[linux-block.git] / mm / page_reporting.c

// SPDX-License-Identifier: GPL-2.0
#include <linux/mm.h>
#include <linux/mmzone.h>
#include <linux/page_reporting.h>
#include <linux/gfp.h>
#include <linux/export.h>
#include <linux/delay.h>
#include <linux/scatterlist.h>

#include "page_reporting.h"
#include "internal.h"

#define PAGE_REPORTING_DELAY	(2 * HZ)
static struct page_reporting_dev_info __rcu *pr_dev_info __read_mostly;

enum {
	PAGE_REPORTING_IDLE = 0,
	PAGE_REPORTING_REQUESTED,
	PAGE_REPORTING_ACTIVE
};

/* request page reporting */
static void
__page_reporting_request(struct page_reporting_dev_info *prdev)
{
	unsigned int state;

	/* Check to see if we are in desired state */
	state = atomic_read(&prdev->state);
	if (state == PAGE_REPORTING_REQUESTED)
		return;

	/*
	 *  If reporting is already active there is nothing we need to do.
	 *  Test against 0 as that represents PAGE_REPORTING_IDLE.
	 */
	state = atomic_xchg(&prdev->state, PAGE_REPORTING_REQUESTED);
	if (state != PAGE_REPORTING_IDLE)
		return;

	/*
	 * Delay the start of work to allow a sizable queue to build. For
	 * now we are limiting this to running no more than once every
	 * couple of seconds.
	 */
	schedule_delayed_work(&prdev->work, PAGE_REPORTING_DELAY);
}

/* notify prdev of free page reporting request */
void __page_reporting_notify(void)
{
	struct page_reporting_dev_info *prdev;

	/*
	 * We use RCU to protect the pr_dev_info pointer. In almost all
	 * cases this should be present, however in the unlikely case of
	 * a shutdown this will be NULL and we should exit.
	 */
	rcu_read_lock();
	prdev = rcu_dereference(pr_dev_info);
	if (likely(prdev))
		__page_reporting_request(prdev);

	rcu_read_unlock();
}

static void
page_reporting_drain(struct page_reporting_dev_info *prdev,
		     struct scatterlist *sgl, unsigned int nents, bool reported)
{
	struct scatterlist *sg = sgl;

	/*
	 * Drain the now reported pages back into their respective
	 * free lists/areas. We assume at least one page is populated.
	 */
	do {
		struct page *page = sg_page(sg);
		int mt = get_pageblock_migratetype(page);
		unsigned int order = get_order(sg->length);

		__putback_isolated_page(page, order, mt);

		/* If the pages were not reported due to error skip flagging */
		if (!reported)
			continue;

		/*
		 * If page was not comingled with another page we can
		 * consider the result to be "reported" since the page
		 * hasn't been modified, otherwise we will need to
		 * report on the new larger page when we make our way
		 * up to that higher order.
		 */
		if (PageBuddy(page) && buddy_order(page) == order)
			__SetPageReported(page);
	} while ((sg = sg_next(sg)));

	/* reinitialize scatterlist now that it is empty */
	sg_init_table(sgl, nents);
}

/*
 * The page reporting cycle consists of 4 stages, fill, report, drain, and
 * idle. We will cycle through the first 3 stages until we cannot obtain a
 * full scatterlist of pages, in that case we will switch to idle.
 */
static int
page_reporting_cycle(struct page_reporting_dev_info *prdev, struct zone *zone,
		     unsigned int order, unsigned int mt,
		     struct scatterlist *sgl, unsigned int *offset)
{
	struct free_area *area = &zone->free_area[order];
	struct list_head *list = &area->free_list[mt];
	unsigned int page_len = PAGE_SIZE << order;
	struct page *page, *next;
	long budget;
	int err = 0;

	/*
	 * Perform early check, if free area is empty there is
	 * nothing to process so we can skip this free_list.
	 */
	if (list_empty(list))
		return err;

	spin_lock_irq(&zone->lock);

