[net-next-2.6.git] / arch / x86 / mm / pgtable_32.c

/*
 *  linux/arch/i386/mm/pgtable.c
 */

#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/mm.h>
#include <linux/nmi.h>
#include <linux/swap.h>
#include <linux/smp.h>
#include <linux/highmem.h>
#include <linux/slab.h>
#include <linux/pagemap.h>
#include <linux/spinlock.h>
#include <linux/module.h>
#include <linux/quicklist.h>

#include <asm/system.h>
#include <asm/pgtable.h>
#include <asm/pgalloc.h>
#include <asm/fixmap.h>
#include <asm/e820.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>

void show_mem(void)
{
	int total = 0, reserved = 0;
	int shared = 0, cached = 0;
	int highmem = 0;
	struct page *page;
	pg_data_t *pgdat;
	unsigned long i;
	unsigned long flags;

	printk(KERN_INFO "Mem-info:\n");
	show_free_areas();
	printk(KERN_INFO "Free swap:       %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
	for_each_online_pgdat(pgdat) {
		pgdat_resize_lock(pgdat, &flags);
		for (i = 0; i < pgdat->node_spanned_pages; ++i) {
			if (unlikely(i % MAX_ORDER_NR_PAGES == 0))
				touch_nmi_watchdog();
			page = pgdat_page_nr(pgdat, i);
			total++;
			if (PageHighMem(page))
				highmem++;
			if (PageReserved(page))
				reserved++;
			else if (PageSwapCache(page))
				cached++;
			else if (page_count(page))
				shared += page_count(page) - 1;
		}
		pgdat_resize_unlock(pgdat, &flags);
	}
	printk(KERN_INFO "%d pages of RAM\n", total);
	printk(KERN_INFO "%d pages of HIGHMEM\n", highmem);
	printk(KERN_INFO "%d reserved pages\n", reserved);
	printk(KERN_INFO "%d pages shared\n", shared);
	printk(KERN_INFO "%d pages swap cached\n", cached);

	printk(KERN_INFO "%lu pages dirty\n", global_page_state(NR_FILE_DIRTY));
	printk(KERN_INFO "%lu pages writeback\n",
					global_page_state(NR_WRITEBACK));
	printk(KERN_INFO "%lu pages mapped\n", global_page_state(NR_FILE_MAPPED));
	printk(KERN_INFO "%lu pages slab\n",
		global_page_state(NR_SLAB_RECLAIMABLE) +
		global_page_state(NR_SLAB_UNRECLAIMABLE));
	printk(KERN_INFO "%lu pages pagetables\n",
					global_page_state(NR_PAGETABLE));
}

/*
 * Associate a virtual page frame with a given physical page frame 
 * and protection flags for that frame.
 */ 
static void set_pte_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

	pgd = swapper_pg_dir + pgd_index(vaddr);
	if (pgd_none(*pgd)) {
		BUG();
		return;
	}
	pud = pud_offset(pgd, vaddr);
	if (pud_none(*pud)) {
		BUG();
		return;
	}
	pmd = pmd_offset(pud, vaddr);
	if (pmd_none(*pmd)) {
		BUG();
		return;
	}
	pte = pte_offset_kernel(pmd, vaddr);
	if (pgprot_val(flags))
		set_pte_present(&init_mm, vaddr, pte, pfn_pte(pfn, flags));
	else
		pte_clear(&init_mm, vaddr, pte);

	/*
	 * It's enough to flush this one mapping.
	 * (PGE mappings get flushed as well)
	 */
	__flush_tlb_one(vaddr);
}

/*
 * Associate a large virtual page frame with a given physical page frame 
 * and protection flags for that frame. pfn is for the base of the page,
 * vaddr is what the page gets mapped to - both must be properly aligned. 
 * The pmd must already be instantiated. Assumes PAE mode.
 */ 
void set_pmd_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;

	if (vaddr & (PMD_SIZE-1)) {		/* vaddr is misaligned */
		printk(KERN_WARNING "set_pmd_pfn: vaddr misaligned\n");
		return; /* BUG(); */
	}
	if (pfn & (PTRS_PER_PTE-1)) {		/* pfn is misaligned */
		printk(KERN_WARNING "set_pmd_pfn: pfn misaligned\n");
		return; /* BUG(); */
	}
	pgd = swapper_pg_dir + pgd_index(vaddr);
	if (pgd_none(*pgd)) {
		printk(KERN_WARNING "set_pmd_pfn: pgd_none\n");
		return; /* BUG(); */
	}
	pud = pud_offset(pgd, vaddr);
	pmd = pmd_offset(pud, vaddr);
	set_pmd(pmd, pfn_pmd(pfn, flags));
	/*
	 * It's enough to flush this one mapping.
	 * (PGE mappings get flushed as well)
	 */
	__flush_tlb_one(vaddr);
}

