[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/smp_lock.h>
#include <linux/slab.h>
#include <linux/sysctl.h>
#include <asm/mman.h>
#include <asm/pgalloc.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>

unsigned int hpage_shift=HPAGE_SHIFT_DEFAULT;

pte_t *
huge_pte_alloc (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);
	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;
}

int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	return 0;
}

#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(unsigned long addr, unsigned long len, pgoff_t pgoff)
{
	if (pgoff & (~HPAGE_MASK >> PAGE_SHIFT))
		return -EINVAL;
	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;
}
struct page *
follow_huge_pmd(struct mm_struct *mm, unsigned long address, pmd_t *pmd, int write)
{
	return NULL;
}

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_area_struct *vmm;

	if (len > RGN_MAP_LIMIT)
		return -ENOMEM;
	if (len & ~HPAGE_MASK)
		return -EINVAL;
	/* This code assumes that RGN_HPAGE != 0. */
	if ((REGION_NUMBER(addr) != RGN_HPAGE) || (addr & (HPAGE_SIZE - 1)))
		addr = HPAGE_REGION_BASE;
	else
		addr = ALIGN(addr, HPAGE_SIZE);
	for (vmm = find_vma(current->mm, addr); ; vmm = vmm->vm_next) {
		/* At this point:  (!vmm || addr < vmm->vm_end). */
		if (REGION_OFFSET(addr) + len > RGN_MAP_LIMIT)
			return -ENOMEM;
		if (!vmm || (addr + len) <= vmm->vm_start)
			return addr;
		addr = ALIGN(vmm->vm_end, HPAGE_SIZE);
	}
}

