[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));
}

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
	102	static inline unsigned long map_new_virtual(struct page *page)
	103	{
	104	unsigned long vaddr;
	105	int count;
	106
	107	start:
	108	count = LAST_PKMAP;
	109	/* Find an empty entry */
	110	for (;;) {
	111	last_pkmap_nr = (last_pkmap_nr + 1) & LAST_PKMAP_MASK;
	112	if (!last_pkmap_nr) {
	113	flush_all_zero_pkmaps();
	114	count = LAST_PKMAP;
	115	}
	116	if (!pkmap_count[last_pkmap_nr])
	117	break; /* Found a usable entry */
	118	if (--count)
	119	continue;
	120
	121	/*
	122	* Sleep for somebody else to unmap their entries
	123	*/
	124	{
	125	DECLARE_WAITQUEUE(wait, current);
	126
	127	__set_current_state(TASK_UNINTERRUPTIBLE);
	128	add_wait_queue(&pkmap_map_wait, &wait);
	129	spin_unlock(&kmap_lock);
	130	schedule();
	131	remove_wait_queue(&pkmap_map_wait, &wait);
	132	spin_lock(&kmap_lock);
	133
	134	/* Somebody else might have mapped it while we slept */
	135	if (page_address(page))
	136	return (unsigned long)page_address(page);
	137
	138	/* Re-start */
	139	goto start;
	140	}
	141	}
	142	vaddr = PKMAP_ADDR(last_pkmap_nr);
	143	set_pte_at(&init_mm, vaddr,
	144	&(pkmap_page_table[last_pkmap_nr]), mk_pte(page, kmap_prot));
	145
	146	pkmap_count[last_pkmap_nr] = 1;
	147	set_page_address(page, (void *)vaddr);
	148
149	return vaddr;
150	}
151
152	void fastcall kmap_high(struct page page)
153	{
154	unsigned long vaddr;
155
156	/*
157	* For highmem pages, we can't trust "virtual" until
158	* after we have the lock.
159	*
160	* We cannot call this from interrupts, as it may block
161	*/
162	spin_lock(&kmap_lock);
163	vaddr = (unsigned long)page_address(page);
164	if (!vaddr)
165	vaddr = map_new_virtual(page);
166	pkmap_count[PKMAP_NR(vaddr)]++;
75babcac	167	BUG_ON(pkmap_count[PKMAP_NR(vaddr)] < 2);
1da177e4 LT	168	spin_unlock(&kmap_lock);
	169	return (void*) vaddr;
	170	}
	171
	172	EXPORT_SYMBOL(kmap_high);
	173
	174	void fastcall kunmap_high(struct page *page)
	175	{
	176	unsigned long vaddr;
	177	unsigned long nr;
	178	int need_wakeup;
	179
	180	spin_lock(&kmap_lock);
	181	vaddr = (unsigned long)page_address(page);
75babcac	182	BUG_ON(!vaddr);
1da177e4 LT	183	nr = PKMAP_NR(vaddr);
	184
	185	/*
	186	* A count must never go down to zero
	187	* without a TLB flush!
	188	*/
	189	need_wakeup = 0;
	190	switch (--pkmap_count[nr]) {
	191	case 0:
	192	BUG();
	193	case 1:
	194	/*
	195	* Avoid an unnecessary wake_up() function call.
	196	* The common case is pkmap_count[] == 1, but
	197	* no waiters.
	198	* The tasks queued in the wait-queue are guarded
	199	* by both the lock in the wait-queue-head and by
	200	* the kmap_lock. As the kmap_lock is held here,
	201	* no need for the wait-queue-head's lock. Simply
	202	* test if the queue is empty.
	203	*/
	204	need_wakeup = waitqueue_active(&pkmap_map_wait);
	205	}
	206	spin_unlock(&kmap_lock);
	207
	208	/* do wake-up, if needed, race-free outside of the spin lock */
	209	if (need_wakeup)
	210	wake_up(&pkmap_map_wait);
	211	}
	212
	213	EXPORT_SYMBOL(kunmap_high);
1da177e4 LT	214	#endif
1da177e4 LT	215
1da177e4 LT	216	#if defined(HASHED_PAGE_VIRTUAL)
	217
	218	#define PA_HASH_ORDER 7
	219
	220	/*
	221	* Describes one page->virtual association
	222	*/
	223	struct page_address_map {
	224	struct page *page;
	225	void *virtual;
	226	struct list_head list;
	227	};
	228
	229	/*
	230	* page_address_map freelist, allocated from page_address_maps.
