#define do_swap_account (0)
#endif
-#define SOFTLIMIT_EVENTS_THRESH (1000)
-#define THRESHOLDS_EVENTS_THRESH (100)
+/*
+ * Per memcg event counter is incremented at every pagein/pageout. This counter
+ * is used for trigger some periodic events. This is straightforward and better
+ * than using jiffies etc. to handle periodic memcg event.
+ *
+ * These values will be used as !((event) & ((1 <<(thresh)) - 1))
+ */
+#define THRESHOLDS_EVENTS_THRESH (7) /* once in 128 */
+#define SOFTLIMIT_EVENTS_THRESH (10) /* once in 1024 */
/*
* Statistics for memory cgroup.
MEM_CGROUP_STAT_PGPGIN_COUNT, /* # of pages paged in */
MEM_CGROUP_STAT_PGPGOUT_COUNT, /* # of pages paged out */
MEM_CGROUP_STAT_SWAPOUT, /* # of pages, swapped out */
- MEM_CGROUP_STAT_SOFTLIMIT, /* decrements on each page in/out.
- used by soft limit implementation */
- MEM_CGROUP_STAT_THRESHOLDS, /* decrements on each page in/out.
- used by threshold implementation */
+ MEM_CGROUP_EVENTS, /* incremented at every pagein/pageout */
MEM_CGROUP_STAT_NSTATS,
};
struct mem_cgroup_threshold entries[0];
};
-static bool mem_cgroup_threshold_check(struct mem_cgroup *mem);
static void mem_cgroup_threshold(struct mem_cgroup *mem);
/*
* Should the accounting and control be hierarchical, per subtree?
*/
bool use_hierarchy;
- unsigned long last_oom_jiffies;
+ atomic_t oom_lock;
atomic_t refcnt;
unsigned int swappiness;
spin_unlock(&mctz->lock);
}
-static bool mem_cgroup_soft_limit_check(struct mem_cgroup *mem)
-{
- bool ret = false;
- s64 val;
-
- val = this_cpu_read(mem->stat->count[MEM_CGROUP_STAT_SOFTLIMIT]);
- if (unlikely(val < 0)) {
- this_cpu_write(mem->stat->count[MEM_CGROUP_STAT_SOFTLIMIT],
- SOFTLIMIT_EVENTS_THRESH);
- ret = true;
- }
- return ret;
-}
static void mem_cgroup_update_tree(struct mem_cgroup *mem, struct page *page)
{
__this_cpu_inc(mem->stat->count[MEM_CGROUP_STAT_PGPGIN_COUNT]);
else
__this_cpu_inc(mem->stat->count[MEM_CGROUP_STAT_PGPGOUT_COUNT]);
- __this_cpu_dec(mem->stat->count[MEM_CGROUP_STAT_SOFTLIMIT]);
- __this_cpu_dec(mem->stat->count[MEM_CGROUP_STAT_THRESHOLDS]);
+ __this_cpu_inc(mem->stat->count[MEM_CGROUP_EVENTS]);
preempt_enable();
}
return total;
}
+static bool __memcg_event_check(struct mem_cgroup *mem, int event_mask_shift)
+{
+ s64 val;
+
+ val = this_cpu_read(mem->stat->count[MEM_CGROUP_EVENTS]);
+
+ return !(val & ((1 << event_mask_shift) - 1));
+}
+
+/*
+ * Check events in order.
+ *
+ */
+static void memcg_check_events(struct mem_cgroup *mem, struct page *page)
+{
+ /* threshold event is triggered in finer grain than soft limit */
+ if (unlikely(__memcg_event_check(mem, THRESHOLDS_EVENTS_THRESH))) {
+ mem_cgroup_threshold(mem);
+ if (unlikely(__memcg_event_check(mem, SOFTLIMIT_EVENTS_THRESH)))
+ mem_cgroup_update_tree(mem, page);
+ }
+}
+
static struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
{
return container_of(cgroup_subsys_state(cont,
return total;
}
-bool mem_cgroup_oom_called(struct task_struct *task)
+static int mem_cgroup_oom_lock_cb(struct mem_cgroup *mem, void *data)
{
- bool ret = false;
- struct mem_cgroup *mem;
- struct mm_struct *mm;
+ int *val = (int *)data;
+ int x;
+ /*
+ * Logically, we can stop scanning immediately when we find
+ * a memcg is already locked. But condidering unlock ops and
+ * creation/removal of memcg, scan-all is simple operation.