static int fixmaps;
unsigned long __FIXADDR_TOP = 0xfffff000;
EXPORT_SYMBOL(__FIXADDR_TOP);

void __set_fixmap (enum fixed_addresses idx, unsigned long phys, pgprot_t flags)
{
	unsigned long address = __fix_to_virt(idx);

	if (idx >= __end_of_fixed_addresses) {
		BUG();
		return;
	}
	set_pte_pfn(address, phys >> PAGE_SHIFT, flags);
	fixmaps++;
}

/**
 * reserve_top_address - reserves a hole in the top of kernel address space
 * @reserve - size of hole to reserve
 *
 * Can be used to relocate the fixmap area and poke a hole in the top
 * of kernel address space to make room for a hypervisor.
 */
void reserve_top_address(unsigned long reserve)
{
	BUG_ON(fixmaps > 0);
	printk(KERN_INFO "Reserving virtual address space above 0x%08x\n",
	       (int)-reserve);
	__FIXADDR_TOP = -reserve - PAGE_SIZE;
	__VMALLOC_RESERVE += reserve;
}

pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
{
	return (pte_t *)__get_free_page(GFP_KERNEL|__GFP_REPEAT|__GFP_ZERO);
}

struct page *pte_alloc_one(struct mm_struct *mm, unsigned long address)
{
	struct page *pte;

#ifdef CONFIG_HIGHPTE
	pte = alloc_pages(GFP_KERNEL|__GFP_HIGHMEM|__GFP_REPEAT|__GFP_ZERO, 0);
#else
	pte = alloc_pages(GFP_KERNEL|__GFP_REPEAT|__GFP_ZERO, 0);
#endif
	return pte;
}

void pmd_ctor(struct kmem_cache *cache, void *pmd)
{
	memset(pmd, 0, PTRS_PER_PMD*sizeof(pmd_t));
}

/*
 * List of all pgd's needed for non-PAE so it can invalidate entries
 * in both cached and uncached pgd's; not needed for PAE since the
 * kernel pmd is shared. If PAE were not to share the pmd a similar
 * tactic would be needed. This is essentially codepath-based locking
 * against pageattr.c; it is the unique case in which a valid change
 * of kernel pagetables can't be lazily synchronized by vmalloc faults.
 * vmalloc faults work because attached pagetables are never freed.
 * -- wli
 */
DEFINE_SPINLOCK(pgd_lock);
struct page *pgd_list;

static inline void pgd_list_add(pgd_t *pgd)
{
	struct page *page = virt_to_page(pgd);
	page->index = (unsigned long)pgd_list;
	if (pgd_list)
		set_page_private(pgd_list, (unsigned long)&page->index);
	pgd_list = page;
	set_page_private(page, (unsigned long)&pgd_list);
}

static inline void pgd_list_del(pgd_t *pgd)
{
	struct page *next, **pprev, *page = virt_to_page(pgd);
	next = (struct page *)page->index;
	pprev = (struct page **)page_private(page);
	*pprev = next;
	if (next)
		set_page_private(next, (unsigned long)pprev);
}


#if (PTRS_PER_PMD == 1)
/* Non-PAE pgd constructor */
static void pgd_ctor(void *pgd)
{
	unsigned long flags;

	/* !PAE, no pagetable sharing */
	memset(pgd, 0, USER_PTRS_PER_PGD*sizeof(pgd_t));

	spin_lock_irqsave(&pgd_lock, flags);

