memory cgroup enhancements: force_empty interface for dropping all account in empty...
KAMEZAWA Hiroyuki [Thu, 7 Feb 2008 08:14:16 +0000 (00:14 -0800)]
This patch adds an interface "memory.force_empty".  Any write to this file
will drop all charges in this cgroup if there is no task under.

%echo 1 > /....../memory.force_empty

will drop all charges of memory cgroup if cgroup's tasks is empty.

This is useful to invoke rmdir() against memory cgroup successfully.

Tested and worked well on x86_64/fake-NUMA system.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcontrol.c

index 9793873..c867612 100644 (file)
@@ -471,6 +471,7 @@ void mem_cgroup_uncharge(struct page_cgroup *pc)
                page = pc->page;
                /*
                 * get page->cgroup and clear it under lock.
+                * force_empty can drop page->cgroup without checking refcnt.
                 */
                if (clear_page_cgroup(page, pc) == pc) {
                        mem = pc->mem_cgroup;
@@ -480,13 +481,6 @@ void mem_cgroup_uncharge(struct page_cgroup *pc)
                        list_del_init(&pc->lru);
                        spin_unlock_irqrestore(&mem->lru_lock, flags);
                        kfree(pc);
-               } else {
-                       /*
-                        * Note:This will be removed when force-empty patch is
-                        * applied. just show warning here.
-                        */
-                       printk(KERN_ERR "Race in mem_cgroup_uncharge() ?");
-                       dump_stack();
                }
        }
 }
@@ -534,6 +528,76 @@ retry:
        return;
 }
 
+/*
+ * This routine traverse page_cgroup in given list and drop them all.
+ * This routine ignores page_cgroup->ref_cnt.
+ * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
+ */
+#define FORCE_UNCHARGE_BATCH   (128)
+static void
+mem_cgroup_force_empty_list(struct mem_cgroup *mem, struct list_head *list)
+{
+       struct page_cgroup *pc;
+       struct page *page;
+       int count;
+       unsigned long flags;
+
+retry:
+       count = FORCE_UNCHARGE_BATCH;
+       spin_lock_irqsave(&mem->lru_lock, flags);
+
+       while (--count && !list_empty(list)) {
+               pc = list_entry(list->prev, struct page_cgroup, lru);
+               page = pc->page;
+               /* Avoid race with charge */
+               atomic_set(&pc->ref_cnt, 0);
+               if (clear_page_cgroup(page, pc) == pc) {
+                       css_put(&mem->css);
+                       res_counter_uncharge(&mem->res, PAGE_SIZE);
+                       list_del_init(&pc->lru);
+                       kfree(pc);
+               } else  /* being uncharged ? ...do relax */
+                       break;
+       }
+       spin_unlock_irqrestore(&mem->lru_lock, flags);
+       if (!list_empty(list)) {
+               cond_resched();
+               goto retry;
+       }
+       return;
+}
+
+/*
+ * make mem_cgroup's charge to be 0 if there is no task.
+ * This enables deleting this mem_cgroup.
+ */
+
+int mem_cgroup_force_empty(struct mem_cgroup *mem)
+{
+       int ret = -EBUSY;
+       css_get(&mem->css);
+       /*
+        * page reclaim code (kswapd etc..) will move pages between
+`       * active_list <-> inactive_list while we don't take a lock.
+        * So, we have to do loop here until all lists are empty.
+        */
+       while (!(list_empty(&mem->active_list) &&
+                list_empty(&mem->inactive_list))) {
+               if (atomic_read(&mem->css.cgroup->count) > 0)
+                       goto out;
+               /* drop all page_cgroup in active_list */
+               mem_cgroup_force_empty_list(mem, &mem->active_list);
+               /* drop all page_cgroup in inactive_list */
+               mem_cgroup_force_empty_list(mem, &mem->inactive_list);
+       }
+       ret = 0;
+out:
+       css_put(&mem->css);
+       return ret;
+}
+
+
+
 int mem_cgroup_write_strategy(char *buf, unsigned long long *tmp)
 {
        *tmp = memparse(buf, &buf);
@@ -619,6 +683,33 @@ static ssize_t mem_control_type_read(struct cgroup *cont,
                        ppos, buf, s - buf);
 }
 
+
+static ssize_t mem_force_empty_write(struct cgroup *cont,
+                               struct cftype *cft, struct file *file,
+                               const char __user *userbuf,
+                               size_t nbytes, loff_t *ppos)
+{
+       struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
+       int ret;
+       ret = mem_cgroup_force_empty(mem);
+       if (!ret)
+               ret = nbytes;
+       return ret;
+}
+
+/*
+ * Note: This should be removed if cgroup supports write-only file.
+ */
+
+static ssize_t mem_force_empty_read(struct cgroup *cont,
+                               struct cftype *cft,
+                               struct file *file, char __user *userbuf,
+                               size_t nbytes, loff_t *ppos)
+{
+       return -EINVAL;
+}
+
+
 static struct cftype mem_cgroup_files[] = {
        {
                .name = "usage_in_bytes",
@@ -641,6 +732,11 @@ static struct cftype mem_cgroup_files[] = {
                .write = mem_control_type_write,
                .read = mem_control_type_read,
        },
+       {
+               .name = "force_empty",
+               .write = mem_force_empty_write,
+               .read = mem_force_empty_read,
+       },
 };
 
 static struct mem_cgroup init_mem_cgroup;