+ */
+ x = atomic_inc_return(&mem->oom_lock);
+ *val = max(x, *val);
+ return 0;
+}
+/*
+ * Check OOM-Killer is already running under our hierarchy.
+ * If someone is running, return false.
+ */
+static bool mem_cgroup_oom_lock(struct mem_cgroup *mem)
+{
+ int lock_count = 0;
- rcu_read_lock();
- mm = task->mm;
- if (!mm)
- mm = &init_mm;
- mem = mem_cgroup_from_task(rcu_dereference(mm->owner));
- if (mem && time_before(jiffies, mem->last_oom_jiffies + HZ/10))
- ret = true;
- rcu_read_unlock();
- return ret;
+ mem_cgroup_walk_tree(mem, &lock_count, mem_cgroup_oom_lock_cb);
+
+ if (lock_count == 1)
+ return true;
+ return false;
}
-static int record_last_oom_cb(struct mem_cgroup *mem, void *data)
+static int mem_cgroup_oom_unlock_cb(struct mem_cgroup *mem, void *data)
{
- mem->last_oom_jiffies = jiffies;
+ /*
+ * When a new child is created while the hierarchy is under oom,
+ * mem_cgroup_oom_lock() may not be called. We have to use
+ * atomic_add_unless() here.
+ */
+ atomic_add_unless(&mem->oom_lock, -1, 0);
return 0;
}
-static void record_last_oom(struct mem_cgroup *mem)
+static void mem_cgroup_oom_unlock(struct mem_cgroup *mem)
{
- mem_cgroup_walk_tree(mem, NULL, record_last_oom_cb);
+ mem_cgroup_walk_tree(mem, NULL, mem_cgroup_oom_unlock_cb);
+}
+
+static DEFINE_MUTEX(memcg_oom_mutex);
+static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);
+
+/*
+ * try to call OOM killer. returns false if we should exit memory-reclaim loop.
+ */
+bool mem_cgroup_handle_oom(struct mem_cgroup *mem, gfp_t mask)
+{
+ DEFINE_WAIT(wait);
+ bool locked;
+
+ /* At first, try to OOM lock hierarchy under mem.*/
+ mutex_lock(&memcg_oom_mutex);
+ locked = mem_cgroup_oom_lock(mem);
+ /*
+ * Even if signal_pending(), we can't quit charge() loop without
+ * accounting. So, UNINTERRUPTIBLE is appropriate. But SIGKILL
+ * under OOM is always welcomed, use TASK_KILLABLE here.
+ */
+ if (!locked)
+ prepare_to_wait(&memcg_oom_waitq, &wait, TASK_KILLABLE);
+ mutex_unlock(&memcg_oom_mutex);
+
+ if (locked)
+ mem_cgroup_out_of_memory(mem, mask);
+ else {
+ schedule();
+ finish_wait(&memcg_oom_waitq, &wait);
+ }
+ mutex_lock(&memcg_oom_mutex);
+ mem_cgroup_oom_unlock(mem);
+ /*
+ * Here, we use global waitq .....more fine grained waitq ?
+ * Assume following hierarchy.
+ * A/
+ * 01
+ * 02
+ * assume OOM happens both in A and 01 at the same time. Tthey are
+ * mutually exclusive by lock. (kill in 01 helps A.)
+ * When we use per memcg waitq, we have to wake up waiters on A and 02
+ * in addtion to waiters on 01. We use global waitq for avoiding mess.
+ * It will not be a big problem.
