[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;

	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);
	}
	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;
	73
	74	flush_cache_kmaps();
	75
	76	for (i = 0; i < LAST_PKMAP; i++) {
	77	struct page *page;
	78
	79	/*
	80	* zero means we don't have anything to do,
	81	* >1 means that it is still in use. Only
	82	* a count of 1 means that it is free but
	83	* needs to be unmapped
	84	*/
	85	if (pkmap_count[i] != 1)
	86	continue;
	87	pkmap_count[i] = 0;
	88
	89	/* sanity check */
75babcac	90	BUG_ON(pte_none(pkmap_page_table[i]));
1da177e4 LT	91
	92	/*
	93	* Don't need an atomic fetch-and-clear op here;
	94	* no-one has the page mapped, and cannot get at
	95	* its virtual address (and hence PTE) without first
	96	* getting the kmap_lock (which is held here).
	97	* So no dangers, even with speculative execution.
	98	*/
	99	page = pte_page(pkmap_page_table[i]);
	100	pte_clear(&init_mm, (unsigned long)page_address(page),
	101	&pkmap_page_table[i]);
	102
	103	set_page_address(page, NULL);
	104	}
	105	flush_tlb_kernel_range(PKMAP_ADDR(0), PKMAP_ADDR(LAST_PKMAP));
	106	}
	107
77f6078a RD	108	/**
	109	* kmap_flush_unused - flush all unused kmap mappings in order to remove stray mappings
	110	*/
ce6234b5 JF	111	void kmap_flush_unused(void)
	112	{
	113	spin_lock(&kmap_lock);
	114	flush_all_zero_pkmaps();
	115	spin_unlock(&kmap_lock);
	116	}
	117
1da177e4 LT	118	static inline unsigned long map_new_virtual(struct page *page)
	119	{
	120	unsigned long vaddr;
	121	int count;
	122
	123	start:
	124	count = LAST_PKMAP;
	125	/* Find an empty entry */
	126	for (;;) {
	127	last_pkmap_nr = (last_pkmap_nr + 1) & LAST_PKMAP_MASK;
	128	if (!last_pkmap_nr) {
	129	flush_all_zero_pkmaps();
	130	count = LAST_PKMAP;
	131	}
	132	if (!pkmap_count[last_pkmap_nr])
	133	break; /* Found a usable entry */
	134	if (--count)
	135	continue;
	136
	137	/*
	138	* Sleep for somebody else to unmap their entries
	139	*/
	140	{
	141	DECLARE_WAITQUEUE(wait, current);
	142
	143	__set_current_state(TASK_UNINTERRUPTIBLE);
	144	add_wait_queue(&pkmap_map_wait, &wait);
	145	spin_unlock(&kmap_lock);
	146	schedule();
	147	remove_wait_queue(&pkmap_map_wait, &wait);
	148	spin_lock(&kmap_lock);
	149
	150	/* Somebody else might have mapped it while we slept */
	151	if (page_address(page))
	152	return (unsigned long)page_address(page);
	153
	154	/* Re-start */
	155	goto start;
	156	}
	157	}
	158	vaddr = PKMAP_ADDR(last_pkmap_nr);
	159	set_pte_at(&init_mm, vaddr,
	160	&(pkmap_page_table[last_pkmap_nr]), mk_pte(page, kmap_prot));
	161
	162	pkmap_count[last_pkmap_nr] = 1;
	163	set_page_address(page, (void *)vaddr);
	164
	165	return vaddr;
	166	}
	167
77f6078a RD	168	/**
	169	* kmap_high - map a highmem page into memory
	170	* @page: &struct page to map
	171	*
	172	* Returns the page's virtual memory address.
	173	*
	174	* We cannot call this from interrupts, as it may block.
	175	*/
920c7a5d	176	void kmap_high(struct page page)
1da177e4 LT	177	{
	178	unsigned long vaddr;
	179
	180	/*
	181	* For highmem pages, we can't trust "virtual" until
	182	* after we have the lock.
1da177e4 LT	183	*/
	184	spin_lock(&kmap_lock);
	185	vaddr = (unsigned long)page_address(page);
	186	if (!vaddr)
	187	vaddr = map_new_virtual(page);
	188	pkmap_count[PKMAP_NR(vaddr)]++;
75babcac	189	BUG_ON(pkmap_count[PKMAP_NR(vaddr)] < 2);
1da177e4 LT	190	spin_unlock(&kmap_lock);
	191	return (void*) vaddr;
	192	}
	193
	194	EXPORT_SYMBOL(kmap_high);
	195
77f6078a RD	196	/**
	197	* kunmap_high - map a highmem page into memory
	198	* @page: &struct page to unmap
	199	*/
920c7a5d	200	void kunmap_high(struct page *page)
1da177e4 LT	201	{
	202	unsigned long vaddr;
	203	unsigned long nr;
	204	int need_wakeup;
	205
	206	spin_lock(&kmap_lock);
	207	vaddr = (unsigned long)page_address(page);
75babcac	208	BUG_ON(!vaddr);
1da177e4 LT	209	nr = PKMAP_NR(vaddr);
	210
	211	/*
	212	* A count must never go down to zero
	213	* without a TLB flush!
	214	*/
	215	need_wakeup = 0;
	216	switch (--pkmap_count[nr]) {
	217	case 0:
	218	BUG();
	219	case 1:
	220	/*
	221	* Avoid an unnecessary wake_up() function call.