	/* must happen under lock */
	clone_pgd_range((pgd_t *)pgd + USER_PTRS_PER_PGD,
			swapper_pg_dir + USER_PTRS_PER_PGD,
			KERNEL_PGD_PTRS);
	paravirt_alloc_pd_clone(__pa(pgd) >> PAGE_SHIFT,
				__pa(swapper_pg_dir) >> PAGE_SHIFT,
				USER_PTRS_PER_PGD,
				KERNEL_PGD_PTRS);
	pgd_list_add(pgd);
	spin_unlock_irqrestore(&pgd_lock, flags);
}
#else  /* PTRS_PER_PMD > 1 */
/* PAE pgd constructor */
static void pgd_ctor(void *pgd)
{
	/* PAE, kernel PMD may be shared */

	if (SHARED_KERNEL_PMD) {
		clone_pgd_range((pgd_t *)pgd + USER_PTRS_PER_PGD,
				swapper_pg_dir + USER_PTRS_PER_PGD,
				KERNEL_PGD_PTRS);
	} else {
		unsigned long flags;

		memset(pgd, 0, USER_PTRS_PER_PGD*sizeof(pgd_t));
		spin_lock_irqsave(&pgd_lock, flags);
		pgd_list_add(pgd);
		spin_unlock_irqrestore(&pgd_lock, flags);
	}
}
#endif	/* PTRS_PER_PMD */

static void pgd_dtor(void *pgd)
{
	unsigned long flags; /* can be called from interrupt context */

	if (SHARED_KERNEL_PMD)
		return;

	paravirt_release_pd(__pa(pgd) >> PAGE_SHIFT);
	spin_lock_irqsave(&pgd_lock, flags);
	pgd_list_del(pgd);
	spin_unlock_irqrestore(&pgd_lock, flags);
}

#define UNSHARED_PTRS_PER_PGD				\
	(SHARED_KERNEL_PMD ? USER_PTRS_PER_PGD : PTRS_PER_PGD)

/* If we allocate a pmd for part of the kernel address space, then
   make sure its initialized with the appropriate kernel mappings.
   Otherwise use a cached zeroed pmd.  */
static pmd_t *pmd_cache_alloc(int idx)
{
	pmd_t *pmd;

	if (idx >= USER_PTRS_PER_PGD) {
		pmd = (pmd_t *)__get_free_page(GFP_KERNEL);

		if (pmd)
			memcpy(pmd,
			       (void *)pgd_page_vaddr(swapper_pg_dir[idx]),
			       sizeof(pmd_t) * PTRS_PER_PMD);
	} else
		pmd = kmem_cache_alloc(pmd_cache, GFP_KERNEL);

	return pmd;
}

static void pmd_cache_free(pmd_t *pmd, int idx)
{
	if (idx >= USER_PTRS_PER_PGD)
		free_page((unsigned long)pmd);
	else
		kmem_cache_free(pmd_cache, pmd);
}

pgd_t *pgd_alloc(struct mm_struct *mm)
{
	int i;
	pgd_t *pgd = quicklist_alloc(0, GFP_KERNEL, pgd_ctor);

	if (PTRS_PER_PMD == 1 || !pgd)
		return pgd;

 	for (i = 0; i < UNSHARED_PTRS_PER_PGD; ++i) {
		pmd_t *pmd = pmd_cache_alloc(i);

		if (!pmd)
			goto out_oom;

		paravirt_alloc_pd(__pa(pmd) >> PAGE_SHIFT);
		set_pgd(&pgd[i], __pgd(1 + __pa(pmd)));
	}
	return pgd;

out_oom:
	for (i--; i >= 0; i--) {
		pgd_t pgdent = pgd[i];
		void* pmd = (void *)__va(pgd_val(pgdent)-1);
		paravirt_release_pd(__pa(pmd) >> PAGE_SHIFT);
		pmd_cache_free(pmd, i);
	}
	quicklist_free(0, pgd_dtor, pgd);
	return NULL;
}

void pgd_free(pgd_t *pgd)
{
	int i;

	/* in the PAE case user pgd entries are overwritten before usage */
	if (PTRS_PER_PMD > 1)
		for (i = 0; i < UNSHARED_PTRS_PER_PGD; ++i) {
			pgd_t pgdent = pgd[i];
			void* pmd = (void *)__va(pgd_val(pgdent)-1);
			paravirt_release_pd(__pa(pmd) >> PAGE_SHIFT);
			pmd_cache_free(pmd, i);
		}
	/* in the non-PAE case, free_pgtables() clears user pgd entries */
	quicklist_free(0, pgd_dtor, pgd);
}

