[net-next-2.6.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;

	BUG_ON(vma->vm_flags & VM_RESERVED);

	pgd = pgd_offset(mm, addr);
	spin_lock(&mm->page_table_lock);
	
	pud = pud_alloc(mm, pgd, addr);
	if (!pud)
		goto err_unlock;

	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		goto err_unlock;

	pte = pte_alloc_map(mm, pmd, addr);
	if (!pte)
		goto err_unlock;

	/*
	 * 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 err_unlock;
	err = -ENOMEM;
	if (page_mapcount(page) > INT_MAX/2)
		goto err_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;
	pte_unmap(pte);
	update_mmu_cache(vma, addr, pte_val);

	err = 0;
err_unlock:
	spin_unlock(&mm->page_table_lock);
	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;

	BUG_ON(vma->vm_flags & VM_RESERVED);

	pgd = pgd_offset(mm, addr);
	spin_lock(&mm->page_table_lock);
	
	pud = pud_alloc(mm, pgd, addr);
	if (!pud)
		goto err_unlock;

	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		goto err_unlock;

	pte = pte_alloc_map(mm, pmd, addr);
	if (!pte)
		goto err_unlock;

	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;
	pte_unmap(pte);
	update_mmu_cache(vma, addr, pte_val);
	spin_unlock(&mm->page_table_lock);
	return 0;