	222	* The common case is pkmap_count[] == 1, but
	223	* no waiters.
	224	* The tasks queued in the wait-queue are guarded
	225	* by both the lock in the wait-queue-head and by
	226	* the kmap_lock. As the kmap_lock is held here,
	227	* no need for the wait-queue-head's lock. Simply
	228	* test if the queue is empty.
	229	*/
	230	need_wakeup = waitqueue_active(&pkmap_map_wait);
	231	}
	232	spin_unlock(&kmap_lock);
	233
	234	/* do wake-up, if needed, race-free outside of the spin lock */
	235	if (need_wakeup)
	236	wake_up(&pkmap_map_wait);
	237	}
	238
	239	EXPORT_SYMBOL(kunmap_high);
1da177e4 LT	240	#endif
1da177e4 LT	241
1da177e4 LT	242	#if defined(HASHED_PAGE_VIRTUAL)
	243
	244	#define PA_HASH_ORDER 7
	245
	246	/*
	247	* Describes one page->virtual association
	248	*/
	249	struct page_address_map {
	250	struct page *page;
	251	void *virtual;
	252	struct list_head list;
	253	};
	254
	255	/*
	256	* page_address_map freelist, allocated from page_address_maps.
	257	*/
	258	static struct list_head page_address_pool; /* freelist */
	259	static spinlock_t pool_lock; /* protects page_address_pool */
	260
	261	/*
	262	* Hash table bucket
	263	*/
	264	static struct page_address_slot {
	265	struct list_head lh; /* List of page_address_maps */
	266	spinlock_t lock; /* Protect this bucket's list */
	267	} ____cacheline_aligned_in_smp page_address_htable[1<<PA_HASH_ORDER];
	268
	269	static struct page_address_slot page_slot(struct page page)
	270	{
	271	return &page_address_htable[hash_ptr(page, PA_HASH_ORDER)];
	272	}
	273
77f6078a RD	274	/**
	275	* page_address - get the mapped virtual address of a page
	276	* @page: &struct page to get the virtual address of
	277	*
	278	* Returns the page's virtual address.
	279	*/
1da177e4 LT	280	void page_address(struct page page)
	281	{
	282	unsigned long flags;
	283	void *ret;
	284	struct page_address_slot *pas;
	285
	286	if (!PageHighMem(page))
	287	return lowmem_page_address(page);
	288
	289	pas = page_slot(page);
	290	ret = NULL;
	291	spin_lock_irqsave(&pas->lock, flags);
	292	if (!list_empty(&pas->lh)) {
	293	struct page_address_map *pam;
	294
	295	list_for_each_entry(pam, &pas->lh, list) {
	296	if (pam->page == page) {
	297	ret = pam->virtual;
	298	goto done;
	299	}
	300	}
	301	}
	302	done:
	303	spin_unlock_irqrestore(&pas->lock, flags);
	304	return ret;
	305	}
	306
	307	EXPORT_SYMBOL(page_address);
	308
77f6078a RD	309	/**
	310	* set_page_address - set a page's virtual address
	311	* @page: &struct page to set
	312	* @virtual: virtual address to use
	313	*/
1da177e4 LT	314	void set_page_address(struct page page, void virtual)
	315	{
	316	unsigned long flags;
	317	struct page_address_slot *pas;
	318	struct page_address_map *pam;
	319
	320	BUG_ON(!PageHighMem(page));
	321
	322	pas = page_slot(page);
	323	if (virtual) { /* Add */
	324	BUG_ON(list_empty(&page_address_pool));
	325
	326	spin_lock_irqsave(&pool_lock, flags);
	327	pam = list_entry(page_address_pool.next,
	328	struct page_address_map, list);
	329	list_del(&pam->list);
	330	spin_unlock_irqrestore(&pool_lock, flags);
	331
	332	pam->page = page;
	333	pam->virtual = virtual;
	334
	335	spin_lock_irqsave(&pas->lock, flags);
	336	list_add_tail(&pam->list, &pas->lh);
	337	spin_unlock_irqrestore(&pas->lock, flags);
	338	} else { /* Remove */
	339	spin_lock_irqsave(&pas->lock, flags);
	340	list_for_each_entry(pam, &pas->lh, list) {
	341	if (pam->page == page) {
	342	list_del(&pam->list);
	343	spin_unlock_irqrestore(&pas->lock, flags);
	344	spin_lock_irqsave(&pool_lock, flags);
	345	list_add_tail(&pam->list, &page_address_pool);
	346	spin_unlock_irqrestore(&pool_lock, flags);
	347	goto done;
	348	}
	349	}
	350	spin_unlock_irqrestore(&pas->lock, flags);
	351	}
	352	done:
	353	return;
	354	}
	355
	356	static struct page_address_map page_address_maps[LAST_PKMAP];
	357
	358	void __init page_address_init(void)
	359	{
	360	int i;
	361
	362	INIT_LIST_HEAD(&page_address_pool);
	363	for (i = 0; i < ARRAY_SIZE(page_address_maps); i++)
	364	list_add(&page_address_maps[i].list, &page_address_pool);
	365	for (i = 0; i < ARRAY_SIZE(page_address_htable); i++) {
	366	INIT_LIST_HEAD(&page_address_htable[i].lh);
	367	spin_lock_init(&page_address_htable[i].lock);
	368	}
	369	spin_lock_init(&pool_lock);
	370	}
	371
	372	#endif /* defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) */