[deliverable/linux.git] / arch / ia64 / mm / hugetlbpage.c

/*
 * IA-64 Huge TLB Page Support for Kernel.
 *
 * Copyright (C) 2002-2004 Rohit Seth <rohit.seth@intel.com>
 * Copyright (C) 2003-2004 Ken Chen <kenneth.w.chen@intel.com>
 *
 * Sep, 2003: add numa support
 * Feb, 2004: dynamic hugetlb page size via boot parameter
 */

#include <linux/init.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/hugetlb.h>
#include <linux/pagemap.h>
#include <linux/module.h>
#include <linux/sysctl.h>
#include <linux/log2.h>
#include <asm/mman.h>
#include <asm/pgalloc.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>

unsigned int hpage_shift = HPAGE_SHIFT_DEFAULT;
EXPORT_SYMBOL(hpage_shift);

pte_t *
huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz)
{
	unsigned long taddr = htlbpage_to_page(addr);
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, taddr);
	pud = pud_alloc(mm, pgd, taddr);
	if (pud) {
		pmd = pmd_alloc(mm, pud, taddr);
		if (pmd)
			pte = pte_alloc_map(mm, pmd, taddr);
	}
	return pte;
}

pte_t *
huge_pte_offset (struct mm_struct *mm, unsigned long addr)
{
	unsigned long taddr = htlbpage_to_page(addr);
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, taddr);
	if (pgd_present(*pgd)) {
		pud = pud_offset(pgd, taddr);
		if (pud_present(*pud)) {
			pmd = pmd_offset(pud, taddr);
			if (pmd_present(*pmd))
				pte = pte_offset_map(pmd, taddr);
		}
	}

	return pte;
}

#define mk_pte_huge(entry) { pte_val(entry) |= _PAGE_P; }

/*
 * Don't actually need to do any preparation, but need to make sure
 * the address is in the right region.
 */
int prepare_hugepage_range(struct file *file,
			unsigned long addr, unsigned long len)
{
	if (len & ~HPAGE_MASK)
		return -EINVAL;
	if (addr & ~HPAGE_MASK)
		return -EINVAL;
	if (REGION_NUMBER(addr) != RGN_HPAGE)
		return -EINVAL;

	return 0;
}

struct page *follow_huge_addr(struct mm_struct *mm, unsigned long addr, int write)
{
	struct page *page;
	pte_t *ptep;

	if (REGION_NUMBER(addr) != RGN_HPAGE)
		return ERR_PTR(-EINVAL);

	ptep = huge_pte_offset(mm, addr);
	if (!ptep || pte_none(*ptep))
		return NULL;
	page = pte_page(*ptep);
	page += ((addr & ~HPAGE_MASK) >> PAGE_SHIFT);
	return page;
}
int pmd_huge(pmd_t pmd)
{
	return 0;
}

int pud_huge(pud_t pud)
{
	return 0;
}

void hugetlb_free_pgd_range(struct mmu_gather *tlb,
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
{
	/*
	 * This is called to free hugetlb page tables.
	 *
	 * The offset of these addresses from the base of the hugetlb
	 * region must be scaled down by HPAGE_SIZE/PAGE_SIZE so that
	 * the standard free_pgd_range will free the right page tables.
	 *
	 * If floor and ceiling are also in the hugetlb region, they
	 * must likewise be scaled down; but if outside, left unchanged.
	 */

	addr = htlbpage_to_page(addr);
	end  = htlbpage_to_page(end);
	if (REGION_NUMBER(floor) == RGN_HPAGE)
		floor = htlbpage_to_page(floor);
	if (REGION_NUMBER(ceiling) == RGN_HPAGE)
		ceiling = htlbpage_to_page(ceiling);

	free_pgd_range(tlb, addr, end, floor, ceiling);
}

unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
		unsigned long pgoff, unsigned long flags)
{
	struct vm_unmapped_area_info info;

	if (len > RGN_MAP_LIMIT)
		return -ENOMEM;
	if (len & ~HPAGE_MASK)
		return -EINVAL;

	/* Handle MAP_FIXED */
	if (flags & MAP_FIXED) {
		if (prepare_hugepage_range(file, addr, len))
			return -EINVAL;
		return addr;
	}

