[linux.git] / mm / highmem.c

/*
 * High memory handling common code and variables.
 *
 * (C) 1999 Andrea Arcangeli, SuSE GmbH, [email protected]
 *          Gerhard Wichert, Siemens AG, [email protected]
 *
 *
 * Redesigned the x86 32-bit VM architecture to deal with
 * 64-bit physical space. With current x86 CPUs this
 * means up to 64 Gigabytes physical RAM.
 *
 * Rewrote high memory support to move the page cache into
 * high memory. Implemented permanent (schedulable) kmaps
 * based on Linus' idea.
 *
 * Copyright (C) 1999 Ingo Molnar <[email protected]>
 */

#include <linux/mm.h>
#include <linux/module.h>
#include <linux/swap.h>
#include <linux/bio.h>
#include <linux/pagemap.h>
#include <linux/mempool.h>
#include <linux/blkdev.h>
#include <linux/init.h>
#include <linux/hash.h>
#include <linux/highmem.h>
#include <linux/blktrace_api.h>
#include <asm/tlbflush.h>

/*
 * Virtual_count is not a pure "count".
 *  0 means that it is not mapped, and has not been mapped
 *    since a TLB flush - it is usable.
 *  1 means that there are no users, but it has been mapped
 *    since the last TLB flush - so we can't use it.
 *  n means that there are (n-1) current users of it.
 */
#ifdef CONFIG_HIGHMEM

unsigned long totalhigh_pages __read_mostly;
EXPORT_SYMBOL(totalhigh_pages);

unsigned int nr_free_highpages (void)
{
	pg_data_t *pgdat;
	unsigned int pages = 0;

	for_each_online_pgdat(pgdat) {
		pages += zone_page_state(&pgdat->node_zones[ZONE_HIGHMEM],
			NR_FREE_PAGES);
		if (zone_movable_is_highmem())
			pages += zone_page_state(
					&pgdat->node_zones[ZONE_MOVABLE],
					NR_FREE_PAGES);
	}

	return pages;
}

static int pkmap_count[LAST_PKMAP];
static unsigned int last_pkmap_nr;
static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(kmap_lock);

pte_t * pkmap_page_table;

static DECLARE_WAIT_QUEUE_HEAD(pkmap_map_wait);

static void flush_all_zero_pkmaps(void)
{
	int i;
	int need_flush = 0;

	flush_cache_kmaps();

	for (i = 0; i < LAST_PKMAP; i++) {
		struct page *page;

		/*
		 * zero means we don't have anything to do,
		 * >1 means that it is still in use. Only
		 * a count of 1 means that it is free but
		 * needs to be unmapped
		 */
		if (pkmap_count[i] != 1)
			continue;
		pkmap_count[i] = 0;

		/* sanity check */
		BUG_ON(pte_none(pkmap_page_table[i]));

		/*
		 * Don't need an atomic fetch-and-clear op here;
		 * no-one has the page mapped, and cannot get at
		 * its virtual address (and hence PTE) without first
		 * getting the kmap_lock (which is held here).
		 * So no dangers, even with speculative execution.
		 */
		page = pte_page(pkmap_page_table[i]);
		pte_clear(&init_mm, (unsigned long)page_address(page),
			  &pkmap_page_table[i]);

		set_page_address(page, NULL);
		need_flush = 1;
	}
	if (need_flush)
		flush_tlb_kernel_range(PKMAP_ADDR(0), PKMAP_ADDR(LAST_PKMAP));
}

/**
 * kmap_flush_unused - flush all unused kmap mappings in order to remove stray mappings
 */
void kmap_flush_unused(void)
{
	spin_lock(&kmap_lock);
	flush_all_zero_pkmaps();
	spin_unlock(&kmap_lock);
}

static inline unsigned long map_new_virtual(struct page *page)
{
	unsigned long vaddr;
	int count;

start:
	count = LAST_PKMAP;
	/* Find an empty entry */
	for (;;) {
		last_pkmap_nr = (last_pkmap_nr + 1) & LAST_PKMAP_MASK;
		if (!last_pkmap_nr) {
			flush_all_zero_pkmaps();
			count = LAST_PKMAP;
		}
		if (!pkmap_count[last_pkmap_nr])
			break;	/* Found a usable entry */
		if (--count)
			continue;

