
index 727b953..e6852f6 100644
@@ -26,6 +26,7 @@
 #include <net/netfilter/nf_conntrack_l4proto.h>
 #include <net/netfilter/nf_conntrack_l3proto.h>
 #include <net/netfilter/nf_conntrack_core.h>
+#include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
 
 static bool ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
 			      struct nf_conntrack_tuple *tuple)
net/netfilter/nf_conntrack_proto_tcp.c +2 -0
index a1edb9c..7d3944f 100644
@@ -25,6 +25,8 @@
 #include <net/netfilter/nf_conntrack_l4proto.h>
 #include <net/netfilter/nf_conntrack_ecache.h>
 #include <net/netfilter/nf_log.h>
+#include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
+#include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
 
 /* Protects ct->proto.tcp */
 static DEFINE_RWLOCK(tcp_lock);
net/netfilter/nf_conntrack_proto_udp.c +2 -0
index 2b8b1f5..d402117 100644
@@ -22,6 +22,8 @@
 #include <net/netfilter/nf_conntrack_l4proto.h>
 #include <net/netfilter/nf_conntrack_ecache.h>
 #include <net/netfilter/nf_log.h>
+#include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
+#include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
 
 static unsigned int nf_ct_udp_timeout __read_mostly = 30*HZ;
 static unsigned int nf_ct_udp_timeout_stream __read_mostly = 180*HZ;
#37 lore raw

netfilter 25/41: Kconfig spelling fixes (trivial)

In reply to Patrick McHardy

commit 67c0d57930ff9a24c6c34abee1b01f7716a9b0e2

Author: Stephen Hemminger <[email protected]>
Date:   Mon Mar 16 15:17:23 2009 +0100

netfilter: Kconfig spelling fixes (trivial)

net/ipv4/netfilter/Kconfig +3 -3
index f8d6180..1833bdb 100644
@@ -31,7 +31,7 @@ config NF_CONNTRACK_PROC_COMPAT
 	default y
 	help
 	  This option enables /proc and sysctl compatibility with the old
-	  layer 3 dependant connection tracking. This is needed to keep
+	  layer 3 dependent connection tracking. This is needed to keep
 	  old programs that have not been adapted to the new names working.
 
 	  If unsure, say Y.
@@ -99,7 +99,7 @@ config IP_NF_MATCH_TTL
 	---help---
 	This is a backwards-compat option for the user's convenience
 	(e.g. when running oldconfig). It selects
-	COFNIG_NETFILTER_XT_MATCH_HL.
+	CONFIG_NETFILTER_XT_MATCH_HL.
 
 # `filter', generic and specific targets
 config IP_NF_FILTER
@@ -329,7 +329,7 @@ config IP_NF_TARGET_TTL
 	---help---
 	This is a backwards-compat option for the user's convenience
 	(e.g. when running oldconfig). It selects
-	COFNIG_NETFILTER_XT_TARGET_HL.
+	CONFIG_NETFILTER_XT_TARGET_HL.
 
 # raw + specific targets
 config IP_NF_RAW
#38 lore raw

Re: netfilter 25/41: Kconfig spelling fixes (trivial)

Quoted 8 lines · Patrick McHardy

commit 67c0d57930ff9a24c6c34abee1b01f7716a9b0e2

Author: Stephen Hemminger <[email protected]>
Date:   Mon Mar 16 15:17:23 2009 +0100

netfilter: Kconfig spelling fixes (trivial)

Please add this patch as a followup commit.

parent 1d45209d89e647e9f27e4afa1f47338df73bc112 (v2.6.29-rc5-881-g1d45209) commit 85982fcfd17fbb1000302e253f82edd88efa180c

Author: Jan Engelhardt <[email protected]>
Date:   Tue Mar 24 15:18:17 2009 +0100

netfilter: trivial Kconfig spelling fixes

Supplements commit 67c0d57930ff9a24c6c34abee1b01f7716a9b0e2.

---
 net/ipv6/netfilter/Kconfig |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
net/ipv6/netfilter/Kconfig +2 -2
index 625353a..29d643b 100644
@@ -101,7 +101,7 @@ config IP6_NF_MATCH_HL
 	---help---
 	This is a backwards-compat option for the user's convenience
 	(e.g. when running oldconfig). It selects
-	COFNIG_NETFILTER_XT_MATCH_HL.
+	CONFIG_NETFILTER_XT_MATCH_HL.
 
 config IP6_NF_MATCH_IPV6HEADER
 	tristate '"ipv6header" IPv6 Extension Headers Match'
@@ -137,7 +137,7 @@ config IP6_NF_TARGET_HL
 	---help---
 	This is a backwards-compat option for the user's convenience
 	(e.g. when running oldconfig). It selects
-	COFNIG_NETFILTER_XT_TARGET_HL.
+	CONFIG_NETFILTER_XT_TARGET_HL.
 
 config IP6_NF_TARGET_LOG
 	tristate "LOG target support"
Signature · 1 line
# Created with git-export-patch
maintainer Applied
#39 lore raw

Re: netfilter 25/41: Kconfig spelling fixes (trivial)

From: Jan Engelhardt <[email protected]>
Date: Tue, 24 Mar 2009 15:20:08 +0100 (CET)

Quoted 5 lines

netfilter: trivial Kconfig spelling fixes

Supplements commit 67c0d57930ff9a24c6c34abee1b01f7716a9b0e2.

Applied.

#40 lore raw

netfilter 26/41: conntrack: increase drop stats if sequence adjustment fails

In reply to Patrick McHardy

commit 1db7a748dfd50d7615913730763c024444900030

Author: Pablo Neira Ayuso <[email protected]>
Date:   Mon Mar 16 15:18:50 2009 +0100

netfilter: conntrack: increase drop stats if sequence adjustment fails

This patch increases the statistics of packets drop if the sequence adjustment fails in ipv4_confirm().

net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c +3 -1
index 4beb04f..8b681f2 100644
@@ -120,8 +120,10 @@ static unsigned int ipv4_confirm(unsigned int hooknum,
 		typeof(nf_nat_seq_adjust_hook) seq_adjust;
 
 		seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
-		if (!seq_adjust || !seq_adjust(skb, ct, ctinfo))
+		if (!seq_adjust || !seq_adjust(skb, ct, ctinfo)) {
+			NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
 			return NF_DROP;
+		}
 	}
 out:
 	/* We've seen it coming out the other side: confirm it */
#41 lore raw

netfilter 27/41: ctnetlink: cleanup master conntrack assignation

In reply to Patrick McHardy

commit 7ec4749675bf33ea639bbcca8a5365ccc5091a6a

Author: Pablo Neira Ayuso <[email protected]>
Date:   Mon Mar 16 15:25:46 2009 +0100

netfilter: ctnetlink: cleanup master conntrack assignation

This patch moves the assignation of the master conntrack to ctnetlink_create_conntrack(), which is where it really belongs. This patch is a cleanup.

net/netfilter/nf_conntrack_netlink.c +20 -29
index cb78aa0..cca22d5 100644
@@ -1128,9 +1128,9 @@ static int
 ctnetlink_create_conntrack(struct nlattr *cda[],
 			   struct nf_conntrack_tuple *otuple,
 			   struct nf_conntrack_tuple *rtuple,
-			   struct nf_conn *master_ct,
 			   u32 pid,
-			   int report)
+			   int report,
+			   u8 u3)
 {
 	struct nf_conn *ct;
 	int err = -EINVAL;
@@ -1241,7 +1241,22 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
 #endif
 