	/*
	 * Limit how many calls we will be making to the page reporting
	 * device for this list. By doing this we avoid processing any
	 * given list for too long.
	 *
	 * The current value used allows us enough calls to process over a
	 * sixteenth of the current list plus one additional call to handle
	 * any pages that may have already been present from the previous
	 * list processed. This should result in us reporting all pages on
	 * an idle system in about 30 seconds.
	 *
	 * The division here should be cheap since PAGE_REPORTING_CAPACITY
	 * should always be a power of 2.
	 */
	budget = DIV_ROUND_UP(area->nr_free, PAGE_REPORTING_CAPACITY * 16);

	/* loop through free list adding unreported pages to sg list */
	list_for_each_entry_safe(page, next, list, lru) {
		/* We are going to skip over the reported pages. */
		if (PageReported(page))
			continue;

		/*
		 * If we fully consumed our budget then update our
		 * state to indicate that we are requesting additional
		 * processing and exit this list.
		 */
		if (budget < 0) {
			atomic_set(&prdev->state, PAGE_REPORTING_REQUESTED);
			next = page;
			break;
		}

		/* Attempt to pull page from list and place in scatterlist */
		if (*offset) {
			if (!__isolate_free_page(page, order)) {
				next = page;
				break;
			}

			/* Add page to scatter list */
			--(*offset);
			sg_set_page(&sgl[*offset], page, page_len, 0);

			continue;
		}

		/*
		 * Make the first non-reported page in the free list
		 * the new head of the free list before we release the
		 * zone lock.
		 */
		if (!list_is_first(&page->lru, list))
			list_rotate_to_front(&page->lru, list);

		/* release lock before waiting on report processing */
		spin_unlock_irq(&zone->lock);

		/* begin processing pages in local list */
		err = prdev->report(prdev, sgl, PAGE_REPORTING_CAPACITY);

		/* reset offset since the full list was reported */
		*offset = PAGE_REPORTING_CAPACITY;

		/* update budget to reflect call to report function */
		budget--;

		/* reacquire zone lock and resume processing */
		spin_lock_irq(&zone->lock);

		/* flush reported pages from the sg list */
		page_reporting_drain(prdev, sgl, PAGE_REPORTING_CAPACITY, !err);

		/*
		 * Reset next to first entry, the old next isn't valid
		 * since we dropped the lock to report the pages
		 */
		next = list_first_entry(list, struct page, lru);

		/* exit on error */
		if (err)
			break;
	}

	/* Rotate any leftover pages to the head of the freelist */
	if (&next->lru != list && !list_is_first(&next->lru, list))
		list_rotate_to_front(&next->lru, list);

	spin_unlock_irq(&zone->lock);

	return err;
}

static int
page_reporting_process_zone(struct page_reporting_dev_info *prdev,
			    struct scatterlist *sgl, struct zone *zone)
{
	unsigned int order, mt, leftover, offset = PAGE_REPORTING_CAPACITY;
	unsigned long watermark;
	int err = 0;

	/* Generate minimum watermark to be able to guarantee progress */
	watermark = low_wmark_pages(zone) +
		    (PAGE_REPORTING_CAPACITY << PAGE_REPORTING_MIN_ORDER);

	/*
	 * Cancel request if insufficient free memory or if we failed
	 * to allocate page reporting statistics for the zone.
	 */
	if (!zone_watermark_ok(zone, 0, watermark, 0, ALLOC_CMA))
		return err;

	/* Process each free list starting from lowest order/mt */
	for (order = PAGE_REPORTING_MIN_ORDER; order < MAX_ORDER; order++) {
		for (mt = 0; mt < MIGRATE_TYPES; mt++) {
			/* We do not pull pages from the isolate free list */
			if (is_migrate_isolate(mt))
				continue;

			err = page_reporting_cycle(prdev, zone, order, mt,
						   sgl, &offset);
			if (err)
				return err;
		}
	}

	/* report the leftover pages before going idle */
	leftover = PAGE_REPORTING_CAPACITY - offset;
	if (leftover) {
		sgl = &sgl[offset];
		err = prdev->report(prdev, sgl, leftover);

		/* flush any remaining pages out from the last report */
		spin_lock_irq(&zone->lock);
		page_reporting_drain(prdev, sgl, leftover, !err);
		spin_unlock_irq(&zone->lock);
	}

	return err;
}

static void page_reporting_process(struct work_struct *work)
{
	struct delayed_work *d_work = to_delayed_work(work);
	struct page_reporting_dev_info *prdev =
		container_of(d_work, struct page_reporting_dev_info, work);
	int err = 0, state = PAGE_REPORTING_ACTIVE;
	struct scatterlist *sgl;
	struct zone *zone;

