[deliverable/linux.git] / mm / page_isolation.c

/*
 * linux/mm/page_isolation.c
 */

#include <linux/mm.h>
#include <linux/page-isolation.h>
#include <linux/pageblock-flags.h>
#include <linux/memory.h>
#include <linux/hugetlb.h>
#include "internal.h"

int set_migratetype_isolate(struct page *page, bool skip_hwpoisoned_pages)
{
	struct zone *zone;
	unsigned long flags, pfn;
	struct memory_isolate_notify arg;
	int notifier_ret;
	int ret = -EBUSY;

	zone = page_zone(page);

	spin_lock_irqsave(&zone->lock, flags);

	pfn = page_to_pfn(page);
	arg.start_pfn = pfn;
	arg.nr_pages = pageblock_nr_pages;
	arg.pages_found = 0;

	/*
	 * It may be possible to isolate a pageblock even if the
	 * migratetype is not MIGRATE_MOVABLE. The memory isolation
	 * notifier chain is used by balloon drivers to return the
	 * number of pages in a range that are held by the balloon
	 * driver to shrink memory. If all the pages are accounted for
	 * by balloons, are free, or on the LRU, isolation can continue.
	 * Later, for example, when memory hotplug notifier runs, these
	 * pages reported as "can be isolated" should be isolated(freed)
	 * by the balloon driver through the memory notifier chain.
	 */
	notifier_ret = memory_isolate_notify(MEM_ISOLATE_COUNT, &arg);
	notifier_ret = notifier_to_errno(notifier_ret);
	if (notifier_ret)
		goto out;
	/*
	 * FIXME: Now, memory hotplug doesn't call shrink_slab() by itself.
	 * We just check MOVABLE pages.
	 */
	if (!has_unmovable_pages(zone, page, arg.pages_found,
				 skip_hwpoisoned_pages))
		ret = 0;

	/*
	 * immobile means "not-on-lru" paes. If immobile is larger than
	 * removable-by-driver pages reported by notifier, we'll fail.
	 */

out:
	if (!ret) {
		unsigned long nr_pages;
		int migratetype = get_pageblock_migratetype(page);

		set_pageblock_migratetype(page, MIGRATE_ISOLATE);
		zone->nr_isolate_pageblock++;
		nr_pages = move_freepages_block(zone, page, MIGRATE_ISOLATE);

		__mod_zone_freepage_state(zone, -nr_pages, migratetype);
	}

	spin_unlock_irqrestore(&zone->lock, flags);
	if (!ret)
		drain_all_pages(zone);
	return ret;
}

void unset_migratetype_isolate(struct page *page, unsigned migratetype)
{
	struct zone *zone;
	unsigned long flags, nr_pages;
	struct page *isolated_page = NULL;
	unsigned int order;
	unsigned long page_idx, buddy_idx;
	struct page *buddy;

	zone = page_zone(page);
	spin_lock_irqsave(&zone->lock, flags);
	if (get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
		goto out;

	/*
	 * Because freepage with more than pageblock_order on isolated
	 * pageblock is restricted to merge due to freepage counting problem,
	 * it is possible that there is free buddy page.
	 * move_freepages_block() doesn't care of merge so we need other
	 * approach in order to merge them. Isolation and free will make
	 * these pages to be merged.
	 */
	if (PageBuddy(page)) {
		order = page_order(page);
		if (order >= pageblock_order) {
			page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);
			buddy_idx = __find_buddy_index(page_idx, order);
			buddy = page + (buddy_idx - page_idx);

			if (!is_migrate_isolate_page(buddy)) {
				__isolate_free_page(page, order);
				set_page_refcounted(page);
				isolated_page = page;
			}
		}
	}