 	/* setup master conntrack: this is a confirmed expectation */
-	if (master_ct) {
+	if (cda[CTA_TUPLE_MASTER]) {
+		struct nf_conntrack_tuple master;
+		struct nf_conntrack_tuple_hash *master_h;
+		struct nf_conn *master_ct;
+
+		err = ctnetlink_parse_tuple(cda, &master, CTA_TUPLE_MASTER, u3);
+		if (err < 0)
+			goto err;
+
+		master_h = __nf_conntrack_find(&init_net, &master);
+		if (master_h == NULL) {
+			err = -ENOENT;
+			goto err;
+		}
+		master_ct = nf_ct_tuplehash_to_ctrack(master_h);
+		nf_conntrack_get(&master_ct->ct_general);
 		__set_bit(IPS_EXPECTED_BIT, &ct->status);
 		ct->master = master_ct;
 	}
@@ -1289,39 +1304,15 @@ ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
 		h = __nf_conntrack_find(&init_net, &rtuple);
 
 	if (h == NULL) {
-		struct nf_conntrack_tuple master;
-		struct nf_conntrack_tuple_hash *master_h = NULL;
-		struct nf_conn *master_ct = NULL;
-
-		if (cda[CTA_TUPLE_MASTER]) {
-			err = ctnetlink_parse_tuple(cda,
-						    &master,
-						    CTA_TUPLE_MASTER,
-						    u3);
-			if (err < 0)
-				goto out_unlock;
-
-			master_h = __nf_conntrack_find(&init_net, &master);
-			if (master_h == NULL) {
-				err = -ENOENT;
-				goto out_unlock;
-			}
-			master_ct = nf_ct_tuplehash_to_ctrack(master_h);
-			nf_conntrack_get(&master_ct->ct_general);
-		}
-
 		err = -ENOENT;
 		if (nlh->nlmsg_flags & NLM_F_CREATE)
 			err = ctnetlink_create_conntrack(cda,
 							 &otuple,
 							 &rtuple,
-							 master_ct,
 							 NETLINK_CB(skb).pid,
-							 nlmsg_report(nlh));
+							 nlmsg_report(nlh),
+							 u3);
 		spin_unlock_bh(&nf_conntrack_lock);
-		if (err < 0 && master_ct)
-			nf_ct_put(master_ct);
-
 		return err;
 	}
 	/* implicit 'else' */
#42 lore raw

netfilter 28/41: ctnetlink: cleanup conntrack update preliminary checkings

In reply to Patrick McHardy

commit e098360f159b3358f085543eb6dc2eb500d6667c

Author: Pablo Neira Ayuso <[email protected]>
Date:   Mon Mar 16 15:27:22 2009 +0100

netfilter: ctnetlink: cleanup conntrack update preliminary checkings

This patch moves the preliminary checkings that must be fulfilled to update a conntrack, which are the following:

* NAT manglings cannot be updated
* Changing the master conntrack is not allowed.

This patch is a cleanup.

net/netfilter/nf_conntrack_netlink.c +4 -11
index cca22d5..b67db69 100644
@@ -1062,6 +1062,10 @@ ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
 {
 	int err;
 
+	/* only allow NAT changes and master assignation for new conntracks */
+	if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST] || cda[CTA_TUPLE_MASTER])
+		return -EOPNOTSUPP;
+
 	if (cda[CTA_HELP]) {
 		err = ctnetlink_change_helper(ct, cda);
 		if (err < 0)
@@ -1323,17 +1327,6 @@ ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
 	if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
 		struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 
-		/* we only allow nat config for new conntracks */
-		if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
-			err = -EOPNOTSUPP;
-			goto out_unlock;
-		}
-		/* can't link an existing conntrack to a master */
-		if (cda[CTA_TUPLE_MASTER]) {
-			err = -EOPNOTSUPP;
-			goto out_unlock;
-		}
-
 		err = ctnetlink_change_conntrack(ct, cda);
 		if (err == 0) {
 			nf_conntrack_get(&ct->ct_general);
#43 lore raw

netfilter 29/41: ctnetlink: move event reporting for new entries outside the lock

In reply to Patrick McHardy

commit f0a3c0869f3b0ef93d9df044e9a41e40086d4c97

Author: Pablo Neira Ayuso <[email protected]>
Date:   Mon Mar 16 15:28:09 2009 +0100

netfilter: ctnetlink: move event reporting for new entries outside the lock

This patch moves the event reporting outside the lock section. With this patch, the creation and update of entries is homogeneous from the event reporting perspective. Moreover, as the event reporting is done outside the lock section, the netlink broadcast delivery can benefit of the yield() call under congestion.

net/netfilter/nf_conntrack_netlink.c +23 -18
index b67db69..9fb7cf7 100644
@@ -1128,12 +1128,10 @@ ctnetlink_event_report(struct nf_conn *ct, u32 pid, int report)
 				  report);
 }
 
-static int
+static struct nf_conn *
 ctnetlink_create_conntrack(struct nlattr *cda[],
 			   struct nf_conntrack_tuple *otuple,
 			   struct nf_conntrack_tuple *rtuple,
-			   u32 pid,
-			   int report,
 			   u8 u3)
 {
 	struct nf_conn *ct;
@@ -1142,7 +1140,7 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
 
 	ct = nf_conntrack_alloc(&init_net, otuple, rtuple, GFP_ATOMIC);
 	if (IS_ERR(ct))
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
 
 	if (!cda[CTA_TIMEOUT])
 		goto err;
@@ -1265,18 +1263,14 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
 		ct->master = master_ct;
 	}
 
-	nf_conntrack_get(&ct->ct_general);
 	add_timer(&ct->timeout);
 	nf_conntrack_hash_insert(ct);
 	rcu_read_unlock();
-	ctnetlink_event_report(ct, pid, report);
-	nf_ct_put(ct);
-
-	return 0;
 
+	return ct;
 err:
 	nf_conntrack_free(ct);
-	return err;
+	return ERR_PTR(err);
 }
 
 static int
@@ -1309,14 +1303,25 @@ ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
 
 	if (h == NULL) {
 		err = -ENOENT;
-		if (nlh->nlmsg_flags & NLM_F_CREATE)
-			err = ctnetlink_create_conntrack(cda,
-							 &otuple,
-							 &rtuple,
-							 NETLINK_CB(skb).pid,
-							 nlmsg_report(nlh),
-							 u3);
-		spin_unlock_bh(&nf_conntrack_lock);
+		if (nlh->nlmsg_flags & NLM_F_CREATE) {
+			struct nf_conn *ct;
+
+			ct = ctnetlink_create_conntrack(cda, &otuple,
+							&rtuple, u3);
+			if (IS_ERR(ct)) {
+				err = PTR_ERR(ct);
+				goto out_unlock;
+			}
+			err = 0;
+			nf_conntrack_get(&ct->ct_general);
+			spin_unlock_bh(&nf_conntrack_lock);
+			ctnetlink_event_report(ct,
+					       NETLINK_CB(skb).pid,
+					       nlmsg_report(nlh));
+			nf_ct_put(ct);
+		} else
+			spin_unlock_bh(&nf_conntrack_lock);
+
 		return err;
 	}
 	/* implicit 'else' */
#44 lore raw

netfilter 30/41: auto-load ip6_queue module when socket opened

In reply to Patrick McHardy

commit 26c3b6780618f09abb5f7e03b09b13dbb8e8aa24

Author: Scott James Remnant <[email protected]>
Date:   Mon Mar 16 15:30:14 2009 +0100

netfilter: auto-load ip6_queue module when socket opened

The ip6_queue module is missing the net-pf-16-proto-13 alias that would cause it to be auto-loaded when a socket of that type is opened. This patch adds the alias.

net/ipv6/netfilter/ip6_queue.c +1 -0
index 5859c04..b693f84 100644
@@ -643,6 +643,7 @@ static void __exit ip6_queue_fini(void)
 
 MODULE_DESCRIPTION("IPv6 packet queue handler");
 MODULE_LICENSE("GPL");
+MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_IP6_FW);
 
 module_init(ip6_queue_init);
 module_exit(ip6_queue_fini);
#45 lore raw

netfilter 31/41: auto-load ip_queue module when socket opened

In reply to Patrick McHardy

commit 95ba434f898c3cb5c7457dce265bf0ab72ba8ce9

Author: Scott James Remnant <[email protected]>
Date:   Mon Mar 16 15:31:10 2009 +0100

netfilter: auto-load ip_queue module when socket opened

The ip_queue module is missing the net-pf-16-proto-3 alias that would causae it to be auto-loaded when a socket of that type is opened. This patch adds the alias.

net/ipv4/netfilter/ip_queue.c +2 -0
index 432ce9d..5f22c91 100644
@@ -24,6 +24,7 @@
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
 #include <linux/security.h>
+#include <linux/net.h>
 #include <linux/mutex.h>
 #include <net/net_namespace.h>
 #include <net/sock.h>
@@ -640,6 +641,7 @@ static void __exit ip_queue_fini(void)
 MODULE_DESCRIPTION("IPv4 packet queue handler");
 MODULE_AUTHOR("James Morris <[email protected]>");
 MODULE_LICENSE("GPL");
+MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_FIREWALL);
 
 module_init(ip_queue_init);
 module_exit(ip_queue_fini);
#46 lore raw

netfilter 32/41: xtables: avoid pointer to self

In reply to Patrick McHardy

commit acc738fec03bdaa5b77340c32a82fbfedaaabef0

Author: Jan Engelhardt <[email protected]>
Date:   Mon Mar 16 15:35:29 2009 +0100

netfilter: xtables: avoid pointer to self

Commit 784544739a25c30637397ace5489eeb6e15d7d49 (netfilter: iptables:
lock free counters) broke a number of modules whose rule data referenced itself. A reallocation would not reestablish the correct references, so it is best to use a separate struct that does not fall under RCU.

