netfilter: xtables: use percpu rule counters

The binary arp/ip/ip6tables ruleset is stored per cpu.

The only reason left as to why we need percpu duplication are the rule
counters embedded into ipt_entry et al -- since each cpu has its own copy
of the rules, all counters can be lockless.

The downside is that the more cpus are supported, the more memory is
required.  Rules are not just duplicated per online cpu but for each
possible cpu, i.e. if maxcpu is 144, then rule is duplicated 144 times,
not for the e.g. 64 cores present.

To save some memory and also improve utilization of shared caches it
would be preferable to only store the rule blob once.

So we first need to separate counters and the rule blob.

Instead of using entry->counters, allocate this percpu and store the
percpu address in entry->counters.pcnt on CONFIG_SMP.

This change makes no sense as-is; it is merely an intermediate step to
remove the percpu duplication of the rule set in a followup patch.

Suggested-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reported-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

2 files changed, 54 insertions, 9 deletions

diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c
index a61200754f4b..0ada09ae6e81 100644
--- a/net/ipv4/netfilter/arp_tables.c
+++ b/net/ipv4/netfilter/arp_tables.c
@@ -289,13 +289,15 @@ unsigned int arpt_do_table(struct sk_buff *skb,
 	arp = arp_hdr(skb);
 	do {
 		const struct xt_entry_target *t;
+		struct xt_counters *counter;
 
 		if (!arp_packet_match(arp, skb->dev, indev, outdev, &e->arp)) {
 			e = arpt_next_entry(e);
 			continue;
 		}
 
-		ADD_COUNTER(e->counters, arp_hdr_len(skb->dev), 1);
+		counter = xt_get_this_cpu_counter(&e->counters);
+		ADD_COUNTER(*counter, arp_hdr_len(skb->dev), 1);
 
 		t = arpt_get_target_c(e);
 
@@ -521,6 +523,10 @@ find_check_entry(struct arpt_entry *e, const char *name, unsigned int size)
 	if (ret)
 		return ret;
 
+	e->counters.pcnt = xt_percpu_counter_alloc();
+	if (IS_ERR_VALUE(e->counters.pcnt))
+		return -ENOMEM;
+
 	t = arpt_get_target(e);
 	target = xt_request_find_target(NFPROTO_ARP, t->u.user.name,
 					t->u.user.revision);
@@ -538,6 +544,8 @@ find_check_entry(struct arpt_entry *e, const char *name, unsigned int size)
 err:
 	module_put(t->u.kernel.target->me);
 out:
+	xt_percpu_counter_free(e->counters.pcnt);
+
 	return ret;
 }
 
@@ -614,6 +622,7 @@ static inline void cleanup_entry(struct arpt_entry *e)
 	if (par.target->destroy != NULL)
 		par.target->destroy(&par);
 	module_put(par.target->me);
+	xt_percpu_counter_free(e->counters.pcnt);
 }
 
 /* Checks and translates the user-supplied table segment (held in
@@ -723,13 +732,15 @@ static void get_counters(const struct xt_table_info *t,
 
 		i = 0;
 		xt_entry_foreach(iter, t->entries[cpu], t->size) {
+			struct xt_counters *tmp;
 			u64 bcnt, pcnt;
 			unsigned int start;
 
+			tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 			do {
 				start = read_seqcount_begin(s);
-				bcnt = iter->counters.bcnt;
-				pcnt = iter->counters.pcnt;
+				bcnt = tmp->bcnt;
+				pcnt = tmp->pcnt;
 			} while (read_seqcount_retry(s, start));
 
 			ADD_COUNTER(counters[i], bcnt, pcnt);
@@ -1186,7 +1197,10 @@ static int do_add_counters(struct net *net, const void __user *user,
 	loc_cpu_entry = private->entries[curcpu];
 	addend = xt_write_recseq_begin();
 	xt_entry_foreach(iter, loc_cpu_entry, private->size) {
-		ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
+		struct xt_counters *tmp;
+
+		tmp = xt_get_this_cpu_counter(&iter->counters);
+		ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
 		++i;
 	}
 	xt_write_recseq_end(addend);
@@ -1416,9 +1430,17 @@ static int translate_compat_table(const char *name,
 
