[linux-2.6-block.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"

#define CREATE_TRACE_POINTS
#include <trace/events/page_isolation.h>

static 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;
}

static 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 (pfn_valid_within(page_to_pfn(buddy)) &&
			    !is_migrate_isolate_page(buddy)) {
				__isolate_free_page(page, order);
				kernel_map_pages(page, (1 << order), 1);
				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(!IS_ALIGNED(start_pfn, pageblock_nr_pages));
	BUG_ON(!IS_ALIGNED(end_pfn, pageblock_nr_pages));

	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(!IS_ALIGNED(start_pfn, pageblock_nr_pages));
	BUG_ON(!IS_ALIGNED(end_pfn, pageblock_nr_pages));

	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 the last tested pfn.
 */
static unsigned long
__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 the page is on a free list, it has to be on
			 * the correct MIGRATE_ISOLATE freelist. There is no
			 * simple way to verify that as VM_BUG_ON(), though.
			 */
			pfn += 1 << page_order(page);
		else if (skip_hwpoisoned_pages && PageHWPoison(page))
			/* A HWPoisoned page cannot be also PageBuddy */
			pfn++;
		else
			break;
	}

	return pfn;
}

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;

	/*
	 * 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);
	pfn = __test_page_isolated_in_pageblock(start_pfn, end_pfn,
						skip_hwpoisoned_pages);
	spin_unlock_irqrestore(&zone->lock, flags);

	trace_test_pages_isolated(start_pfn, end_pfn, pfn);

	return pfn < end_pfn ? -EBUSY : 0;
}

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)) {
		int node = next_online_node(page_to_nid(page));
		if (node == MAX_NUMNODES)
			node = first_online_node;
		return alloc_huge_page_node(page_hstate(compound_head(page)),
					    node);
	}

