[mirror_ubuntu-bionic-kernel.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 <linux/page_owner.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);
				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) {
		kernel_map_pages(page, (1 << order), 1);
		set_page_refcounted(page);
		set_page_owner(page, order, __GFP_MOVABLE);
		__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;
}

/* Caller should ensure that requested range is in a single zone */
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))
		return alloc_huge_page_node(page_hstate(compound_head(page)),
					    next_node_in(page_to_nid(page),
							 node_online_map));

	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>
83358ece	10	#include <linux/page_owner.h>
a5d76b54 KH	11	#include "internal.h"
a5d76b54 KH	12
0f0848e5 JK	13	#define CREATE_TRACE_POINTS
	14	#include <trace/events/page_isolation.h>
	15
c5b4e1b0 NH	16	static int set_migratetype_isolate(struct page *page,
c5b4e1b0 NH	17	bool skip_hwpoisoned_pages)
ee6f509c MK	18	{
	19	struct zone *zone;
	20	unsigned long flags, pfn;
	21	struct memory_isolate_notify arg;
	22	int notifier_ret;
	23	int ret = -EBUSY;
	24
	25	zone = page_zone(page);
	26
	27	spin_lock_irqsave(&zone->lock, flags);
	28
	29	pfn = page_to_pfn(page);
	30	arg.start_pfn = pfn;
	31	arg.nr_pages = pageblock_nr_pages;
	32	arg.pages_found = 0;
	33
	34	/*
	35	* It may be possible to isolate a pageblock even if the
	36	* migratetype is not MIGRATE_MOVABLE. The memory isolation
	37	* notifier chain is used by balloon drivers to return the
	38	* number of pages in a range that are held by the balloon
	39	* driver to shrink memory. If all the pages are accounted for
	40	* by balloons, are free, or on the LRU, isolation can continue.
	41	* Later, for example, when memory hotplug notifier runs, these
	42	* pages reported as "can be isolated" should be isolated(freed)
	43	* by the balloon driver through the memory notifier chain.
	44	*/
	45	notifier_ret = memory_isolate_notify(MEM_ISOLATE_COUNT, &arg);
	46	notifier_ret = notifier_to_errno(notifier_ret);
	47	if (notifier_ret)
	48	goto out;
	49	/*
	50	* FIXME: Now, memory hotplug doesn't call shrink_slab() by itself.
	51	* We just check MOVABLE pages.
	52	*/
b023f468 WC	53	if (!has_unmovable_pages(zone, page, arg.pages_found,
b023f468 WC	54	skip_hwpoisoned_pages))
ee6f509c MK	55	ret = 0;
	56
	57	/*
	58	* immobile means "not-on-lru" paes. If immobile is larger than
	59	* removable-by-driver pages reported by notifier, we'll fail.
	60	*/
	61
	62	out:
	63	if (!ret) {
2139cbe6	64	unsigned long nr_pages;
d1ce749a	65	int migratetype = get_pageblock_migratetype(page);
2139cbe6	66
a458431e	67	set_pageblock_migratetype(page, MIGRATE_ISOLATE);
ad53f92e	68	zone->nr_isolate_pageblock++;
2139cbe6 BZ	69	nr_pages = move_freepages_block(zone, page, MIGRATE_ISOLATE);
2139cbe6 BZ	70
d1ce749a	71	__mod_zone_freepage_state(zone, -nr_pages, migratetype);
ee6f509c MK	72	}
	73
	74	spin_unlock_irqrestore(&zone->lock, flags);
	75	if (!ret)
ec25af84	76	drain_all_pages(zone);
ee6f509c MK	77	return ret;
	78	}
	79
c5b4e1b0	80	static void unset_migratetype_isolate(struct page *page, unsigned migratetype)
ee6f509c MK	81	{
ee6f509c MK	82	struct zone *zone;
2139cbe6	83	unsigned long flags, nr_pages;
3c605096 JK	84	struct page *isolated_page = NULL;
	85	unsigned int order;
	86	unsigned long page_idx, buddy_idx;
	87	struct page *buddy;
2139cbe6	88
ee6f509c MK	89	zone = page_zone(page);
	90	spin_lock_irqsave(&zone->lock, flags);
	91	if (get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
	92	goto out;
3c605096 JK	93
	94	/*
	95	* Because freepage with more than pageblock_order on isolated
	96	* pageblock is restricted to merge due to freepage counting problem,
	97	* it is possible that there is free buddy page.
	98	* move_freepages_block() doesn't care of merge so we need other
	99	* approach in order to merge them. Isolation and free will make
	100	* these pages to be merged.