+ * (And a task may be moved to other groups while it's waiting for OOM.)
+ */
+ wake_up_all(&memcg_oom_waitq);
+ mutex_unlock(&memcg_oom_mutex);
+
+ if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
+ return false;
+ /* Give chance to dying process */
+ schedule_timeout(1);
+ return true;
}
/*
lock_page_cgroup(pc);
mem = pc->mem_cgroup;
- if (!mem)
- goto done;
-
- if (!PageCgroupUsed(pc))
+ if (!mem || !PageCgroupUsed(pc))
goto done;
/*
* Preemption is already disabled. We can use __this_cpu_xxx
*/
- __this_cpu_add(mem->stat->count[MEM_CGROUP_STAT_FILE_MAPPED], val);
+ if (val > 0) {
+ __this_cpu_inc(mem->stat->count[MEM_CGROUP_STAT_FILE_MAPPED]);
+ SetPageCgroupFileMapped(pc);
+ } else {
+ __this_cpu_dec(mem->stat->count[MEM_CGROUP_STAT_FILE_MAPPED]);
+ ClearPageCgroupFileMapped(pc);
+ }
done:
unlock_page_cgroup(pc);
/*
* Cache charges(val) which is from res_counter, to local per_cpu area.
- * This will be consumed by consumt_stock() function, later.
+ * This will be consumed by consume_stock() function, later.
*/
static void refill_stock(struct mem_cgroup *mem, int val)
{
struct res_counter *fail_res;
int csize = CHARGE_SIZE;
- if (unlikely(test_thread_flag(TIF_MEMDIE))) {
- /* Don't account this! */
- *memcg = NULL;
- return 0;
- }
+ /*
+ * Unlike gloval-vm's OOM-kill, we're not in memory shortage
+ * in system level. So, allow to go ahead dying process in addition to
+ * MEMDIE process.
+ */
+ if (unlikely(test_thread_flag(TIF_MEMDIE)
+ || fatal_signal_pending(current)))
+ goto bypass;
/*
* We always charge the cgroup the mm_struct belongs to.
* There is a small race that "from" or "to" can be
* freed by rmdir, so we use css_tryget().
*/
- rcu_read_lock();
from = mc.from;
to = mc.to;
if (from && css_tryget(&from->css)) {
do_continue = (to == mem_over_limit);
css_put(&to->css);
}
- rcu_read_unlock();
if (do_continue) {
DEFINE_WAIT(wait);
prepare_to_wait(&mc.waitq, &wait,
}
if (!nr_retries--) {
- if (oom) {
- mem_cgroup_out_of_memory(mem_over_limit, gfp_mask);
- record_last_oom(mem_over_limit);
+ if (!oom)
+ goto nomem;
+ if (mem_cgroup_handle_oom(mem_over_limit, gfp_mask)) {
+ nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
+ continue;
}
- goto nomem;
+ /* When we reach here, current task is dying .*/
+ css_put(&mem->css);
+ goto bypass;
}
}
if (csize > PAGE_SIZE)
nomem:
css_put(&mem->css);
return -ENOMEM;
+bypass:
+ *memcg = NULL;
+ return 0;
}
/*
* Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
* if they exceeds softlimit.