	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
0f0848e5 JK	12	#define CREATE_TRACE_POINTS
	13	#include <trace/events/page_isolation.h>
	14
c5b4e1b0 NH	15	static int set_migratetype_isolate(struct page *page,
c5b4e1b0 NH	16	bool skip_hwpoisoned_pages)
ee6f509c MK	17	{
	18	struct zone *zone;
	19	unsigned long flags, pfn;
	20	struct memory_isolate_notify arg;
	21	int notifier_ret;
	22	int ret = -EBUSY;
	23
	24	zone = page_zone(page);
	25
	26	spin_lock_irqsave(&zone->lock, flags);
	27
	28	pfn = page_to_pfn(page);
	29	arg.start_pfn = pfn;
	30	arg.nr_pages = pageblock_nr_pages;
	31	arg.pages_found = 0;
	32
	33	/*
	34	* It may be possible to isolate a pageblock even if the
	35	* migratetype is not MIGRATE_MOVABLE. The memory isolation
	36	* notifier chain is used by balloon drivers to return the
	37	* number of pages in a range that are held by the balloon
	38	* driver to shrink memory. If all the pages are accounted for
	39	* by balloons, are free, or on the LRU, isolation can continue.
	40	* Later, for example, when memory hotplug notifier runs, these
	41	* pages reported as "can be isolated" should be isolated(freed)
	42	* by the balloon driver through the memory notifier chain.
	43	*/
	44	notifier_ret = memory_isolate_notify(MEM_ISOLATE_COUNT, &arg);
	45	notifier_ret = notifier_to_errno(notifier_ret);
	46	if (notifier_ret)
	47	goto out;
	48	/*
	49	* FIXME: Now, memory hotplug doesn't call shrink_slab() by itself.
	50	* We just check MOVABLE pages.
	51	*/
b023f468 WC	52	if (!has_unmovable_pages(zone, page, arg.pages_found,
b023f468 WC	53	skip_hwpoisoned_pages))
ee6f509c MK	54	ret = 0;
	55
	56	/*
	57	* immobile means "not-on-lru" paes. If immobile is larger than
	58	* removable-by-driver pages reported by notifier, we'll fail.
	59	*/
	60
	61	out:
	62	if (!ret) {
2139cbe6	63	unsigned long nr_pages;
d1ce749a	64	int migratetype = get_pageblock_migratetype(page);
2139cbe6	65
a458431e	66	set_pageblock_migratetype(page, MIGRATE_ISOLATE);
ad53f92e	67	zone->nr_isolate_pageblock++;
2139cbe6 BZ	68	nr_pages = move_freepages_block(zone, page, MIGRATE_ISOLATE);
2139cbe6 BZ	69
d1ce749a	70	__mod_zone_freepage_state(zone, -nr_pages, migratetype);
ee6f509c MK	71	}
	72
	73	spin_unlock_irqrestore(&zone->lock, flags);
	74	if (!ret)
ec25af84	75	drain_all_pages(zone);
ee6f509c MK	76	return ret;
	77	}
	78
c5b4e1b0	79	static void unset_migratetype_isolate(struct page *page, unsigned migratetype)
ee6f509c MK	80	{
ee6f509c MK	81	struct zone *zone;
2139cbe6	82	unsigned long flags, nr_pages;
3c605096 JK	83	struct page *isolated_page = NULL;
	84	unsigned int order;
	85	unsigned long page_idx, buddy_idx;
	86	struct page *buddy;
2139cbe6	87
ee6f509c MK	88	zone = page_zone(page);
	89	spin_lock_irqsave(&zone->lock, flags);
	90	if (get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
	91	goto out;
3c605096 JK	92
	93	/*
	94	* Because freepage with more than pageblock_order on isolated
	95	* pageblock is restricted to merge due to freepage counting problem,
	96	* it is possible that there is free buddy page.
	97	* move_freepages_block() doesn't care of merge so we need other
	98	* approach in order to merge them. Isolation and free will make
	99	* these pages to be merged.
	100	*/
	101	if (PageBuddy(page)) {
	102	order = page_order(page);
	103	if (order >= pageblock_order) {
	104	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);
	105	buddy_idx = __find_buddy_index(page_idx, order);
	106	buddy = page + (buddy_idx - page_idx);
	107
1ae7013d HZ	108	if (pfn_valid_within(page_to_pfn(buddy)) &&
1ae7013d HZ	109	!is_migrate_isolate_page(buddy)) {
3c605096	110	__isolate_free_page(page, order);
cfa86943	111	kernel_map_pages(page, (1 << order), 1);
3c605096 JK	112	set_page_refcounted(page);
	113	isolated_page = page;
	114	}
	115	}
	116	}
	117
	118	/*
	119	* If we isolate freepage with more than pageblock_order, there
	120	* should be no freepage in the range, so we could avoid costly
	121	* pageblock scanning for freepage moving.
	122	*/
	123	if (!isolated_page) {
	124	nr_pages = move_freepages_block(zone, page, migratetype);
	125	__mod_zone_freepage_state(zone, nr_pages, migratetype);
	126	}
a458431e	127	set_pageblock_migratetype(page, migratetype);
ad53f92e	128	zone->nr_isolate_pageblock--;
ee6f509c MK	129	out:
ee6f509c MK	130	spin_unlock_irqrestore(&zone->lock, flags);
3c605096 JK	131	if (isolated_page)
3c605096 JK	132	__free_pages(isolated_page, order);
ee6f509c MK	133	}
ee6f509c MK	134
a5d76b54 KH	135	static inline struct page *
	136	__first_valid_page(unsigned long pfn, unsigned long nr_pages)
	137	{
	138	int i;
	139	for (i = 0; i < nr_pages; i++)
	140	if (pfn_valid_within(pfn + i))
	141	break;
	142	if (unlikely(i == nr_pages))
	143	return NULL;
	144	return pfn_to_page(pfn + i);
	145	}
	146
	147	/*
	148	* start_isolate_page_range() -- make page-allocation-type of range of pages
	149	* to be MIGRATE_ISOLATE.
	150	* @start_pfn: The lower PFN of the range to be isolated.
	151	* @end_pfn: The upper PFN of the range to be isolated.
0815f3d8	152	* @migratetype: migrate type to set in error recovery.
a5d76b54 KH	153	*
	154	* Making page-allocation-type to be MIGRATE_ISOLATE means free pages in
	155	* the range will never be allocated. Any free pages and pages freed in the
