[PATCH nf-next] x_tables: remove XT_TABLE_INFO_SZ and a dereference.

[PATCH nf-next] x_tables: remove XT_TABLE_INFO_SZ and a dereference.

[Eric Dumazet]

From: Eric Dumazet <edumazet@google.com>

After Florian patches, there is no need for XT_TABLE_INFO_SZ anymore :
Only one copy of table is kept, instead of one copy per cpu.

We also can avoid a dereference if we put table data right after
xt_table_info. It reduces register pressure and helps compiler.

Then, we attempt a kmalloc() if total size is under order-3 allocation,
to reduce TLB pressure, as in many cases, rules fit in 32 KB.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
---
 include/linux/netfilter/x_tables.h |    5 ----
 net/ipv4/netfilter/arp_tables.c    |    4 +--
 net/ipv4/netfilter/ip_tables.c     |    4 +--
 net/ipv6/netfilter/ip6_tables.c    |    4 +--
 net/netfilter/x_tables.c           |   32 ++++++++++-----------------
 5 files changed, 19 insertions(+), 30 deletions(-)

diff --git a/include/linux/netfilter/x_tables.h b/include/linux/netfilter/x_tables.h
index 9969d79dcde1..95693c4cebdd 100644
--- a/include/linux/netfilter/x_tables.h
+++ b/include/linux/netfilter/x_tables.h
@@ -225,12 +225,9 @@ struct xt_table_info {
 	unsigned int __percpu *stackptr;
 	void ***jumpstack;
 
-	/* Note : this field MUST be the last one, see XT_TABLE_INFO_SZ */
-	void *entries;
+	unsigned char entries[0] __aligned(8);
 };
 
-#define XT_TABLE_INFO_SZ (offsetof(struct xt_table_info, entries) \
-			  + nr_cpu_ids * sizeof(char *))
 int xt_register_target(struct xt_target *target);
 void xt_unregister_target(struct xt_target *target);
 int xt_register_targets(struct xt_target *target, unsigned int n);
diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c
index d75c139393fc..95c9b6eece25 100644
--- a/net/ipv4/netfilter/arp_tables.c
+++ b/net/ipv4/netfilter/arp_tables.c
@@ -256,7 +256,7 @@ unsigned int arpt_do_table(struct sk_buff *skb,
 	const struct arphdr *arp;
 	struct arpt_entry *e, *back;
 	const char *indev, *outdev;
-	void *table_base;
+	const void *table_base;
 	const struct xt_table_info *private;
 	struct xt_action_param acpar;
 	unsigned int addend;
@@ -868,7 +868,7 @@ static int compat_table_info(const struct xt_table_info *info,
 			     struct xt_table_info *newinfo)
 {
 	struct arpt_entry *iter;
-	void *loc_cpu_entry;
+	const void *loc_cpu_entry;
 	int ret;
 
 	if (!newinfo || !info)
diff --git a/net/ipv4/netfilter/ip_tables.c b/net/ipv4/netfilter/ip_tables.c
index 6151500c2df8..6c72fbb7b49e 100644
--- a/net/ipv4/netfilter/ip_tables.c
+++ b/net/ipv4/netfilter/ip_tables.c
@@ -938,7 +938,7 @@ copy_entries_to_user(unsigned int total_size,
 	struct xt_counters *counters;
 	const struct xt_table_info *private = table->private;
 	int ret = 0;
-	void *loc_cpu_entry;
+	const void *loc_cpu_entry;
 
 	counters = alloc_counters(table);
 	if (IS_ERR(counters))
@@ -1052,7 +1052,7 @@ static int compat_table_info(const struct xt_table_info *info,
 			     struct xt_table_info *newinfo)
 {
 	struct ipt_entry *iter;
-	void *loc_cpu_entry;
+	const void *loc_cpu_entry;
 	int ret;
 