void check_pgt_cache(void)
{
	quicklist_trim(0, pgd_dtor, 25, 16);
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/arch/i386/mm/pgtable.c
	3	*/
	4
1da177e4 LT	5	#include <linux/sched.h>
	6	#include <linux/kernel.h>
	7	#include <linux/errno.h>
	8	#include <linux/mm.h>
27eb0b28	9	#include <linux/nmi.h>
1da177e4 LT	10	#include <linux/swap.h>
	11	#include <linux/smp.h>
	12	#include <linux/highmem.h>
	13	#include <linux/slab.h>
	14	#include <linux/pagemap.h>
	15	#include <linux/spinlock.h>
052e7994	16	#include <linux/module.h>
f1d1a842	17	#include <linux/quicklist.h>
1da177e4 LT	18
	19	#include <asm/system.h>
	20	#include <asm/pgtable.h>
	21	#include <asm/pgalloc.h>
	22	#include <asm/fixmap.h>
	23	#include <asm/e820.h>
	24	#include <asm/tlb.h>
	25	#include <asm/tlbflush.h>
	26
	27	void show_mem(void)
	28	{
	29	int total = 0, reserved = 0;
	30	int shared = 0, cached = 0;
	31	int highmem = 0;
	32	struct page *page;
	33	pg_data_t *pgdat;
	34	unsigned long i;
208d54e5	35	unsigned long flags;
1da177e4	36
f90e7185	37	printk(KERN_INFO "Mem-info:\n");
1da177e4	38	show_free_areas();
f90e7185	39	printk(KERN_INFO "Free swap: %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
ec936fc5	40	for_each_online_pgdat(pgdat) {
208d54e5	41	pgdat_resize_lock(pgdat, &flags);
1da177e4	42	for (i = 0; i < pgdat->node_spanned_pages; ++i) {
27eb0b28 PB	43	if (unlikely(i % MAX_ORDER_NR_PAGES == 0))
27eb0b28 PB	44	touch_nmi_watchdog();
408fde81	45	page = pgdat_page_nr(pgdat, i);
1da177e4 LT	46	total++;
	47	if (PageHighMem(page))
	48	highmem++;
	49	if (PageReserved(page))
	50	reserved++;
	51	else if (PageSwapCache(page))
	52	cached++;
	53	else if (page_count(page))
	54	shared += page_count(page) - 1;
	55	}
208d54e5	56	pgdat_resize_unlock(pgdat, &flags);
1da177e4	57	}
f90e7185 CL	58	printk(KERN_INFO "%d pages of RAM\n", total);
	59	printk(KERN_INFO "%d pages of HIGHMEM\n", highmem);
	60	printk(KERN_INFO "%d reserved pages\n", reserved);
	61	printk(KERN_INFO "%d pages shared\n", shared);
	62	printk(KERN_INFO "%d pages swap cached\n", cached);
6f4e1e50	63
b1e7a8fd	64	printk(KERN_INFO "%lu pages dirty\n", global_page_state(NR_FILE_DIRTY));
ce866b34 CL	65	printk(KERN_INFO "%lu pages writeback\n",
ce866b34 CL	66	global_page_state(NR_WRITEBACK));
65ba55f5	67	printk(KERN_INFO "%lu pages mapped\n", global_page_state(NR_FILE_MAPPED));
972d1a7b CL	68	printk(KERN_INFO "%lu pages slab\n",
	69	global_page_state(NR_SLAB_RECLAIMABLE) +
	70	global_page_state(NR_SLAB_UNRECLAIMABLE));
df849a15 CL	71	printk(KERN_INFO "%lu pages pagetables\n",
df849a15 CL	72	global_page_state(NR_PAGETABLE));
1da177e4 LT	73	}
	74
	75	/*
	76	* Associate a virtual page frame with a given physical page frame
	77	* and protection flags for that frame.
	78	*/
	79	static void set_pte_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags)
	80	{
	81	pgd_t *pgd;
	82	pud_t *pud;
	83	pmd_t *pmd;
	84	pte_t *pte;
	85
	86	pgd = swapper_pg_dir + pgd_index(vaddr);
	87	if (pgd_none(*pgd)) {
	88	BUG();
	89	return;
	90	}
	91	pud = pud_offset(pgd, vaddr);