	/*
	 * Change the state to "Active" so that we can track if there is
	 * anyone requests page reporting after we complete our pass. If
	 * the state is not altered by the end of the pass we will switch
	 * to idle and quit scheduling reporting runs.
	 */
	atomic_set(&prdev->state, state);

	/* allocate scatterlist to store pages being reported on */
	sgl = kmalloc_array(PAGE_REPORTING_CAPACITY, sizeof(*sgl), GFP_KERNEL);
	if (!sgl)
		goto err_out;

	sg_init_table(sgl, PAGE_REPORTING_CAPACITY);

	for_each_zone(zone) {
		err = page_reporting_process_zone(prdev, sgl, zone);
		if (err)
			break;
	}

	kfree(sgl);
err_out:
	/*
	 * If the state has reverted back to requested then there may be
	 * additional pages to be processed. We will defer for 2s to allow
	 * more pages to accumulate.
	 */
	state = atomic_cmpxchg(&prdev->state, state, PAGE_REPORTING_IDLE);
	if (state == PAGE_REPORTING_REQUESTED)
		schedule_delayed_work(&prdev->work, PAGE_REPORTING_DELAY);
}

static DEFINE_MUTEX(page_reporting_mutex);
DEFINE_STATIC_KEY_FALSE(page_reporting_enabled);

int page_reporting_register(struct page_reporting_dev_info *prdev)
{
	int err = 0;

	mutex_lock(&page_reporting_mutex);

	/* nothing to do if already in use */
	if (rcu_access_pointer(pr_dev_info)) {
		err = -EBUSY;
		goto err_out;
	}

	/* initialize state and work structures */
	atomic_set(&prdev->state, PAGE_REPORTING_IDLE);
	INIT_DELAYED_WORK(&prdev->work, &page_reporting_process);

	/* Begin initial flush of zones */
	__page_reporting_request(prdev);

	/* Assign device to allow notifications */
	rcu_assign_pointer(pr_dev_info, prdev);

	/* enable page reporting notification */
	if (!static_key_enabled(&page_reporting_enabled)) {
		static_branch_enable(&page_reporting_enabled);
		pr_info("Free page reporting enabled\n");
	}
err_out:
	mutex_unlock(&page_reporting_mutex);

	return err;
}
EXPORT_SYMBOL_GPL(page_reporting_register);

void page_reporting_unregister(struct page_reporting_dev_info *prdev)
{
	mutex_lock(&page_reporting_mutex);

	if (rcu_access_pointer(pr_dev_info) == prdev) {
		/* Disable page reporting notification */
		RCU_INIT_POINTER(pr_dev_info, NULL);
		synchronize_rcu();

		/* Flush any existing work, and lock it out */
		cancel_delayed_work_sync(&prdev->work);
	}