	156	* future will not be allocated again.
	157	*
	158	* start_pfn/end_pfn must be aligned to pageblock_order.
	159	* Returns 0 on success and -EBUSY if any part of range cannot be isolated.
	160	*/
0815f3d8	161	int start_isolate_page_range(unsigned long start_pfn, unsigned long end_pfn,
b023f468	162	unsigned migratetype, bool skip_hwpoisoned_pages)
a5d76b54 KH	163	{
	164	unsigned long pfn;
	165	unsigned long undo_pfn;
	166	struct page *page;
	167
fec174d6 NH	168	BUG_ON(!IS_ALIGNED(start_pfn, pageblock_nr_pages));
fec174d6 NH	169	BUG_ON(!IS_ALIGNED(end_pfn, pageblock_nr_pages));
a5d76b54 KH	170
	171	for (pfn = start_pfn;
	172	pfn < end_pfn;
	173	pfn += pageblock_nr_pages) {
	174	page = __first_valid_page(pfn, pageblock_nr_pages);
b023f468 WC	175	if (page &&
b023f468 WC	176	set_migratetype_isolate(page, skip_hwpoisoned_pages)) {
a5d76b54 KH	177	undo_pfn = pfn;
	178	goto undo;
	179	}
	180	}
	181	return 0;
	182	undo:
	183	for (pfn = start_pfn;
dbc0e4ce	184	pfn < undo_pfn;
a5d76b54	185	pfn += pageblock_nr_pages)
0815f3d8	186	unset_migratetype_isolate(pfn_to_page(pfn), migratetype);
a5d76b54 KH	187
	188	return -EBUSY;
	189	}
	190
	191	/*
	192	* Make isolated pages available again.
	193	*/
0815f3d8 MN	194	int undo_isolate_page_range(unsigned long start_pfn, unsigned long end_pfn,
0815f3d8 MN	195	unsigned migratetype)
a5d76b54 KH	196	{
	197	unsigned long pfn;
	198	struct page *page;
6f8d2b8a WX	199
	200	BUG_ON(!IS_ALIGNED(start_pfn, pageblock_nr_pages));
	201	BUG_ON(!IS_ALIGNED(end_pfn, pageblock_nr_pages));
	202
a5d76b54 KH	203	for (pfn = start_pfn;
	204	pfn < end_pfn;
	205	pfn += pageblock_nr_pages) {
	206	page = __first_valid_page(pfn, pageblock_nr_pages);
dbc0e4ce	207	if (!page \|\| get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
a5d76b54	208	continue;
0815f3d8	209	unset_migratetype_isolate(page, migratetype);
a5d76b54 KH	210	}
	211	return 0;
	212	}
	213	/*
	214	* Test all pages in the range is free(means isolated) or not.
	215	* all pages in [start_pfn...end_pfn) must be in the same zone.
	216	* zone->lock must be held before call this.
	217	*
ec3b6882	218	* Returns the last tested pfn.
a5d76b54	219	*/
fea85cff	220	static unsigned long
b023f468 WC	221	__test_page_isolated_in_pageblock(unsigned long pfn, unsigned long end_pfn,
b023f468 WC	222	bool skip_hwpoisoned_pages)
a5d76b54 KH	223	{
	224	struct page *page;
	225
	226	while (pfn < end_pfn) {
	227	if (!pfn_valid_within(pfn)) {
	228	pfn++;
	229	continue;
	230	}
	231	page = pfn_to_page(pfn);
aa016d14	232	if (PageBuddy(page))
435b405c	233	/*
aa016d14 VB	234	* If the page is on a free list, it has to be on
	235	* the correct MIGRATE_ISOLATE freelist. There is no
	236	* simple way to verify that as VM_BUG_ON(), though.
435b405c	237	*/
a5d76b54	238	pfn += 1 << page_order(page);
aa016d14 VB	239	else if (skip_hwpoisoned_pages && PageHWPoison(page))
aa016d14 VB	240	/* A HWPoisoned page cannot be also PageBuddy */
b023f468	241	pfn++;
a5d76b54 KH	242	else
	243	break;
	244	}
fea85cff JK	245
fea85cff JK	246	return pfn;
a5d76b54 KH	247	}
a5d76b54 KH	248
b023f468 WC	249	int test_pages_isolated(unsigned long start_pfn, unsigned long end_pfn,
b023f468 WC	250	bool skip_hwpoisoned_pages)
a5d76b54	251	{
6c1b7f68	252	unsigned long pfn, flags;
a5d76b54	253	struct page *page;
6c1b7f68	254	struct zone *zone;
a5d76b54	255
a5d76b54	256	/*
85dbe706 TC	257	* Note: pageblock_nr_pages != MAX_ORDER. Then, chunks of free pages
	258	* are not aligned to pageblock_nr_pages.
	259	* Then we just check migratetype first.
a5d76b54 KH	260	*/
	261	for (pfn = start_pfn; pfn < end_pfn; pfn += pageblock_nr_pages) {
	262	page = __first_valid_page(pfn, pageblock_nr_pages);
dbc0e4ce	263	if (page && get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
a5d76b54 KH	264	break;
a5d76b54 KH	265	}
a70dcb96 GS	266	page = __first_valid_page(start_pfn, end_pfn - start_pfn);
a70dcb96 GS	267	if ((pfn < end_pfn) \|\| !page)
a5d76b54	268	return -EBUSY;
85dbe706	269	/* Check all pages are free or marked as ISOLATED */
a70dcb96	270	zone = page_zone(page);
6c1b7f68	271	spin_lock_irqsave(&zone->lock, flags);
fea85cff	272	pfn = __test_page_isolated_in_pageblock(start_pfn, end_pfn,
b023f468	273	skip_hwpoisoned_pages);
6c1b7f68	274	spin_unlock_irqrestore(&zone->lock, flags);
fea85cff	275
0f0848e5 JK	276	trace_test_pages_isolated(start_pfn, end_pfn, pfn);
0f0848e5 JK	277
fea85cff	278	return pfn < end_pfn ? -EBUSY : 0;
a5d76b54	279	}
723a0644 MK	280
	281	struct page alloc_migrate_target(struct page page, unsigned long private,
	282	int **resultp)
	283	{
	284	gfp_t gfp_mask = GFP_USER \| __GFP_MOVABLE;
	285
c8721bbb NH	286	/*
	287	* TODO: allocate a destination hugepage from a nearest neighbor node,
	288	* accordance with memory policy of the user process if possible. For
	289	* now as a simple work-around, we use the next node for destination.
	290	*/
	291	if (PageHuge(page)) {
6f25a14a XQ	292	int node = next_online_node(page_to_nid(page));
	293	if (node == MAX_NUMNODES)
	294	node = first_online_node;
c8721bbb	295	return alloc_huge_page_node(page_hstate(compound_head(page)),
6f25a14a	296	node);
c8721bbb NH	297	}
c8721bbb NH	298
723a0644 MK	299	if (PageHighMem(page))
	300	gfp_mask \|= __GFP_HIGHMEM;
	301
	302	return alloc_page(gfp_mask);
	303	}