	/*
	 * If we isolate freepage with more than pageblock_order, there
	 * should be no freepage in the range, so we could avoid costly
	 * pageblock scanning for freepage moving.
	 */
	if (!isolated_page) {
		nr_pages = move_freepages_block(zone, page, migratetype);
		__mod_zone_freepage_state(zone, nr_pages, migratetype);
	}
	set_pageblock_migratetype(page, migratetype);
	zone->nr_isolate_pageblock--;
out:
	spin_unlock_irqrestore(&zone->lock, flags);
	if (isolated_page)
		__free_pages(isolated_page, order);
}

static inline struct page *
__first_valid_page(unsigned long pfn, unsigned long nr_pages)
{
	int i;
	for (i = 0; i < nr_pages; i++)
		if (pfn_valid_within(pfn + i))
			break;
	if (unlikely(i == nr_pages))
		return NULL;
	return pfn_to_page(pfn + i);
}

/*
 * start_isolate_page_range() -- make page-allocation-type of range of pages
 * to be MIGRATE_ISOLATE.
 * @start_pfn: The lower PFN of the range to be isolated.
 * @end_pfn: The upper PFN of the range to be isolated.
 * @migratetype: migrate type to set in error recovery.
 *
 * Making page-allocation-type to be MIGRATE_ISOLATE means free pages in
 * the range will never be allocated. Any free pages and pages freed in the
 * future will not be allocated again.
 *
 * start_pfn/end_pfn must be aligned to pageblock_order.
 * Returns 0 on success and -EBUSY if any part of range cannot be isolated.
 */
int start_isolate_page_range(unsigned long start_pfn, unsigned long end_pfn,
			     unsigned migratetype, bool skip_hwpoisoned_pages)
{
	unsigned long pfn;
	unsigned long undo_pfn;
	struct page *page;

	BUG_ON((start_pfn) & (pageblock_nr_pages - 1));
	BUG_ON((end_pfn) & (pageblock_nr_pages - 1));

	for (pfn = start_pfn;
	     pfn < end_pfn;
	     pfn += pageblock_nr_pages) {
		page = __first_valid_page(pfn, pageblock_nr_pages);
		if (page &&
		    set_migratetype_isolate(page, skip_hwpoisoned_pages)) {
			undo_pfn = pfn;
			goto undo;
		}
	}
	return 0;
undo:
	for (pfn = start_pfn;
	     pfn < undo_pfn;
	     pfn += pageblock_nr_pages)
		unset_migratetype_isolate(pfn_to_page(pfn), migratetype);

	return -EBUSY;
}

/*
 * Make isolated pages available again.
 */
int undo_isolate_page_range(unsigned long start_pfn, unsigned long end_pfn,
			    unsigned migratetype)
{
	unsigned long pfn;
	struct page *page;
	BUG_ON((start_pfn) & (pageblock_nr_pages - 1));
	BUG_ON((end_pfn) & (pageblock_nr_pages - 1));
	for (pfn = start_pfn;
	     pfn < end_pfn;
	     pfn += pageblock_nr_pages) {
		page = __first_valid_page(pfn, pageblock_nr_pages);
		if (!page || get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
			continue;
		unset_migratetype_isolate(page, migratetype);
	}
	return 0;
}
/*
 * Test all pages in the range is free(means isolated) or not.
 * all pages in [start_pfn...end_pfn) must be in the same zone.
 * zone->lock must be held before call this.
 *
 * Returns 1 if all pages in the range are isolated.
 */
static int
__test_page_isolated_in_pageblock(unsigned long pfn, unsigned long end_pfn,
				  bool skip_hwpoisoned_pages)
{
	struct page *page;

	while (pfn < end_pfn) {
		if (!pfn_valid_within(pfn)) {
			pfn++;
			continue;
		}
		page = pfn_to_page(pfn);
		if (PageBuddy(page)) {
			/*
			 * If race between isolatation and allocation happens,
			 * some free pages could be in MIGRATE_MOVABLE list
			 * although pageblock's migratation type of the page
			 * is MIGRATE_ISOLATE. Catch it and move the page into
			 * MIGRATE_ISOLATE list.
			 */
			if (get_freepage_migratetype(page) != MIGRATE_ISOLATE) {
				struct page *end_page;