	mutex_unlock(&page_reporting_mutex);
}
EXPORT_SYMBOL_GPL(page_reporting_unregister);
Commit	Line	Data
36e66c55 AD	1	// SPDX-License-Identifier: GPL-2.0
	2	#include <linux/mm.h>
	3	#include <linux/mmzone.h>
	4	#include <linux/page_reporting.h>
	5	#include <linux/gfp.h>
	6	#include <linux/export.h>
	7	#include <linux/delay.h>
	8	#include <linux/scatterlist.h>
	9
	10	#include "page_reporting.h"
	11	#include "internal.h"
	12
	13	#define PAGE_REPORTING_DELAY (2 * HZ)
	14	static struct page_reporting_dev_info __rcu *pr_dev_info __read_mostly;
	15
	16	enum {
	17	PAGE_REPORTING_IDLE = 0,
	18	PAGE_REPORTING_REQUESTED,
	19	PAGE_REPORTING_ACTIVE
	20	};
	21
	22	/* request page reporting */
	23	static void
	24	__page_reporting_request(struct page_reporting_dev_info *prdev)
	25	{
	26	unsigned int state;
	27
	28	/* Check to see if we are in desired state */
	29	state = atomic_read(&prdev->state);
	30	if (state == PAGE_REPORTING_REQUESTED)
	31	return;
	32
	33	/*
	34	* If reporting is already active there is nothing we need to do.
	35	* Test against 0 as that represents PAGE_REPORTING_IDLE.
	36	*/
	37	state = atomic_xchg(&prdev->state, PAGE_REPORTING_REQUESTED);
	38	if (state != PAGE_REPORTING_IDLE)
	39	return;
	40
	41	/*
	42	* Delay the start of work to allow a sizable queue to build. For
	43	* now we are limiting this to running no more than once every
	44	* couple of seconds.
	45	*/
	46	schedule_delayed_work(&prdev->work, PAGE_REPORTING_DELAY);
	47	}
	48
	49	/* notify prdev of free page reporting request */
	50	void __page_reporting_notify(void)
	51	{
	52	struct page_reporting_dev_info *prdev;
	53
	54	/*
	55	* We use RCU to protect the pr_dev_info pointer. In almost all
	56	* cases this should be present, however in the unlikely case of
	57	* a shutdown this will be NULL and we should exit.
	58	*/
	59	rcu_read_lock();
	60	prdev = rcu_dereference(pr_dev_info);
	61	if (likely(prdev))
	62	__page_reporting_request(prdev);
	63
	64	rcu_read_unlock();
65	}
66
67	static void
68	page_reporting_drain(struct page_reporting_dev_info *prdev,
69	struct scatterlist *sgl, unsigned int nents, bool reported)
70	{
71	struct scatterlist *sg = sgl;
72
73	/*
74	* Drain the now reported pages back into their respective
75	* free lists/areas. We assume at least one page is populated.
76	*/
77	do {
78	struct page *page = sg_page(sg);
79	int mt = get_pageblock_migratetype(page);
80	unsigned int order = get_order(sg->length);
81
82	__putback_isolated_page(page, order, mt);
83
84	/* If the pages were not reported due to error skip flagging */
85	if (!reported)
86	continue;
87
88	/*
89	* If page was not comingled with another page we can
90	* consider the result to be "reported" since the page
91	* hasn't been modified, otherwise we will need to
92	* report on the new larger page when we make our way
93	* up to that higher order.
94	*/
ab130f91	95	if (PageBuddy(page) && buddy_order(page) == order)
36e66c55 AD	96	__SetPageReported(page);
	97	} while ((sg = sg_next(sg)));
	98
	99	/* reinitialize scatterlist now that it is empty */
	100	sg_init_table(sgl, nents);
	101	}
	102
	103	/*
	104	* The page reporting cycle consists of 4 stages, fill, report, drain, and
	105	* idle. We will cycle through the first 3 stages until we cannot obtain a
	106	* full scatterlist of pages, in that case we will switch to idle.
	107	*/
	108	static int
	109	page_reporting_cycle(struct page_reporting_dev_info prdev, struct zone zone,
	110	unsigned int order, unsigned int mt,
	111	struct scatterlist sgl, unsigned int offset)
	112	{
	113	struct free_area *area = &zone->free_area[order];
	114	struct list_head *list = &area->free_list[mt];
	115	unsigned int page_len = PAGE_SIZE << order;
	116	struct page page, next;
43b76f29	117	long budget;
36e66c55 AD	118	int err = 0;
	119
	120	/*
	121	* Perform early check, if free area is empty there is