		/*
		 * Sleep for somebody else to unmap their entries
		 */
		{
			DECLARE_WAITQUEUE(wait, current);

			__set_current_state(TASK_UNINTERRUPTIBLE);
			add_wait_queue(&pkmap_map_wait, &wait);
			spin_unlock(&kmap_lock);
			schedule();
			remove_wait_queue(&pkmap_map_wait, &wait);
			spin_lock(&kmap_lock);

			/* Somebody else might have mapped it while we slept */
			if (page_address(page))
				return (unsigned long)page_address(page);

			/* Re-start */
			goto start;
		}
	}
	vaddr = PKMAP_ADDR(last_pkmap_nr);
	set_pte_at(&init_mm, vaddr,
		   &(pkmap_page_table[last_pkmap_nr]), mk_pte(page, kmap_prot));

	pkmap_count[last_pkmap_nr] = 1;
	set_page_address(page, (void *)vaddr);

	return vaddr;
}

/**
 * kmap_high - map a highmem page into memory
 * @page: &struct page to map
 *
 * Returns the page's virtual memory address.
 *
 * We cannot call this from interrupts, as it may block.
 */
void *kmap_high(struct page *page)
{
	unsigned long vaddr;

	/*
	 * For highmem pages, we can't trust "virtual" until
	 * after we have the lock.
	 */
	spin_lock(&kmap_lock);
	vaddr = (unsigned long)page_address(page);
	if (!vaddr)
		vaddr = map_new_virtual(page);
	pkmap_count[PKMAP_NR(vaddr)]++;
	BUG_ON(pkmap_count[PKMAP_NR(vaddr)] < 2);
	spin_unlock(&kmap_lock);
	return (void*) vaddr;
}

EXPORT_SYMBOL(kmap_high);

/**
 * kunmap_high - map a highmem page into memory
 * @page: &struct page to unmap
 */
void kunmap_high(struct page *page)
{
	unsigned long vaddr;
	unsigned long nr;
	int need_wakeup;

	spin_lock(&kmap_lock);
	vaddr = (unsigned long)page_address(page);
	BUG_ON(!vaddr);
	nr = PKMAP_NR(vaddr);

	/*
	 * A count must never go down to zero
	 * without a TLB flush!
	 */
	need_wakeup = 0;
	switch (--pkmap_count[nr]) {
	case 0:
		BUG();
	case 1:
		/*
		 * Avoid an unnecessary wake_up() function call.
		 * The common case is pkmap_count[] == 1, but
		 * no waiters.
		 * The tasks queued in the wait-queue are guarded
		 * by both the lock in the wait-queue-head and by
		 * the kmap_lock.  As the kmap_lock is held here,
		 * no need for the wait-queue-head's lock.  Simply
		 * test if the queue is empty.
		 */
		need_wakeup = waitqueue_active(&pkmap_map_wait);
	}
	spin_unlock(&kmap_lock);

	/* do wake-up, if needed, race-free outside of the spin lock */
	if (need_wakeup)
		wake_up(&pkmap_map_wait);
}

EXPORT_SYMBOL(kunmap_high);
#endif

#if defined(HASHED_PAGE_VIRTUAL)

#define PA_HASH_ORDER	7

/*
 * Describes one page->virtual association
 */
struct page_address_map {
	struct page *page;
	void *virtual;
	struct list_head list;
};

/*
 * page_address_map freelist, allocated from page_address_maps.
 */
static struct list_head page_address_pool;	/* freelist */
static spinlock_t pool_lock;			/* protects page_address_pool */

/*
 * Hash table bucket
 */
static struct page_address_slot {
	struct list_head lh;			/* List of page_address_maps */
	spinlock_t lock;			/* Protect this bucket's list */
} ____cacheline_aligned_in_smp page_address_htable[1<<PA_HASH_ORDER];

static struct page_address_slot *page_slot(struct page *page)
{
	return &page_address_htable[hash_ptr(page, PA_HASH_ORDER)];
}

