[net-next-2.6.git] / mm / highmem.c

/*
 * High memory handling common code and variables.
 *
 * (C) 1999 Andrea Arcangeli, SuSE GmbH, andrea@suse.de
 *          Gerhard Wichert, Siemens AG, Gerhard.Wichert@pdb.siemens.de
 *
 *
 * 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 <mingo@redhat.com>
 */

#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;

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));
}

/* 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;
}

void *kmap_high(struct page *page)
{
	unsigned long vaddr;

	/*
	 * For highmem pages, we can't trust "virtual" until
	 * after we have the lock.
	 *
	 * We cannot call this from interrupts, as it may block
	 */
	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);

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)];
}

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);

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, andrea@suse.de
	5	* Gerhard Wichert, Siemens AG, Gerhard.Wichert@pdb.siemens.de
	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 <mingo@redhat.com>
	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 CL	42	unsigned long totalhigh_pages __read_mostly;
	43
	44	unsigned int nr_free_highpages (void)
	45	{
	46	pg_data_t *pgdat;
	47	unsigned int pages = 0;
	48
2a1e274a	49	for_each_online_pgdat(pgdat) {
d23ad423 CL	50	pages += zone_page_state(&pgdat->node_zones[ZONE_HIGHMEM],
d23ad423 CL	51	NR_FREE_PAGES);
2a1e274a MG	52	if (zone_movable_is_highmem())
	53	pages += zone_page_state(
	54	&pgdat->node_zones[ZONE_MOVABLE],
	55	NR_FREE_PAGES);
	56	}
c1f60a5a CL	57
	58	return pages;
	59	}
	60
1da177e4 LT	61	static int pkmap_count[LAST_PKMAP];
	62	static unsigned int last_pkmap_nr;
	63	static __cacheline_aligned_in_smp DEFINE_SPINLOCK(kmap_lock);
	64
	65	pte_t * pkmap_page_table;
	66
	67	static DECLARE_WAIT_QUEUE_HEAD(pkmap_map_wait);
	68
	69	static void flush_all_zero_pkmaps(void)
	70	{
	71	int i;
	72
	73	flush_cache_kmaps();
	74
	75	for (i = 0; i < LAST_PKMAP; i++) {
	76	struct page *page;
	77
	78	/*
	79	* zero means we don't have anything to do,
	80	* >1 means that it is still in use. Only
	81	* a count of 1 means that it is free but
	82	* needs to be unmapped
	83	*/
	84	if (pkmap_count[i] != 1)
	85	continue;
	86	pkmap_count[i] = 0;
	87
	88	/* sanity check */
75babcac	89	BUG_ON(pte_none(pkmap_page_table[i]));
1da177e4 LT	90
	91	/*
	92	* Don't need an atomic fetch-and-clear op here;
	93	* no-one has the page mapped, and cannot get at
	94	* its virtual address (and hence PTE) without first
	95	* getting the kmap_lock (which is held here).
	96	* So no dangers, even with speculative execution.
	97	*/
	98	page = pte_page(pkmap_page_table[i]);
	99	pte_clear(&init_mm, (unsigned long)page_address(page),
	100	&pkmap_page_table[i]);
	101
	102	set_page_address(page, NULL);
	103	}
	104	flush_tlb_kernel_range(PKMAP_ADDR(0), PKMAP_ADDR(LAST_PKMAP));
	105	}
	106
ce6234b5 JF	107	/* Flush all unused kmap mappings in order to remove stray
	108	mappings. */
	109	void kmap_flush_unused(void)
	110	{
	111	spin_lock(&kmap_lock);
	112	flush_all_zero_pkmaps();
	113	spin_unlock(&kmap_lock);
	114	}
	115