	122	* nothing to process so we can skip this free_list.
	123	*/
	124	if (list_empty(list))
	125	return err;
	126
	127	spin_lock_irq(&zone->lock);
	128
43b76f29 AD	129	/*
	130	* Limit how many calls we will be making to the page reporting
	131	* device for this list. By doing this we avoid processing any
	132	* given list for too long.
	133	*
	134	* The current value used allows us enough calls to process over a
	135	* sixteenth of the current list plus one additional call to handle
	136	* any pages that may have already been present from the previous
	137	* list processed. This should result in us reporting all pages on
	138	* an idle system in about 30 seconds.
	139	*
	140	* The division here should be cheap since PAGE_REPORTING_CAPACITY
	141	* should always be a power of 2.
	142	*/
	143	budget = DIV_ROUND_UP(area->nr_free, PAGE_REPORTING_CAPACITY * 16);
	144
36e66c55 AD	145	/* loop through free list adding unreported pages to sg list */
	146	list_for_each_entry_safe(page, next, list, lru) {
	147	/* We are going to skip over the reported pages. */
	148	if (PageReported(page))
	149	continue;
	150
43b76f29 AD	151	/*
	152	* If we fully consumed our budget then update our
	153	* state to indicate that we are requesting additional
	154	* processing and exit this list.
	155	*/
	156	if (budget < 0) {
	157	atomic_set(&prdev->state, PAGE_REPORTING_REQUESTED);
	158	next = page;
	159	break;
	160	}
	161
02cf8719 AD	162	/* Attempt to pull page from list and place in scatterlist */
	163	if (*offset) {
	164	if (!__isolate_free_page(page, order)) {
	165	next = page;
	166	break;
	167	}
36e66c55	168
02cf8719 AD	169	/* Add page to scatter list */
	170	--(*offset);
	171	sg_set_page(&sgl[*offset], page, page_len, 0);
36e66c55	172
36e66c55	173	continue;
02cf8719 AD	174	}
	175
	176	/*
43b76f29	177	* Make the first non-reported page in the free list
02cf8719 AD	178	* the new head of the free list before we release the
	179	* zone lock.
	180	*/
58f6f034	181	if (!list_is_first(&page->lru, list))
02cf8719	182	list_rotate_to_front(&page->lru, list);
36e66c55 AD	183
	184	/* release lock before waiting on report processing */
	185	spin_unlock_irq(&zone->lock);
	186
	187	/* begin processing pages in local list */
	188	err = prdev->report(prdev, sgl, PAGE_REPORTING_CAPACITY);
	189
	190	/* reset offset since the full list was reported */
	191	*offset = PAGE_REPORTING_CAPACITY;
	192
43b76f29 AD	193	/* update budget to reflect call to report function */
	194	budget--;
	195
36e66c55 AD	196	/* reacquire zone lock and resume processing */
	197	spin_lock_irq(&zone->lock);
	198
	199	/* flush reported pages from the sg list */
	200	page_reporting_drain(prdev, sgl, PAGE_REPORTING_CAPACITY, !err);
	201
	202	/*
	203	* Reset next to first entry, the old next isn't valid
	204	* since we dropped the lock to report the pages
	205	*/
	206	next = list_first_entry(list, struct page, lru);
	207
	208	/* exit on error */
	209	if (err)
	210	break;
	211	}
	212
02cf8719 AD	213	/* Rotate any leftover pages to the head of the freelist */
	214	if (&next->lru != list && !list_is_first(&next->lru, list))
	215	list_rotate_to_front(&next->lru, list);
	216
36e66c55 AD	217	spin_unlock_irq(&zone->lock);
	218
	219	return err;
	220	}
	221
	222	static int
	223	page_reporting_process_zone(struct page_reporting_dev_info *prdev,
	224	struct scatterlist sgl, struct zone zone)
	225	{
	226	unsigned int order, mt, leftover, offset = PAGE_REPORTING_CAPACITY;
	227	unsigned long watermark;
	228	int err = 0;
	229
	230	/* Generate minimum watermark to be able to guarantee progress */
	231	watermark = low_wmark_pages(zone) +
	232	(PAGE_REPORTING_CAPACITY << PAGE_REPORTING_MIN_ORDER);
	233
	234	/*
	235	* Cancel request if insufficient free memory or if we failed
	236	* to allocate page reporting statistics for the zone.
	237	*/
	238	if (!zone_watermark_ok(zone, 0, watermark, 0, ALLOC_CMA))