err_unlock:
	spin_unlock(&mm->page_table_lock);
	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;
	66
b5810039 NP	67	BUG_ON(vma->vm_flags & VM_RESERVED);
b5810039 NP	68
1da177e4 LT	69	pgd = pgd_offset(mm, addr);
	70	spin_lock(&mm->page_table_lock);
	71
	72	pud = pud_alloc(mm, pgd, addr);
	73	if (!pud)
	74	goto err_unlock;
	75
	76	pmd = pmd_alloc(mm, pud, addr);
	77	if (!pmd)
	78	goto err_unlock;
	79
	80	pte = pte_alloc_map(mm, pmd, addr);
	81	if (!pte)
	82	goto err_unlock;
	83
	84	/*
	85	* This page may have been truncated. Tell the
	86	* caller about it.
	87	*/
	88	err = -EINVAL;
	89	inode = vma->vm_file->f_mapping->host;
	90	size = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
	91	if (!page->mapping \|\| page->index >= size)
	92	goto err_unlock;
f5154a98 HD	93	err = -ENOMEM;
	94	if (page_mapcount(page) > INT_MAX/2)
	95	goto err_unlock;
1da177e4	96
861f2fb8 HD	97	if (pte_none(*pte) \|\| !zap_pte(mm, vma, addr, pte))
861f2fb8 HD	98	inc_mm_counter(mm, file_rss);
1da177e4	99
1da177e4 LT	100	flush_icache_page(vma, page);
	101	set_pte_at(mm, addr, pte, mk_pte(page, prot));
	102	page_add_file_rmap(page);
	103	pte_val = *pte;
	104	pte_unmap(pte);
	105	update_mmu_cache(vma, addr, pte_val);
	106
	107	err = 0;
	108	err_unlock:
	109	spin_unlock(&mm->page_table_lock);
	110	return err;
	111	}
	112	EXPORT_SYMBOL(install_page);
	113
	114
	115	/*
	116	* Install a file pte to a given virtual memory address, release any
	117	* previously existing mapping.
	118	*/
	119	int install_file_pte(struct mm_struct mm, struct vm_area_struct vma,
	120	unsigned long addr, unsigned long pgoff, pgprot_t prot)
	121	{
	122	int err = -ENOMEM;
	123	pte_t *pte;
	124	pmd_t *pmd;
	125	pud_t *pud;
	126	pgd_t *pgd;
	127	pte_t pte_val;
	128
b5810039 NP	129	BUG_ON(vma->vm_flags & VM_RESERVED);
b5810039 NP	130
1da177e4 LT	131	pgd = pgd_offset(mm, addr);
	132	spin_lock(&mm->page_table_lock);
	133
	134	pud = pud_alloc(mm, pgd, addr);
	135	if (!pud)
	136	goto err_unlock;
	137
	138	pmd = pmd_alloc(mm, pud, addr);
	139	if (!pmd)
	140	goto err_unlock;
	141
	142	pte = pte_alloc_map(mm, pmd, addr);
	143	if (!pte)
	144	goto err_unlock;
	145
365e9c87 HD	146	if (!pte_none(*pte) && zap_pte(mm, vma, addr, pte)) {
365e9c87 HD	147	update_hiwater_rss(mm);
861f2fb8	148	dec_mm_counter(mm, file_rss);
365e9c87	149	}
1da177e4 LT	150
	151	set_pte_at(mm, addr, pte, pgoff_to_pte(pgoff));
	152	pte_val = *pte;
	153	pte_unmap(pte);
	154	update_mmu_cache(vma, addr, pte_val);
	155	spin_unlock(&mm->page_table_lock);
	156	return 0;
	157
	158	err_unlock:
	159	spin_unlock(&mm->page_table_lock);
	160	return err;
	161	}
	162
	163
	164	/***
	165	* sys_remap_file_pages - remap arbitrary pages of a shared backing store
	166	* file within an existing vma.
	167	* @start: start of the remapped virtual memory range
	168	* @size: size of the remapped virtual memory range
	169	* @prot: new protection bits of the range
	170	* @pgoff: to be mapped page of the backing store file
	171	* @flags: 0 or MAP_NONBLOCKED - the later will cause no IO.
	172	*
	173	* this syscall works purely via pagetables, so it's the most efficient
	174	* way to map the same (large) file into a given virtual window. Unlike
	175	* mmap()/mremap() it does not create any new vmas. The new mappings are
	176	* also safe across swapout.
	177	*
	178	* NOTE: the 'prot' parameter right now is ignored, and the vma's default
	179	* protection is used. Arbitrary protections might be implemented in the
	180	* future.
	181	*/
	182	asmlinkage long sys_remap_file_pages(unsigned long start, unsigned long size,
	183	unsigned long __prot, unsigned long pgoff, unsigned long flags)
	184	{
	185	struct mm_struct *mm = current->mm;
	186	struct address_space *mapping;
	187	unsigned long end = start + size;
	188	struct vm_area_struct *vma;
	189	int err = -EINVAL;
	190	int has_write_lock = 0;
	191
	192	if (__prot)
	193	return err;
	194	/*
	195	* Sanitize the syscall parameters:
	196	*/
	197	start = start & PAGE_MASK;
	198	size = size & PAGE_MASK;
	199
	200	/* Does the address range wrap, or is the span zero-sized? */
	201	if (start + size <= start)
	202	return err;
	203
	204	/* Can we represent this offset inside this architecture's pte's? */
	205	#if PTE_FILE_MAX_BITS < BITS_PER_LONG
	206	if (pgoff + (size >> PAGE_SHIFT) >= (1UL << PTE_FILE_MAX_BITS))
	207	return err;
	208	#endif
	209
	210	/* We need down_write() to change vma->vm_flags. */
	211	down_read(&mm->mmap_sem);
	212	retry:
	213	vma = find_vma(mm, start);
214
215	/*
216	* Make sure the vma is shared, that it supports prefaulting,
217	* and that the remapped range is valid and fully within
218	* the single existing vma. vm_private_data is used as a
219	* swapout cursor in a VM_NONLINEAR vma (unless VM_RESERVED
220	* or VM_LOCKED, but VM_LOCKED could be revoked later on).
221	*/
222	if (vma && (vma->vm_flags & VM_SHARED) &&
223	(!vma->vm_private_data \|\|
224	(vma->vm_flags & (VM_NONLINEAR\|VM_RESERVED))) &&
225	vma->vm_ops && vma->vm_ops->populate &&
226	end > start && start >= vma->vm_start &&
227	end <= vma->vm_end) {
228
229	/* Must set VM_NONLINEAR before any pages are populated. */
230	if (pgoff != linear_page_index(vma, start) &&
231	!(vma->vm_flags & VM_NONLINEAR)) {
232	if (!has_write_lock) {
233	up_read(&mm->mmap_sem);
234	down_write(&mm->mmap_sem);
235	has_write_lock = 1;
236	goto retry;
237	}
238	mapping = vma->vm_file->f_mapping;
239	spin_lock(&mapping->i_mmap_lock);
240	flush_dcache_mmap_lock(mapping);
241	vma->vm_flags \|= VM_NONLINEAR;
242	vma_prio_tree_remove(vma, &mapping->i_mmap);
243	vma_nonlinear_insert(vma, &mapping->i_mmap_nonlinear);
244	flush_dcache_mmap_unlock(mapping);
245	spin_unlock(&mapping->i_mmap_lock);
246	}
247
248	err = vma->vm_ops->populate(vma, start, size,
249	vma->vm_page_prot,
250	pgoff, flags & MAP_NONBLOCK);
251
252	/*
253	* We can't clear VM_NONLINEAR because we'd have to do
254	* it after ->populate completes, and that would prevent
255	* downgrading the lock. (Locks can't be upgraded).
256	*/
257	}
258	if (likely(!has_write_lock))
259	up_read(&mm->mmap_sem);
260	else
261	up_write(&mm->mmap_sem);
262
263	return err;
264	}
265