include/linux/netfilter/xt_limit.h +5 -4
index b3ce653..fda222c 100644
@@ -4,6 +4,8 @@
 /* timings are in milliseconds. */
 #define XT_LIMIT_SCALE 10000
 
+struct xt_limit_priv;
+
 /* 1/10,000 sec period => max of 10,000/sec.  Min rate is then 429490
    seconds, or one every 59 hours. */
 struct xt_rateinfo {
@@ -11,11 +13,10 @@ struct xt_rateinfo {
 	u_int32_t burst;  /* Period multiplier for upper limit. */
 
 	/* Used internally by the kernel */
-	unsigned long prev;
-	u_int32_t credit;
+	unsigned long prev; /* moved to xt_limit_priv */
+	u_int32_t credit; /* moved to xt_limit_priv */
 	u_int32_t credit_cap, cost;
 
-	/* Ugly, ugly fucker. */
-	struct xt_rateinfo *master;
+	struct xt_limit_priv *master;
 };
 #endif /*_XT_RATE_H*/
include/linux/netfilter/xt_quota.h +3 -1
index 4c8368d..8dc89df 100644
@@ -6,13 +6,15 @@ enum xt_quota_flags {
 };
 #define XT_QUOTA_MASK		0x1
 
+struct xt_quota_priv;
+
 struct xt_quota_info {
 	u_int32_t		flags;
 	u_int32_t		pad;
 
 	/* Used internally by the kernel */
 	aligned_u64		quota;
-	struct xt_quota_info	*master;
+	struct xt_quota_priv	*master;
 };
 
 #endif /* _XT_QUOTA_H */
include/linux/netfilter/xt_statistic.h +4 -3
index 3d38bc9..8f521ab 100644
@@ -13,6 +13,8 @@ enum xt_statistic_flags {
 };
 #define XT_STATISTIC_MASK		0x1
 
+struct xt_statistic_priv;
+
 struct xt_statistic_info {
 	u_int16_t			mode;
 	u_int16_t			flags;
@@ -23,11 +25,10 @@ struct xt_statistic_info {
 		struct {
 			u_int32_t	every;
 			u_int32_t	packet;
-			/* Used internally by the kernel */
-			u_int32_t	count;
+			u_int32_t	count; /* unused */
 		} nth;
 	} u;
-	struct xt_statistic_info	*master __attribute__((aligned(8)));
+	struct xt_statistic_priv *master __attribute__((aligned(8)));
 };
 
 #endif /* _XT_STATISTIC_H */
net/netfilter/xt_limit.c +29 -11
index c908d69..2e8089e 100644
@@ -14,6 +14,11 @@
 #include <linux/netfilter/x_tables.h>
 #include <linux/netfilter/xt_limit.h>
 
+struct xt_limit_priv {
+	unsigned long prev;
+	uint32_t credit;
+};
+
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Herve Eychenne <[email protected]>");
 MODULE_DESCRIPTION("Xtables: rate-limit match");
@@ -60,18 +65,18 @@ static DEFINE_SPINLOCK(limit_lock);
 static bool
 limit_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 {
-	struct xt_rateinfo *r =
-		((const struct xt_rateinfo *)par->matchinfo)->master;
+	const struct xt_rateinfo *r = par->matchinfo;
+	struct xt_limit_priv *priv = r->master;
 	unsigned long now = jiffies;
 
 	spin_lock_bh(&limit_lock);
-	r->credit += (now - xchg(&r->prev, now)) * CREDITS_PER_JIFFY;
-	if (r->credit > r->credit_cap)
-		r->credit = r->credit_cap;
+	priv->credit += (now - xchg(&priv->prev, now)) * CREDITS_PER_JIFFY;
+	if (priv->credit > r->credit_cap)
+		priv->credit = r->credit_cap;
 
-	if (r->credit >= r->cost) {
+	if (priv->credit >= r->cost) {
 		/* We're not limited. */
-		r->credit -= r->cost;
+		priv->credit -= r->cost;
 		spin_unlock_bh(&limit_lock);
 		return true;
 	}
@@ -95,6 +100,7 @@ user2credits(u_int32_t user)
 static bool limit_mt_check(const struct xt_mtchk_param *par)
 {
 	struct xt_rateinfo *r = par->matchinfo;
+	struct xt_limit_priv *priv;
 
 	/* Check for overflow. */
 	if (r->burst == 0
@@ -104,19 +110,30 @@ static bool limit_mt_check(const struct xt_mtchk_param *par)
 		return false;
 	}
 
-	/* For SMP, we only want to use one set of counters. */
-	r->master = r;
+	priv = kmalloc(sizeof(*priv), GFP_KERNEL);
+	if (priv == NULL)
+		return -ENOMEM;
+
+	/* For SMP, we only want to use one set of state. */
+	r->master = priv;
 	if (r->cost == 0) {
 		/* User avg in seconds * XT_LIMIT_SCALE: convert to jiffies *
 		   128. */
-		r->prev = jiffies;
-		r->credit = user2credits(r->avg * r->burst);	 /* Credits full. */
+		priv->prev = jiffies;
+		priv->credit = user2credits(r->avg * r->burst); /* Credits full. */
 		r->credit_cap = user2credits(r->avg * r->burst); /* Credits full. */
 		r->cost = user2credits(r->avg);
 	}
 	return true;
 }
 
+static void limit_mt_destroy(const struct xt_mtdtor_param *par)
+{
+	const struct xt_rateinfo *info = par->matchinfo;
+
+	kfree(info->master);
+}
+
 #ifdef CONFIG_COMPAT
 struct compat_xt_rateinfo {
 	u_int32_t avg;
@@ -167,6 +184,7 @@ static struct xt_match limit_mt_reg __read_mostly = {
 	.family           = NFPROTO_UNSPEC,
 	.match            = limit_mt,
 	.checkentry       = limit_mt_check,
+	.destroy          = limit_mt_destroy,
 	.matchsize        = sizeof(struct xt_rateinfo),
 #ifdef CONFIG_COMPAT
 	.compatsize       = sizeof(struct compat_xt_rateinfo),
net/netfilter/xt_quota.c +24 -7
index c84fce5..01dd07b 100644
@@ -9,6 +9,10 @@
 #include <linux/netfilter/x_tables.h>
 #include <linux/netfilter/xt_quota.h>
 
+struct xt_quota_priv {
+	uint64_t quota;
+};
+
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Sam Johnston <[email protected]>");
 MODULE_DESCRIPTION("Xtables: countdown quota match");
@@ -20,18 +24,20 @@ static DEFINE_SPINLOCK(quota_lock);
 static bool
 quota_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 {
-	struct xt_quota_info *q =
-		((const struct xt_quota_info *)par->matchinfo)->master;
+	struct xt_quota_info *q = (void *)par->matchinfo;
+	struct xt_quota_priv *priv = q->master;
 	bool ret = q->flags & XT_QUOTA_INVERT;
 
 	spin_lock_bh(&quota_lock);
-	if (q->quota >= skb->len) {
-		q->quota -= skb->len;
+	if (priv->quota >= skb->len) {
+		priv->quota -= skb->len;
 		ret = !ret;
 	} else {
 		/* we do not allow even small packets from now on */
-		q->quota = 0;
+		priv->quota = 0;
 	}
+	/* Copy quota back to matchinfo so that iptables can display it */
+	q->quota = priv->quota;
 	spin_unlock_bh(&quota_lock);
 
 	return ret;
@@ -43,17 +49,28 @@ static bool quota_mt_check(const struct xt_mtchk_param *par)
 
 	if (q->flags & ~XT_QUOTA_MASK)
 		return false;
-	/* For SMP, we only want to use one set of counters. */
-	q->master = q;
+
+	q->master = kmalloc(sizeof(*q->master), GFP_KERNEL);
+	if (q->master == NULL)
+		return -ENOMEM;
+
 	return true;
 }
 