	/* This code assumes that RGN_HPAGE != 0. */
	if ((REGION_NUMBER(addr) != RGN_HPAGE) || (addr & (HPAGE_SIZE - 1)))
		addr = HPAGE_REGION_BASE;

	info.flags = 0;
	info.length = len;
	info.low_limit = addr;
	info.high_limit = HPAGE_REGION_BASE + RGN_MAP_LIMIT;
	info.align_mask = PAGE_MASK & (HPAGE_SIZE - 1);
	info.align_offset = 0;
	return vm_unmapped_area(&info);
}

static int __init hugetlb_setup_sz(char *str)
{
	u64 tr_pages;
	unsigned long long size;

	if (ia64_pal_vm_page_size(&tr_pages, NULL) != 0)
		/*
		 * shouldn't happen, but just in case.
		 */
		tr_pages = 0x15557000UL;

	size = memparse(str, &str);
	if (*str || !is_power_of_2(size) || !(tr_pages & size) ||
		size <= PAGE_SIZE ||
		size >= (1UL << PAGE_SHIFT << MAX_ORDER)) {
		printk(KERN_WARNING "Invalid huge page size specified\n");
		return 1;
	}

	hpage_shift = __ffs(size);
	/*
	 * boot cpu already executed ia64_mmu_init, and has HPAGE_SHIFT_DEFAULT
	 * override here with new page shift.
	 */
	ia64_set_rr(HPAGE_REGION_BASE, hpage_shift << 2);
	return 0;
}
early_param("hugepagesz", hugetlb_setup_sz);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* IA-64 Huge TLB Page Support for Kernel.
	3	*
	4	* Copyright (C) 2002-2004 Rohit Seth <rohit.seth@intel.com>
	5	* Copyright (C) 2003-2004 Ken Chen <kenneth.w.chen@intel.com>
	6	*
	7	* Sep, 2003: add numa support
	8	* Feb, 2004: dynamic hugetlb page size via boot parameter
	9	*/
	10
1da177e4 LT	11	#include <linux/init.h>
	12	#include <linux/fs.h>
	13	#include <linux/mm.h>
	14	#include <linux/hugetlb.h>
	15	#include <linux/pagemap.h>
34d8a380	16	#include <linux/module.h>
1da177e4	17	#include <linux/sysctl.h>
9be26f4c	18	#include <linux/log2.h>
1da177e4 LT	19	#include <asm/mman.h>
	20	#include <asm/pgalloc.h>
	21	#include <asm/tlb.h>
	22	#include <asm/tlbflush.h>
	23
34d8a380 JS	24	unsigned int hpage_shift = HPAGE_SHIFT_DEFAULT;
34d8a380 JS	25	EXPORT_SYMBOL(hpage_shift);
1da177e4	26
63551ae0	27	pte_t *
a5516438	28	huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz)
1da177e4 LT	29	{
	30	unsigned long taddr = htlbpage_to_page(addr);
	31	pgd_t *pgd;
	32	pud_t *pud;
	33	pmd_t *pmd;
	34	pte_t *pte = NULL;
	35
	36	pgd = pgd_offset(mm, taddr);
	37	pud = pud_alloc(mm, pgd, taddr);
	38	if (pud) {
	39	pmd = pmd_alloc(mm, pud, taddr);
	40	if (pmd)
3ed3a4f0	41	pte = pte_alloc_map(mm, pmd, taddr);
1da177e4 LT	42	}
	43	return pte;
	44	}
	45
63551ae0	46	pte_t *
1da177e4 LT	47	huge_pte_offset (struct mm_struct *mm, unsigned long addr)
	48	{
	49	unsigned long taddr = htlbpage_to_page(addr);
	50	pgd_t *pgd;
	51	pud_t *pud;
	52	pmd_t *pmd;
	53	pte_t *pte = NULL;
	54
	55	pgd = pgd_offset(mm, taddr);
	56	if (pgd_present(*pgd)) {
	57	pud = pud_offset(pgd, taddr);
	58	if (pud_present(*pud)) {
	59	pmd = pmd_offset(pud, taddr);
	60	if (pmd_present(*pmd))
	61	pte = pte_offset_map(pmd, taddr);
	62	}
	63	}
	64
	65	return pte;
	66	}
	67
	68	#define mk_pte_huge(entry) { pte_val(entry) \|= _PAGE_P; }
	69
1da177e4	70	/*
42b88bef DG	71	* Don't actually need to do any preparation, but need to make sure
42b88bef DG	72	* the address is in the right region.
1da177e4	73	*/
a5516438 AK	74	int prepare_hugepage_range(struct file *file,
a5516438 AK	75	unsigned long addr, unsigned long len)
1da177e4 LT	76	{
	77	if (len & ~HPAGE_MASK)
	78	return -EINVAL;
	79	if (addr & ~HPAGE_MASK)
	80	return -EINVAL;
0a41e250	81	if (REGION_NUMBER(addr) != RGN_HPAGE)
1da177e4 LT	82	return -EINVAL;
	83
