[net-next-2.6.git] / mm / msync.c

/*
 *	linux/mm/msync.c
 *
 * Copyright (C) 1994-1999  Linus Torvalds
 */

/*
 * The msync() system call.
 */
#include <linux/slab.h>
#include <linux/pagemap.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/hugetlb.h>
#include <linux/syscalls.h>

#include <asm/pgtable.h>
#include <asm/tlbflush.h>

static void msync_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
				unsigned long addr, unsigned long end)
{
	pte_t *pte;
	spinlock_t *ptl;
	int progress = 0;

again:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	do {
		struct page *page;

		if (progress >= 64) {
			progress = 0;
			if (need_resched() || need_lockbreak(ptl))
				break;
		}
		progress++;
		if (!pte_present(*pte))
			continue;
		if (!pte_maybe_dirty(*pte))
			continue;
		page = vm_normal_page(vma, addr, *pte);
		if (!page)
			continue;
		if (ptep_clear_flush_dirty(vma, addr, pte) ||
		    page_test_and_clear_dirty(page))
			set_page_dirty(page);
		progress += 3;
	} while (pte++, addr += PAGE_SIZE, addr != end);
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();
	if (addr != end)
		goto again;
}

static inline void msync_pmd_range(struct vm_area_struct *vma, pud_t *pud,
				unsigned long addr, unsigned long end)
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
		msync_pte_range(vma, pmd, addr, next);
	} while (pmd++, addr = next, addr != end);
}

static inline void msync_pud_range(struct vm_area_struct *vma, pgd_t *pgd,
				unsigned long addr, unsigned long end)
{
	pud_t *pud;
	unsigned long next;

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
		msync_pmd_range(vma, pud, addr, next);
	} while (pud++, addr = next, addr != end);
}

static void msync_page_range(struct vm_area_struct *vma,
				unsigned long addr, unsigned long end)
{
	pgd_t *pgd;
	unsigned long next;

	/* For hugepages we can't go walking the page table normally,
	 * but that's ok, hugetlbfs is memory based, so we don't need
	 * to do anything more on an msync().
	 */
	if (vma->vm_flags & VM_HUGETLB)
		return;

	BUG_ON(addr >= end);
	pgd = pgd_offset(vma->vm_mm, addr);
	flush_cache_range(vma, addr, end);
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
		msync_pud_range(vma, pgd, addr, next);
	} while (pgd++, addr = next, addr != end);
}

/*
 * MS_SYNC syncs the entire file - including mappings.
 *
 * MS_ASYNC does not start I/O (it used to, up to 2.5.67).  Instead, it just
 * marks the relevant pages dirty.  The application may now run fsync() to
 * write out the dirty pages and wait on the writeout and check the result.
 * Or the application may run fadvise(FADV_DONTNEED) against the fd to start
 * async writeout immediately.
 * So my _not_ starting I/O in MS_ASYNC we provide complete flexibility to
 * applications.
 */
static int msync_interval(struct vm_area_struct *vma,
			unsigned long addr, unsigned long end, int flags)
{
	int ret = 0;
	struct file *file = vma->vm_file;

	if ((flags & MS_INVALIDATE) && (vma->vm_flags & VM_LOCKED))
		return -EBUSY;

	if (file && (vma->vm_flags & VM_SHARED)) {
		msync_page_range(vma, addr, end);

		if (flags & MS_SYNC) {
			struct address_space *mapping = file->f_mapping;
			int err;

			ret = filemap_fdatawrite(mapping);
			if (file->f_op && file->f_op->fsync) {
				/*
				 * We don't take i_mutex here because mmap_sem
				 * is already held.
				 */
				err = file->f_op->fsync(file,file->f_dentry,1);
				if (err && !ret)
					ret = err;
			}
			err = filemap_fdatawait(mapping);
			if (!ret)
				ret = err;
		}
	}
	return ret;
}

asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
{
	unsigned long end;
	struct vm_area_struct *vma;
	int unmapped_error, error = -EINVAL;

	if (flags & MS_SYNC)
		current->flags |= PF_SYNCWRITE;