+static void quota_mt_destroy(const struct xt_mtdtor_param *par)
+{
+	const struct xt_quota_info *q = par->matchinfo;
+
+	kfree(q->master);
+}
+
 static struct xt_match quota_mt_reg __read_mostly = {
 	.name       = "quota",
 	.revision   = 0,
 	.family     = NFPROTO_UNSPEC,
 	.match      = quota_mt,
 	.checkentry = quota_mt_check,
+	.destroy    = quota_mt_destroy,
 	.matchsize  = sizeof(struct xt_quota_info),
 	.me         = THIS_MODULE,
 };
net/netfilter/xt_statistic.c +23 -5
index 0d75141..d8c0f8f 100644
@@ -16,6 +16,10 @@
 #include <linux/netfilter/xt_statistic.h>
 #include <linux/netfilter/x_tables.h>
 
+struct xt_statistic_priv {
+	uint32_t count;
+};
+
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Patrick McHardy <[email protected]>");
 MODULE_DESCRIPTION("Xtables: statistics-based matching (\"Nth\", random)");
@@ -27,7 +31,7 @@ static DEFINE_SPINLOCK(nth_lock);
 static bool
 statistic_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 {
-	struct xt_statistic_info *info = (void *)par->matchinfo;
+	const struct xt_statistic_info *info = par->matchinfo;
 	bool ret = info->flags & XT_STATISTIC_INVERT;
 
 	switch (info->mode) {
@@ -36,10 +40,9 @@ statistic_mt(const struct sk_buff *skb, const struct xt_match_param *par)
 			ret = !ret;
 		break;
 	case XT_STATISTIC_MODE_NTH:
-		info = info->master;
 		spin_lock_bh(&nth_lock);
-		if (info->u.nth.count++ == info->u.nth.every) {
-			info->u.nth.count = 0;
+		if (info->master->count++ == info->u.nth.every) {
+			info->master->count = 0;
 			ret = !ret;
 		}
 		spin_unlock_bh(&nth_lock);
@@ -56,16 +59,31 @@ static bool statistic_mt_check(const struct xt_mtchk_param *par)
 	if (info->mode > XT_STATISTIC_MODE_MAX ||
 	    info->flags & ~XT_STATISTIC_MASK)
 		return false;
-	info->master = info;
+
+	info->master = kzalloc(sizeof(*info->master), GFP_KERNEL);
+	if (info->master == NULL) {
+		printk(KERN_ERR KBUILD_MODNAME ": Out of memory\n");
+		return false;
+	}
+	info->master->count = info->u.nth.count;
+
 	return true;
 }
 
+static void statistic_mt_destroy(const struct xt_mtdtor_param *par)
+{
+	const struct xt_statistic_info *info = par->matchinfo;
+
+	kfree(info->master);
+}
+
 static struct xt_match xt_statistic_mt_reg __read_mostly = {
 	.name       = "statistic",
 	.revision   = 0,
 	.family     = NFPROTO_UNSPEC,
 	.match      = statistic_mt,
 	.checkentry = statistic_mt_check,
+	.destroy    = statistic_mt_destroy,
 	.matchsize  = sizeof(struct xt_statistic_info),
 	.me         = THIS_MODULE,
 };
#47 lore raw

net 33/41: sysctl_net - use net_eq to compare nets

In reply to Patrick McHardy

commit 81a1d3c31e3517f9939b3e04d21cf4a6b0997419

Author: Cyrill Gorcunov <[email protected]>
Date:   Mon Mar 16 16:23:30 2009 +0100

net: sysctl_net - use net_eq to compare nets

net/sysctl_net.c +1 -1
index 972201c..0b15d72 100644
@@ -61,7 +61,7 @@ static struct ctl_table_root net_sysctl_root = {
 static int net_ctl_ro_header_perms(struct ctl_table_root *root,
 		struct nsproxy *namespaces, struct ctl_table *table)
 {
-	if (namespaces->net_ns == &init_net)
+	if (net_eq(namespaces->net_ns, &init_net))
 		return table->mode;
 	else
 		return table->mode & ~0222;
#48 lore raw

net 34/41: netfilter conntrack - add per-net functionality for DCCP protocol

In reply to Patrick McHardy

commit 1546000fe8db0d3f47b0ef1dd487ec23fbd95313

Author: Cyrill Gorcunov <[email protected]>
Date:   Mon Mar 16 16:30:49 2009 +0100

net: netfilter conntrack - add per-net functionality for DCCP protocol

Module specific data moved into per-net site and being allocated/freed during net namespace creation/deletion.

net/netfilter/nf_conntrack_proto_dccp.c +108 -37
index 8fcf176..d3d5a7f 100644
@@ -16,6 +16,9 @@
 #include <linux/skbuff.h>
 #include <linux/dccp.h>
 
+#include <net/net_namespace.h>
+#include <net/netns/generic.h>
+
 #include <linux/netfilter/nfnetlink_conntrack.h>
 #include <net/netfilter/nf_conntrack.h>
 #include <net/netfilter/nf_conntrack_l4proto.h>
@@ -23,8 +26,6 @@
 
 static DEFINE_RWLOCK(dccp_lock);
 
-static int nf_ct_dccp_loose __read_mostly = 1;
-
 /* Timeouts are based on values from RFC4340:
  *
  * - REQUEST:
@@ -72,16 +73,6 @@ static int nf_ct_dccp_loose __read_mostly = 1;
 
 #define DCCP_MSL (2 * 60 * HZ)
 
-static unsigned int dccp_timeout[CT_DCCP_MAX + 1] __read_mostly = {
-	[CT_DCCP_REQUEST]	= 2 * DCCP_MSL,
-	[CT_DCCP_RESPOND]	= 4 * DCCP_MSL,
-	[CT_DCCP_PARTOPEN]	= 4 * DCCP_MSL,
-	[CT_DCCP_OPEN]		= 12 * 3600 * HZ,
-	[CT_DCCP_CLOSEREQ]	= 64 * HZ,
-	[CT_DCCP_CLOSING]	= 64 * HZ,
-	[CT_DCCP_TIMEWAIT]	= 2 * DCCP_MSL,
-};
-
 static const char * const dccp_state_names[] = {
 	[CT_DCCP_NONE]		= "NONE",
 	[CT_DCCP_REQUEST]	= "REQUEST",
@@ -393,6 +384,22 @@ dccp_state_table[CT_DCCP_ROLE_MAX + 1][DCCP_PKT_SYNCACK + 1][CT_DCCP_MAX + 1] =
 	},
 };
 
+/* this module per-net specifics */
+static int dccp_net_id;
+struct dccp_net {
+	int dccp_loose;
+	unsigned int dccp_timeout[CT_DCCP_MAX + 1];
+#ifdef CONFIG_SYSCTL
+	struct ctl_table_header *sysctl_header;
+	struct ctl_table *sysctl_table;
+#endif
+};
+
+static inline struct dccp_net *dccp_pernet(struct net *net)
+{
+	return net_generic(net, dccp_net_id);
+}
+
 static bool dccp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
 			      struct nf_conntrack_tuple *tuple)
 {
@@ -419,6 +426,7 @@ static bool dccp_new(struct nf_conn *ct, const struct sk_buff *skb,
 		     unsigned int dataoff)
 {
 	struct net *net = nf_ct_net(ct);
+	struct dccp_net *dn;
 	struct dccp_hdr _dh, *dh;
 	const char *msg;
 	u_int8_t state;
@@ -429,7 +437,8 @@ static bool dccp_new(struct nf_conn *ct, const struct sk_buff *skb,
 	state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE];
 	switch (state) {
 	default:
-		if (nf_ct_dccp_loose == 0) {
+		dn = dccp_pernet(net);
+		if (dn->dccp_loose == 0) {
 			msg = "nf_ct_dccp: not picking up existing connection ";
 			goto out_invalid;
 		}
@@ -465,6 +474,7 @@ static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb,
 		       u_int8_t pf, unsigned int hooknum)
 {
 	struct net *net = nf_ct_net(ct);
+	struct dccp_net *dn;
 	enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 	struct dccp_hdr _dh, *dh;
 	u_int8_t type, old_state, new_state;
@@ -542,7 +552,9 @@ static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb,
 	ct->proto.dccp.last_pkt = type;
 	ct->proto.dccp.state = new_state;
 	write_unlock_bh(&dccp_lock);
-	nf_ct_refresh_acct(ct, ctinfo, skb, dccp_timeout[new_state]);
+
+	dn = dccp_pernet(net);
+	nf_ct_refresh_acct(ct, ctinfo, skb, dn->dccp_timeout[new_state]);
 
 	return NF_ACCEPT;
 }
@@ -660,13 +672,11 @@ static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct)
 #endif
 
