Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757371AbZDNSSJ (ORCPT ); Tue, 14 Apr 2009 14:18:09 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1757468AbZDNSRj (ORCPT ); Tue, 14 Apr 2009 14:17:39 -0400 Received: from mail.vyatta.com ([76.74.103.46]:58025 "EHLO mail.vyatta.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754931AbZDNSRf (ORCPT ); Tue, 14 Apr 2009 14:17:35 -0400 Date: Tue, 14 Apr 2009 11:17:16 -0700 From: Stephen Hemminger To: Jeff Chua Cc: Eric Dumazet , Patrick McHardy , paulmck@linux.vnet.ibm.com, David Miller , paulus@samba.org, mingo@elte.hu, torvalds@linux-foundation.org, laijs@cn.fujitsu.com, jengelh@medozas.de, r000n@r000n.net, linux-kernel@vger.kernel.org, netfilter-devel@vger.kernel.org, netdev@vger.kernel.org, benh@kernel.crashing.org Subject: [PATCH] netfilter: use per-cpu spinlock rather than RCU (v2) Message-ID: <20090414111716.28585806@nehalam> In-Reply-To: References: <20090411174801.GG6822@linux.vnet.ibm.com> <18913.53699.544083.320542@cargo.ozlabs.ibm.com> <20090412173108.GO6822@linux.vnet.ibm.com> <20090412.181330.23529546.davem@davemloft.net> <20090413040413.GQ6822@linux.vnet.ibm.com> <20090413095309.631cf395@nehalam> <49E48136.5060700@trash.net> <49E49C65.8060808@cosmosbay.com> <20090414074554.7fa73e2f@nehalam> <49E4B0A5.70404@cosmosbay.com> Organization: Vyatta X-Mailer: Claws Mail 3.6.1 (GTK+ 2.16.1; x86_64-pc-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 20358 Lines: 746 Subject: iptables: This is an alternative version of ip/ip6/arp tables locking using per-cpu locks. This avoids the overhead of synchronize_net() during update but still removes the expensive rwlock in earlier versions. The idea for this came from an earlier version done by Eric Duzamet. Locking is done per-cpu, the fast path locks on the current cpu and updates counters. The slow case involves acquiring the locks on all cpu's. The mutex that was added for 2.6.30 in xt_table is unnecessary since there already is a mutex for xt[af].mutex that is held. Signed-off-by: Stephen Hemminger --- include/linux/netfilter/x_tables.h | 5 - net/ipv4/netfilter/arp_tables.c | 112 +++++++++------------------------ net/ipv4/netfilter/ip_tables.c | 123 +++++++++++-------------------------- net/ipv6/netfilter/ip6_tables.c | 119 +++++++++++------------------------ net/netfilter/x_tables.c | 28 -------- 5 files changed, 110 insertions(+), 277 deletions(-) --- a/include/linux/netfilter/x_tables.h 2009-04-14 10:13:59.932292529 -0700 +++ b/include/linux/netfilter/x_tables.h 2009-04-14 10:14:04.121043331 -0700 @@ -354,9 +354,6 @@ struct xt_table /* What hooks you will enter on */ unsigned int valid_hooks; - /* Lock for the curtain */ - struct mutex lock; - /* Man behind the curtain... */ struct xt_table_info *private; @@ -434,8 +431,6 @@ extern void xt_proto_fini(struct net *ne 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); /* * This helper is performance critical and must be inlined --- a/net/ipv4/netfilter/ip_tables.c 2009-04-14 10:13:59.923167357 -0700 +++ b/net/ipv4/netfilter/ip_tables.c 2009-04-14 11:13:51.066830533 -0700 @@ -297,6 +297,8 @@ static void trace_packet(struct sk_buff } #endif +static DEFINE_PER_CPU(spinlock_t, ip_tables_lock); + /* Returns one of the generic firewall policies, like NF_ACCEPT. */ unsigned