	239	return err;
	240
	241	/* Process each free list starting from lowest order/mt */
	242	for (order = PAGE_REPORTING_MIN_ORDER; order < MAX_ORDER; order++) {
	243	for (mt = 0; mt < MIGRATE_TYPES; mt++) {
	244	/* We do not pull pages from the isolate free list */
	245	if (is_migrate_isolate(mt))
	246	continue;
	247
	248	err = page_reporting_cycle(prdev, zone, order, mt,
	249	sgl, &offset);
	250	if (err)
	251	return err;
	252	}
	253	}
	254
	255	/* report the leftover pages before going idle */
	256	leftover = PAGE_REPORTING_CAPACITY - offset;
	257	if (leftover) {
	258	sgl = &sgl[offset];
	259	err = prdev->report(prdev, sgl, leftover);
	260
	261	/* flush any remaining pages out from the last report */
	262	spin_lock_irq(&zone->lock);
	263	page_reporting_drain(prdev, sgl, leftover, !err);
	264	spin_unlock_irq(&zone->lock);
	265	}
	266
	267	return err;
	268	}
	269
	270	static void page_reporting_process(struct work_struct *work)
	271	{
	272	struct delayed_work *d_work = to_delayed_work(work);
	273	struct page_reporting_dev_info *prdev =
	274	container_of(d_work, struct page_reporting_dev_info, work);
	275	int err = 0, state = PAGE_REPORTING_ACTIVE;
	276	struct scatterlist *sgl;
	277	struct zone *zone;
	278
	279	/*
	280	* Change the state to "Active" so that we can track if there is
281	* anyone requests page reporting after we complete our pass. If
282	* the state is not altered by the end of the pass we will switch
283	* to idle and quit scheduling reporting runs.
284	*/
285	atomic_set(&prdev->state, state);
286
287	/* allocate scatterlist to store pages being reported on */
288	sgl = kmalloc_array(PAGE_REPORTING_CAPACITY, sizeof(*sgl), GFP_KERNEL);
289	if (!sgl)
290	goto err_out;
291
292	sg_init_table(sgl, PAGE_REPORTING_CAPACITY);
293
294	for_each_zone(zone) {
295	err = page_reporting_process_zone(prdev, sgl, zone);
296	if (err)
297	break;
298	}
299
300	kfree(sgl);
301	err_out:
302	/*
303	* If the state has reverted back to requested then there may be
304	* additional pages to be processed. We will defer for 2s to allow
305	* more pages to accumulate.
306	*/
307	state = atomic_cmpxchg(&prdev->state, state, PAGE_REPORTING_IDLE);
308	if (state == PAGE_REPORTING_REQUESTED)
309	schedule_delayed_work(&prdev->work, PAGE_REPORTING_DELAY);
310	}
311
312	static DEFINE_MUTEX(page_reporting_mutex);
313	DEFINE_STATIC_KEY_FALSE(page_reporting_enabled);
314
315	int page_reporting_register(struct page_reporting_dev_info *prdev)
316	{
317	int err = 0;
318
319	mutex_lock(&page_reporting_mutex);
320
321	/* nothing to do if already in use */
322	if (rcu_access_pointer(pr_dev_info)) {
323	err = -EBUSY;
324	goto err_out;
325	}
326
327	/* initialize state and work structures */
328	atomic_set(&prdev->state, PAGE_REPORTING_IDLE);
329	INIT_DELAYED_WORK(&prdev->work, &page_reporting_process);
330
331	/* Begin initial flush of zones */
332	__page_reporting_request(prdev);
333
334	/* Assign device to allow notifications */
335	rcu_assign_pointer(pr_dev_info, prdev);
336
337	/* enable page reporting notification */
338	if (!static_key_enabled(&page_reporting_enabled)) {
339	static_branch_enable(&page_reporting_enabled);
340	pr_info("Free page reporting enabled\n");
341	}
342	err_out:
343	mutex_unlock(&page_reporting_mutex);
344
345	return err;
346	}
347	EXPORT_SYMBOL_GPL(page_reporting_register);
348
349	void page_reporting_unregister(struct page_reporting_dev_info *prdev)
350	{
351	mutex_lock(&page_reporting_mutex);
352
353	if (rcu_access_pointer(pr_dev_info) == prdev) {
354	/* Disable page reporting notification */
355	RCU_INIT_POINTER(pr_dev_info, NULL);
356	synchronize_rcu();
357
358	/* Flush any existing work, and lock it out */
359	cancel_delayed_work_sync(&prdev->work);
360	}
361
362	mutex_unlock(&page_reporting_mutex);
363	}
364	EXPORT_SYMBOL_GPL(page_reporting_unregister);