[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 inline void zap_pte(struct mm_struct *mm, struct vm_area_struct *vma,
			unsigned long addr, pte_t *ptep)
{
	pte_t pte = *ptep;

	if (pte_none(pte))
		return;
	if (pte_present(pte)) {
		unsigned long pfn = pte_pfn(pte);

		flush_cache_page(vma, addr, pfn);
		pte = ptep_clear_flush(vma, addr, ptep);
		if (pfn_valid(pfn)) {
			struct page *page = pfn_to_page(pfn);
			if (!PageReserved(page)) {
				if (pte_dirty(pte))
					set_page_dirty(page);
				page_remove_rmap(page);
				page_cache_release(page);
				dec_mm_counter(mm, rss);
			}
		}
	} else {
		if (!pte_file(pte))
			free_swap_and_cache(pte_to_swp_entry(pte));
		pte_clear(mm, addr, ptep);
	}
}

/*
 * 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;

	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;

	zap_pte(mm, vma, addr, pte);

	inc_mm_counter(mm,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;

	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;

	zap_pte(mm, vma, addr, pte);

	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
	23	static inline void zap_pte(struct mm_struct mm, struct vm_area_struct vma,
	24	unsigned long addr, pte_t *ptep)
	25	{
	26	pte_t pte = *ptep;
	27
	28	if (pte_none(pte))
	29	return;
	30	if (pte_present(pte)) {
	31	unsigned long pfn = pte_pfn(pte);
	32
	33	flush_cache_page(vma, addr, pfn);
	34	pte = ptep_clear_flush(vma, addr, ptep);
	35	if (pfn_valid(pfn)) {
	36	struct page *page = pfn_to_page(pfn);
	37	if (!PageReserved(page)) {
	38	if (pte_dirty(pte))
	39	set_page_dirty(page);
	40	page_remove_rmap(page);
	41	page_cache_release(page);
	42	dec_mm_counter(mm, rss);
	43	}
	44	}
	45	} else {
	46	if (!pte_file(pte))
	47	free_swap_and_cache(pte_to_swp_entry(pte));
	48	pte_clear(mm, addr, ptep);
	49	}
	50	}
	51
	52	/*
	53	* Install a file page to a given virtual memory address, release any
	54	* previously existing mapping.
	55	*/
	56	int install_page(struct mm_struct mm, struct vm_area_struct vma,
	57	unsigned long addr, struct page *page, pgprot_t prot)
	58	{
	59	struct inode *inode;
	60	pgoff_t size;
	61	int err = -ENOMEM;
	62	pte_t *pte;
	63	pmd_t *pmd;
	64	pud_t *pud;
65	pgd_t *pgd;
66	pte_t pte_val;
67
68	pgd = pgd_offset(mm, addr);
69	spin_lock(&mm->page_table_lock);
70
71	pud = pud_alloc(mm, pgd, addr);
72	if (!pud)
73	goto err_unlock;
74
75	pmd = pmd_alloc(mm, pud, addr);
76	if (!pmd)
77	goto err_unlock;
78
79	pte = pte_alloc_map(mm, pmd, addr);
80	if (!pte)
81	goto err_unlock;
82
83	/*
84	* This page may have been truncated. Tell the
85	* caller about it.
86	*/
87	err = -EINVAL;
88	inode = vma->vm_file->f_mapping->host;
89	size = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
90	if (!page->mapping \|\| page->index >= size)
91	goto err_unlock;
f5154a98 HD	92	err = -ENOMEM;
	93	if (page_mapcount(page) > INT_MAX/2)
	94	goto err_unlock;
1da177e4 LT	95
	96	zap_pte(mm, vma, addr, pte);
	97
	98	inc_mm_counter(mm,rss);
	99	flush_icache_page(vma, page);
	100	set_pte_at(mm, addr, pte, mk_pte(page, prot));
	101	page_add_file_rmap(page);
	102	pte_val = *pte;
	103	pte_unmap(pte);
	104	update_mmu_cache(vma, addr, pte_val);
	105
	106	err = 0;
	107	err_unlock:
	108	spin_unlock(&mm->page_table_lock);
	109	return err;
	110	}
	111	EXPORT_SYMBOL(install_page);
	112
	113
	114	/*
	115	* Install a file pte to a given virtual memory address, release any
	116	* previously existing mapping.
	117	*/
	118	int install_file_pte(struct mm_struct mm, struct vm_area_struct vma,
	119	unsigned long addr, unsigned long pgoff, pgprot_t prot)
	120	{
	121	int err = -ENOMEM;
	122	pte_t *pte;
	123	pmd_t *pmd;
	124	pud_t *pud;
	125	pgd_t *pgd;
	126	pte_t pte_val;
	127
	128	pgd = pgd_offset(mm, addr);
	129	spin_lock(&mm->page_table_lock);
	130
	131	pud = pud_alloc(mm, pgd, addr);
	132	if (!pud)
	133	goto err_unlock;
	134
	135	pmd = pmd_alloc(mm, pud, addr);
	136	if (!pmd)
