[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);

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