	92	if (pud_none(*pud)) {
	93	BUG();
	94	return;
	95	}
	96	pmd = pmd_offset(pud, vaddr);
	97	if (pmd_none(*pmd)) {
	98	BUG();
	99	return;
	100	}
	101	pte = pte_offset_kernel(pmd, vaddr);
b0bfece4	102	if (pgprot_val(flags))
aa506dc7	103	set_pte_present(&init_mm, vaddr, pte, pfn_pte(pfn, flags));
b0bfece4 JB	104	else
b0bfece4 JB	105	pte_clear(&init_mm, vaddr, pte);
1da177e4 LT	106
	107	/*
	108	* It's enough to flush this one mapping.
	109	* (PGE mappings get flushed as well)
	110	*/
	111	__flush_tlb_one(vaddr);
	112	}
	113
	114	/*
	115	* Associate a large virtual page frame with a given physical page frame
	116	* and protection flags for that frame. pfn is for the base of the page,
	117	* vaddr is what the page gets mapped to - both must be properly aligned.
	118	* The pmd must already be instantiated. Assumes PAE mode.
	119	*/
	120	void set_pmd_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags)
	121	{
	122	pgd_t *pgd;
	123	pud_t *pud;
	124	pmd_t *pmd;
	125
	126	if (vaddr & (PMD_SIZE-1)) { /* vaddr is misaligned */
f90e7185	127	printk(KERN_WARNING "set_pmd_pfn: vaddr misaligned\n");
1da177e4 LT	128	return; /* BUG(); */
	129	}
	130	if (pfn & (PTRS_PER_PTE-1)) { /* pfn is misaligned */
f90e7185	131	printk(KERN_WARNING "set_pmd_pfn: pfn misaligned\n");
1da177e4 LT	132	return; /* BUG(); */
	133	}
	134	pgd = swapper_pg_dir + pgd_index(vaddr);
	135	if (pgd_none(*pgd)) {
f90e7185	136	printk(KERN_WARNING "set_pmd_pfn: pgd_none\n");
1da177e4 LT	137	return; /* BUG(); */
	138	}
	139	pud = pud_offset(pgd, vaddr);
	140	pmd = pmd_offset(pud, vaddr);
	141	set_pmd(pmd, pfn_pmd(pfn, flags));
	142	/*
	143	* It's enough to flush this one mapping.
	144	* (PGE mappings get flushed as well)
	145	*/
	146	__flush_tlb_one(vaddr);
	147	}
	148
052e7994	149	static int fixmaps;
052e7994 JF	150	unsigned long __FIXADDR_TOP = 0xfffff000;
052e7994 JF	151	EXPORT_SYMBOL(__FIXADDR_TOP);
052e7994	152
1da177e4 LT	153	void __set_fixmap (enum fixed_addresses idx, unsigned long phys, pgprot_t flags)
	154	{
	155	unsigned long address = __fix_to_virt(idx);
	156
	157	if (idx >= __end_of_fixed_addresses) {
	158	BUG();
	159	return;
	160	}
	161	set_pte_pfn(address, phys >> PAGE_SHIFT, flags);
052e7994 JF	162	fixmaps++;
	163	}
	164
	165	/**
	166	* reserve_top_address - reserves a hole in the top of kernel address space
	167	* @reserve - size of hole to reserve
	168	*
	169	* Can be used to relocate the fixmap area and poke a hole in the top
	170	* of kernel address space to make room for a hypervisor.
	171	*/
	172	void reserve_top_address(unsigned long reserve)
	173	{
	174	BUG_ON(fixmaps > 0);
7ce0bcfd ZA	175	printk(KERN_INFO "Reserving virtual address space above 0x%08x\n",
7ce0bcfd ZA	176	(int)-reserve);
052e7994 JF	177	__FIXADDR_TOP = -reserve - PAGE_SIZE;
052e7994 JF	178	__VMALLOC_RESERVE += reserve;
1da177e4 LT	179	}
	180
	181	pte_t pte_alloc_one_kernel(struct mm_struct mm, unsigned long address)
	182	{
	183	return (pte_t *)__get_free_page(GFP_KERNEL\|__GFP_REPEAT\|__GFP_ZERO);
	184	}
	185
