net-Fix-netfilter-percpu-assumptions-for-real.patch

From da72b214fd92086186c95fa11bc35ffd36a14044 Mon Sep 17 00:00:00 2001
From: Thomas Gleixner <tglx@linutronix.de>
Date: Sun, 4 Oct 2009 16:58:07 +0200
Subject: [PATCH] net: Fix netfilter percpu assumptions for real

commit 00ef66ebb37437ca205d06224e7b956d205f7886 in tip.

commit 21ece08 (net: fix the xtables smp_processor_id assumptions for
-rt) fixed only half of the problem. The filter functions might run in
thread context and can be preempted and migrated on -RT.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

diff --git a/include/linux/netfilter/x_tables.h b/include/linux/netfilter/x_tables.h
index 5c588ca..dcc03d7 100644
--- a/include/linux/netfilter/x_tables.h
+++ b/include/linux/netfilter/x_tables.h
@@ -482,22 +482,35 @@ DECLARE_PER_CPU(struct xt_info_lock, xt_info_locks);
  * _Only_ that special combination of being per-cpu and never getting
  * re-entered asynchronously means that the count is safe.
  */
-static inline void xt_info_rdlock_bh(void)
+static inline int xt_info_rdlock_bh(void)
 {
 	struct xt_info_lock *lock;
+	int cpu;
 
 	local_bh_disable();
-	lock = &__raw_get_cpu_var(xt_info_locks);
-	if (likely(!lock->readers++))
+	preempt_disable_rt();
+	cpu = smp_processor_id();
+	lock = &per_cpu(xt_info_locks, cpu);
+	if (likely(!lock->readers++)) {
+		preempt_enable_rt();
 		spin_lock(&lock->lock);
+	} else
+		preempt_enable_rt();
+	return cpu;
 }
 
-static inline void xt_info_rdunlock_bh(void)
+static inline void xt_info_rdunlock_bh(int cpu)
 {
-	struct xt_info_lock *lock = &__raw_get_cpu_var(xt_info_locks);
+	struct xt_info_lock *lock = &per_cpu(xt_info_locks, cpu);
 
-	if (likely(!--lock->readers))
+	preempt_disable_rt();
+
+	if (likely(!--lock->readers)) {
+		preempt_enable_rt();
 		spin_unlock(&lock->lock);
+	} else
+		preempt_enable_rt();
+
 	local_bh_enable();
 }
 
diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c
index 9438d18..4afb6c8 100644
--- a/net/ipv4/netfilter/arp_tables.c
+++ b/net/ipv4/netfilter/arp_tables.c
@@ -266,6 +266,7 @@ unsigned int arpt_do_table(struct sk_buff *skb,
 	void *table_base;
 	const struct xt_table_info *private;
 	struct xt_target_param tgpar;
+	int cpu;
 
 	if (!pskb_may_pull(skb, arp_hdr_len(skb->dev)))
 		return NF_DROP;
@@ -273,9 +274,9 @@ unsigned int arpt_do_table(struct sk_buff *skb,
 	indev = in ? in->name : nulldevname;
 	outdev = out ? out->name : nulldevname;
 
-	xt_info_rdlock_bh();
+	cpu = xt_info_rdlock_bh();
 	private = table->private;
-	table_base = private->entries[raw_smp_processor_id()];
+	table_base = private->entries[cpu];
 
 	e = get_entry(table_base, private->hook_entry[hook]);
 	back = get_entry(table_base, private->underflow[hook]);
@@ -346,7 +347,7 @@ unsigned int arpt_do_table(struct sk_buff *skb,
 			/* Verdict */
 			break;
 	} while (!hotdrop);
-	xt_info_rdunlock_bh();
+	xt_info_rdunlock_bh(cpu);
 
 	if (hotdrop)
 		return NF_DROP;
diff --git a/net/ipv4/netfilter/ip_tables.c b/net/ipv4/netfilter/ip_tables.c
index 83501d8..1c33431 100644
--- a/net/ipv4/netfilter/ip_tables.c
+++ b/net/ipv4/netfilter/ip_tables.c
@@ -335,6 +335,7 @@ ipt_do_table(struct sk_buff *skb,
 	const struct xt_table_info *private;
 	struct xt_match_param mtpar;
 	struct xt_target_param tgpar;
+	int cpu;
 
 	/* Initialization */
 	ip = ip_hdr(skb);
@@ -355,9 +356,9 @@ ipt_do_table(struct sk_buff *skb,
 	mtpar.hooknum = tgpar.hooknum = hook;
 
 	IP_NF_ASSERT(table->valid_hooks & (1 << hook));
-	xt_info_rdlock_bh();
+	cpu = xt_info_rdlock_bh();
 	private = table->private;
-	table_base = private->entries[raw_smp_processor_id()];
+	table_base = private->entries[cpu];
 
 	e = get_entry(table_base, private->hook_entry[hook]);
 
@@ -447,7 +448,7 @@ ipt_do_table(struct sk_buff *skb,
 			/* Verdict */
 			break;
 	} while (!hotdrop);
-	xt_info_rdunlock_bh();
+	xt_info_rdunlock_bh(cpu);
 
 #ifdef DEBUG_ALLOW_ALL
 	return NF_ACCEPT;
diff --git a/net/ipv6/netfilter/ip6_tables.c b/net/ipv6/netfilter/ip6_tables.c
index 66c78d0..6d5b78b 100644
--- a/net/ipv6/netfilter/ip6_tables.c
+++ b/net/ipv6/netfilter/ip6_tables.c
@@ -364,6 +364,7 @@ ip6t_do_table(struct sk_buff *skb,
 	const struct xt_table_info *private;
 	struct xt_match_param mtpar;
 	struct xt_target_param tgpar;
+	int cpu;
 
 	/* Initialization */
 	indev = in ? in->name : nulldevname;
@@ -382,9 +383,9 @@ ip6t_do_table(struct sk_buff *skb,
 
 	IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 
-	xt_info_rdlock_bh();
+	cpu = xt_info_rdlock_bh();
 	private = table->private;
-	table_base = private->entries[raw_smp_processor_id()];
+	table_base = private->entries[cpu];
 
 	e = get_entry(table_base, private->hook_entry[hook]);
 
@@ -478,7 +479,7 @@ ip6t_do_table(struct sk_buff *skb,
 #ifdef CONFIG_NETFILTER_DEBUG
 	tb_comefrom = NETFILTER_LINK_POISON;
 #endif
-	xt_info_rdunlock_bh();
+	xt_info_rdunlock_bh(cpu);
 
 #ifdef DEBUG_ALLOW_ALL
 	return NF_ACCEPT;
-- 
1.7.1.1