memcg: reduce size of mem_cgroup by using nr_cpu_ids
authorJan Blunck <jblunck@suse.de>
Thu, 8 Jan 2009 02:07:53 +0000 (18:07 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 8 Jan 2009 16:31:04 +0000 (08:31 -0800)
As Jan Blunck <jblunck@suse.de> pointed out, allocating per-cpu stat for
memcg to the size of NR_CPUS is not good.

This patch changes mem_cgroup's cpustat allocation not based on NR_CPUS
but based on nr_cpu_ids.

Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memcontrol.c

index 49234d93988acf13b2a699408ed6200be2a5b6aa..e00f25e6545f15aedd00b9b4f2627106b513188e 100644 (file)
@@ -60,7 +60,7 @@ struct mem_cgroup_stat_cpu {
 } ____cacheline_aligned_in_smp;
 
 struct mem_cgroup_stat {
-       struct mem_cgroup_stat_cpu cpustat[NR_CPUS];
+       struct mem_cgroup_stat_cpu cpustat[0];
 };
 
 /*
@@ -129,11 +129,10 @@ struct mem_cgroup {
 
        int     prev_priority;  /* for recording reclaim priority */
        /*
-        * statistics.
+        * statistics. This must be placed at the end of memcg.
         */
        struct mem_cgroup_stat stat;
 };
-static struct mem_cgroup init_mem_cgroup;
 
 enum charge_type {
        MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
@@ -1293,23 +1292,30 @@ static void free_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
        kfree(mem->info.nodeinfo[node]);
 }
 
+static int mem_cgroup_size(void)
+{
+       int cpustat_size = nr_cpu_ids * sizeof(struct mem_cgroup_stat_cpu);
+       return sizeof(struct mem_cgroup) + cpustat_size;
+}
+
 static struct mem_cgroup *mem_cgroup_alloc(void)
 {
        struct mem_cgroup *mem;
+       int size = mem_cgroup_size();
 
-       if (sizeof(*mem) < PAGE_SIZE)
-               mem = kmalloc(sizeof(*mem), GFP_KERNEL);
+       if (size < PAGE_SIZE)
+               mem = kmalloc(size, GFP_KERNEL);
        else
-               mem = vmalloc(sizeof(*mem));
+               mem = vmalloc(size);
 
        if (mem)
-               memset(mem, 0, sizeof(*mem));
+               memset(mem, 0, size);
        return mem;
 }
 
 static void mem_cgroup_free(struct mem_cgroup *mem)
 {
-       if (sizeof(*mem) < PAGE_SIZE)
+       if (mem_cgroup_size() < PAGE_SIZE)
                kfree(mem);
        else
                vfree(mem);
@@ -1322,13 +1328,9 @@ mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
        struct mem_cgroup *mem;
        int node;
 
-       if (unlikely((cont->parent) == NULL)) {
-               mem = &init_mem_cgroup;
-       } else {
-               mem = mem_cgroup_alloc();
-               if (!mem)
-                       return ERR_PTR(-ENOMEM);
-       }
+       mem = mem_cgroup_alloc();
+       if (!mem)
+               return ERR_PTR(-ENOMEM);
 
        res_counter_init(&mem->res);
 
@@ -1340,8 +1342,7 @@ mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
 free_out:
        for_each_node_state(node, N_POSSIBLE)
                free_mem_cgroup_per_zone_info(mem, node);
-       if (cont->parent != NULL)
-               mem_cgroup_free(mem);
+       mem_cgroup_free(mem);
        return ERR_PTR(-ENOMEM);
 }