/**
 * page_address - get the mapped virtual address of a page
 * @page: &struct page to get the virtual address of
 *
 * Returns the page's virtual address.
 */
void *page_address(struct page *page)
{
	unsigned long flags;
	void *ret;
	struct page_address_slot *pas;

	if (!PageHighMem(page))
		return lowmem_page_address(page);

	pas = page_slot(page);
	ret = NULL;
	spin_lock_irqsave(&pas->lock, flags);
	if (!list_empty(&pas->lh)) {
		struct page_address_map *pam;

		list_for_each_entry(pam, &pas->lh, list) {
			if (pam->page == page) {
				ret = pam->virtual;
				goto done;
			}
		}
	}
done:
	spin_unlock_irqrestore(&pas->lock, flags);
	return ret;
}

EXPORT_SYMBOL(page_address);

/**
 * set_page_address - set a page's virtual address
 * @page: &struct page to set
 * @virtual: virtual address to use
 */
void set_page_address(struct page *page, void *virtual)
{
	unsigned long flags;
	struct page_address_slot *pas;
	struct page_address_map *pam;

	BUG_ON(!PageHighMem(page));

	pas = page_slot(page);
	if (virtual) {		/* Add */
		BUG_ON(list_empty(&page_address_pool));

		spin_lock_irqsave(&pool_lock, flags);
		pam = list_entry(page_address_pool.next,
				struct page_address_map, list);
		list_del(&pam->list);
		spin_unlock_irqrestore(&pool_lock, flags);

		pam->page = page;
		pam->virtual = virtual;

		spin_lock_irqsave(&pas->lock, flags);
		list_add_tail(&pam->list, &pas->lh);
		spin_unlock_irqrestore(&pas->lock, flags);
	} else {		/* Remove */
		spin_lock_irqsave(&pas->lock, flags);
		list_for_each_entry(pam, &pas->lh, list) {
			if (pam->page == page) {
				list_del(&pam->list);
				spin_unlock_irqrestore(&pas->lock, flags);
				spin_lock_irqsave(&pool_lock, flags);
				list_add_tail(&pam->list, &page_address_pool);
				spin_unlock_irqrestore(&pool_lock, flags);
				goto done;
			}
		}
		spin_unlock_irqrestore(&pas->lock, flags);
	}
done:
	return;
}

static struct page_address_map page_address_maps[LAST_PKMAP];

void __init page_address_init(void)
{
	int i;

	INIT_LIST_HEAD(&page_address_pool);
	for (i = 0; i < ARRAY_SIZE(page_address_maps); i++)
		list_add(&page_address_maps[i].list, &page_address_pool);
	for (i = 0; i < ARRAY_SIZE(page_address_htable); i++) {
		INIT_LIST_HEAD(&page_address_htable[i].lh);
		spin_lock_init(&page_address_htable[i].lock);
	}
	spin_lock_init(&pool_lock);
}