 	if (!newinfo || !info)
diff --git a/net/ipv6/netfilter/ip6_tables.c b/net/ipv6/netfilter/ip6_tables.c
index 80a7f0d38aa4..3c35ced39b42 100644
--- a/net/ipv6/netfilter/ip6_tables.c
+++ b/net/ipv6/netfilter/ip6_tables.c
@@ -951,7 +951,7 @@ copy_entries_to_user(unsigned int total_size,
 	struct xt_counters *counters;
 	const struct xt_table_info *private = table->private;
 	int ret = 0;
-	void *loc_cpu_entry;
+	const void *loc_cpu_entry;
 
 	counters = alloc_counters(table);
 	if (IS_ERR(counters))
@@ -1065,7 +1065,7 @@ static int compat_table_info(const struct xt_table_info *info,
 			     struct xt_table_info *newinfo)
 {
 	struct ip6t_entry *iter;
-	void *loc_cpu_entry;
+	const void *loc_cpu_entry;
 	int ret;
 
 	if (!newinfo || !info)
diff --git a/net/netfilter/x_tables.c b/net/netfilter/x_tables.c
index 6062ce3e862c..d324fe71260c 100644
--- a/net/netfilter/x_tables.c
+++ b/net/netfilter/x_tables.c
@@ -658,29 +658,23 @@ EXPORT_SYMBOL_GPL(xt_compat_target_to_user);
 
 struct xt_table_info *xt_alloc_table_info(unsigned int size)
 {
-	struct xt_table_info *newinfo;
+	struct xt_table_info *info = NULL;
+	size_t sz = sizeof(*info) + size;
 
 	/* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */
 	if ((SMP_ALIGN(size) >> PAGE_SHIFT) + 2 > totalram_pages)
 		return NULL;
 
-	newinfo = kzalloc(XT_TABLE_INFO_SZ, GFP_KERNEL);
-	if (!newinfo)
-		return NULL;
-
-	newinfo->size = size;
-
-	if (size <= PAGE_SIZE)
-		newinfo->entries = kmalloc(size, GFP_KERNEL);
-	else
-		newinfo->entries = vmalloc(size);
-
-	if (newinfo->entries == NULL) {
-		xt_free_table_info(newinfo);
-		return NULL;
+	if (sz <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
+		info = kmalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY);
+	if (!info) {
+		info = vmalloc(sz);
+		if (!info)
+			return NULL;
 	}
-
-	return newinfo;
+	memset(info, 0, sizeof(*info));
+	info->size = size;
+	return info;
 }
 EXPORT_SYMBOL(xt_alloc_table_info);
 
@@ -688,8 +682,6 @@ void xt_free_table_info(struct xt_table_info *info)
 {
 	int cpu;
 
-	kvfree(info->entries);
-
 	if (info->jumpstack != NULL) {
 		for_each_possible_cpu(cpu)
 			kvfree(info->jumpstack[cpu]);
@@ -698,7 +690,7 @@ void xt_free_table_info(struct xt_table_info *info)
 
 	free_percpu(info->stackptr);
 
-	kfree(info);
+	kvfree(info);
 }
 EXPORT_SYMBOL(xt_free_table_info);
 

Re: [PATCH nf-next] x_tables: remove XT_TABLE_INFO_SZ and a dereference.

[Pablo Neira Ayuso]

On Mon, Jun 15, 2015 at 09:57:30AM -0700, Eric Dumazet wrote:
> From: Eric Dumazet <edumazet@google.com>
> 
> After Florian patches, there is no need for XT_TABLE_INFO_SZ anymore :
> Only one copy of table is kept, instead of one copy per cpu.
> 
> We also can avoid a dereference if we put table data right after
> xt_table_info. It reduces register pressure and helps compiler.
> 
> Then, we attempt a kmalloc() if total size is under order-3 allocation,
> to reduce TLB pressure, as in many cases, rules fit in 32 KB.

Applied, thanks a lot Eric.