 #ifdef CONFIG_SYSCTL
-static unsigned int dccp_sysctl_table_users;
-static struct ctl_table_header *dccp_sysctl_header;
-static ctl_table dccp_sysctl_table[] = {
+/* template, data assigned later */
+static struct ctl_table dccp_sysctl_table[] = {
 	{
 		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "nf_conntrack_dccp_timeout_request",
-		.data		= &dccp_timeout[CT_DCCP_REQUEST],
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_jiffies,
@@ -674,7 +684,6 @@ static ctl_table dccp_sysctl_table[] = {
 	{
 		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "nf_conntrack_dccp_timeout_respond",
-		.data		= &dccp_timeout[CT_DCCP_RESPOND],
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_jiffies,
@@ -682,7 +691,6 @@ static ctl_table dccp_sysctl_table[] = {
 	{
 		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "nf_conntrack_dccp_timeout_partopen",
-		.data		= &dccp_timeout[CT_DCCP_PARTOPEN],
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_jiffies,
@@ -690,7 +698,6 @@ static ctl_table dccp_sysctl_table[] = {
 	{
 		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "nf_conntrack_dccp_timeout_open",
-		.data		= &dccp_timeout[CT_DCCP_OPEN],
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_jiffies,
@@ -698,7 +705,6 @@ static ctl_table dccp_sysctl_table[] = {
 	{
 		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "nf_conntrack_dccp_timeout_closereq",
-		.data		= &dccp_timeout[CT_DCCP_CLOSEREQ],
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_jiffies,
@@ -706,7 +712,6 @@ static ctl_table dccp_sysctl_table[] = {
 	{
 		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "nf_conntrack_dccp_timeout_closing",
-		.data		= &dccp_timeout[CT_DCCP_CLOSING],
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_jiffies,
@@ -714,7 +719,6 @@ static ctl_table dccp_sysctl_table[] = {
 	{
 		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "nf_conntrack_dccp_timeout_timewait",
-		.data		= &dccp_timeout[CT_DCCP_TIMEWAIT],
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_jiffies,
@@ -722,8 +726,7 @@ static ctl_table dccp_sysctl_table[] = {
 	{
 		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "nf_conntrack_dccp_loose",
-		.data		= &nf_ct_dccp_loose,
-		.maxlen		= sizeof(nf_ct_dccp_loose),
+		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
 	},
@@ -751,11 +754,6 @@ static struct nf_conntrack_l4proto dccp_proto4 __read_mostly = {
 	.nlattr_to_tuple	= nf_ct_port_nlattr_to_tuple,
 	.nla_policy		= nf_ct_port_nla_policy,
 #endif
-#ifdef CONFIG_SYSCTL
-	.ctl_table_users	= &dccp_sysctl_table_users,
-	.ctl_table_header	= &dccp_sysctl_header,
-	.ctl_table		= dccp_sysctl_table,
-#endif
 };
 
 static struct nf_conntrack_l4proto dccp_proto6 __read_mostly = {
@@ -776,34 +774,107 @@ static struct nf_conntrack_l4proto dccp_proto6 __read_mostly = {
 	.nlattr_to_tuple	= nf_ct_port_nlattr_to_tuple,
 	.nla_policy		= nf_ct_port_nla_policy,
 #endif
+};
+
+static __net_init int dccp_net_init(struct net *net)
+{
+	struct dccp_net *dn;
+	int err;
+
+	dn = kmalloc(sizeof(*dn), GFP_KERNEL);
+	if (!dn)
+		return -ENOMEM;
+
+	/* default values */
+	dn->dccp_loose = 1;
+	dn->dccp_timeout[CT_DCCP_REQUEST]	= 2 * DCCP_MSL;
+	dn->dccp_timeout[CT_DCCP_RESPOND]	= 4 * DCCP_MSL;
+	dn->dccp_timeout[CT_DCCP_PARTOPEN]	= 4 * DCCP_MSL;
+	dn->dccp_timeout[CT_DCCP_OPEN]		= 12 * 3600 * HZ;
+	dn->dccp_timeout[CT_DCCP_CLOSEREQ]	= 64 * HZ;
+	dn->dccp_timeout[CT_DCCP_CLOSING]	= 64 * HZ;
+	dn->dccp_timeout[CT_DCCP_TIMEWAIT]	= 2 * DCCP_MSL;
+
+	err = net_assign_generic(net, dccp_net_id, dn);
+	if (err)
+		goto out;
+
 #ifdef CONFIG_SYSCTL
-	.ctl_table_users	= &dccp_sysctl_table_users,
-	.ctl_table_header	= &dccp_sysctl_header,
-	.ctl_table		= dccp_sysctl_table,
+	err = -ENOMEM;
+	dn->sysctl_table = kmemdup(dccp_sysctl_table,
+			sizeof(dccp_sysctl_table), GFP_KERNEL);
+	if (!dn->sysctl_table)
+		goto out;
+
+	dn->sysctl_table[0].data = &dn->dccp_timeout[CT_DCCP_REQUEST];
+	dn->sysctl_table[1].data = &dn->dccp_timeout[CT_DCCP_RESPOND];
+	dn->sysctl_table[2].data = &dn->dccp_timeout[CT_DCCP_PARTOPEN];
+	dn->sysctl_table[3].data = &dn->dccp_timeout[CT_DCCP_OPEN];
+	dn->sysctl_table[4].data = &dn->dccp_timeout[CT_DCCP_CLOSEREQ];
+	dn->sysctl_table[5].data = &dn->dccp_timeout[CT_DCCP_CLOSING];
+	dn->sysctl_table[6].data = &dn->dccp_timeout[CT_DCCP_TIMEWAIT];
+	dn->sysctl_table[7].data = &dn->dccp_loose;
+
+	dn->sysctl_header = register_net_sysctl_table(net,
+			nf_net_netfilter_sysctl_path, dn->sysctl_table);
+	if (!dn->sysctl_header) {
+		kfree(dn->sysctl_table);
+		goto out;
+	}
 #endif
+
+	return 0;
+
+out:
+	kfree(dn);
+	return err;
+}
+
+static __net_exit void dccp_net_exit(struct net *net)
+{
+	struct dccp_net *dn = dccp_pernet(net);
+#ifdef CONFIG_SYSCTL
+	unregister_net_sysctl_table(dn->sysctl_header);
+	kfree(dn->sysctl_table);
+#endif
+	kfree(dn);
+
+	net_assign_generic(net, dccp_net_id, NULL);
+}
+
+static struct pernet_operations dccp_net_ops = {
+	.init = dccp_net_init,
+	.exit = dccp_net_exit,
 };
 
 static int __init nf_conntrack_proto_dccp_init(void)
 {
 	int err;
 
-	err = nf_conntrack_l4proto_register(&dccp_proto4);
+	err = register_pernet_gen_subsys(&dccp_net_id, &dccp_net_ops);
 	if (err < 0)
 		goto err1;
 
-	err = nf_conntrack_l4proto_register(&dccp_proto6);
+	err = nf_conntrack_l4proto_register(&dccp_proto4);
 	if (err < 0)
 		goto err2;
+
+	err = nf_conntrack_l4proto_register(&dccp_proto6);
+	if (err < 0)
+		goto err3;
 	return 0;
 
-err2:
+err3:
 	nf_conntrack_l4proto_unregister(&dccp_proto4);
+err2:
+	unregister_pernet_gen_subsys(dccp_net_id, &dccp_net_ops);
 err1:
 	return err;
 }
 
 static void __exit nf_conntrack_proto_dccp_fini(void)
 {
+	unregister_pernet_gen_subsys(dccp_net_id, &dccp_net_ops);
 	nf_conntrack_l4proto_unregister(&dccp_proto6);
 	nf_conntrack_l4proto_unregister(&dccp_proto4);
 }
#49 lore raw

netfilter 35/41: xtables: add cluster match

In reply to Patrick McHardy

commit 0269ea4937343536ec7e85649932bc8c9686ea78

Author: Pablo Neira Ayuso <[email protected]>
Date:   Mon Mar 16 17:10:36 2009 +0100

netfilter: xtables: add cluster match

This patch adds the iptables cluster match. This match can be used to deploy gateway and back-end load-sharing clusters. The cluster can be composed of 32 nodes maximum (although I have only tested this with two nodes, so I cannot tell what is the real scalability limit of this solution in terms of cluster nodes).