	137	goto err_unlock;
	138
	139	pte = pte_alloc_map(mm, pmd, addr);
	140	if (!pte)
	141	goto err_unlock;
	142
	143	zap_pte(mm, vma, addr, pte);
	144
	145	set_pte_at(mm, addr, pte, pgoff_to_pte(pgoff));
	146	pte_val = *pte;
	147	pte_unmap(pte);
	148	update_mmu_cache(vma, addr, pte_val);
	149	spin_unlock(&mm->page_table_lock);
	150	return 0;
	151
	152	err_unlock:
	153	spin_unlock(&mm->page_table_lock);
	154	return err;
	155	}
	156
	157
	158	/***
159	* sys_remap_file_pages - remap arbitrary pages of a shared backing store
160	* file within an existing vma.
161	* @start: start of the remapped virtual memory range
162	* @size: size of the remapped virtual memory range
163	* @prot: new protection bits of the range
164	* @pgoff: to be mapped page of the backing store file
165	* @flags: 0 or MAP_NONBLOCKED - the later will cause no IO.
166	*
167	* this syscall works purely via pagetables, so it's the most efficient
168	* way to map the same (large) file into a given virtual window. Unlike
169	* mmap()/mremap() it does not create any new vmas. The new mappings are
170	* also safe across swapout.
171	*
172	* NOTE: the 'prot' parameter right now is ignored, and the vma's default
173	* protection is used. Arbitrary protections might be implemented in the
174	* future.
175	*/
176	asmlinkage long sys_remap_file_pages(unsigned long start, unsigned long size,
177	unsigned long __prot, unsigned long pgoff, unsigned long flags)
178	{
179	struct mm_struct *mm = current->mm;
180	struct address_space *mapping;
181	unsigned long end = start + size;
182	struct vm_area_struct *vma;
183	int err = -EINVAL;
184	int has_write_lock = 0;
185
186	if (__prot)
187	return err;
188	/*
189	* Sanitize the syscall parameters:
190	*/
191	start = start & PAGE_MASK;
192	size = size & PAGE_MASK;
193
194	/* Does the address range wrap, or is the span zero-sized? */
195	if (start + size <= start)
196	return err;
197
198	/* Can we represent this offset inside this architecture's pte's? */
199	#if PTE_FILE_MAX_BITS < BITS_PER_LONG
200	if (pgoff + (size >> PAGE_SHIFT) >= (1UL << PTE_FILE_MAX_BITS))
201	return err;
202	#endif
203
204	/* We need down_write() to change vma->vm_flags. */
205	down_read(&mm->mmap_sem);
206	retry:
207	vma = find_vma(mm, start);
208
209	/*
210	* Make sure the vma is shared, that it supports prefaulting,
211	* and that the remapped range is valid and fully within
212	* the single existing vma. vm_private_data is used as a
213	* swapout cursor in a VM_NONLINEAR vma (unless VM_RESERVED
214	* or VM_LOCKED, but VM_LOCKED could be revoked later on).
215	*/
216	if (vma && (vma->vm_flags & VM_SHARED) &&
217	(!vma->vm_private_data \|\|
218	(vma->vm_flags & (VM_NONLINEAR\|VM_RESERVED))) &&
219	vma->vm_ops && vma->vm_ops->populate &&
220	end > start && start >= vma->vm_start &&
221	end <= vma->vm_end) {
222
223	/* Must set VM_NONLINEAR before any pages are populated. */
224	if (pgoff != linear_page_index(vma, start) &&
225	!(vma->vm_flags & VM_NONLINEAR)) {
226	if (!has_write_lock) {
227	up_read(&mm->mmap_sem);
228	down_write(&mm->mmap_sem);
229	has_write_lock = 1;
230	goto retry;
231	}
232	mapping = vma->vm_file->f_mapping;
233	spin_lock(&mapping->i_mmap_lock);
234	flush_dcache_mmap_lock(mapping);
235	vma->vm_flags \|= VM_NONLINEAR;
236	vma_prio_tree_remove(vma, &mapping->i_mmap);
237	vma_nonlinear_insert(vma, &mapping->i_mmap_nonlinear);
238	flush_dcache_mmap_unlock(mapping);
239	spin_unlock(&mapping->i_mmap_lock);
240	}
241
242	err = vma->vm_ops->populate(vma, start, size,
243	vma->vm_page_prot,
244	pgoff, flags & MAP_NONBLOCK);
245
246	/*
247	* We can't clear VM_NONLINEAR because we'd have to do
248	* it after ->populate completes, and that would prevent
249	* downgrading the lock. (Locks can't be upgraded).
250	*/
251	}
252	if (likely(!has_write_lock))
253	up_read(&mm->mmap_sem);
254	else
255	up_write(&mm->mmap_sem);
256
257	return err;
258	}
259