*/
- if (mem_cgroup_soft_limit_check(mem))
- mem_cgroup_update_tree(mem, pc->page);
- if (mem_cgroup_threshold_check(mem))
- mem_cgroup_threshold(mem);
-
+ memcg_check_events(mem, pc->page);
}
/**
static void __mem_cgroup_move_account(struct page_cgroup *pc,
struct mem_cgroup *from, struct mem_cgroup *to, bool uncharge)
{
- struct page *page;
-
VM_BUG_ON(from == to);
VM_BUG_ON(PageLRU(pc->page));
VM_BUG_ON(!PageCgroupLocked(pc));
VM_BUG_ON(!PageCgroupUsed(pc));
VM_BUG_ON(pc->mem_cgroup != from);
- page = pc->page;
- if (page_mapped(page) && !PageAnon(page)) {
+ if (PageCgroupFileMapped(pc)) {
/* Update mapped_file data for mem_cgroup */
preempt_disable();
__this_cpu_dec(from->stat->count[MEM_CGROUP_STAT_FILE_MAPPED]);
ret = 0;
}
unlock_page_cgroup(pc);
+ /*
+ * check events
+ */
+ memcg_check_events(to, pc->page);
+ memcg_check_events(from, pc->page);
return ret;
}
mz = page_cgroup_zoneinfo(pc);
unlock_page_cgroup(pc);
- if (mem_cgroup_soft_limit_check(mem))
- mem_cgroup_update_tree(mem, page);
- if (mem_cgroup_threshold_check(mem))
- mem_cgroup_threshold(mem);
+ memcg_check_events(mem, page);
/* at swapout, this memcg will be accessed to record to swap */
if (ctype != MEM_CGROUP_CHARGE_TYPE_SWAPOUT)
css_put(&mem->css);
}
unlock_page_cgroup(pc);
+ *ptr = mem;
if (mem) {
- ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, &mem, false);
+ ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, ptr, false);
css_put(&mem->css);
}
- *ptr = mem;
return ret;
}
return 0;
}
-static bool mem_cgroup_threshold_check(struct mem_cgroup *mem)
-{
- bool ret = false;
- s64 val;
-
- val = this_cpu_read(mem->stat->count[MEM_CGROUP_STAT_THRESHOLDS]);
- if (unlikely(val < 0)) {
- this_cpu_write(mem->stat->count[MEM_CGROUP_STAT_THRESHOLDS],
- THRESHOLDS_EVENTS_THRESH);
- ret = true;
- }
- return ret;
-}
-
static void __mem_cgroup_threshold(struct mem_cgroup *memcg, bool swap)
{
struct mem_cgroup_threshold_ary *t;
}
}
- /*
- * We need to increment refcnt to be sure that all thresholds
- * will be unregistered before calling __mem_cgroup_free()
- */
- mem_cgroup_get(memcg);
-
if (type == _MEM)
rcu_assign_pointer(memcg->thresholds, thresholds_new);
else
/* To be sure that nobody uses thresholds before freeing it */
synchronize_rcu();
- for (i = 0; i < thresholds->size - size; i++)
- mem_cgroup_put(memcg);
-
kfree(thresholds);
unlock:
mutex_unlock(&memcg->thresholds_lock);
else
mem = vmalloc(size);
- if (mem)
- memset(mem, 0, size);
+ if (!mem)
+ return NULL;
+
+ memset(mem, 0, size);
mem->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
if (!mem->stat) {
if (size < PAGE_SIZE)
}
return ret;
}
-#else /* !CONFIG_MMU */
-static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
- struct cgroup *cgroup,
- struct task_struct *p,
- bool threadgroup)
-{
- return 0;
-}
-static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
- struct cgroup *cgroup,
- struct task_struct *p,
- bool threadgroup)
-{
-}
-static void mem_cgroup_move_task(struct cgroup_subsys *ss,
- struct cgroup *cont,
- struct cgroup *old_cont,
- struct task_struct *p,
- bool threadgroup)
-{
-}
-#endif
/**
* is_target_pte_for_mc - check a pte whether it is valid for move charge
}
mem_cgroup_clear_mc();
}
+#else /* !CONFIG_MMU */
+static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
+ struct cgroup *cgroup,
+ struct task_struct *p,
+ bool threadgroup)
+{
+ return 0;
+}
+static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
+ struct cgroup *cgroup,
+ struct task_struct *p,
+ bool threadgroup)
+{
+}
+static void mem_cgroup_move_task(struct cgroup_subsys *ss,
+ struct cgroup *cont,
+ struct cgroup *old_cont,
+ struct task_struct *p,
+ bool threadgroup)
+{
+}
+#endif
struct cgroup_subsys mem_cgroup_subsys = {
.name = "memory",