 	i = 0;
 	xt_entry_foreach(iter1, entry1, newinfo->size) {
+		iter1->counters.pcnt = xt_percpu_counter_alloc();
+		if (IS_ERR_VALUE(iter1->counters.pcnt)) {
+			ret = -ENOMEM;
+			break;
+		}
+
 		ret = check_target(iter1, name);
-		if (ret != 0)
+		if (ret != 0) {
+			xt_percpu_counter_free(iter1->counters.pcnt);
 			break;
+		}
 		++i;
 		if (strcmp(arpt_get_target(iter1)->u.user.name,
 		    XT_ERROR_TARGET) == 0)
diff --git a/net/ipv4/netfilter/ip_tables.c b/net/ipv4/netfilter/ip_tables.c
index e7abf5145edc..d190b10dabfe 100644
--- a/net/ipv4/netfilter/ip_tables.c
+++ b/net/ipv4/netfilter/ip_tables.c
@@ -345,6 +345,7 @@ ipt_do_table(struct sk_buff *skb,
 	do {
 		const struct xt_entry_target *t;
 		const struct xt_entry_match *ematch;
+		struct xt_counters *counter;
 
 		IP_NF_ASSERT(e);
 		if (!ip_packet_match(ip, indev, outdev,
@@ -361,7 +362,8 @@ ipt_do_table(struct sk_buff *skb,
 				goto no_match;
 		}
 
-		ADD_COUNTER(e->counters, skb->len, 1);
+		counter = xt_get_this_cpu_counter(&e->counters);
+		ADD_COUNTER(*counter, skb->len, 1);
 
 		t = ipt_get_target(e);
 		IP_NF_ASSERT(t->u.kernel.target);
@@ -665,6 +667,10 @@ find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 	if (ret)
 		return ret;
 
+	e->counters.pcnt = xt_percpu_counter_alloc();
+	if (IS_ERR_VALUE(e->counters.pcnt))
+		return -ENOMEM;
+
 	j = 0;
 	mtpar.net	= net;
 	mtpar.table     = name;
@@ -691,6 +697,7 @@ find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 	ret = check_target(e, net, name);
 	if (ret)
 		goto err;
+
 	return 0;
  err:
 	module_put(t->u.kernel.target->me);
@@ -700,6 +707,9 @@ find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 			break;
 		cleanup_match(ematch, net);
 	}
+
+	xt_percpu_counter_free(e->counters.pcnt);
+
 	return ret;
 }
 
@@ -784,6 +794,7 @@ cleanup_entry(struct ipt_entry *e, struct net *net)
 	if (par.target->destroy != NULL)
 		par.target->destroy(&par);
 	module_put(par.target->me);
+	xt_percpu_counter_free(e->counters.pcnt);
 }
 
 /* Checks and translates the user-supplied table segment (held in
@@ -888,13 +899,15 @@ get_counters(const struct xt_table_info *t,
 
 		i = 0;
 		xt_entry_foreach(iter, t->entries[cpu], t->size) {
+			struct xt_counters *tmp;
 			u64 bcnt, pcnt;
 			unsigned int start;
 
+			tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 			do {
 				start = read_seqcount_begin(s);
-				bcnt = iter->counters.bcnt;
-				pcnt = iter->counters.pcnt;
+				bcnt = tmp->bcnt;
+				pcnt = tmp->pcnt;
 			} while (read_seqcount_retry(s, start));
 
 			ADD_COUNTER(counters[i], bcnt, pcnt);
@@ -1374,7 +1387,10 @@ do_add_counters(struct net *net, const void __user *user,
 	loc_cpu_entry = private->entries[curcpu];
 	addend = xt_write_recseq_begin();
 	xt_entry_foreach(iter, loc_cpu_entry, private->size) {
-		ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
+		struct xt_counters *tmp;
+
+		tmp = xt_get_this_cpu_counter(&iter->counters);
+		ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
 		++i;
 	}
 	xt_write_recseq_end(addend);
@@ -1608,6 +1624,10 @@ compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
 	unsigned int j;
 	int ret = 0;
 
+	e->counters.pcnt = xt_percpu_counter_alloc();
+	if (IS_ERR_VALUE(e->counters.pcnt))
+		return -ENOMEM;
+
 	j = 0;
 	mtpar.net	= net;
 	mtpar.table     = name;
@@ -1632,6 +1652,9 @@ compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
 			break;
 		cleanup_match(ematch, net);
 	}
+
+	xt_percpu_counter_free(e->counters.pcnt);
+
 	return ret;
 }