				end_page = page + (1 << page_order(page)) - 1;
				move_freepages(page_zone(page), page, end_page,
						MIGRATE_ISOLATE);
			}
			pfn += 1 << page_order(page);
		}
		else if (page_count(page) == 0 &&
			get_freepage_migratetype(page) == MIGRATE_ISOLATE)
			pfn += 1;
		else if (skip_hwpoisoned_pages && PageHWPoison(page)) {
			/*
			 * The HWPoisoned page may be not in buddy
			 * system, and page_count() is not 0.
			 */
			pfn++;
			continue;
		}
		else
			break;
	}
	if (pfn < end_pfn)
		return 0;
	return 1;
}

int test_pages_isolated(unsigned long start_pfn, unsigned long end_pfn,
			bool skip_hwpoisoned_pages)
{
	unsigned long pfn, flags;
	struct page *page;
	struct zone *zone;
	int ret;

	/*
	 * Note: pageblock_nr_pages != MAX_ORDER. Then, chunks of free pages
	 * are not aligned to pageblock_nr_pages.
	 * Then we just check migratetype first.
	 */
	for (pfn = start_pfn; pfn < end_pfn; pfn += pageblock_nr_pages) {
		page = __first_valid_page(pfn, pageblock_nr_pages);
		if (page && get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
			break;
	}
	page = __first_valid_page(start_pfn, end_pfn - start_pfn);
	if ((pfn < end_pfn) || !page)
		return -EBUSY;
	/* Check all pages are free or marked as ISOLATED */
	zone = page_zone(page);
	spin_lock_irqsave(&zone->lock, flags);
	ret = __test_page_isolated_in_pageblock(start_pfn, end_pfn,
						skip_hwpoisoned_pages);
	spin_unlock_irqrestore(&zone->lock, flags);
	return ret ? 0 : -EBUSY;
}

struct page *alloc_migrate_target(struct page *page, unsigned long private,
				  int **resultp)
{
	gfp_t gfp_mask = GFP_USER | __GFP_MOVABLE;

	/*
	 * TODO: allocate a destination hugepage from a nearest neighbor node,
	 * accordance with memory policy of the user process if possible. For
	 * now as a simple work-around, we use the next node for destination.
	 */
	if (PageHuge(page)) {
		nodemask_t src = nodemask_of_node(page_to_nid(page));
		nodemask_t dst;
		nodes_complement(dst, src);
		return alloc_huge_page_node(page_hstate(compound_head(page)),
					    next_node(page_to_nid(page), dst));
	}

	if (PageHighMem(page))
		gfp_mask |= __GFP_HIGHMEM;