#endif	/* defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) */
Commit	Line	Data
1da177e4 LT	1	/*
	2	* High memory handling common code and variables.
	3	*
	4	* (C) 1999 Andrea Arcangeli, SuSE GmbH, [email protected]
	5	* Gerhard Wichert, Siemens AG, [email protected]
	6	*
	7	*
	8	* Redesigned the x86 32-bit VM architecture to deal with
	9	* 64-bit physical space. With current x86 CPUs this
	10	* means up to 64 Gigabytes physical RAM.
	11	*
	12	* Rewrote high memory support to move the page cache into
	13	* high memory. Implemented permanent (schedulable) kmaps
	14	* based on Linus' idea.
	15	*
	16	* Copyright (C) 1999 Ingo Molnar <[email protected]>
	17	*/
	18
	19	#include <linux/mm.h>
	20	#include <linux/module.h>
	21	#include <linux/swap.h>
	22	#include <linux/bio.h>
	23	#include <linux/pagemap.h>
	24	#include <linux/mempool.h>
	25	#include <linux/blkdev.h>
	26	#include <linux/init.h>
	27	#include <linux/hash.h>
	28	#include <linux/highmem.h>
2056a782	29	#include <linux/blktrace_api.h>
1da177e4 LT	30	#include <asm/tlbflush.h>
1da177e4 LT	31
1da177e4 LT	32	/*
	33	* Virtual_count is not a pure "count".
	34	* 0 means that it is not mapped, and has not been mapped
	35	* since a TLB flush - it is usable.
	36	* 1 means that there are no users, but it has been mapped
	37	* since the last TLB flush - so we can't use it.
	38	* n means that there are (n-1) current users of it.
	39	*/
	40	#ifdef CONFIG_HIGHMEM
260b2367	41
c1f60a5a	42	unsigned long totalhigh_pages __read_mostly;
db7a94d6	43	EXPORT_SYMBOL(totalhigh_pages);
c1f60a5a CL	44
	45	unsigned int nr_free_highpages (void)
	46	{
	47	pg_data_t *pgdat;
	48	unsigned int pages = 0;
	49
2a1e274a	50	for_each_online_pgdat(pgdat) {
d23ad423 CL	51	pages += zone_page_state(&pgdat->node_zones[ZONE_HIGHMEM],
d23ad423 CL	52	NR_FREE_PAGES);
2a1e274a MG	53	if (zone_movable_is_highmem())
	54	pages += zone_page_state(
	55	&pgdat->node_zones[ZONE_MOVABLE],
	56	NR_FREE_PAGES);
	57	}
c1f60a5a CL	58
	59	return pages;
	60	}
	61
1da177e4 LT	62	static int pkmap_count[LAST_PKMAP];
	63	static unsigned int last_pkmap_nr;
	64	static __cacheline_aligned_in_smp DEFINE_SPINLOCK(kmap_lock);
	65
	66	pte_t * pkmap_page_table;
	67
	68	static DECLARE_WAIT_QUEUE_HEAD(pkmap_map_wait);
	69
	70	static void flush_all_zero_pkmaps(void)
	71	{
	72	int i;
5843d9a4	73	int need_flush = 0;
1da177e4 LT	74
	75	flush_cache_kmaps();
	76
	77	for (i = 0; i < LAST_PKMAP; i++) {
	78	struct page *page;
	79
	80	/*
	81	* zero means we don't have anything to do,
	82	* >1 means that it is still in use. Only
	83	* a count of 1 means that it is free but
	84	* needs to be unmapped
	85	*/
	86	if (pkmap_count[i] != 1)
	87	continue;
	88	pkmap_count[i] = 0;
	89
	90	/* sanity check */
75babcac	91	BUG_ON(pte_none(pkmap_page_table[i]));
1da177e4 LT	92
	93	/*
	94	* Don't need an atomic fetch-and-clear op here;
	95	* no-one has the page mapped, and cannot get at
	96	* its virtual address (and hence PTE) without first
	97	* getting the kmap_lock (which is held here).
	98	* So no dangers, even with speculative execution.
	99	*/
	100	page = pte_page(pkmap_page_table[i]);
	101	pte_clear(&init_mm, (unsigned long)page_address(page),
	102	&pkmap_page_table[i]);
	103
	104	set_page_address(page, NULL);
5843d9a4	105	need_flush = 1;
1da177e4	106	}
5843d9a4 NP	107	if (need_flush)
5843d9a4 NP	108	flush_tlb_kernel_range(PKMAP_ADDR(0), PKMAP_ADDR(LAST_PKMAP));
1da177e4 LT	109	}
1da177e4 LT	110
77f6078a RD	111	/**
	112	* kmap_flush_unused - flush all unused kmap mappings in order to remove stray mappings
	113	*/
ce6234b5 JF	114	void kmap_flush_unused(void)
