netfilter: xtables: stackptr should be percpu
commit f3c5c1bfd4 (netfilter: xtables: make ip_tables reentrant)
introduced a performance regression, because stackptr array is shared by
all cpus, adding cache line ping pongs. (16 cpus share a 64 bytes cache
line)
Fix this using alloc_percpu()
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-By: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
diff --git a/net/netfilter/x_tables.c b/net/netfilter/x_tables.c
index 47b1e79..e34622f 100644
--- a/net/netfilter/x_tables.c
+++ b/net/netfilter/x_tables.c
@@ -699,10 +699,8 @@
vfree(info->jumpstack);
else
kfree(info->jumpstack);
- if (sizeof(unsigned int) * nr_cpu_ids > PAGE_SIZE)
- vfree(info->stackptr);
- else
- kfree(info->stackptr);
+
+ free_percpu(info->stackptr);
kfree(info);
}
@@ -753,14 +751,9 @@
unsigned int size;
int cpu;
- size = sizeof(unsigned int) * nr_cpu_ids;
- if (size > PAGE_SIZE)
- i->stackptr = vmalloc(size);
- else
- i->stackptr = kmalloc(size, GFP_KERNEL);
+ i->stackptr = alloc_percpu(unsigned int);
if (i->stackptr == NULL)
return -ENOMEM;
- memset(i->stackptr, 0, size);
size = sizeof(void **) * nr_cpu_ids;
if (size > PAGE_SIZE)
