[linux.git] / mm / fremap.c

/*
 *   linux/mm/fremap.c
 * 
 * Explicit pagetable population and nonlinear (random) mappings support.
 *
 * started by Ingo Molnar, Copyright (C) 2002, 2003
 */

#include <linux/mm.h>
#include <linux/swap.h>
#include <linux/file.h>
#include <linux/mman.h>
#include <linux/pagemap.h>
#include <linux/swapops.h>
#include <linux/rmap.h>
#include <linux/module.h>
#include <linux/syscalls.h>

#include <asm/mmu_context.h>
#include <asm/cacheflush.h>
#include <asm/tlbflush.h>

static int zap_pte(struct mm_struct *mm, struct vm_area_struct *vma,
			unsigned long addr, pte_t *ptep)
{
	pte_t pte = *ptep;
	struct page *page = NULL;

	if (pte_present(pte)) {
		unsigned long pfn = pte_pfn(pte);
		flush_cache_page(vma, addr, pfn);
		pte = ptep_clear_flush(vma, addr, ptep);
		if (unlikely(!pfn_valid(pfn))) {
			print_bad_pte(vma, pte, addr);
			goto out;
		}
		page = pfn_to_page(pfn);
		if (pte_dirty(pte))
			set_page_dirty(page);
		page_remove_rmap(page);
		page_cache_release(page);
	} else {
		if (!pte_file(pte))
			free_swap_and_cache(pte_to_swp_entry(pte));
		pte_clear(mm, addr, ptep);
	}
out:
	return !!page;
}

/*
 * Install a file page to a given virtual memory address, release any
 * previously existing mapping.
 */
int install_page(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long addr, struct page *page, pgprot_t prot)
{
	struct inode *inode;
	pgoff_t size;
	int err = -ENOMEM;
	pte_t *pte;
	pmd_t *pmd;
	pud_t *pud;
	pgd_t *pgd;
	pte_t pte_val;
	spinlock_t *ptl;

	BUG_ON(vma->vm_flags & VM_RESERVED);

	pgd = pgd_offset(mm, addr);
	pud = pud_alloc(mm, pgd, addr);
	if (!pud)
		goto out;
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		goto out;
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
	if (!pte)
		goto out;

	/*
	 * This page may have been truncated. Tell the
	 * caller about it.
	 */
	err = -EINVAL;
	inode = vma->vm_file->f_mapping->host;
	size = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
	if (!page->mapping || page->index >= size)
		goto unlock;
	err = -ENOMEM;
	if (page_mapcount(page) > INT_MAX/2)
		goto unlock;

	if (pte_none(*pte) || !zap_pte(mm, vma, addr, pte))
		inc_mm_counter(mm, file_rss);

	flush_icache_page(vma, page);
	set_pte_at(mm, addr, pte, mk_pte(page, prot));
	page_add_file_rmap(page);
	pte_val = *pte;
	update_mmu_cache(vma, addr, pte_val);
	err = 0;
unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return err;
}
EXPORT_SYMBOL(install_page);

/*
 * Install a file pte to a given virtual memory address, release any
 * previously existing mapping.
 */
int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long addr, unsigned long pgoff, pgprot_t prot)
{
	int err = -ENOMEM;
	pte_t *pte;
	pmd_t *pmd;
	pud_t *pud;
	pgd_t *pgd;
	pte_t pte_val;
	spinlock_t *ptl;

	BUG_ON(vma->vm_flags & VM_RESERVED);

	pgd = pgd_offset(mm, addr);
	pud = pud_alloc(mm, pgd, addr);
	if (!pud)
		goto out;
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		goto out;
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
	if (!pte)
		goto out;

	if (!pte_none(*pte) && zap_pte(mm, vma, addr, pte)) {
		update_hiwater_rss(mm);
		dec_mm_counter(mm, file_rss);
	}

	set_pte_at(mm, addr, pte, pgoff_to_pte(pgoff));
	pte_val = *pte;
	update_mmu_cache(vma, addr, pte_val);
	pte_unmap_unlock(pte, ptl);
	err = 0;
out:
	return err;
}