int ipt_do_table(struct sk_buff *skb, @@ -339,9 +341,13 @@ ipt_do_table(struct sk_buff *skb, IP_NF_ASSERT(table->valid_hooks & (1 << hook)); - rcu_read_lock_bh(); - private = rcu_dereference(table->private); - table_base = rcu_dereference(private->entries[smp_processor_id()]); + /* Not spin_lock_bh() because get_cpu_var + * needs to happen after preempt is disabled + */ + local_bh_disable(); + spin_lock(&__get_cpu_var(ip_tables_lock)); + private = table->private; + table_base = private->entries[smp_processor_id()]; e = get_entry(table_base, private->hook_entry[hook]); @@ -436,8 +442,8 @@ ipt_do_table(struct sk_buff *skb, e = (void *)e + e->next_offset; } } while (!hotdrop); - - rcu_read_unlock_bh(); + spin_unlock(&__get_cpu_var(ip_tables_lock)); + local_bh_enable(); #ifdef DEBUG_ALLOW_ALL return NF_ACCEPT; @@ -902,75 +908,25 @@ get_counters(const struct xt_table_info curcpu = raw_smp_processor_id(); i = 0; + spin_lock_bh(&per_cpu(ip_tables_lock, curcpu)); IPT_ENTRY_ITERATE(t->entries[curcpu], t->size, set_entry_to_counter, counters, &i); + spin_unlock_bh(&per_cpu(ip_tables_lock, curcpu)); for_each_possible_cpu(cpu) { if (cpu == curcpu) continue; i = 0; + spin_lock_bh(&per_cpu(ip_tables_lock, cpu)); IPT_ENTRY_ITERATE(t->entries[cpu], t->size, add_entry_to_counter, counters, &i); - } - -} - -/* We're lazy, and add to the first CPU; overflow works its fey magic - * and everything is OK. */ -static int -add_counter_to_entry(struct ipt_entry *e, - const struct xt_counters addme[], - unsigned int *i) -{ - ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt); - - (*i)++; - return 0; -} - -/* Take values from counters and add them back onto the current cpu */ -static void put_counters(struct xt_table_info *t, - const struct xt_counters counters[]) -{ - unsigned int i, cpu; - - local_bh_disable(); - cpu = smp_processor_id(); - i = 0; - IPT_ENTRY_ITERATE(t->entries[cpu], - t->size, - add_counter_to_entry, - counters, - &i); - local_bh_enable(); -} - - -static inline int -zero_entry_counter(struct ipt_entry *e, void *arg) -{ - e->counters.bcnt = 0; - e->counters.pcnt = 0; - return 0; -} - -static void -clone_counters(struct xt_table_info *newinfo, const struct xt_table_info *info) -{ - unsigned int cpu; - const void *loc_cpu_entry = info->entries[raw_smp_processor_id()]; - - memcpy(newinfo, info, offsetof(struct xt_table_info, entries)); - for_each_possible_cpu(cpu) { - memcpy(newinfo->entries[cpu], loc_cpu_entry, info->size); - IPT_ENTRY_ITERATE(newinfo->entries[cpu], newinfo->size, - zero_entry_counter, NULL); + spin_unlock_bh(&per_cpu(ip_tables_lock, cpu)); } } @@ -979,7 +935,6 @@ static struct xt_counters * alloc_counte unsigned int countersize; struct xt_counters *counters; struct xt_table_info *private = table->private; - struct xt_table_info *info; /* We need atomic snapshot of counters: rest doesn't change (other than comefrom, which userspace doesn't care @@ -990,26 +945,10 @@ static struct xt_counters * alloc_counte if (counters == NULL) goto nomem; - info = xt_alloc_table_info(private->size); - if (!info) - goto free_counters; - - clone_counters(info, private); - - mutex_lock(&table->lock); - xt_table_entry_swap_rcu(private, info); - synchronize_net(); /* Wait until smoke has cleared */ - - get_counters(info, counters); - put_counters(private, counters); - mutex_unlock(&table->lock); - - xt_free_table_info(info); + get_counters(private, counters); return counters; - free_counters: - vfree(counters); nomem: return ERR_PTR(-ENOMEM); } @@ -1377,6 +1316,18 @@ do_replace(struct net *net, void __user return ret; } +/* We're lazy, and add to the first CPU; overflow works its fey magic + * and everything is OK. */ +static int +add_counter_to_entry(struct ipt_entry *e, + const struct xt_counters addme[], + unsigned int *i) +{ + ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt); + + (*i)++; + return 0; +} static int do_add_counters(struct net *net, void __user *user, unsigned int len, int compat) @@ -1437,25 +1388,26 @@ do_add_counters(struct net *net, void __ goto free; } - mutex_lock(&t->lock); private = t->private; if (private->number != num_counters) { ret = -EINVAL; goto unlock_up_free; } - preempt_disable(); - i = 0; /* Choose the copy that is on our node */ - loc_cpu_entry = private->entries[raw_smp_processor_id()]; + local_bh_disable(); + spin_lock(&__get_cpu_var(ip_tables_lock)); + loc_cpu_entry = private->entries[smp_processor_id()]; + i = 0; IPT_ENTRY_ITERATE(loc_cpu_entry, private->size, add_counter_to_entry, paddc, &i); - preempt_enable(); + spin_unlock(&__get_cpu_var(ip_tables_lock)); + local_bh_enable(); + unlock_up_free: - mutex_unlock(&t->lock); xt_table_unlock(t); module_put(t->me); free: @@ -2272,7 +2224,10 @@ static struct pernet_operations ip_table static int __init ip_tables_init(void) { - int ret; + int cpu, ret; + + for_each_possible_cpu(cpu) + spin_lock_init(&per_cpu(ip_tables_lock, cpu)); ret = register_pernet_subsys(&ip_tables_net_ops); if (ret < 0) --- a/net/netfilter/x_tables.c 2009-04-14 10:13:59.901168540 -0700 +++ b/net/netfilter/x_tables.c 2009-04-14 10:14:04.123043143 -0700 @@ -625,20 +625,6 @@ void xt_free_table_info(struct xt_table_ } 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) @@ -682,26 +668,19 @@ xt_replace_table(struct xt_table *table, struct xt_table_info *newinfo, int *error) { - struct xt_table_info *oldinfo, *private; + struct xt_table_info *private; /* Do the substitution. */ - 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); - mutex_unlock(&table->lock); *error = -EAGAIN; return NULL; } - oldinfo = private; - rcu_assign_pointer(table->private, newinfo); - newinfo->initial_entries = oldinfo->initial_entries; - mutex_unlock(&table->lock); - - synchronize_net(); - return oldinfo; + newinfo->initial_entries = private->initial_entries; + return xchg(&table->private, newinfo); } EXPORT_SYMBOL_GPL(xt_replace_table); @@ -734,7 +713,6 @@ struct xt_table *xt_register_table(struc /* Simplifies replace_table code. */ table->private = bootstrap; - mutex_init(&table->lock); if (!xt_replace_table(table, 0, newinfo, &ret)) goto unlock; --- a/net/ipv6/netfilter/ip6_tables.c 2009-04-14 10:13:59.909167832 -0700 +++ b/net/ipv6/netfilter/ip6_tables.c 2009-04-14 11:15:15.405141464 -0700 @@ -329,6 +329,8 @@ static void trace_packet(struct sk_buff } #endif +static DEFINE_PER_CPU(spinlock_t, ip6_tables_lock); + /* Returns one of the generic firewall policies, like NF_ACCEPT. */ unsigned int ip6t_do_table(struct sk_buff *skb, @@ -365,9 +367,10 @@ ip6t_do_table(struct sk_buff *skb, IP_NF_ASSERT(table->valid_hooks & (1 << hook)); - rcu_read_lock_bh(); - private = rcu_dereference(table->private); - table_base = rcu_dereference(private->entries[smp_processor_id()]); + local_bh_disable(); + spin_lock(&__get_cpu_var(ip6_tables_lock)); + private = table->private; + table_base = private->entries[smp_processor_id()]; e = get_entry(table_base, private->hook_entry[hook]); @@ -466,7 +469,8 @@ ip6t_do_table(struct sk_buff *skb, #ifdef CONFIG_NETFILTER_DEBUG ((struct ip6t_entry *)table_base)->comefrom = NETFILTER_LINK_POISON; #endif - rcu_read_unlock_bh(); + spin_unlock(&__get_cpu_var(ip6_tables_lock)); + local_bh_enable(); #ifdef DEBUG_ALLOW_ALL return NF_ACCEPT; @@ -931,73 +935,25 @@ get_counters(const struct xt_table_info curcpu = raw_smp_processor_id(); i = 0; + spin_lock_bh(&per_cpu(ip6_tables_lock, curcpu)); IP6T_ENTRY_ITERATE(t->entries[curcpu], t->size, set_entry_to_counter, counters, &i); + spin_unlock_bh(&per_cpu(ip6_tables_lock, curcpu)); for_each_possible_cpu(cpu) { if (cpu == curcpu) continue; i = 0; + spin_lock_bh(&per_cpu(ip6_tables_lock, cpu)); IP6T_ENTRY_ITERATE(t->entries[cpu], t->size, add_entry_to_counter, counters, &i); - } -} - -/* We're lazy, and add to the first CPU; overflow works its fey magic - * and everything is OK. */ -static int -add_counter_to_entry(struct ip6t_entry *e, - const struct xt_counters addme[], - unsigned int *i) -{ - ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt); - - (*i)++; - return 0; -} - -/* Take values from counters and add them back onto the current cpu */ -static void put_counters(struct xt_table_info *t, - const struct xt_counters counters[]) -{ - unsigned int i, cpu; - - local_bh_disable(); - cpu = smp_processor_id(); - i = 0; - IP6T_ENTRY_ITERATE(t->entries[cpu], - t->size, - add_counter_to_entry, - counters, - &i); - local_bh_enable(); -} - -static inline int -zero_entry_counter(struct ip6t_entry *e, void *arg) -{ - e->counters.bcnt = 0; - e->counters.pcnt = 0; - return 0; -} - -static void -clone_counters(struct xt_table_info *newinfo, const struct xt_table_info *info) -{ - unsigned int cpu; - const void *loc_cpu_entry = info->entries[raw_smp_processor_id()]; - - memcpy(newinfo, info, offsetof(struct xt_table_info, entries)); - for_each_possible_cpu(cpu) { - memcpy(newinfo->entries[cpu], loc_cpu_entry, info->size); - IP6T_ENTRY_ITERATE(newinfo->entries[cpu], newinfo->size, - zero_entry_counter, NULL); + spin_unlock_bh(&per_cpu(ip6_tables_lock, cpu)); } } @@ -1006,7 +962,6 @@ static struct xt_counters *alloc_counter unsigned int countersize; struct xt_counters *counters; struct xt_table_info *private = table->private; - struct xt_table_info *info; /* We need atomic snapshot of counters: rest doesn't change (other than comefrom, which userspace doesn't care @@ -1017,26 +972,9 @@ static struct xt_counters *alloc_counter if (counters == NULL) goto nomem; - info = xt_alloc_table_info(private->size); - if (!info) - goto free_counters; - - clone_counters(info, private); - - mutex_lock(&table->lock); - xt_table_entry_swap_rcu(private, info); - synchronize_net(); /* Wait until smoke has cleared */ - - get_counters(info, counters); - put_counters(private, counters); - mutex_unlock(&table->lock); - - xt_free_table_info(info); - + get_counters(private, counters); return counters; - free_counters: - vfree(counters); nomem: return ERR_PTR(-ENOMEM); } @@ -1405,6 +1343,19 @@ do_replace(struct net *net, void __user return ret; } +/* We're lazy, and add to the first CPU; overflow works its fey magic + * and everything is OK. */ +static int +add_counter_to_entry(struct ip6t_entry *e, + const struct xt_counters addme[], + unsigned int *i) +{ + ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt); + + (*i)++; + return 0; +} + static int do_add_counters(struct net *net, void __user *user, unsigned int len, int compat) @@ -1465,25 +1416,26 @@ do_add_counters(struct net *net, void __ goto free; } - mutex_lock(&t->lock); private = t->private; if (private->number != num_counters) { ret = -EINVAL; goto unlock_up_free; } - preempt_disable(); - i = 0; + local_bh_disable(); /* Choose the copy that is on our node */ - loc_cpu_entry = private->entries[raw_smp_processor_id()]; + spin_lock(&__get_cpu_var(ip6_tables_lock)); + loc_cpu_entry = private->entries[smp_processor_id()]; + i = 0; IP6T_ENTRY_ITERATE(loc_cpu_entry, private->size, add_counter_to_entry, paddc, &i); - preempt_enable(); + spin_unlock(&__get_cpu_var(ip6_tables_lock)); + local_bh_enable(); + unlock_up_free: - mutex_unlock(&t->lock); xt_table_unlock(t); module_put(t->me); free: @@ -2298,7 +2250,10 @@ static struct pernet_operations ip6_tabl static int __init ip6_tables_init(void) { - int ret; + int cpu, ret; + + for_each_possible_cpu(cpu) + spin_lock_init(&per_cpu(ip6_tables_lock, cpu)); ret = register_pernet_subsys(&ip6_tables_net_ops); if (ret < 0) --- a/net/ipv4/netfilter/arp_tables.c 2009-04-14 10:13:59.915167635 -0700 +++ b/net/ipv4/netfilter/arp_tables.c 2009-04-14 11:14:36.203830871 -0700 @@ -231,6 +231,8 @@ static inline struct arpt_entry *get_ent return (struct arpt_entry *)(base + offset); } +static DEFINE_PER_CPU(spinlock_t, arp_tables_lock); + unsigned int arpt_do_table(struct sk_buff *skb, unsigned int hook, const struct net_device *in, @@ -253,9 +255,10 @@ unsigned int arpt_do_table(struct sk_buf indev = in ? in->name : nulldevname; outdev = out ? out->name : nulldevname; - rcu_read_lock_bh(); - private = rcu_dereference(table->private); - table_base = rcu_dereference(private->entries[smp_processor_id()]); + local_bh_disable(); + spin_lock(&__get_cpu_var(arp_tables_lock)); + private = table->private; + table_base = private->entries[smp_processor_id()]; e = get_entry(table_base, private->hook_entry[hook]); back = get_entry(table_base, private->underflow[hook]); @@ -328,8 +331,8 @@ unsigned int arpt_do_table(struct sk_buf e = (void *)e + e->next_offset; } } while (!hotdrop); - - rcu_read_unlock_bh(); + spin_unlock(&__get_cpu_var(arp_tables_lock)); + local_bh_enable(); if (hotdrop) return NF_DROP; @@ -716,74 +719,25 @@ static void get_counters(const struct xt curcpu = raw_smp_processor_id(); i = 0; + spin_lock_bh(&per_cpu(arp_tables_lock, curcpu)); ARPT_ENTRY_ITERATE(t->entries[curcpu], t->size, set_entry_to_counter, counters, &i); + spin_unlock_bh(&per_cpu(arp_tables_lock, curcpu)); for_each_possible_cpu(cpu) { if (cpu == curcpu) continue; i = 0; + spin_lock_bh(&per_cpu(arp_tables_lock, cpu)); ARPT_ENTRY_ITERATE(t->entries[cpu], t->size, add_entry_to_counter, counters, &i); - } -} - - -/* We're lazy, and add to the first CPU; overflow works its fey magic - * and everything