	115	{
	116	spin_lock(&kmap_lock);
	117	flush_all_zero_pkmaps();
	118	spin_unlock(&kmap_lock);
	119	}
	120
1da177e4 LT	121	static inline unsigned long map_new_virtual(struct page *page)
	122	{
	123	unsigned long vaddr;
	124	int count;
	125
	126	start:
	127	count = LAST_PKMAP;
	128	/* Find an empty entry */
	129	for (;;) {
	130	last_pkmap_nr = (last_pkmap_nr + 1) & LAST_PKMAP_MASK;
	131	if (!last_pkmap_nr) {
	132	flush_all_zero_pkmaps();
	133	count = LAST_PKMAP;
	134	}
	135	if (!pkmap_count[last_pkmap_nr])
	136	break; /* Found a usable entry */
	137	if (--count)
	138	continue;
	139
	140	/*
	141	* Sleep for somebody else to unmap their entries
	142	*/
	143	{
	144	DECLARE_WAITQUEUE(wait, current);
	145
	146	__set_current_state(TASK_UNINTERRUPTIBLE);
	147	add_wait_queue(&pkmap_map_wait, &wait);
	148	spin_unlock(&kmap_lock);
	149	schedule();
	150	remove_wait_queue(&pkmap_map_wait, &wait);
	151	spin_lock(&kmap_lock);
	152
	153	/* Somebody else might have mapped it while we slept */
	154	if (page_address(page))
	155	return (unsigned long)page_address(page);
	156
	157	/* Re-start */
	158	goto start;
	159	}
	160	}
	161	vaddr = PKMAP_ADDR(last_pkmap_nr);
	162	set_pte_at(&init_mm, vaddr,
	163	&(pkmap_page_table[last_pkmap_nr]), mk_pte(page, kmap_prot));
	164
	165	pkmap_count[last_pkmap_nr] = 1;
	166	set_page_address(page, (void *)vaddr);
	167
	168	return vaddr;
	169	}
	170
77f6078a RD	171	/**
	172	* kmap_high - map a highmem page into memory
	173	* @page: &struct page to map
	174	*
	175	* Returns the page's virtual memory address.
	176	*
	177	* We cannot call this from interrupts, as it may block.
	178	*/
920c7a5d	179	void kmap_high(struct page page)
1da177e4 LT	180	{
	181	unsigned long vaddr;
	182
	183	/*
	184	* For highmem pages, we can't trust "virtual" until
	185	* after we have the lock.
1da177e4 LT	186	*/
	187	spin_lock(&kmap_lock);
	188	vaddr = (unsigned long)page_address(page);
	189	if (!vaddr)
	190	vaddr = map_new_virtual(page);
	191	pkmap_count[PKMAP_NR(vaddr)]++;
75babcac	192	BUG_ON(pkmap_count[PKMAP_NR(vaddr)] < 2);
1da177e4 LT	193	spin_unlock(&kmap_lock);
	194	return (void*) vaddr;
	195	}
	196
	197	EXPORT_SYMBOL(kmap_high);
	198
77f6078a RD	199	/**
	200	* kunmap_high - map a highmem page into memory
	201	* @page: &struct page to unmap
	202	*/
920c7a5d	203	void kunmap_high(struct page *page)
1da177e4 LT	204	{
	205	unsigned long vaddr;
	206	unsigned long nr;
	207	int need_wakeup;
	208
	209	spin_lock(&kmap_lock);
	210	vaddr = (unsigned long)page_address(page);
75babcac	211	BUG_ON(!vaddr);
1da177e4 LT	212	nr = PKMAP_NR(vaddr);
	213
	214	/*
	215	* A count must never go down to zero
	216	* without a TLB flush!
	217	*/
	218	need_wakeup = 0;
	219	switch (--pkmap_count[nr]) {
	220	case 0:
	221	BUG();
	222	case 1:
	223	/*
	224	* Avoid an unnecessary wake_up() function call.
	225	* The common case is pkmap_count[] == 1, but
	226	* no waiters.
	227	* The tasks queued in the wait-queue are guarded
	228	* by both the lock in the wait-queue-head and by
	229	* the kmap_lock. As the kmap_lock is held here,
	230	* no need for the wait-queue-head's lock. Simply
	231	* test if the queue is empty.
	232	*/
	233	need_wakeup = waitqueue_active(&pkmap_map_wait);
	234	}
	235	spin_unlock(&kmap_lock);
	236
	237	/* do wake-up, if needed, race-free outside of the spin lock */
	238	if (need_wakeup)
	239	wake_up(&pkmap_map_wait);
	240	}
	241
	242	EXPORT_SYMBOL(kunmap_high);
1da177e4 LT	243	#endif
1da177e4 LT	244
1da177e4 LT	245	#if defined(HASHED_PAGE_VIRTUAL)
	246
	247	#define PA_HASH_ORDER 7
	248
	249	/*
	250	* Describes one page->virtual association