/***
 * sys_remap_file_pages - remap arbitrary pages of a shared backing store
 *                        file within an existing vma.
 * @start: start of the remapped virtual memory range
 * @size: size of the remapped virtual memory range
 * @prot: new protection bits of the range
 * @pgoff: to be mapped page of the backing store file
 * @flags: 0 or MAP_NONBLOCKED - the later will cause no IO.
 *
 * this syscall works purely via pagetables, so it's the most efficient
 * way to map the same (large) file into a given virtual window. Unlike
 * mmap()/mremap() it does not create any new vmas. The new mappings are
 * also safe across swapout.
 *
 * NOTE: the 'prot' parameter right now is ignored, and the vma's default
 * protection is used. Arbitrary protections might be implemented in the
 * future.
 */
asmlinkage long sys_remap_file_pages(unsigned long start, unsigned long size,
	unsigned long __prot, unsigned long pgoff, unsigned long flags)
{
	struct mm_struct *mm = current->mm;
	struct address_space *mapping;
	unsigned long end = start + size;
	struct vm_area_struct *vma;
	int err = -EINVAL;
	int has_write_lock = 0;

	if (__prot)
		return err;
	/*
	 * Sanitize the syscall parameters:
	 */
	start = start & PAGE_MASK;
	size = size & PAGE_MASK;

	/* Does the address range wrap, or is the span zero-sized? */
	if (start + size <= start)
		return err;

	/* Can we represent this offset inside this architecture's pte's? */
#if PTE_FILE_MAX_BITS < BITS_PER_LONG
	if (pgoff + (size >> PAGE_SHIFT) >= (1UL << PTE_FILE_MAX_BITS))
		return err;
#endif

	/* We need down_write() to change vma->vm_flags. */
	down_read(&mm->mmap_sem);
 retry:
	vma = find_vma(mm, start);

	/*
	 * Make sure the vma is shared, that it supports prefaulting,
	 * and that the remapped range is valid and fully within
	 * the single existing vma.  vm_private_data is used as a
	 * swapout cursor in a VM_NONLINEAR vma (unless VM_RESERVED
	 * or VM_LOCKED, but VM_LOCKED could be revoked later on).
	 */
	if (vma && (vma->vm_flags & VM_SHARED) &&
		(!vma->vm_private_data ||
			(vma->vm_flags & (VM_NONLINEAR|VM_RESERVED))) &&
		vma->vm_ops && vma->vm_ops->populate &&
			end > start && start >= vma->vm_start &&
				end <= vma->vm_end) {

		/* Must set VM_NONLINEAR before any pages are populated. */
		if (pgoff != linear_page_index(vma, start) &&
		    !(vma->vm_flags & VM_NONLINEAR)) {
			if (!has_write_lock) {
				up_read(&mm->mmap_sem);
				down_write(&mm->mmap_sem);
				has_write_lock = 1;
				goto retry;
			}
			mapping = vma->vm_file->f_mapping;
			spin_lock(&mapping->i_mmap_lock);
			flush_dcache_mmap_lock(mapping);
			vma->vm_flags |= VM_NONLINEAR;
			vma_prio_tree_remove(vma, &mapping->i_mmap);
			vma_nonlinear_insert(vma, &mapping->i_mmap_nonlinear);
			flush_dcache_mmap_unlock(mapping);
			spin_unlock(&mapping->i_mmap_lock);
		}

		err = vma->vm_ops->populate(vma, start, size,
					    vma->vm_page_prot,
					    pgoff, flags & MAP_NONBLOCK);

		/*
		 * We can't clear VM_NONLINEAR because we'd have to do
		 * it after ->populate completes, and that would prevent
		 * downgrading the lock.  (Locks can't be upgraded).
		 */
	}
	if (likely(!has_write_lock))
		up_read(&mm->mmap_sem);
	else
		up_write(&mm->mmap_sem);