	186	struct page pte_alloc_one(struct mm_struct mm, unsigned long address)
	187	{
	188	struct page *pte;
	189
	190	#ifdef CONFIG_HIGHPTE
	191	pte = alloc_pages(GFP_KERNEL\|__GFP_HIGHMEM\|__GFP_REPEAT\|__GFP_ZERO, 0);
	192	#else
	193	pte = alloc_pages(GFP_KERNEL\|__GFP_REPEAT\|__GFP_ZERO, 0);
	194	#endif
	195	return pte;
	196	}
	197
4ba9b9d0	198	void pmd_ctor(struct kmem_cache cache, void pmd)
1da177e4 LT	199	{
	200	memset(pmd, 0, PTRS_PER_PMD*sizeof(pmd_t));
	201	}
	202
	203	/*
	204	* List of all pgd's needed for non-PAE so it can invalidate entries
	205	* in both cached and uncached pgd's; not needed for PAE since the
	206	* kernel pmd is shared. If PAE were not to share the pmd a similar
	207	* tactic would be needed. This is essentially codepath-based locking
	208	* against pageattr.c; it is the unique case in which a valid change
	209	* of kernel pagetables can't be lazily synchronized by vmalloc faults.
	210	* vmalloc faults work because attached pagetables are never freed.
1da177e4 LT	211	* -- wli
	212	*/
	213	DEFINE_SPINLOCK(pgd_lock);
	214	struct page *pgd_list;
	215
	216	static inline void pgd_list_add(pgd_t *pgd)
	217	{
	218	struct page *page = virt_to_page(pgd);
	219	page->index = (unsigned long)pgd_list;
	220	if (pgd_list)
4c21e2f2	221	set_page_private(pgd_list, (unsigned long)&page->index);
1da177e4	222	pgd_list = page;
4c21e2f2	223	set_page_private(page, (unsigned long)&pgd_list);
1da177e4 LT	224	}
	225
	226	static inline void pgd_list_del(pgd_t *pgd)
	227	{
	228	struct page next, pprev, page = virt_to_page(pgd);
	229	next = (struct page *)page->index;
4c21e2f2	230	pprev = (struct page **)page_private(page);
1da177e4 LT	231	*pprev = next;
1da177e4 LT	232	if (next)
4c21e2f2	233	set_page_private(next, (unsigned long)pprev);
1da177e4 LT	234	}
1da177e4 LT	235
f1d1a842 CL	236
f1d1a842 CL	237
5311ab62 JF	238	#if (PTRS_PER_PMD == 1)
5311ab62 JF	239	/* Non-PAE pgd constructor */
2378569d	240	static void pgd_ctor(void *pgd)
1da177e4 LT	241	{
	242	unsigned long flags;
	243
5311ab62 JF	244	/* !PAE, no pagetable sharing */
	245	memset(pgd, 0, USER_PTRS_PER_PGD*sizeof(pgd_t));
	246
	247	spin_lock_irqsave(&pgd_lock, flags);
1da177e4	248
5311ab62	249	/* must happen under lock */
d7271b14	250	clone_pgd_range((pgd_t *)pgd + USER_PTRS_PER_PGD,
1da177e4	251	swapper_pg_dir + USER_PTRS_PER_PGD,
d7271b14	252	KERNEL_PGD_PTRS);
c119ecce	253	paravirt_alloc_pd_clone(__pa(pgd) >> PAGE_SHIFT,
5311ab62 JF	254	__pa(swapper_pg_dir) >> PAGE_SHIFT,
	255	USER_PTRS_PER_PGD,
	256	KERNEL_PGD_PTRS);
1da177e4 LT	257	pgd_list_add(pgd);
1da177e4 LT	258	spin_unlock_irqrestore(&pgd_lock, flags);
1da177e4	259	}
5311ab62 JF	260	#else /* PTRS_PER_PMD > 1 */
5311ab62 JF	261	/* PAE pgd constructor */
2378569d	262	static void pgd_ctor(void *pgd)
5311ab62 JF	263	{
	264	/* PAE, kernel PMD may be shared */
	265
	266	if (SHARED_KERNEL_PMD) {
	267	clone_pgd_range((pgd_t *)pgd + USER_PTRS_PER_PGD,
	268	swapper_pg_dir + USER_PTRS_PER_PGD,
	269	KERNEL_PGD_PTRS);
	270	} else {
	271	unsigned long flags;