	84	return 0;
	85	}
	86
1da177e4 LT	87	struct page follow_huge_addr(struct mm_struct mm, unsigned long addr, int write)
	88	{
	89	struct page *page;
	90	pte_t *ptep;
	91
0a41e250	92	if (REGION_NUMBER(addr) != RGN_HPAGE)
1da177e4 LT	93	return ERR_PTR(-EINVAL);
	94
	95	ptep = huge_pte_offset(mm, addr);
	96	if (!ptep \|\| pte_none(*ptep))
	97	return NULL;
	98	page = pte_page(*ptep);
	99	page += ((addr & ~HPAGE_MASK) >> PAGE_SHIFT);
	100	return page;
	101	}
	102	int pmd_huge(pmd_t pmd)
	103	{
	104	return 0;
	105	}
ceb86879 AK	106
	107	int pud_huge(pud_t pud)
	108	{
	109	return 0;
	110	}
	111
42b77728	112	void hugetlb_free_pgd_range(struct mmu_gather *tlb,
3bf5ee95 HD	113	unsigned long addr, unsigned long end,
3bf5ee95 HD	114	unsigned long floor, unsigned long ceiling)
1da177e4	115	{
3bf5ee95	116	/*
2332c9ae	117	* This is called to free hugetlb page tables.
3bf5ee95 HD	118	*
	119	* The offset of these addresses from the base of the hugetlb
	120	* region must be scaled down by HPAGE_SIZE/PAGE_SIZE so that
	121	* the standard free_pgd_range will free the right page tables.
	122	*
	123	* If floor and ceiling are also in the hugetlb region, they
	124	* must likewise be scaled down; but if outside, left unchanged.
	125	*/
	126
	127	addr = htlbpage_to_page(addr);
	128	end = htlbpage_to_page(end);
2332c9ae	129	if (REGION_NUMBER(floor) == RGN_HPAGE)
3bf5ee95	130	floor = htlbpage_to_page(floor);
2332c9ae	131	if (REGION_NUMBER(ceiling) == RGN_HPAGE)
3bf5ee95 HD	132	ceiling = htlbpage_to_page(ceiling);
	133
	134	free_pgd_range(tlb, addr, end, floor, ceiling);
1da177e4 LT	135	}
1da177e4 LT	136
1da177e4 LT	137	unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
	138	unsigned long pgoff, unsigned long flags)
	139	{
e31048af	140	struct vm_unmapped_area_info info;
1da177e4 LT	141
	142	if (len > RGN_MAP_LIMIT)
	143	return -ENOMEM;
	144	if (len & ~HPAGE_MASK)
	145	return -EINVAL;
afa37394 BH	146
	147	/* Handle MAP_FIXED */
	148	if (flags & MAP_FIXED) {
a5516438	149	if (prepare_hugepage_range(file, addr, len))
afa37394 BH	150	return -EINVAL;
	151	return addr;
	152	}
	153
0a41e250 PC	154	/* This code assumes that RGN_HPAGE != 0. */
0a41e250 PC	155	if ((REGION_NUMBER(addr) != RGN_HPAGE) \|\| (addr & (HPAGE_SIZE - 1)))
1da177e4	156	addr = HPAGE_REGION_BASE;
e31048af ML	157
	158	info.flags = 0;
	159	info.length = len;
	160	info.low_limit = addr;
	161	info.high_limit = HPAGE_REGION_BASE + RGN_MAP_LIMIT;
	162	info.align_mask = PAGE_MASK & (HPAGE_SIZE - 1);
	163	info.align_offset = 0;
	164	return vm_unmapped_area(&info);
1da177e4 LT	165	}
	166
	167	static int __init hugetlb_setup_sz(char *str)
	168	{
	169	u64 tr_pages;
	170	unsigned long long size;
	171
	172	if (ia64_pal_vm_page_size(&tr_pages, NULL) != 0)
	173	/*
	174	* shouldn't happen, but just in case.
	175	*/
	176	tr_pages = 0x15557000UL;
	177
	178	size = memparse(str, &str);
9be26f4c	179	if (*str \|\| !is_power_of_2(size) \|\| !(tr_pages & size) \|\|
1da177e4 LT	180	size <= PAGE_SIZE \|\|
	181	size >= (1UL << PAGE_SHIFT << MAX_ORDER)) {
	182	printk(KERN_WARNING "Invalid huge page size specified\n");
	183	return 1;
	184	}
	185
	186	hpage_shift = __ffs(size);
	187	/*
	188	* boot cpu already executed ia64_mmu_init, and has HPAGE_SHIFT_DEFAULT
	189	* override here with new page shift.
	190	*/
	191	ia64_set_rr(HPAGE_REGION_BASE, hpage_shift << 2);
d9c23400	192	return 0;
1da177e4	193	}
d9c23400	194	early_param("hugepagesz", hugetlb_setup_sz);