Assuming that all the nodes see all packets (see below for an example on how to do that if your switch does not allow this), the cluster match decides if this node has to handle a packet given:

(jhash(source IP) % total_nodes) & node_mask

For related connections, the master conntrack is used. The following is an example of its use to deploy a gateway cluster composed of two nodes (where this is the node 1):

iptables -I PREROUTING -t mangle -i eth1 -m cluster \
--cluster-total-nodes 2 --cluster-local-node 1 \
--cluster-proc-name eth1 -j MARK --set-mark 0xffff
iptables -A PREROUTING -t mangle -i eth1 \
-m mark ! --mark 0xffff -j DROP
iptables -A PREROUTING -t mangle -i eth2 -m cluster \
--cluster-total-nodes 2 --cluster-local-node 1 \
--cluster-proc-name eth2 -j MARK --set-mark 0xffff
iptables -A PREROUTING -t mangle -i eth2 \
-m mark ! --mark 0xffff -j DROP

And the following commands to make all nodes see the same packets:

ip maddr add 01:00:5e:00:01:01 dev eth1 ip maddr add 01:00:5e:00:01:02 dev eth2 arptables -I OUTPUT -o eth1 --h-length 6 \
-j mangle --mangle-mac-s 01:00:5e:00:01:01
arptables -I INPUT -i eth1 --h-length 6 \
--destination-mac 01:00:5e:00:01:01 \
-j mangle --mangle-mac-d 00:zz:yy:xx:5a:27
arptables -I OUTPUT -o eth2 --h-length 6 \
-j mangle --mangle-mac-s 01:00:5e:00:01:02
arptables -I INPUT -i eth2 --h-length 6 \
--destination-mac 01:00:5e:00:01:02 \
-j mangle --mangle-mac-d 00:zz:yy:xx:5a:27

In the case of TCP connections, pickup facility has to be disabled to avoid marking TCP ACK packets coming in the reply direction as valid.

echo 0 > /proc/sys/net/netfilter/nf_conntrack_tcp_loose

BTW, some final notes:

* This match mangles the skbuff pkt_type in case that it detects
PACKET_MULTICAST for a non-multicast address. This may be done in a PKTTYPE target for this sole purpose.
* This match supersedes the CLUSTERIP target.

include/linux/netfilter/Kbuild +1 -0
index 947b47d..af9d2fb 100644
@@ -21,6 +21,7 @@ header-y += xt_connbytes.h
 header-y += xt_connlimit.h
 header-y += xt_connmark.h
 header-y += xt_conntrack.h
+header-y += xt_cluster.h
 header-y += xt_dccp.h
 header-y += xt_dscp.h
 header-y += xt_esp.h
include/linux/netfilter/xt_cluster.h new file +15 -0
new file mode 100644
index 0000000..5e0a0d0
@@ -0,0 +1,15 @@
+#ifndef _XT_CLUSTER_MATCH_H
+#define _XT_CLUSTER_MATCH_H
+
+enum xt_cluster_flags {
+	XT_CLUSTER_F_INV	= (1 << 0)
+};
+
+struct xt_cluster_match_info {
+	u_int32_t		total_nodes;
+	u_int32_t		node_mask;
+	u_int32_t		hash_seed;
+	u_int32_t		flags;
+};
+
+#endif /* _XT_CLUSTER_MATCH_H */
net/netfilter/Kconfig +16 -0
index cdbaaff..2562d05 100644
@@ -527,6 +527,22 @@ config NETFILTER_XT_TARGET_TCPOPTSTRIP
 	  This option adds a "TCPOPTSTRIP" target, which allows you to strip
 	  TCP options from TCP packets.
 
+config NETFILTER_XT_MATCH_CLUSTER
+	tristate '"cluster" match support'
+	depends on NF_CONNTRACK
+	depends on NETFILTER_ADVANCED
+	---help---
+	  This option allows you to build work-load-sharing clusters of
+	  network servers/stateful firewalls without having a dedicated
+	  load-balancing router/server/switch. Basically, this match returns
+	  true when the packet must be handled by this cluster node. Thus,
+	  all nodes see all packets and this match decides which node handles
+	  what packets. The work-load sharing algorithm is based on source
+	  address hashing.
+
+	  If you say Y or M here, try `iptables -m cluster --help` for
+	  more information.
+
 config NETFILTER_XT_MATCH_COMMENT
 	tristate  '"comment" match support'
 	depends on NETFILTER_ADVANCED
net/netfilter/Makefile +1 -0
index 7a9b839..6282060 100644
@@ -59,6 +59,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_TCPOPTSTRIP) += xt_TCPOPTSTRIP.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_TRACE) += xt_TRACE.o
 
 # matches
+obj-$(CONFIG_NETFILTER_XT_MATCH_CLUSTER) += xt_cluster.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_COMMENT) += xt_comment.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CONNBYTES) += xt_connbytes.o
 obj-$(CONFIG_NETFILTER_XT_MATCH_CONNLIMIT) += xt_connlimit.o
net/netfilter/xt_cluster.c new file +164 -0

164 lines not on this page · raw message

#50 lore raw

netfilter 36/41: ctnetlink: remove remaining module refcounting

In reply to Patrick McHardy

commit cd91566e4bdbcb8841385e4b2eacc8d0c29c9208

Author: Florian Westphal <[email protected]>
Date:   Wed Mar 18 17:28:37 2009 +0100

netfilter: ctnetlink: remove remaining module refcounting

Convert the remaining refcount users.

As pointed out by Patrick McHardy, the protocols can be accessed safely using RCU.

net/netfilter/nf_conntrack_netlink.c +9 -6
index 9fb7cf7..735ea9c 100644
@@ -599,7 +599,8 @@ ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
 
 	nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
 
-	l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
+	rcu_read_lock();
+	l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
 
 	if (likely(l3proto->nlattr_to_tuple)) {
 		ret = nla_validate_nested(attr, CTA_IP_MAX,
@@ -608,7 +609,7 @@ ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
 			ret = l3proto->nlattr_to_tuple(tb, tuple);
 	}
 
-	nf_ct_l3proto_put(l3proto);
+	rcu_read_unlock();
 
 	return ret;
 }
@@ -633,7 +634,8 @@ ctnetlink_parse_tuple_proto(struct nlattr *attr,
 		return -EINVAL;
 	tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
 
-	l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
+	rcu_read_lock();
+	l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
 
 	if (likely(l4proto->nlattr_to_tuple)) {
 		ret = nla_validate_nested(attr, CTA_PROTO_MAX,
@@ -642,7 +644,7 @@ ctnetlink_parse_tuple_proto(struct nlattr *attr,
 			ret = l4proto->nlattr_to_tuple(tb, tuple);
 	}
 
-	nf_ct_l4proto_put(l4proto);
+	rcu_read_unlock();
 
 	return ret;
 }
@@ -989,10 +991,11 @@ ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
 
 	nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
 
-	l4proto = nf_ct_l4proto_find_get(nf_ct_l3num(ct), nf_ct_protonum(ct));
+	rcu_read_lock();
+	l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
 	if (l4proto->from_nlattr)
 		err = l4proto->from_nlattr(tb, ct);
-	nf_ct_l4proto_put(l4proto);
+	rcu_read_unlock();
 
 	return err;
 }
#51 lore raw

netfilter 37/41: remove nf_ct_l4proto_find_get/nf_ct_l4proto_put

In reply to Patrick McHardy

commit 711d60a9e7f88e394ccca10f5fc83f95f0cea5b1

Author: Florian Westphal <[email protected]>
Date:   Wed Mar 18 17:30:50 2009 +0100

netfilter: remove nf_ct_l4proto_find_get/nf_ct_l4proto_put

users have been moved to __nf_ct_l4proto_find.

include/net/netfilter/nf_conntrack_l4proto.h +0 -5
index 16ab604..b01070b 100644
@@ -98,11 +98,6 @@ extern struct nf_conntrack_l4proto nf_conntrack_l4proto_generic;
 extern struct nf_conntrack_l4proto *
 __nf_ct_l4proto_find(u_int16_t l3proto, u_int8_t l4proto);
 
