Revision 03049269de433cb5fe2859be9ae4469ceb1163ed authored by Michal Hocko on 25 March 2016, 21:20:33 UTC, committed by Linus Torvalds on 25 March 2016, 23:37:42 UTC
wake_oom_reaper has allowed only 1 oom victim to be queued.  The main
reason for that was the simplicity as other solutions would require some
way of queuing.  The current approach is racy and that was deemed
sufficient as the oom_reaper is considered a best effort approach to
help with oom handling when the OOM victim cannot terminate in a
reasonable time.  The race could lead to missing an oom victim which can
get stuck

out_of_memory
  wake_oom_reaper
    cmpxchg // OK
    			oom_reaper
			  oom_reap_task
			    __oom_reap_task
oom_victim terminates
			      atomic_inc_not_zero // fail
out_of_memory
  wake_oom_reaper
    cmpxchg // fails
			  task_to_reap = NULL

This race requires 2 OOM invocations in a short time period which is not
very likely but certainly not impossible.  E.g.  the original victim
might have not released a lot of memory for some reason.

The situation would improve considerably if wake_oom_reaper used a more
robust queuing.  This is what this patch implements.  This means adding
oom_reaper_list list_head into task_struct (eat a hole before embeded
thread_struct for that purpose) and a oom_reaper_lock spinlock for
queuing synchronization.  wake_oom_reaper will then add the task on the
queue and oom_reaper will dequeue it.

Signed-off-by: Michal Hocko <mhocko@suse.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: Andrea Argangeli <andrea@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent bc448e8
Raw File
list_sort.c

#define pr_fmt(fmt) "list_sort_test: " fmt

#include <linux/kernel.h>
#include <linux/bug.h>
#include <linux/compiler.h>
#include <linux/export.h>
#include <linux/string.h>
#include <linux/list_sort.h>
#include <linux/list.h>

#define MAX_LIST_LENGTH_BITS 20

/*
 * Returns a list organized in an intermediate format suited
 * to chaining of merge() calls: null-terminated, no reserved or
 * sentinel head node, "prev" links not maintained.
 */
static struct list_head *merge(void *priv,
				int (*cmp)(void *priv, struct list_head *a,
					struct list_head *b),
				struct list_head *a, struct list_head *b)
{
	struct list_head head, *tail = &head;

	while (a && b) {
		/* if equal, take 'a' -- important for sort stability */
		if ((*cmp)(priv, a, b) <= 0) {
			tail->next = a;
			a = a->next;
		} else {
			tail->next = b;
			b = b->next;
		}
		tail = tail->next;
	}
	tail->next = a?:b;
	return head.next;
}

/*
 * Combine final list merge with restoration of standard doubly-linked
 * list structure.  This approach duplicates code from merge(), but
 * runs faster than the tidier alternatives of either a separate final
 * prev-link restoration pass, or maintaining the prev links
 * throughout.
 */
static void merge_and_restore_back_links(void *priv,
				int (*cmp)(void *priv, struct list_head *a,
					struct list_head *b),
				struct list_head *head,
				struct list_head *a, struct list_head *b)
{
	struct list_head *tail = head;
	u8 count = 0;

	while (a && b) {
		/* if equal, take 'a' -- important for sort stability */
		if ((*cmp)(priv, a, b) <= 0) {
			tail->next = a;
			a->prev = tail;
			a = a->next;
		} else {
			tail->next = b;
			b->prev = tail;
			b = b->next;
		}
		tail = tail->next;
	}
	tail->next = a ? : b;

	do {
		/*
		 * In worst cases this loop may run many iterations.
		 * Continue callbacks to the client even though no
		 * element comparison is needed, so the client's cmp()
		 * routine can invoke cond_resched() periodically.
		 */
		if (unlikely(!(++count)))
			(*cmp)(priv, tail->next, tail->next);

		tail->next->prev = tail;
		tail = tail->next;
	} while (tail->next);

	tail->next = head;
	head->prev = tail;
}

/**
 * list_sort - sort a list
 * @priv: private data, opaque to list_sort(), passed to @cmp
 * @head: the list to sort
 * @cmp: the elements comparison function
 *
 * This function implements "merge sort", which has O(nlog(n))
 * complexity.
 *
 * The comparison function @cmp must return a negative value if @a
 * should sort before @b, and a positive value if @a should sort after
 * @b. If @a and @b are equivalent, and their original relative
 * ordering is to be preserved, @cmp must return 0.
 */
void list_sort(void *priv, struct list_head *head,
		int (*cmp)(void *priv, struct list_head *a,
			struct list_head *b))
{
	struct list_head *part[MAX_LIST_LENGTH_BITS+1]; /* sorted partial lists
						-- last slot is a sentinel */
	int lev;  /* index into part[] */
	int max_lev = 0;
	struct list_head *list;

	if (list_empty(head))
		return;

	memset(part, 0, sizeof(part));

	head->prev->next = NULL;
	list = head->next;