	return err;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/mm/fremap.c
	3	*
	4	* Explicit pagetable population and nonlinear (random) mappings support.
	5	*
	6	* started by Ingo Molnar, Copyright (C) 2002, 2003
	7	*/
	8
	9	#include <linux/mm.h>
	10	#include <linux/swap.h>
	11	#include <linux/file.h>
	12	#include <linux/mman.h>
	13	#include <linux/pagemap.h>
	14	#include <linux/swapops.h>
	15	#include <linux/rmap.h>
	16	#include <linux/module.h>
	17	#include <linux/syscalls.h>
	18
	19	#include <asm/mmu_context.h>
	20	#include <asm/cacheflush.h>
	21	#include <asm/tlbflush.h>
	22
861f2fb8	23	static int zap_pte(struct mm_struct mm, struct vm_area_struct vma,
1da177e4 LT	24	unsigned long addr, pte_t *ptep)
	25	{
	26	pte_t pte = *ptep;
861f2fb8	27	struct page *page = NULL;
1da177e4	28
1da177e4 LT	29	if (pte_present(pte)) {
1da177e4 LT	30	unsigned long pfn = pte_pfn(pte);
1da177e4 LT	31	flush_cache_page(vma, addr, pfn);
1da177e4 LT	32	pte = ptep_clear_flush(vma, addr, ptep);
b5810039 NP	33	if (unlikely(!pfn_valid(pfn))) {
b5810039 NP	34	print_bad_pte(vma, pte, addr);
861f2fb8	35	goto out;
1da177e4	36	}
b5810039 NP	37	page = pfn_to_page(pfn);
	38	if (pte_dirty(pte))
	39	set_page_dirty(page);
	40	page_remove_rmap(page);
	41	page_cache_release(page);
1da177e4 LT	42	} else {
	43	if (!pte_file(pte))
	44	free_swap_and_cache(pte_to_swp_entry(pte));
	45	pte_clear(mm, addr, ptep);
	46	}
861f2fb8 HD	47	out:
861f2fb8 HD	48	return !!page;
1da177e4 LT	49	}
	50
	51	/*
	52	* Install a file page to a given virtual memory address, release any
	53	* previously existing mapping.
	54	*/
	55	int install_page(struct mm_struct mm, struct vm_area_struct vma,
	56	unsigned long addr, struct page *page, pgprot_t prot)
	57	{
	58	struct inode *inode;
	59	pgoff_t size;
	60	int err = -ENOMEM;
	61	pte_t *pte;
	62	pmd_t *pmd;
	63	pud_t *pud;
	64	pgd_t *pgd;
	65	pte_t pte_val;
c74df32c	66	spinlock_t *ptl;
1da177e4	67
b5810039 NP	68	BUG_ON(vma->vm_flags & VM_RESERVED);
b5810039 NP	69
1da177e4	70	pgd = pgd_offset(mm, addr);
1da177e4 LT	71	pud = pud_alloc(mm, pgd, addr);
1da177e4 LT	72	if (!pud)
c74df32c	73	goto out;
1da177e4 LT	74	pmd = pmd_alloc(mm, pud, addr);
1da177e4 LT	75	if (!pmd)
c74df32c HD	76	goto out;
c74df32c HD	77	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
1da177e4	78	if (!pte)
c74df32c	79	goto out;
1da177e4 LT	80
	81	/*
	82	* This page may have been truncated. Tell the
	83	* caller about it.
	84	*/
	85	err = -EINVAL;
	86	inode = vma->vm_file->f_mapping->host;
	87	size = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
	88	if (!page->mapping \|\| page->index >= size)
c74df32c	89	goto unlock;
f5154a98 HD	90	err = -ENOMEM;
f5154a98 HD	91	if (page_mapcount(page) > INT_MAX/2)
c74df32c	92	goto unlock;
1da177e4	93
861f2fb8 HD	94	if (pte_none(*pte) \|\| !zap_pte(mm, vma, addr, pte))
861f2fb8 HD	95	inc_mm_counter(mm, file_rss);
1da177e4	96
1da177e4 LT	97	flush_icache_page(vma, page);
	98	set_pte_at(mm, addr, pte, mk_pte(page, prot));
	99	page_add_file_rmap(page);
	100	pte_val = *pte;
1da177e4	101	update_mmu_cache(vma, addr, pte_val);
1da177e4	102	err = 0;
c74df32c HD	103	unlock:
	104	pte_unmap_unlock(pte, ptl);
	105	out:
1da177e4 LT	106	return err;
	107	}
	108	EXPORT_SYMBOL(install_page);
	109
1da177e4 LT	110	/*
	111	* Install a file pte to a given virtual memory address, release any
	112	* previously existing mapping.
	113	*/
	114	int install_file_pte(struct mm_struct mm, struct vm_area_struct vma,
	115	unsigned long addr, unsigned long pgoff, pgprot_t prot)
	116	{
	117	int err = -ENOMEM;
	118	pte_t *pte;
	119	pmd_t *pmd;
	120	pud_t *pud;
	121	pgd_t *pgd;
	122	pte_t pte_val;
c74df32c	123	spinlock_t *ptl;
1da177e4	124
b5810039 NP	125	BUG_ON(vma->vm_flags & VM_RESERVED);
b5810039 NP	126
1da177e4	127	pgd = pgd_offset(mm, addr);
1da177e4 LT	128	pud = pud_alloc(mm, pgd, addr);
1da177e4 LT	129	if (!pud)
c74df32c	130	goto out;
1da177e4 LT	131	pmd = pmd_alloc(mm, pud, addr);
1da177e4 LT	132	if (!pmd)
c74df32c HD	133	goto out;
c74df32c HD	134	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