-extern struct nf_conntrack_l4proto *
-nf_ct_l4proto_find_get(u_int16_t l3proto, u_int8_t protocol);
-
-extern void nf_ct_l4proto_put(struct nf_conntrack_l4proto *p);
-
 /* Protocol registration. */
 extern int nf_conntrack_l4proto_register(struct nf_conntrack_l4proto *proto);
 extern void nf_conntrack_l4proto_unregister(struct nf_conntrack_l4proto *proto);
net/netfilter/nf_conntrack_proto.c +0 -21
index 592d733..9a62b4e 100644
@@ -74,27 +74,6 @@ EXPORT_SYMBOL_GPL(__nf_ct_l4proto_find);
 
 /* this is guaranteed to always return a valid protocol helper, since
  * it falls back to generic_protocol */
-struct nf_conntrack_l4proto *
-nf_ct_l4proto_find_get(u_int16_t l3proto, u_int8_t l4proto)
-{
-	struct nf_conntrack_l4proto *p;
-
-	rcu_read_lock();
-	p = __nf_ct_l4proto_find(l3proto, l4proto);
-	if (!try_module_get(p->me))
-		p = &nf_conntrack_l4proto_generic;
-	rcu_read_unlock();
-
-	return p;
-}
-EXPORT_SYMBOL_GPL(nf_ct_l4proto_find_get);
-
-void nf_ct_l4proto_put(struct nf_conntrack_l4proto *p)
-{
-	module_put(p->me);
-}
-EXPORT_SYMBOL_GPL(nf_ct_l4proto_put);
-
 struct nf_conntrack_l3proto *
 nf_ct_l3proto_find_get(u_int16_t l3proto)
 {
#52 lore raw

netfilter 38/41: ctnetlink: fix rcu context imbalance

In reply to Patrick McHardy

commit 0f5b3e85a3716efebb0150ebb7c6d022e2bf17d7

Author: Patrick McHardy <[email protected]>
Date:   Wed Mar 18 17:36:40 2009 +0100

netfilter: ctnetlink: fix rcu context imbalance

Introduced by 7ec47496 (netfilter: ctnetlink: cleanup master conntrack assignation):

net/netfilter/nf_conntrack_netlink.c:1275:2: warning: context imbalance in 'ctnetlink_create_conntrack' - different lock contexts for basic block

net/netfilter/nf_conntrack_netlink.c +23 -34
index 735ea9c..d1fe9d1 100644
@@ -1146,7 +1146,7 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
 		return ERR_PTR(-ENOMEM);
 
 	if (!cda[CTA_TIMEOUT])
-		goto err;
+		goto err1;
 	ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
 
 	ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
@@ -1157,10 +1157,8 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
  		char *helpname;
  
  		err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
- 		if (err < 0) {
-			rcu_read_unlock();
-			goto err;
-		}
+ 		if (err < 0)
+			goto err2;
 
 		helper = __nf_conntrack_helper_find_byname(helpname);
 		if (helper == NULL) {
@@ -1168,28 +1166,26 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
 #ifdef CONFIG_MODULES
 			if (request_module("nfct-helper-%s", helpname) < 0) {
 				err = -EOPNOTSUPP;
-				goto err;
+				goto err1;
 			}
 
 			rcu_read_lock();
 			helper = __nf_conntrack_helper_find_byname(helpname);
 			if (helper) {
-				rcu_read_unlock();
 				err = -EAGAIN;
-				goto err;
+				goto err2;
 			}
 			rcu_read_unlock();
 #endif
 			err = -EOPNOTSUPP;
-			goto err;
+			goto err1;
 		} else {
 			struct nf_conn_help *help;
 
 			help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
 			if (help == NULL) {
-				rcu_read_unlock();
 				err = -ENOMEM;
-				goto err;
+				goto err2;
 			}
 
 			/* not in hash table yet so not strictly necessary */
@@ -1198,44 +1194,34 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
 	} else {
 		/* try an implicit helper assignation */
 		err = __nf_ct_try_assign_helper(ct, GFP_ATOMIC);
-		if (err < 0) {
-			rcu_read_unlock();
-			goto err;
-		}
+		if (err < 0)
+			goto err2;
 	}
 
 	if (cda[CTA_STATUS]) {
 		err = ctnetlink_change_status(ct, cda);
-		if (err < 0) {
-			rcu_read_unlock();
-			goto err;
-		}
+		if (err < 0)
+			goto err2;
 	}
 
 	if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
 		err = ctnetlink_change_nat(ct, cda);
-		if (err < 0) {
-			rcu_read_unlock();
-			goto err;
-		}
+		if (err < 0)
+			goto err2;
 	}
 
 #ifdef CONFIG_NF_NAT_NEEDED
 	if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
 		err = ctnetlink_change_nat_seq_adj(ct, cda);
-		if (err < 0) {
-			rcu_read_unlock();
-			goto err;
-		}
+		if (err < 0)
+			goto err2;
 	}
 #endif
 
 	if (cda[CTA_PROTOINFO]) {
 		err = ctnetlink_change_protoinfo(ct, cda);
-		if (err < 0) {
-			rcu_read_unlock();
-			goto err;
-		}
+		if (err < 0)
+			goto err2;
 	}
 
 	nf_ct_acct_ext_add(ct, GFP_ATOMIC);
@@ -1253,12 +1239,12 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
 
 		err = ctnetlink_parse_tuple(cda, &master, CTA_TUPLE_MASTER, u3);
 		if (err < 0)
-			goto err;
+			goto err2;
 
 		master_h = __nf_conntrack_find(&init_net, &master);
 		if (master_h == NULL) {
 			err = -ENOENT;
-			goto err;
+			goto err2;
 		}
 		master_ct = nf_ct_tuplehash_to_ctrack(master_h);
 		nf_conntrack_get(&master_ct->ct_general);
@@ -1271,7 +1257,10 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
 	rcu_read_unlock();
 
 	return ct;
-err:
+
+err2:
+	rcu_read_unlock();
+err1:
 	nf_conntrack_free(ct);
 	return ERR_PTR(err);
 }
#53 lore raw

netfilter 39/41: sysctl support of logger choice

In reply to Patrick McHardy

commit 176252746ebbc8db97e304345af1f2563c7dc139

Author: Eric Leblond <[email protected]>
Date:   Mon Mar 23 13:16:53 2009 +0100

netfilter: sysctl support of logger choice

This patchs adds support of modification of the used logger via sysctl. It can be used to change the logger to module that can not use the bind operation (ipt_LOG and ipt_ULOG). For this purpose, it creates a directory /proc/sys/net/netfilter/nf_log which contains a file per-protocol. The content of the file is the name current logger (NONE if not set) and a logger can be setup by simply echoing its name to the file. By echoing "NONE" to a /proc/sys/net/netfilter/nf_log/PROTO file, the logger corresponding to this PROTO is set to NULL.

net/netfilter/nf_log.c +84 -1
index 4fcbcc7..8bb998f 100644
@@ -14,6 +14,7 @@
    LOG target modules */
 
 #define NF_LOG_PREFIXLEN		128
+#define NFLOGGER_NAME_LEN		64
 
 static const struct nf_logger *nf_loggers[NFPROTO_NUMPROTO] __read_mostly;
 static struct list_head nf_loggers_l[NFPROTO_NUMPROTO] __read_mostly;
@@ -207,18 +208,100 @@ static const struct file_operations nflog_file_ops = {
 	.release = seq_release,
 };
 
+
 #endif /* PROC_FS */
 
+#ifdef CONFIG_SYSCTL
+struct ctl_path nf_log_sysctl_path[] = {
+	{ .procname = "net", .ctl_name = CTL_NET, },
+	{ .procname = "netfilter", .ctl_name = NET_NETFILTER, },
+	{ .procname = "nf_log", .ctl_name = CTL_UNNUMBERED, },
+	{ }
+};
+
+static char nf_log_sysctl_fnames[NFPROTO_NUMPROTO-NFPROTO_UNSPEC][3];
+static struct ctl_table nf_log_sysctl_table[NFPROTO_NUMPROTO+1];
+static struct ctl_table_header *nf_log_dir_header;
 