	while (list) {
		struct list_head *cur = list;
		list = list->next;
		cur->next = NULL;

		for (lev = 0; part[lev]; lev++) {
			cur = merge(priv, cmp, part[lev], cur);
			part[lev] = NULL;
		}
		if (lev > max_lev) {
			if (unlikely(lev >= ARRAY_SIZE(part)-1)) {
				printk_once(KERN_DEBUG "list too long for efficiency\n");
				lev--;
			}
			max_lev = lev;
		}
		part[lev] = cur;
	}

	for (lev = 0; lev < max_lev; lev++)
		if (part[lev])
			list = merge(priv, cmp, part[lev], list);

	merge_and_restore_back_links(priv, cmp, head, part[max_lev], list);
}
EXPORT_SYMBOL(list_sort);

#ifdef CONFIG_TEST_LIST_SORT

#include <linux/slab.h>
#include <linux/random.h>

/*
 * The pattern of set bits in the list length determines which cases
 * are hit in list_sort().
 */
#define TEST_LIST_LEN (512+128+2) /* not including head */

#define TEST_POISON1 0xDEADBEEF
#define TEST_POISON2 0xA324354C

struct debug_el {
	unsigned int poison1;
	struct list_head list;
	unsigned int poison2;
	int value;
	unsigned serial;
};

/* Array, containing pointers to all elements in the test list */
static struct debug_el **elts __initdata;

static int __init check(struct debug_el *ela, struct debug_el *elb)
{
	if (ela->serial >= TEST_LIST_LEN) {
		pr_err("error: incorrect serial %d\n", ela->serial);
		return -EINVAL;
	}
	if (elb->serial >= TEST_LIST_LEN) {
		pr_err("error: incorrect serial %d\n", elb->serial);
		return -EINVAL;
	}
	if (elts[ela->serial] != ela || elts[elb->serial] != elb) {
		pr_err("error: phantom element\n");
		return -EINVAL;
	}
	if (ela->poison1 != TEST_POISON1 || ela->poison2 != TEST_POISON2) {
		pr_err("error: bad poison: %#x/%#x\n",
			ela->poison1, ela->poison2);
		return -EINVAL;
	}
	if (elb->poison1 != TEST_POISON1 || elb->poison2 != TEST_POISON2) {
		pr_err("error: bad poison: %#x/%#x\n",
			elb->poison1, elb->poison2);
		return -EINVAL;
	}
	return 0;
}

static int __init cmp(void *priv, struct list_head *a, struct list_head *b)
{
	struct debug_el *ela, *elb;

	ela = container_of(a, struct debug_el, list);
	elb = container_of(b, struct debug_el, list);

	check(ela, elb);
	return ela->value - elb->value;
}

static int __init list_sort_test(void)
{
	int i, count = 1, err = -ENOMEM;
	struct debug_el *el;
	struct list_head *cur;
	LIST_HEAD(head);

	pr_debug("start testing list_sort()\n");

	elts = kcalloc(TEST_LIST_LEN, sizeof(*elts), GFP_KERNEL);
	if (!elts) {
		pr_err("error: cannot allocate memory\n");
		return err;
	}

	for (i = 0; i < TEST_LIST_LEN; i++) {
		el = kmalloc(sizeof(*el), GFP_KERNEL);
		if (!el) {
			pr_err("error: cannot allocate memory\n");
			goto exit;
		}
		 /* force some equivalencies */
		el->value = prandom_u32() % (TEST_LIST_LEN / 3);
		el->serial = i;
		el->poison1 = TEST_POISON1;
		el->poison2 = TEST_POISON2;
		elts[i] = el;
		list_add_tail(&el->list, &head);
	}

	list_sort(NULL, &head, cmp);

	err = -EINVAL;
	for (cur = head.next; cur->next != &head; cur = cur->next) {
		struct debug_el *el1;
		int cmp_result;

		if (cur->next->prev != cur) {
			pr_err("error: list is corrupted\n");
			goto exit;
		}

		cmp_result = cmp(NULL, cur, cur->next);
		if (cmp_result > 0) {
			pr_err("error: list is not sorted\n");
			goto exit;
		}

		el = container_of(cur, struct debug_el, list);
		el1 = container_of(cur->next, struct debug_el, list);
		if (cmp_result == 0 && el->serial >= el1->serial) {
			pr_err("error: order of equivalent elements not "
				"preserved\n");
			goto exit;
		}

		if (check(el, el1)) {
			pr_err("error: element check failed\n");
			goto exit;
		}
		count++;
	}
	if (head.prev != cur) {
		pr_err("error: list is corrupted\n");
		goto exit;
	}


	if (count != TEST_LIST_LEN) {
		pr_err("error: bad list length %d", count);
		goto exit;
	}

	err = 0;
exit:
	for (i = 0; i < TEST_LIST_LEN; i++)
		kfree(elts[i]);
	kfree(elts);
	return err;
}
late_initcall(list_sort_test);
#endif /* CONFIG_TEST_LIST_SORT */
back to top