is OK. */ -static int -add_counter_to_entry(struct arpt_entry *e, - const struct xt_counters addme[], - unsigned int *i) -{ - ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt); - - (*i)++; - return 0; -} - -/* Take values from counters and add them back onto the current cpu */ -static void put_counters(struct xt_table_info *t, - const struct xt_counters counters[]) -{ - unsigned int i, cpu; - - local_bh_disable(); - cpu = smp_processor_id(); - i = 0; - ARPT_ENTRY_ITERATE(t->entries[cpu], - t->size, - add_counter_to_entry, - counters, - &i); - local_bh_enable(); -} - -static inline int -zero_entry_counter(struct arpt_entry *e, void *arg) -{ - e->counters.bcnt = 0; - e->counters.pcnt = 0; - return 0; -} - -static void -clone_counters(struct xt_table_info *newinfo, const struct xt_table_info *info) -{ - unsigned int cpu; - const void *loc_cpu_entry = info->entries[raw_smp_processor_id()]; - - memcpy(newinfo, info, offsetof(struct xt_table_info, entries)); - for_each_possible_cpu(cpu) { - memcpy(newinfo->entries[cpu], loc_cpu_entry, info->size); - ARPT_ENTRY_ITERATE(newinfo->entries[cpu], newinfo->size, - zero_entry_counter, NULL); + spin_unlock_bh(&per_cpu(arp_tables_lock, cpu)); } } @@ -792,7 +746,6 @@ static struct xt_counters *alloc_counter unsigned int countersize; struct xt_counters *counters; struct xt_table_info *private = table->private; - struct xt_table_info *info; /* We need atomic snapshot of counters: rest doesn't change * (other than comefrom, which userspace doesn't care @@ -804,26 +757,10 @@ static struct xt_counters *alloc_counter if (counters == NULL) goto nomem; - info = xt_alloc_table_info(private->size); - if (!info) - goto free_counters; - - clone_counters(info, private); - - mutex_lock(&table->lock); - xt_table_entry_swap_rcu(private, info); - synchronize_net(); /* Wait until smoke has cleared */ - - get_counters(info, counters); - put_counters(private, counters); - mutex_unlock(&table->lock); - - xt_free_table_info(info); + get_counters(private, counters); return counters; - free_counters: - vfree(counters); nomem: return ERR_PTR(-ENOMEM); } @@ -1165,6 +1102,19 @@ static int do_replace(struct net *net, v return ret; } +/* We're lazy, and add to the first CPU; overflow works its fey magic + * and everything is OK. */ +static int +add_counter_to_entry(struct arpt_entry *e, + const struct xt_counters addme[], + unsigned int *i) +{ + ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt); + + (*i)++; + return 0; +} + static int do_add_counters(struct net *net, void __user *user, unsigned int len, int compat) { @@ -1224,25 +1174,25 @@ static int do_add_counters(struct net *n goto free; } - mutex_lock(&t->lock); private = t->private; if (private->number != num_counters) { ret = -EINVAL; goto unlock_up_free; } - preempt_disable(); - i = 0; /* Choose the copy that is on our node */ + local_bh_disable(); + spin_lock(&__get_cpu_var(arp_tables_lock)); loc_cpu_entry = private->entries[smp_processor_id()]; + i = 0; ARPT_ENTRY_ITERATE(loc_cpu_entry, private->size, add_counter_to_entry, paddc, &i); - preempt_enable(); + spin_unlock(&__get_cpu_var(arp_tables_lock)); + local_bh_enable(); unlock_up_free: - mutex_unlock(&t->lock); xt_table_unlock(t); module_put(t->me); -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/