1da177e4 LT	116	static inline unsigned long map_new_virtual(struct page *page)
	117	{
	118	unsigned long vaddr;
	119	int count;
	120
	121	start:
	122	count = LAST_PKMAP;
	123	/* Find an empty entry */
	124	for (;;) {
	125	last_pkmap_nr = (last_pkmap_nr + 1) & LAST_PKMAP_MASK;
	126	if (!last_pkmap_nr) {
	127	flush_all_zero_pkmaps();
	128	count = LAST_PKMAP;
	129	}
	130	if (!pkmap_count[last_pkmap_nr])
	131	break; /* Found a usable entry */
	132	if (--count)
	133	continue;
	134
	135	/*
	136	* Sleep for somebody else to unmap their entries
	137	*/
	138	{
	139	DECLARE_WAITQUEUE(wait, current);
	140
	141	__set_current_state(TASK_UNINTERRUPTIBLE);
	142	add_wait_queue(&pkmap_map_wait, &wait);
	143	spin_unlock(&kmap_lock);
	144	schedule();
	145	remove_wait_queue(&pkmap_map_wait, &wait);
	146	spin_lock(&kmap_lock);
	147
	148	/* Somebody else might have mapped it while we slept */
	149	if (page_address(page))
	150	return (unsigned long)page_address(page);
	151
	152	/* Re-start */
	153	goto start;
	154	}
	155	}
	156	vaddr = PKMAP_ADDR(last_pkmap_nr);
	157	set_pte_at(&init_mm, vaddr,
	158	&(pkmap_page_table[last_pkmap_nr]), mk_pte(page, kmap_prot));
	159
	160	pkmap_count[last_pkmap_nr] = 1;
	161	set_page_address(page, (void *)vaddr);
	162
	163	return vaddr;
	164	}
	165
920c7a5d	166	void kmap_high(struct page page)
1da177e4 LT	167	{
	168	unsigned long vaddr;
	169
	170	/*
	171	* For highmem pages, we can't trust "virtual" until
	172	* after we have the lock.
	173	*
	174	* We cannot call this from interrupts, as it may block
	175	*/
	176	spin_lock(&kmap_lock);
	177	vaddr = (unsigned long)page_address(page);
	178	if (!vaddr)
	179	vaddr = map_new_virtual(page);
	180	pkmap_count[PKMAP_NR(vaddr)]++;
75babcac	181	BUG_ON(pkmap_count[PKMAP_NR(vaddr)] < 2);
1da177e4 LT	182	spin_unlock(&kmap_lock);
	183	return (void*) vaddr;
	184	}
	185
	186	EXPORT_SYMBOL(kmap_high);
	187
920c7a5d	188	void kunmap_high(struct page *page)
1da177e4 LT	189	{
	190	unsigned long vaddr;
	191	unsigned long nr;
	192	int need_wakeup;
	193
	194	spin_lock(&kmap_lock);
	195	vaddr = (unsigned long)page_address(page);
75babcac	196	BUG_ON(!vaddr);
1da177e4 LT	197	nr = PKMAP_NR(vaddr);
	198
	199	/*
	200	* A count must never go down to zero
	201	* without a TLB flush!
	202	*/
	203	need_wakeup = 0;
	204	switch (--pkmap_count[nr]) {
	205	case 0:
	206	BUG();
	207	case 1:
	208	/*
	209	* Avoid an unnecessary wake_up() function call.
	210	* The common case is pkmap_count[] == 1, but
	211	* no waiters.
	212	* The tasks queued in the wait-queue are guarded
	213	* by both the lock in the wait-queue-head and by
	214	* the kmap_lock. As the kmap_lock is held here,
	215	* no need for the wait-queue-head's lock. Simply
	216	* test if the queue is empty.
	217	*/
	218	need_wakeup = waitqueue_active(&pkmap_map_wait);
	219	}
	220	spin_unlock(&kmap_lock);
	221
	222	/* do wake-up, if needed, race-free outside of the spin lock */
	223	if (need_wakeup)
	224	wake_up(&pkmap_map_wait);
	225	}
	226
	227	EXPORT_SYMBOL(kunmap_high);
1da177e4 LT	228	#endif
1da177e4 LT	229
1da177e4 LT	230	#if defined(HASHED_PAGE_VIRTUAL)
	231
	232	#define PA_HASH_ORDER 7
	233
	234	/*