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 || (size & (size-1)) || !(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 1;
}
__setup("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>
	16	#include <linux/smp_lock.h>
	17	#include <linux/slab.h>
	18	#include <linux/sysctl.h>
	19	#include <asm/mman.h>
	20	#include <asm/pgalloc.h>
	21	#include <asm/tlb.h>
	22	#include <asm/tlbflush.h>
	23
	24	unsigned int hpage_shift=HPAGE_SHIFT_DEFAULT;
	25
63551ae0	26	pte_t *
1da177e4 LT	27	huge_pte_alloc (struct mm_struct *mm, unsigned long addr)
	28	{
	29	unsigned long taddr = htlbpage_to_page(addr);
	30	pgd_t *pgd;
	31	pud_t *pud;
	32	pmd_t *pmd;
	33	pte_t *pte = NULL;
	34
	35	pgd = pgd_offset(mm, taddr);
	36	pud = pud_alloc(mm, pgd, taddr);
	37	if (pud) {
	38	pmd = pmd_alloc(mm, pud, taddr);
	39	if (pmd)
	40	pte = pte_alloc_map(mm, pmd, taddr);
	41	}
	42	return pte;
	43	}
	44
63551ae0	45	pte_t *
1da177e4 LT	46	huge_pte_offset (struct mm_struct *mm, unsigned long addr)
	47	{
	48	unsigned long taddr = htlbpage_to_page(addr);
	49	pgd_t *pgd;
	50	pud_t *pud;
	51	pmd_t *pmd;
	52	pte_t *pte = NULL;
	53
	54	pgd = pgd_offset(mm, taddr);
	55	if (pgd_present(*pgd)) {
	56	pud = pud_offset(pgd, taddr);
	57	if (pud_present(*pud)) {
	58	pmd = pmd_offset(pud, taddr);
	59	if (pmd_present(*pmd))
	60	pte = pte_offset_map(pmd, taddr);
	61	}
	62	}
	63
	64	return pte;
	65	}
	66
39dde65c CK	67	int huge_pmd_unshare(struct mm_struct mm, unsigned long addr, pte_t *ptep)
	68	{
	69	return 0;
	70	}
	71
1da177e4 LT	72	#define mk_pte_huge(entry) { pte_val(entry) \|= _PAGE_P; }
1da177e4 LT	73
1da177e4	74	/*
42b88bef DG	75	* Don't actually need to do any preparation, but need to make sure
42b88bef DG	76	* the address is in the right region.
1da177e4	77	*/
68589bc3	78	int prepare_hugepage_range(unsigned long addr, unsigned long len, pgoff_t pgoff)
1da177e4	79	{
68589bc3 HD	80	if (pgoff & (~HPAGE_MASK >> PAGE_SHIFT))
68589bc3 HD	81	return -EINVAL;
1da177e4 LT	82	if (len & ~HPAGE_MASK)
	83	return -EINVAL;
	84	if (addr & ~HPAGE_MASK)
	85	return -EINVAL;
0a41e250	86	if (REGION_NUMBER(addr) != RGN_HPAGE)
1da177e4 LT	87	return -EINVAL;
	88
	89	return 0;
	90	}
	91
1da177e4 LT	92	struct page follow_huge_addr(struct mm_struct mm, unsigned long addr, int write)
	93	{
	94	struct page *page;
	95	pte_t *ptep;
	96
0a41e250	97	if (REGION_NUMBER(addr) != RGN_HPAGE)
1da177e4 LT	98	return ERR_PTR(-EINVAL);
	99
	100	ptep = huge_pte_offset(mm, addr);
	101	if (!ptep \|\| pte_none(*ptep))
	102	return NULL;
	103	page = pte_page(*ptep);
	104	page += ((addr & ~HPAGE_MASK) >> PAGE_SHIFT);
	105	return page;
	106	}
	107	int pmd_huge(pmd_t pmd)
	108	{
	109	return 0;
	110	}
	111	struct page *
	112	follow_huge_pmd(struct mm_struct mm, unsigned long address, pmd_t pmd, int write)
	113	{
	114	return NULL;
	115	}
	116
3bf5ee95 HD	117	void hugetlb_free_pgd_range(struct mmu_gather **tlb,
	118	unsigned long addr, unsigned long end,
	119	unsigned long floor, unsigned long ceiling)
1da177e4	120	{
3bf5ee95	121	/*
2332c9ae	122	* This is called to free hugetlb page tables.
3bf5ee95 HD	123	*
	124	* The offset of these addresses from the base of the hugetlb
	125	* region must be scaled down by HPAGE_SIZE/PAGE_SIZE so that
	126	* the standard free_pgd_range will free the right page tables.
	127	*
	128	* If floor and ceiling are also in the hugetlb region, they
	129	* must likewise be scaled down; but if outside, left unchanged.
	130	*/
	131
	132	addr = htlbpage_to_page(addr);
	133	end = htlbpage_to_page(end);
2332c9ae	134	if (REGION_NUMBER(floor) == RGN_HPAGE)
3bf5ee95	135	floor = htlbpage_to_page(floor);
2332c9ae	136	if (REGION_NUMBER(ceiling) == RGN_HPAGE)
3bf5ee95 HD	137	ceiling = htlbpage_to_page(ceiling);
	138
	139	free_pgd_range(tlb, addr, end, floor, ceiling);
1da177e4 LT	140	}
1da177e4 LT	141
1da177e4 LT	142	unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
	143	unsigned long pgoff, unsigned long flags)
	144	{
	145	struct vm_area_struct *vmm;
	146
	147	if (len > RGN_MAP_LIMIT)
	148	return -ENOMEM;
	149	if (len & ~HPAGE_MASK)
	150	return -EINVAL;
0a41e250 PC	151	/* This code assumes that RGN_HPAGE != 0. */
0a41e250 PC	152	if ((REGION_NUMBER(addr) != RGN_HPAGE) \|\| (addr & (HPAGE_SIZE - 1)))
1da177e4 LT	153	addr = HPAGE_REGION_BASE;
	154	else
	155	addr = ALIGN(addr, HPAGE_SIZE);
	156	for (vmm = find_vma(current->mm, addr); ; vmm = vmm->vm_next) {
	157	/* At this point: (!vmm \|\| addr < vmm->vm_end). */
	158	if (REGION_OFFSET(addr) + len > RGN_MAP_LIMIT)
	159	return -ENOMEM;
	160	if (!vmm \|\| (addr + len) <= vmm->vm_start)
	161	return addr;
	162	addr = ALIGN(vmm->vm_end, HPAGE_SIZE);
	163	}
	164	}
	165
	166	static int __init hugetlb_setup_sz(char *str)
	167	{
	168	u64 tr_pages;
	169	unsigned long long size;
	170
	171	if (ia64_pal_vm_page_size(&tr_pages, NULL) != 0)
	172	/*
	173	* shouldn't happen, but just in case.
	174	*/
	175	tr_pages = 0x15557000UL;
	176
	177	size = memparse(str, &str);
	178	if (*str \|\| (size & (size-1)) \|\| !(tr_pages & size) \|\|
	179	size <= PAGE_SIZE \|\|
	180	size >= (1UL << PAGE_SHIFT << MAX_ORDER)) {
	181	printk(KERN_WARNING "Invalid huge page size specified\n");
	182	return 1;
	183	}
	184
	185	hpage_shift = __ffs(size);
	186	/*
	187	* boot cpu already executed ia64_mmu_init, and has HPAGE_SHIFT_DEFAULT
	188	* override here with new page shift.
	189	*/
	190	ia64_set_rr(HPAGE_REGION_BASE, hpage_shift << 2);
	191	return 1;
	192	}
	193	__setup("hugepagesz=", hugetlb_setup_sz);