	return alloc_page(gfp_mask);
}
Commit	Line	Data
a5d76b54 KH	1	/*
	2	* linux/mm/page_isolation.c
	3	*/
	4
a5d76b54 KH	5	#include <linux/mm.h>
	6	#include <linux/page-isolation.h>
	7	#include <linux/pageblock-flags.h>
ee6f509c	8	#include <linux/memory.h>
c8721bbb	9	#include <linux/hugetlb.h>
a5d76b54 KH	10	#include "internal.h"
a5d76b54 KH	11
b023f468	12	int set_migratetype_isolate(struct page *page, bool skip_hwpoisoned_pages)
ee6f509c MK	13	{
	14	struct zone *zone;
	15	unsigned long flags, pfn;
	16	struct memory_isolate_notify arg;
	17	int notifier_ret;
	18	int ret = -EBUSY;
	19
	20	zone = page_zone(page);
	21
	22	spin_lock_irqsave(&zone->lock, flags);
	23
	24	pfn = page_to_pfn(page);
	25	arg.start_pfn = pfn;
	26	arg.nr_pages = pageblock_nr_pages;
	27	arg.pages_found = 0;
	28
	29	/*
	30	* It may be possible to isolate a pageblock even if the
	31	* migratetype is not MIGRATE_MOVABLE. The memory isolation
	32	* notifier chain is used by balloon drivers to return the
	33	* number of pages in a range that are held by the balloon
	34	* driver to shrink memory. If all the pages are accounted for
	35	* by balloons, are free, or on the LRU, isolation can continue.
	36	* Later, for example, when memory hotplug notifier runs, these
	37	* pages reported as "can be isolated" should be isolated(freed)
	38	* by the balloon driver through the memory notifier chain.
	39	*/
	40	notifier_ret = memory_isolate_notify(MEM_ISOLATE_COUNT, &arg);
	41	notifier_ret = notifier_to_errno(notifier_ret);
	42	if (notifier_ret)
	43	goto out;
	44	/*
	45	* FIXME: Now, memory hotplug doesn't call shrink_slab() by itself.
	46	* We just check MOVABLE pages.
	47	*/
b023f468 WC	48	if (!has_unmovable_pages(zone, page, arg.pages_found,
b023f468 WC	49	skip_hwpoisoned_pages))
ee6f509c MK	50	ret = 0;
	51
	52	/*
	53	* immobile means "not-on-lru" paes. If immobile is larger than
	54	* removable-by-driver pages reported by notifier, we'll fail.
	55	*/
	56
	57	out:
	58	if (!ret) {
2139cbe6	59	unsigned long nr_pages;
d1ce749a	60	int migratetype = get_pageblock_migratetype(page);
2139cbe6	61
a458431e	62	set_pageblock_migratetype(page, MIGRATE_ISOLATE);
ad53f92e	63	zone->nr_isolate_pageblock++;
2139cbe6 BZ	64	nr_pages = move_freepages_block(zone, page, MIGRATE_ISOLATE);
2139cbe6 BZ	65
d1ce749a	66	__mod_zone_freepage_state(zone, -nr_pages, migratetype);
ee6f509c MK	67	}
	68
	69	spin_unlock_irqrestore(&zone->lock, flags);
	70	if (!ret)
ec25af84	71	drain_all_pages(zone);
ee6f509c MK	72	return ret;
	73	}
	74
	75	void unset_migratetype_isolate(struct page *page, unsigned migratetype)
	76	{
	77	struct zone *zone;
2139cbe6	78	unsigned long flags, nr_pages;
3c605096 JK	79	struct page *isolated_page = NULL;
	80	unsigned int order;
	81	unsigned long page_idx, buddy_idx;
	82	struct page *buddy;
2139cbe6	83
ee6f509c MK	84	zone = page_zone(page);
	85	spin_lock_irqsave(&zone->lock, flags);
	86	if (get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
	87	goto out;
3c605096 JK	88
	89	/*
	90	* Because freepage with more than pageblock_order on isolated
	91	* pageblock is restricted to merge due to freepage counting problem,
	92	* it is possible that there is free buddy page.
	93	* move_freepages_block() doesn't care of merge so we need other
	94	* approach in order to merge them. Isolation and free will make
	95	* these pages to be merged.
	96	*/
	97	if (PageBuddy(page)) {
	98	order = page_order(page);
	99	if (order >= pageblock_order) {
	100	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);
	101	buddy_idx = __find_buddy_index(page_idx, order);
	102	buddy = page + (buddy_idx - page_idx);
	103