1da177e4	135	if (!pte)
c74df32c	136	goto out;
1da177e4	137
365e9c87 HD	138	if (!pte_none(*pte) && zap_pte(mm, vma, addr, pte)) {
365e9c87 HD	139	update_hiwater_rss(mm);
861f2fb8	140	dec_mm_counter(mm, file_rss);
365e9c87	141	}
1da177e4 LT	142
	143	set_pte_at(mm, addr, pte, pgoff_to_pte(pgoff));
	144	pte_val = *pte;
1da177e4	145	update_mmu_cache(vma, addr, pte_val);
c74df32c HD	146	pte_unmap_unlock(pte, ptl);
	147	err = 0;
	148	out:
1da177e4 LT	149	return err;
	150	}
	151
1da177e4 LT	152	/***
	153	* sys_remap_file_pages - remap arbitrary pages of a shared backing store
	154	* file within an existing vma.
	155	* @start: start of the remapped virtual memory range
	156	* @size: size of the remapped virtual memory range
	157	* @prot: new protection bits of the range
	158	* @pgoff: to be mapped page of the backing store file
	159	* @flags: 0 or MAP_NONBLOCKED - the later will cause no IO.
	160	*
	161	* this syscall works purely via pagetables, so it's the most efficient
	162	* way to map the same (large) file into a given virtual window. Unlike
	163	* mmap()/mremap() it does not create any new vmas. The new mappings are
	164	* also safe across swapout.
	165	*
	166	* NOTE: the 'prot' parameter right now is ignored, and the vma's default
	167	* protection is used. Arbitrary protections might be implemented in the
	168	* future.
	169	*/
	170	asmlinkage long sys_remap_file_pages(unsigned long start, unsigned long size,
	171	unsigned long __prot, unsigned long pgoff, unsigned long flags)
	172	{
	173	struct mm_struct *mm = current->mm;
	174	struct address_space *mapping;
	175	unsigned long end = start + size;
	176	struct vm_area_struct *vma;
	177	int err = -EINVAL;
	178	int has_write_lock = 0;
	179
	180	if (__prot)
	181	return err;
	182	/*
	183	* Sanitize the syscall parameters:
	184	*/
	185	start = start & PAGE_MASK;
	186	size = size & PAGE_MASK;
	187
	188	/* Does the address range wrap, or is the span zero-sized? */
	189	if (start + size <= start)
	190	return err;
	191
	192	/* Can we represent this offset inside this architecture's pte's? */
	193	#if PTE_FILE_MAX_BITS < BITS_PER_LONG
	194	if (pgoff + (size >> PAGE_SHIFT) >= (1UL << PTE_FILE_MAX_BITS))
	195	return err;
	196	#endif
	197
	198	/* We need down_write() to change vma->vm_flags. */
	199	down_read(&mm->mmap_sem);
	200	retry:
	201	vma = find_vma(mm, start);
	202
	203	/*
	204	* Make sure the vma is shared, that it supports prefaulting,
	205	* and that the remapped range is valid and fully within
	206	* the single existing vma. vm_private_data is used as a
	207	* swapout cursor in a VM_NONLINEAR vma (unless VM_RESERVED
	208	* or VM_LOCKED, but VM_LOCKED could be revoked later on).
	209	*/
	210	if (vma && (vma->vm_flags & VM_SHARED) &&
	211	(!vma->vm_private_data \|\|
	212	(vma->vm_flags & (VM_NONLINEAR\|VM_RESERVED))) &&
	213	vma->vm_ops && vma->vm_ops->populate &&
	214	end > start && start >= vma->vm_start &&
	215	end <= vma->vm_end) {
216
217	/* Must set VM_NONLINEAR before any pages are populated. */
218	if (pgoff != linear_page_index(vma, start) &&
219	!(vma->vm_flags & VM_NONLINEAR)) {
220	if (!has_write_lock) {
221	up_read(&mm->mmap_sem);
222	down_write(&mm->mmap_sem);
223	has_write_lock = 1;
224	goto retry;
225	}
226	mapping = vma->vm_file->f_mapping;
227	spin_lock(&mapping->i_mmap_lock);
228	flush_dcache_mmap_lock(mapping);
229	vma->vm_flags \|= VM_NONLINEAR;
230	vma_prio_tree_remove(vma, &mapping->i_mmap);
231	vma_nonlinear_insert(vma, &mapping->i_mmap_nonlinear);
232	flush_dcache_mmap_unlock(mapping);
233	spin_unlock(&mapping->i_mmap_lock);
234	}
235
236	err = vma->vm_ops->populate(vma, start, size,
237	vma->vm_page_prot,
238	pgoff, flags & MAP_NONBLOCK);
239
240	/*
241	* We can't clear VM_NONLINEAR because we'd have to do
242	* it after ->populate completes, and that would prevent
243	* downgrading the lock. (Locks can't be upgraded).
244	*/
245	}
246	if (likely(!has_write_lock))
247	up_read(&mm->mmap_sem);
248	else
249	up_write(&mm->mmap_sem);
250
251	return err;
252	}
253