-int __init netfilter_log_init(void)
+static int nf_log_proc_dostring(ctl_table *table, int write, struct file *filp,
+			 void *buffer, size_t *lenp, loff_t *ppos)
+{
+	const struct nf_logger *logger;
+	int r = 0;
+	int tindex = (unsigned long)table->extra1;
+
+	if (write) {
+		if (!strcmp(buffer, "NONE")) {
+			nf_log_unbind_pf(tindex);
+			return 0;
+		}
+		mutex_lock(&nf_log_mutex);
+		logger = __find_logger(tindex, buffer);
+		if (logger == NULL) {
+			mutex_unlock(&nf_log_mutex);
+			return -ENOENT;
+		}
+		rcu_assign_pointer(nf_loggers[tindex], logger);
+		mutex_unlock(&nf_log_mutex);
+	} else {
+		rcu_read_lock();
+		logger = rcu_dereference(nf_loggers[tindex]);
+		if (!logger)
+			table->data = "NONE";
+		else
+			table->data = logger->name;
+		r = proc_dostring(table, write, filp, buffer, lenp, ppos);
+		rcu_read_unlock();
+	}
+
+	return r;
+}
+
+static __init int netfilter_log_sysctl_init(void)
 {
 	int i;
+
+	for (i = NFPROTO_UNSPEC; i < NFPROTO_NUMPROTO; i++) {
+		snprintf(nf_log_sysctl_fnames[i-NFPROTO_UNSPEC], 3, "%d", i);
+		nf_log_sysctl_table[i].ctl_name	= CTL_UNNUMBERED;
+		nf_log_sysctl_table[i].procname	=
+			nf_log_sysctl_fnames[i-NFPROTO_UNSPEC];
+		nf_log_sysctl_table[i].data = NULL;
+		nf_log_sysctl_table[i].maxlen =
+			NFLOGGER_NAME_LEN * sizeof(char);
+		nf_log_sysctl_table[i].mode = 0644;
+		nf_log_sysctl_table[i].proc_handler = nf_log_proc_dostring;
+		nf_log_sysctl_table[i].extra1 = (void *)(unsigned long) i;
+	}
+
+	nf_log_dir_header = register_sysctl_paths(nf_log_sysctl_path,
+				       nf_log_sysctl_table);
+	if (!nf_log_dir_header)
+		return -ENOMEM;
+
+	return 0;
+}
+#else
+static __init int netfilter_log_sysctl_init(void)
+{
+	return 0;
+}
+#endif /* CONFIG_SYSCTL */
+
+int __init netfilter_log_init(void)
+{
+	int i, r;
 #ifdef CONFIG_PROC_FS
 	if (!proc_create("nf_log", S_IRUGO,
 			 proc_net_netfilter, &nflog_file_ops))
 		return -1;
 #endif
 
+	/* Errors will trigger panic, unroll on error is unnecessary. */
+	r = netfilter_log_sysctl_init();
+	if (r < 0)
+		return r;
+
 	for (i = NFPROTO_UNSPEC; i < NFPROTO_NUMPROTO; i++)
 		INIT_LIST_HEAD(&(nf_loggers_l[i]));
 
#54 lore raw

nefilter 40/41: nfnetlink: add nfnetlink_set_err and use it in ctnetlink

In reply to Patrick McHardy

commit dd5b6ce6fd465eab90357711c8e8124dc3a31ff0

Author: Pablo Neira Ayuso <[email protected]>
Date:   Mon Mar 23 13:21:06 2009 +0100

nefilter: nfnetlink: add nfnetlink_set_err and use it in ctnetlink

This patch adds nfnetlink_set_err() to propagate the error to netlink broadcast listener in case of memory allocation errors in the message building.

include/linux/netfilter/nfnetlink.h +1 -0
index 7d8e045..135e5cf 100644
@@ -76,6 +76,7 @@ extern int nfnetlink_subsys_unregister(const struct nfnetlink_subsystem *n);
 extern int nfnetlink_has_listeners(unsigned int group);
 extern int nfnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, 
 			  int echo);
+extern void nfnetlink_set_err(u32 pid, u32 group, int error);
 extern int nfnetlink_unicast(struct sk_buff *skb, u_int32_t pid, int flags);
 
 extern void nfnl_lock(void);
net/netfilter/nf_conntrack_netlink.c +2 -0
index d1fe9d1..1b75c9e 100644
@@ -518,6 +518,7 @@ static int ctnetlink_conntrack_event(struct notifier_block *this,
 nla_put_failure:
 	rcu_read_unlock();
 nlmsg_failure:
+	nfnetlink_set_err(0, group, -ENOBUFS);
 	kfree_skb(skb);
 	return NOTIFY_DONE;
 }
@@ -1514,6 +1515,7 @@ static int ctnetlink_expect_event(struct notifier_block *this,
 nla_put_failure:
 	rcu_read_unlock();
 nlmsg_failure:
+	nfnetlink_set_err(0, 0, -ENOBUFS);
 	kfree_skb(skb);
 	return NOTIFY_DONE;
 }
net/netfilter/nfnetlink.c +6 -0
index 9c0ba17..2785d66 100644
@@ -113,6 +113,12 @@ int nfnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
 }
 EXPORT_SYMBOL_GPL(nfnetlink_send);
 
+void nfnetlink_set_err(u32 pid, u32 group, int error)
+{
+	netlink_set_err(nfnl, pid, group, error);
+}
+EXPORT_SYMBOL_GPL(nfnetlink_set_err);
+
 int nfnetlink_unicast(struct sk_buff *skb, u_int32_t pid, int flags)
 {
 	return netlink_unicast(nfnl, skb, pid, flags);
net/netlink/af_netlink.c +1 -0
index 6ee69c2..5b33879 100644
@@ -1106,6 +1106,7 @@ void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code)
 
 	read_unlock(&nl_table_lock);
 }
+EXPORT_SYMBOL(netlink_set_err);
 
 /* must be called with netlink table grabbed */
 static void netlink_update_socket_mc(struct netlink_sock *nlk,
#55 lore raw

netfilter 41/41: nf_conntrack: Reduce conntrack count in nf_conntrack_free()

In reply to Patrick McHardy

commit 1d45209d89e647e9f27e4afa1f47338df73bc112

Author: Eric Dumazet <[email protected]>
Date:   Tue Mar 24 14:26:50 2009 +0100

netfilter: nf_conntrack: Reduce conntrack count in nf_conntrack_free()

We use RCU to defer freeing of conntrack structures. In DOS situation, RCU might accumulate about 10.000 elements per CPU in its internal queues. To get accurate conntrack counts (at the expense of slightly more RAM used), we might consider conntrack counter not taking into account "about to be freed elements, waiting in RCU queues". We thus decrement it in nf_conntrack_free(), not in the RCU callback.

net/netfilter/nf_conntrack_core.c +3 -2
index ebc2756..55befe5 100644
@@ -517,16 +517,17 @@ EXPORT_SYMBOL_GPL(nf_conntrack_alloc);
 static void nf_conntrack_free_rcu(struct rcu_head *head)
 {
 	struct nf_conn *ct = container_of(head, struct nf_conn, rcu);
-	struct net *net = nf_ct_net(ct);
 
 	nf_ct_ext_free(ct);
 	kmem_cache_free(nf_conntrack_cachep, ct);
-	atomic_dec(&net->ct.count);
 }
 
 void nf_conntrack_free(struct nf_conn *ct)
 {
+	struct net *net = nf_ct_net(ct);
+
 	nf_ct_ext_destroy(ct);
+	atomic_dec(&net->ct.count);
 	call_rcu(&ct->rcu, nf_conntrack_free_rcu);
 }
 EXPORT_SYMBOL_GPL(nf_conntrack_free);
maintainer
#56 lore raw
In reply to Patrick McHardy

From: Patrick McHardy <[email protected]>
Date: Tue, 24 Mar 2009 15:03:06 +0100 (MET)

Quoted 3 lines

Please apply or pull from:

git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6.git

I'm going to pull this all in, thanks Patrick.

I'll reply to a few patches for which I'd like to see some followup discussion and change.

Thanks.

#57 lore raw
Quoted 11 lines · David Miller

From: Patrick McHardy <[email protected]>
Date: Tue, 24 Mar 2009 15:03:06 +0100 (MET)

Quoted 3 lines

Please apply or pull from:

git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6.git

I'm going to pull this all in, thanks Patrick.

I'll reply to a few patches for which I'd like to see some followup discussion and change.

Thanks. The discussions seems to be over already, so I'll just collect the resulting patches :)