	104	if (!is_migrate_isolate_page(buddy)) {
	105	__isolate_free_page(page, order);
	106	set_page_refcounted(page);
	107	isolated_page = page;
	108	}
	109	}
	110	}
	111
	112	/*
	113	* If we isolate freepage with more than pageblock_order, there
	114	* should be no freepage in the range, so we could avoid costly
	115	* pageblock scanning for freepage moving.
	116	*/
	117	if (!isolated_page) {
	118	nr_pages = move_freepages_block(zone, page, migratetype);
	119	__mod_zone_freepage_state(zone, nr_pages, migratetype);
	120	}
a458431e	121	set_pageblock_migratetype(page, migratetype);
ad53f92e	122	zone->nr_isolate_pageblock--;
ee6f509c MK	123	out:
ee6f509c MK	124	spin_unlock_irqrestore(&zone->lock, flags);
3c605096 JK	125	if (isolated_page)
3c605096 JK	126	__free_pages(isolated_page, order);
ee6f509c MK	127	}
ee6f509c MK	128
a5d76b54 KH	129	static inline struct page *
	130	__first_valid_page(unsigned long pfn, unsigned long nr_pages)
	131	{
	132	int i;
	133	for (i = 0; i < nr_pages; i++)
	134	if (pfn_valid_within(pfn + i))
	135	break;
	136	if (unlikely(i == nr_pages))
	137	return NULL;
	138	return pfn_to_page(pfn + i);
	139	}
	140
	141	/*
	142	* start_isolate_page_range() -- make page-allocation-type of range of pages
	143	* to be MIGRATE_ISOLATE.
	144	* @start_pfn: The lower PFN of the range to be isolated.
	145	* @end_pfn: The upper PFN of the range to be isolated.
0815f3d8	146	* @migratetype: migrate type to set in error recovery.
a5d76b54 KH	147	*
	148	* Making page-allocation-type to be MIGRATE_ISOLATE means free pages in
	149	* the range will never be allocated. Any free pages and pages freed in the
	150	* future will not be allocated again.
	151	*
	152	* start_pfn/end_pfn must be aligned to pageblock_order.
	153	* Returns 0 on success and -EBUSY if any part of range cannot be isolated.
	154	*/
0815f3d8	155	int start_isolate_page_range(unsigned long start_pfn, unsigned long end_pfn,
b023f468	156	unsigned migratetype, bool skip_hwpoisoned_pages)
a5d76b54 KH	157	{
	158	unsigned long pfn;
	159	unsigned long undo_pfn;
	160	struct page *page;
	161
	162	BUG_ON((start_pfn) & (pageblock_nr_pages - 1));
	163	BUG_ON((end_pfn) & (pageblock_nr_pages - 1));
	164
	165	for (pfn = start_pfn;
	166	pfn < end_pfn;
	167	pfn += pageblock_nr_pages) {
	168	page = __first_valid_page(pfn, pageblock_nr_pages);
b023f468 WC	169	if (page &&
b023f468 WC	170	set_migratetype_isolate(page, skip_hwpoisoned_pages)) {
a5d76b54 KH	171	undo_pfn = pfn;
	172	goto undo;
	173	}
	174	}
	175	return 0;
	176	undo:
	177	for (pfn = start_pfn;
dbc0e4ce	178	pfn < undo_pfn;
a5d76b54	179	pfn += pageblock_nr_pages)
0815f3d8	180	unset_migratetype_isolate(pfn_to_page(pfn), migratetype);
a5d76b54 KH	181
	182	return -EBUSY;
	183	}
	184
	185	/*
	186	* Make isolated pages available again.
	187	*/
0815f3d8 MN	188	int undo_isolate_page_range(unsigned long start_pfn, unsigned long end_pfn,
0815f3d8 MN	189	unsigned migratetype)
a5d76b54 KH	190	{
	191	unsigned long pfn;
	192	struct page *page;
	193	BUG_ON((start_pfn) & (pageblock_nr_pages - 1));
	194	BUG_ON((end_pfn) & (pageblock_nr_pages - 1));
	195	for (pfn = start_pfn;
	196	pfn < end_pfn;
	197	pfn += pageblock_nr_pages) {
	198	page = __first_valid_page(pfn, pageblock_nr_pages);
dbc0e4ce	199	if (!page \|\| get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
a5d76b54	200	continue;
0815f3d8	201	unset_migratetype_isolate(page, migratetype);
a5d76b54 KH	202	}
	203	return 0;
	204	}
	205	/*
	206	* Test all pages in the range is free(means isolated) or not.
	207	* all pages in [start_pfn...end_pfn) must be in the same zone.
	208	* zone->lock must be held before call this.