	272
	273	memset(pgd, 0, USER_PTRS_PER_PGD*sizeof(pgd_t));
	274	spin_lock_irqsave(&pgd_lock, flags);
	275	pgd_list_add(pgd);
	276	spin_unlock_irqrestore(&pgd_lock, flags);
	277	}
	278	}
	279	#endif /* PTRS_PER_PMD */
1da177e4	280
2378569d	281	static void pgd_dtor(void *pgd)
1da177e4 LT	282	{
	283	unsigned long flags; /* can be called from interrupt context */
	284
f1d1a842 CL	285	if (SHARED_KERNEL_PMD)
f1d1a842 CL	286	return;
5311ab62	287
c119ecce	288	paravirt_release_pd(__pa(pgd) >> PAGE_SHIFT);
1da177e4 LT	289	spin_lock_irqsave(&pgd_lock, flags);
	290	pgd_list_del(pgd);
	291	spin_unlock_irqrestore(&pgd_lock, flags);
	292	}
	293
5311ab62 JF	294	#define UNSHARED_PTRS_PER_PGD \
	295	(SHARED_KERNEL_PMD ? USER_PTRS_PER_PGD : PTRS_PER_PGD)
	296
	297	/* If we allocate a pmd for part of the kernel address space, then
	298	make sure its initialized with the appropriate kernel mappings.
	299	Otherwise use a cached zeroed pmd. */
	300	static pmd_t *pmd_cache_alloc(int idx)
	301	{
	302	pmd_t *pmd;
	303
	304	if (idx >= USER_PTRS_PER_PGD) {
	305	pmd = (pmd_t *)__get_free_page(GFP_KERNEL);
	306
	307	if (pmd)
	308	memcpy(pmd,
	309	(void *)pgd_page_vaddr(swapper_pg_dir[idx]),
	310	sizeof(pmd_t) * PTRS_PER_PMD);
	311	} else
	312	pmd = kmem_cache_alloc(pmd_cache, GFP_KERNEL);
	313
	314	return pmd;
	315	}
	316
	317	static void pmd_cache_free(pmd_t *pmd, int idx)
	318	{
	319	if (idx >= USER_PTRS_PER_PGD)
	320	free_page((unsigned long)pmd);
	321	else
	322	kmem_cache_free(pmd_cache, pmd);
	323	}
	324
1da177e4 LT	325	pgd_t pgd_alloc(struct mm_struct mm)
	326	{
	327	int i;
f1d1a842	328	pgd_t *pgd = quicklist_alloc(0, GFP_KERNEL, pgd_ctor);
1da177e4 LT	329
	330	if (PTRS_PER_PMD == 1 \|\| !pgd)
	331	return pgd;
	332
5311ab62 JF	333	for (i = 0; i < UNSHARED_PTRS_PER_PGD; ++i) {
	334	pmd_t *pmd = pmd_cache_alloc(i);
	335
1da177e4 LT	336	if (!pmd)
1da177e4 LT	337	goto out_oom;
5311ab62	338
c119ecce	339	paravirt_alloc_pd(__pa(pmd) >> PAGE_SHIFT);
1da177e4 LT	340	set_pgd(&pgd[i], __pgd(1 + __pa(pmd)));
	341	}
	342	return pgd;
	343
	344	out_oom:
c119ecce ZA	345	for (i--; i >= 0; i--) {
	346	pgd_t pgdent = pgd[i];
	347	void* pmd = (void *)__va(pgd_val(pgdent)-1);
	348	paravirt_release_pd(__pa(pmd) >> PAGE_SHIFT);
5311ab62	349	pmd_cache_free(pmd, i);
c119ecce	350	}
f1d1a842	351	quicklist_free(0, pgd_dtor, pgd);
1da177e4 LT	352	return NULL;
	353	}
	354
	355	void pgd_free(pgd_t *pgd)
	356	{
	357	int i;
	358
	359	/* in the PAE case user pgd entries are overwritten before usage */
	360	if (PTRS_PER_PMD > 1)
5311ab62	361	for (i = 0; i < UNSHARED_PTRS_PER_PGD; ++i) {
c119ecce ZA	362	pgd_t pgdent = pgd[i];
	363	void* pmd = (void *)__va(pgd_val(pgdent)-1);
	364	paravirt_release_pd(__pa(pmd) >> PAGE_SHIFT);
5311ab62	365	pmd_cache_free(pmd, i);
c119ecce	366	}
e0da382c	367	/* in the non-PAE case, free_pgtables() clears user pgd entries */
f1d1a842	368	quicklist_free(0, pgd_dtor, pgd);
1da177e4	369	}
f1d1a842 CL	370
	371	void check_pgt_cache(void)
	372	{
	373	quicklist_trim(0, pgd_dtor, 25, 16);
	374	}
	375