	101	*/
	102	if (PageBuddy(page)) {
	103	order = page_order(page);
	104	if (order >= pageblock_order) {
	105	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);
	106	buddy_idx = __find_buddy_index(page_idx, order);
	107	buddy = page + (buddy_idx - page_idx);
	108
1ae7013d HZ	109	if (pfn_valid_within(page_to_pfn(buddy)) &&
1ae7013d HZ	110	!is_migrate_isolate_page(buddy)) {
3c605096	111	__isolate_free_page(page, order);
3c605096 JK	112	isolated_page = page;
	113	}
	114	}
	115	}
	116
	117	/*
	118	* If we isolate freepage with more than pageblock_order, there
	119	* should be no freepage in the range, so we could avoid costly
	120	* pageblock scanning for freepage moving.
	121	*/
	122	if (!isolated_page) {
	123	nr_pages = move_freepages_block(zone, page, migratetype);
	124	__mod_zone_freepage_state(zone, nr_pages, migratetype);
	125	}
a458431e	126	set_pageblock_migratetype(page, migratetype);
ad53f92e	127	zone->nr_isolate_pageblock--;
ee6f509c MK	128	out:
ee6f509c MK	129	spin_unlock_irqrestore(&zone->lock, flags);
83358ece JK	130	if (isolated_page) {
	131	kernel_map_pages(page, (1 << order), 1);
	132	set_page_refcounted(page);
	133	set_page_owner(page, order, __GFP_MOVABLE);
3c605096	134	__free_pages(isolated_page, order);
83358ece	135	}
ee6f509c MK	136	}
ee6f509c MK	137
a5d76b54 KH	138	static inline struct page *
	139	__first_valid_page(unsigned long pfn, unsigned long nr_pages)
	140	{
	141	int i;
	142	for (i = 0; i < nr_pages; i++)
	143	if (pfn_valid_within(pfn + i))
	144	break;
	145	if (unlikely(i == nr_pages))
	146	return NULL;
	147	return pfn_to_page(pfn + i);
	148	}
	149
	150	/*
	151	* start_isolate_page_range() -- make page-allocation-type of range of pages
	152	* to be MIGRATE_ISOLATE.
	153	* @start_pfn: The lower PFN of the range to be isolated.
	154	* @end_pfn: The upper PFN of the range to be isolated.
0815f3d8	155	* @migratetype: migrate type to set in error recovery.
a5d76b54 KH	156	*
	157	* Making page-allocation-type to be MIGRATE_ISOLATE means free pages in
	158	* the range will never be allocated. Any free pages and pages freed in the
	159	* future will not be allocated again.
	160	*
	161	* start_pfn/end_pfn must be aligned to pageblock_order.
	162	* Returns 0 on success and -EBUSY if any part of range cannot be isolated.
	163	*/
0815f3d8	164	int start_isolate_page_range(unsigned long start_pfn, unsigned long end_pfn,
b023f468	165	unsigned migratetype, bool skip_hwpoisoned_pages)
a5d76b54 KH	166	{
	167	unsigned long pfn;
	168	unsigned long undo_pfn;
	169	struct page *page;
	170
fec174d6 NH	171	BUG_ON(!IS_ALIGNED(start_pfn, pageblock_nr_pages));
fec174d6 NH	172	BUG_ON(!IS_ALIGNED(end_pfn, pageblock_nr_pages));
a5d76b54 KH	173
	174	for (pfn = start_pfn;
	175	pfn < end_pfn;
	176	pfn += pageblock_nr_pages) {
	177	page = __first_valid_page(pfn, pageblock_nr_pages);
b023f468 WC	178	if (page &&
b023f468 WC	179	set_migratetype_isolate(page, skip_hwpoisoned_pages)) {
a5d76b54 KH	180	undo_pfn = pfn;
	181	goto undo;
	182	}
	183	}
	184	return 0;
	185	undo:
	186	for (pfn = start_pfn;
dbc0e4ce	187	pfn < undo_pfn;
a5d76b54	188	pfn += pageblock_nr_pages)
0815f3d8	189	unset_migratetype_isolate(pfn_to_page(pfn), migratetype);
a5d76b54 KH	190
	191	return -EBUSY;
	192	}
	193
	194	/*
	195	* Make isolated pages available again.
	196	*/
0815f3d8 MN	197	int undo_isolate_page_range(unsigned long start_pfn, unsigned long end_pfn,
0815f3d8 MN	198	unsigned migratetype)
a5d76b54 KH	199	{
	200	unsigned long pfn;
	201	struct page *page;
6f8d2b8a WX	202
	203	BUG_ON(!IS_ALIGNED(start_pfn, pageblock_nr_pages));
	204	BUG_ON(!IS_ALIGNED(end_pfn, pageblock_nr_pages));
	205
a5d76b54 KH	206	for (pfn = start_pfn;
	207	pfn < end_pfn;
	208	pfn += pageblock_nr_pages) {
	209	page = __first_valid_page(pfn, pageblock_nr_pages);
dbc0e4ce	210	if (!page \|\| get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
a5d76b54	211	continue;
0815f3d8	212	unset_migratetype_isolate(page, migratetype);
a5d76b54 KH	213	}
	214	return 0;
	215	}
	216	/*
	217	* Test all pages in the range is free(means isolated) or not.