	231	*/
	232	static struct list_head page_address_pool; /* freelist */
	233	static spinlock_t pool_lock; /* protects page_address_pool */
	234
	235	/*
	236	* Hash table bucket
	237	*/
	238	static struct page_address_slot {
	239	struct list_head lh; /* List of page_address_maps */
	240	spinlock_t lock; /* Protect this bucket's list */
	241	} ____cacheline_aligned_in_smp page_address_htable[1<<PA_HASH_ORDER];
	242
	243	static struct page_address_slot page_slot(struct page page)
	244	{
	245	return &page_address_htable[hash_ptr(page, PA_HASH_ORDER)];
	246	}
	247
	248	void page_address(struct page page)
	249	{
	250	unsigned long flags;
	251	void *ret;
	252	struct page_address_slot *pas;
	253
	254	if (!PageHighMem(page))
	255	return lowmem_page_address(page);
	256
	257	pas = page_slot(page);
	258	ret = NULL;
	259	spin_lock_irqsave(&pas->lock, flags);
	260	if (!list_empty(&pas->lh)) {
	261	struct page_address_map *pam;
	262
	263	list_for_each_entry(pam, &pas->lh, list) {
	264	if (pam->page == page) {
	265	ret = pam->virtual;
	266	goto done;
	267	}
	268	}
	269	}
	270	done:
	271	spin_unlock_irqrestore(&pas->lock, flags);
	272	return ret;
	273	}
	274
	275	EXPORT_SYMBOL(page_address);
	276
	277	void set_page_address(struct page page, void virtual)
	278	{
	279	unsigned long flags;
280	struct page_address_slot *pas;
281	struct page_address_map *pam;
282
283	BUG_ON(!PageHighMem(page));
284
285	pas = page_slot(page);
286	if (virtual) { /* Add */
287	BUG_ON(list_empty(&page_address_pool));
288
289	spin_lock_irqsave(&pool_lock, flags);
290	pam = list_entry(page_address_pool.next,
291	struct page_address_map, list);
292	list_del(&pam->list);
293	spin_unlock_irqrestore(&pool_lock, flags);
294
295	pam->page = page;
296	pam->virtual = virtual;
297
298	spin_lock_irqsave(&pas->lock, flags);
299	list_add_tail(&pam->list, &pas->lh);
300	spin_unlock_irqrestore(&pas->lock, flags);
301	} else { /* Remove */
302	spin_lock_irqsave(&pas->lock, flags);
303	list_for_each_entry(pam, &pas->lh, list) {
304	if (pam->page == page) {
305	list_del(&pam->list);
306	spin_unlock_irqrestore(&pas->lock, flags);
307	spin_lock_irqsave(&pool_lock, flags);
308	list_add_tail(&pam->list, &page_address_pool);
309	spin_unlock_irqrestore(&pool_lock, flags);
310	goto done;
311	}
312	}
313	spin_unlock_irqrestore(&pas->lock, flags);
314	}
315	done:
316	return;
317	}
318
319	static struct page_address_map page_address_maps[LAST_PKMAP];
320
321	void __init page_address_init(void)
322	{
323	int i;
324
325	INIT_LIST_HEAD(&page_address_pool);
326	for (i = 0; i < ARRAY_SIZE(page_address_maps); i++)
327	list_add(&page_address_maps[i].list, &page_address_pool);
328	for (i = 0; i < ARRAY_SIZE(page_address_htable); i++) {
329	INIT_LIST_HEAD(&page_address_htable[i].lh);
330	spin_lock_init(&page_address_htable[i].lock);
331	}
332	spin_lock_init(&pool_lock);
333	}
334
335	#endif /* defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) */