	235	* Describes one page->virtual association
	236	*/
	237	struct page_address_map {
	238	struct page *page;
	239	void *virtual;
	240	struct list_head list;
	241	};
	242
	243	/*
	244	* page_address_map freelist, allocated from page_address_maps.
	245	*/
	246	static struct list_head page_address_pool; /* freelist */
	247	static spinlock_t pool_lock; /* protects page_address_pool */
	248
	249	/*
	250	* Hash table bucket
	251	*/
	252	static struct page_address_slot {
	253	struct list_head lh; /* List of page_address_maps */
	254	spinlock_t lock; /* Protect this bucket's list */
	255	} ____cacheline_aligned_in_smp page_address_htable[1<<PA_HASH_ORDER];
	256
	257	static struct page_address_slot page_slot(struct page page)
	258	{
	259	return &page_address_htable[hash_ptr(page, PA_HASH_ORDER)];
	260	}
	261
	262	void page_address(struct page page)
	263	{
	264	unsigned long flags;
	265	void *ret;
	266	struct page_address_slot *pas;
	267
	268	if (!PageHighMem(page))
	269	return lowmem_page_address(page);
	270
	271	pas = page_slot(page);
	272	ret = NULL;
	273	spin_lock_irqsave(&pas->lock, flags);
	274	if (!list_empty(&pas->lh)) {
	275	struct page_address_map *pam;
	276
	277	list_for_each_entry(pam, &pas->lh, list) {
	278	if (pam->page == page) {
	279	ret = pam->virtual;
	280	goto done;
	281	}
	282	}
	283	}
	284	done:
	285	spin_unlock_irqrestore(&pas->lock, flags);
	286	return ret;
	287	}
	288
	289	EXPORT_SYMBOL(page_address);
	290
	291	void set_page_address(struct page page, void virtual)
	292	{
	293	unsigned long flags;
294	struct page_address_slot *pas;
295	struct page_address_map *pam;
296
297	BUG_ON(!PageHighMem(page));
298
299	pas = page_slot(page);
300	if (virtual) { /* Add */
301	BUG_ON(list_empty(&page_address_pool));
302
303	spin_lock_irqsave(&pool_lock, flags);
304	pam = list_entry(page_address_pool.next,
305	struct page_address_map, list);
306	list_del(&pam->list);
307	spin_unlock_irqrestore(&pool_lock, flags);
308
309	pam->page = page;
310	pam->virtual = virtual;
311
312	spin_lock_irqsave(&pas->lock, flags);
313	list_add_tail(&pam->list, &pas->lh);
314	spin_unlock_irqrestore(&pas->lock, flags);
315	} else { /* Remove */
316	spin_lock_irqsave(&pas->lock, flags);
317	list_for_each_entry(pam, &pas->lh, list) {
318	if (pam->page == page) {
319	list_del(&pam->list);
320	spin_unlock_irqrestore(&pas->lock, flags);
321	spin_lock_irqsave(&pool_lock, flags);
322	list_add_tail(&pam->list, &page_address_pool);
323	spin_unlock_irqrestore(&pool_lock, flags);
324	goto done;
325	}
326	}
327	spin_unlock_irqrestore(&pas->lock, flags);
328	}
329	done:
330	return;
331	}
332
333	static struct page_address_map page_address_maps[LAST_PKMAP];
334
335	void __init page_address_init(void)
336	{
337	int i;
338
339	INIT_LIST_HEAD(&page_address_pool);
340	for (i = 0; i < ARRAY_SIZE(page_address_maps); i++)
341	list_add(&page_address_maps[i].list, &page_address_pool);
342	for (i = 0; i < ARRAY_SIZE(page_address_htable); i++) {
343	INIT_LIST_HEAD(&page_address_htable[i].lh);
344	spin_lock_init(&page_address_htable[i].lock);
345	}
346	spin_lock_init(&pool_lock);
347	}
348
349	#endif /* defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) */