	209	*
0815f3d8	210	* Returns 1 if all pages in the range are isolated.
a5d76b54 KH	211	*/
a5d76b54 KH	212	static int
b023f468 WC	213	__test_page_isolated_in_pageblock(unsigned long pfn, unsigned long end_pfn,
b023f468 WC	214	bool skip_hwpoisoned_pages)
a5d76b54 KH	215	{
	216	struct page *page;
	217
	218	while (pfn < end_pfn) {
	219	if (!pfn_valid_within(pfn)) {
	220	pfn++;
	221	continue;
	222	}
	223	page = pfn_to_page(pfn);
41d575ad	224	if (PageBuddy(page)) {
435b405c MK	225	/*
	226	* If race between isolatation and allocation happens,
	227	* some free pages could be in MIGRATE_MOVABLE list
	228	* although pageblock's migratation type of the page
	229	* is MIGRATE_ISOLATE. Catch it and move the page into
	230	* MIGRATE_ISOLATE list.
	231	*/
	232	if (get_freepage_migratetype(page) != MIGRATE_ISOLATE) {
	233	struct page *end_page;
	234
	235	end_page = page + (1 << page_order(page)) - 1;
	236	move_freepages(page_zone(page), page, end_page,
	237	MIGRATE_ISOLATE);
	238	}
a5d76b54	239	pfn += 1 << page_order(page);
41d575ad	240	}
a5d76b54	241	else if (page_count(page) == 0 &&
b12c4ad1	242	get_freepage_migratetype(page) == MIGRATE_ISOLATE)
a5d76b54	243	pfn += 1;
b023f468 WC	244	else if (skip_hwpoisoned_pages && PageHWPoison(page)) {
	245	/*
	246	* The HWPoisoned page may be not in buddy
	247	* system, and page_count() is not 0.
	248	*/
	249	pfn++;
	250	continue;
	251	}
a5d76b54 KH	252	else
	253	break;
	254	}
	255	if (pfn < end_pfn)
	256	return 0;
	257	return 1;
	258	}
	259
b023f468 WC	260	int test_pages_isolated(unsigned long start_pfn, unsigned long end_pfn,
b023f468 WC	261	bool skip_hwpoisoned_pages)
a5d76b54	262	{
6c1b7f68	263	unsigned long pfn, flags;
a5d76b54	264	struct page *page;
6c1b7f68 GS	265	struct zone *zone;
6c1b7f68 GS	266	int ret;
a5d76b54	267
a5d76b54	268	/*
85dbe706 TC	269	* Note: pageblock_nr_pages != MAX_ORDER. Then, chunks of free pages
	270	* are not aligned to pageblock_nr_pages.
	271	* Then we just check migratetype first.
a5d76b54 KH	272	*/
	273	for (pfn = start_pfn; pfn < end_pfn; pfn += pageblock_nr_pages) {
	274	page = __first_valid_page(pfn, pageblock_nr_pages);
dbc0e4ce	275	if (page && get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
a5d76b54 KH	276	break;
a5d76b54 KH	277	}
a70dcb96 GS	278	page = __first_valid_page(start_pfn, end_pfn - start_pfn);
a70dcb96 GS	279	if ((pfn < end_pfn) \|\| !page)
a5d76b54	280	return -EBUSY;
85dbe706	281	/* Check all pages are free or marked as ISOLATED */
a70dcb96	282	zone = page_zone(page);
6c1b7f68	283	spin_lock_irqsave(&zone->lock, flags);
b023f468 WC	284	ret = __test_page_isolated_in_pageblock(start_pfn, end_pfn,
b023f468 WC	285	skip_hwpoisoned_pages);
6c1b7f68 GS	286	spin_unlock_irqrestore(&zone->lock, flags);
6c1b7f68 GS	287	return ret ? 0 : -EBUSY;
a5d76b54	288	}
723a0644 MK	289
	290	struct page alloc_migrate_target(struct page page, unsigned long private,
	291	int **resultp)
	292	{
	293	gfp_t gfp_mask = GFP_USER \| __GFP_MOVABLE;
	294
c8721bbb NH	295	/*
	296	* TODO: allocate a destination hugepage from a nearest neighbor node,
	297	* accordance with memory policy of the user process if possible. For
	298	* now as a simple work-around, we use the next node for destination.
	299	*/
	300	if (PageHuge(page)) {
	301	nodemask_t src = nodemask_of_node(page_to_nid(page));
	302	nodemask_t dst;
	303	nodes_complement(dst, src);
	304	return alloc_huge_page_node(page_hstate(compound_head(page)),
	305	next_node(page_to_nid(page), dst));
	306	}
	307
723a0644 MK	308	if (PageHighMem(page))
	309	gfp_mask \|= __GFP_HIGHMEM;
	310
	311	return alloc_page(gfp_mask);
	312	}