	218	* all pages in [start_pfn...end_pfn) must be in the same zone.
	219	* zone->lock must be held before call this.
	220	*
ec3b6882	221	* Returns the last tested pfn.
a5d76b54	222	*/
fea85cff	223	static unsigned long
b023f468 WC	224	__test_page_isolated_in_pageblock(unsigned long pfn, unsigned long end_pfn,
b023f468 WC	225	bool skip_hwpoisoned_pages)
a5d76b54 KH	226	{
	227	struct page *page;
	228
	229	while (pfn < end_pfn) {
	230	if (!pfn_valid_within(pfn)) {
	231	pfn++;
	232	continue;
	233	}
	234	page = pfn_to_page(pfn);
aa016d14	235	if (PageBuddy(page))
435b405c	236	/*
aa016d14 VB	237	* If the page is on a free list, it has to be on
	238	* the correct MIGRATE_ISOLATE freelist. There is no
	239	* simple way to verify that as VM_BUG_ON(), though.
435b405c	240	*/
a5d76b54	241	pfn += 1 << page_order(page);
aa016d14 VB	242	else if (skip_hwpoisoned_pages && PageHWPoison(page))
aa016d14 VB	243	/* A HWPoisoned page cannot be also PageBuddy */
b023f468	244	pfn++;
a5d76b54 KH	245	else
	246	break;
	247	}
fea85cff JK	248
fea85cff JK	249	return pfn;
a5d76b54 KH	250	}
a5d76b54 KH	251
b9eb6319	252	/* Caller should ensure that requested range is in a single zone */
b023f468 WC	253	int test_pages_isolated(unsigned long start_pfn, unsigned long end_pfn,
b023f468 WC	254	bool skip_hwpoisoned_pages)
a5d76b54	255	{
6c1b7f68	256	unsigned long pfn, flags;
a5d76b54	257	struct page *page;
6c1b7f68	258	struct zone *zone;
a5d76b54	259
a5d76b54	260	/*
85dbe706 TC	261	* Note: pageblock_nr_pages != MAX_ORDER. Then, chunks of free pages
	262	* are not aligned to pageblock_nr_pages.
	263	* Then we just check migratetype first.
a5d76b54 KH	264	*/
	265	for (pfn = start_pfn; pfn < end_pfn; pfn += pageblock_nr_pages) {
	266	page = __first_valid_page(pfn, pageblock_nr_pages);
dbc0e4ce	267	if (page && get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
a5d76b54 KH	268	break;
a5d76b54 KH	269	}
a70dcb96 GS	270	page = __first_valid_page(start_pfn, end_pfn - start_pfn);
a70dcb96 GS	271	if ((pfn < end_pfn) \|\| !page)
a5d76b54	272	return -EBUSY;
85dbe706	273	/* Check all pages are free or marked as ISOLATED */
a70dcb96	274	zone = page_zone(page);
6c1b7f68	275	spin_lock_irqsave(&zone->lock, flags);
fea85cff	276	pfn = __test_page_isolated_in_pageblock(start_pfn, end_pfn,
b023f468	277	skip_hwpoisoned_pages);
6c1b7f68	278	spin_unlock_irqrestore(&zone->lock, flags);
fea85cff	279
0f0848e5 JK	280	trace_test_pages_isolated(start_pfn, end_pfn, pfn);
0f0848e5 JK	281
fea85cff	282	return pfn < end_pfn ? -EBUSY : 0;
a5d76b54	283	}
723a0644 MK	284
	285	struct page alloc_migrate_target(struct page page, unsigned long private,
	286	int **resultp)
	287	{
	288	gfp_t gfp_mask = GFP_USER \| __GFP_MOVABLE;
	289
c8721bbb NH	290	/*
	291	* TODO: allocate a destination hugepage from a nearest neighbor node,
	292	* accordance with memory policy of the user process if possible. For
	293	* now as a simple work-around, we use the next node for destination.
	294	*/
0edaf86c	295	if (PageHuge(page))
c8721bbb	296	return alloc_huge_page_node(page_hstate(compound_head(page)),
0edaf86c AM	297	next_node_in(page_to_nid(page),
0edaf86c AM	298	node_online_map));
c8721bbb	299
723a0644 MK	300	if (PageHighMem(page))
	301	gfp_mask \|= __GFP_HIGHMEM;
	302
	303	return alloc_page(gfp_mask);
	304	}