	251	*/
	252	struct page_address_map {
	253	struct page *page;
	254	void *virtual;
	255	struct list_head list;
	256	};
	257
	258	/*
	259	* page_address_map freelist, allocated from page_address_maps.
	260	*/
	261	static struct list_head page_address_pool; /* freelist */
	262	static spinlock_t pool_lock; /* protects page_address_pool */
	263
	264	/*
	265	* Hash table bucket
	266	*/
	267	static struct page_address_slot {
	268	struct list_head lh; /* List of page_address_maps */
	269	spinlock_t lock; /* Protect this bucket's list */
	270	} ____cacheline_aligned_in_smp page_address_htable[1<<PA_HASH_ORDER];
	271
	272	static struct page_address_slot page_slot(struct page page)
	273	{
	274	return &page_address_htable[hash_ptr(page, PA_HASH_ORDER)];
	275	}
	276
77f6078a RD	277	/**
	278	* page_address - get the mapped virtual address of a page
	279	* @page: &struct page to get the virtual address of
	280	*
	281	* Returns the page's virtual address.
	282	*/
1da177e4 LT	283	void page_address(struct page page)
	284	{
	285	unsigned long flags;
	286	void *ret;
	287	struct page_address_slot *pas;
	288
	289	if (!PageHighMem(page))
	290	return lowmem_page_address(page);
	291
	292	pas = page_slot(page);
	293	ret = NULL;
	294	spin_lock_irqsave(&pas->lock, flags);
	295	if (!list_empty(&pas->lh)) {
	296	struct page_address_map *pam;
	297
	298	list_for_each_entry(pam, &pas->lh, list) {
	299	if (pam->page == page) {
	300	ret = pam->virtual;
	301	goto done;
	302	}
	303	}
	304	}
	305	done:
	306	spin_unlock_irqrestore(&pas->lock, flags);
	307	return ret;
	308	}
	309
	310	EXPORT_SYMBOL(page_address);
	311
77f6078a RD	312	/**
	313	* set_page_address - set a page's virtual address
	314	* @page: &struct page to set
	315	* @virtual: virtual address to use
	316	*/
1da177e4 LT	317	void set_page_address(struct page page, void virtual)
	318	{
	319	unsigned long flags;
	320	struct page_address_slot *pas;
	321	struct page_address_map *pam;
	322
	323	BUG_ON(!PageHighMem(page));
	324
	325	pas = page_slot(page);
	326	if (virtual) { /* Add */
	327	BUG_ON(list_empty(&page_address_pool));
	328
	329	spin_lock_irqsave(&pool_lock, flags);
	330	pam = list_entry(page_address_pool.next,
	331	struct page_address_map, list);
	332	list_del(&pam->list);
	333	spin_unlock_irqrestore(&pool_lock, flags);
	334
	335	pam->page = page;
	336	pam->virtual = virtual;
	337
	338	spin_lock_irqsave(&pas->lock, flags);
	339	list_add_tail(&pam->list, &pas->lh);
	340	spin_unlock_irqrestore(&pas->lock, flags);
	341	} else { /* Remove */
	342	spin_lock_irqsave(&pas->lock, flags);
	343	list_for_each_entry(pam, &pas->lh, list) {
	344	if (pam->page == page) {
	345	list_del(&pam->list);
	346	spin_unlock_irqrestore(&pas->lock, flags);
	347	spin_lock_irqsave(&pool_lock, flags);
	348	list_add_tail(&pam->list, &page_address_pool);
	349	spin_unlock_irqrestore(&pool_lock, flags);
	350	goto done;
	351	}
	352	}
	353	spin_unlock_irqrestore(&pas->lock, flags);
	354	}
	355	done:
	356	return;
	357	}
	358
	359	static struct page_address_map page_address_maps[LAST_PKMAP];
	360
	361	void __init page_address_init(void)
	362	{
	363	int i;
	364
	365	INIT_LIST_HEAD(&page_address_pool);
	366	for (i = 0; i < ARRAY_SIZE(page_address_maps); i++)
	367	list_add(&page_address_maps[i].list, &page_address_pool);
	368	for (i = 0; i < ARRAY_SIZE(page_address_htable); i++) {
	369	INIT_LIST_HEAD(&page_address_htable[i].lh);
	370	spin_lock_init(&page_address_htable[i].lock);
	371	}
	372	spin_lock_init(&pool_lock);
	373	}
	374
	375	#endif /* defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) */