	down_read(&current->mm->mmap_sem);
	if (flags & ~(MS_ASYNC | MS_INVALIDATE | MS_SYNC))
		goto out;
	if (start & ~PAGE_MASK)
		goto out;
	if ((flags & MS_ASYNC) && (flags & MS_SYNC))
		goto out;
	error = -ENOMEM;
	len = (len + ~PAGE_MASK) & PAGE_MASK;
	end = start + len;
	if (end < start)
		goto out;
	error = 0;
	if (end == start)
		goto out;
	/*
	 * If the interval [start,end) covers some unmapped address ranges,
	 * just ignore them, but return -ENOMEM at the end.
	 */
	vma = find_vma(current->mm, start);
	unmapped_error = 0;
	for (;;) {
		/* Still start < end. */
		error = -ENOMEM;
		if (!vma)
			goto out;
		/* Here start < vma->vm_end. */
		if (start < vma->vm_start) {
			unmapped_error = -ENOMEM;
			start = vma->vm_start;
		}
		/* Here vma->vm_start <= start < vma->vm_end. */
		if (end <= vma->vm_end) {
			if (start < end) {
				error = msync_interval(vma, start, end, flags);
				if (error)
					goto out;
			}
			error = unmapped_error;
			goto out;
		}
		/* Here vma->vm_start <= start < vma->vm_end < end. */
		error = msync_interval(vma, start, vma->vm_end, flags);
		if (error)
			goto out;
		start = vma->vm_end;
		vma = vma->vm_next;
	}
out:
	up_read(&current->mm->mmap_sem);
	current->flags &= ~PF_SYNCWRITE;
	return error;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/mm/msync.c
	3	*
	4	* Copyright (C) 1994-1999 Linus Torvalds
	5	*/
	6
	7	/*
	8	* The msync() system call.
	9	*/
	10	#include <linux/slab.h>
	11	#include <linux/pagemap.h>
	12	#include <linux/mm.h>
	13	#include <linux/mman.h>
	14	#include <linux/hugetlb.h>
	15	#include <linux/syscalls.h>
	16
	17	#include <asm/pgtable.h>
	18	#include <asm/tlbflush.h>
	19
b57b98d1	20	static void msync_pte_range(struct vm_area_struct vma, pmd_t pmd,
1da177e4 LT	21	unsigned long addr, unsigned long end)
	22	{
	23	pte_t *pte;
705e87c0	24	spinlock_t *ptl;
0c942a45	25	int progress = 0;
1da177e4	26
0c942a45	27	again:
705e87c0	28	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
1da177e4	29	do {
1da177e4 LT	30	struct page *page;
1da177e4 LT	31
0c942a45 HD	32	if (progress >= 64) {
0c942a45 HD	33	progress = 0;
705e87c0	34	if (need_resched() \|\| need_lockbreak(ptl))
0c942a45 HD	35	break;
	36	}
	37	progress++;
1da177e4 LT	38	if (!pte_present(*pte))
1da177e4 LT	39	continue;
b4955ce3 AK	40	if (!pte_maybe_dirty(*pte))
b4955ce3 AK	41	continue;
6aab341e LT	42	page = vm_normal_page(vma, addr, *pte);
6aab341e LT	43	if (!page)
1da177e4	44	continue;
1da177e4 LT	45	if (ptep_clear_flush_dirty(vma, addr, pte) \|\|
	46	page_test_and_clear_dirty(page))
	47	set_page_dirty(page);
0c942a45	48	progress += 3;
1da177e4	49	} while (pte++, addr += PAGE_SIZE, addr != end);
705e87c0 HD	50	pte_unmap_unlock(pte - 1, ptl);
705e87c0 HD	51	cond_resched();
0c942a45 HD	52	if (addr != end)
0c942a45 HD	53	goto again;
1da177e4 LT	54	}
1da177e4 LT	55
b57b98d1	56	static inline void msync_pmd_range(struct vm_area_struct vma, pud_t pud,
1da177e4 LT	57	unsigned long addr, unsigned long end)
	58	{
	59	pmd_t *pmd;
	60	unsigned long next;
	61
	62	pmd = pmd_offset(pud, addr);
	63	do {
	64	next = pmd_addr_end(addr, end);
	65	if (pmd_none_or_clear_bad(pmd))
	66	continue;
b57b98d1	67	msync_pte_range(vma, pmd, addr, next);
1da177e4 LT	68	} while (pmd++, addr = next, addr != end);
	69	}
	70
b57b98d1	71	static inline void msync_pud_range(struct vm_area_struct vma, pgd_t pgd,
1da177e4 LT	72	unsigned long addr, unsigned long end)
	73	{
	74	pud_t *pud;
	75	unsigned long next;
	76
	77	pud = pud_offset(pgd, addr);
	78	do {
	79	next = pud_addr_end(addr, end);
	80	if (pud_none_or_clear_bad(pud))
	81	continue;
b57b98d1	82	msync_pmd_range(vma, pud, addr, next);
1da177e4 LT	83	} while (pud++, addr = next, addr != end);
	84	}
	85
b57b98d1	86	static void msync_page_range(struct vm_area_struct *vma,
1da177e4 LT	87	unsigned long addr, unsigned long end)
1da177e4 LT	88	{
1da177e4 LT	89	pgd_t *pgd;
	90	unsigned long next;
	91
	92	/* For hugepages we can't go walking the page table normally,
	93	* but that's ok, hugetlbfs is memory based, so we don't need
b5810039	94	* to do anything more on an msync().
b5810039	95	*/
6aab341e	96	if (vma->vm_flags & VM_HUGETLB)
1da177e4 LT	97	return;
	98
	99	BUG_ON(addr >= end);
705e87c0	100	pgd = pgd_offset(vma->vm_mm, addr);
1da177e4	101	flush_cache_range(vma, addr, end);
1da177e4 LT	102	do {
	103	next = pgd_addr_end(addr, end);
	104	if (pgd_none_or_clear_bad(pgd))
	105	continue;
b57b98d1	106	msync_pud_range(vma, pgd, addr, next);
1da177e4	107	} while (pgd++, addr = next, addr != end);
1da177e4 LT	108	}
1da177e4 LT	109
1da177e4 LT	110	/*
	111	* MS_SYNC syncs the entire file - including mappings.
	112	*
	113	* MS_ASYNC does not start I/O (it used to, up to 2.5.67). Instead, it just
	114	* marks the relevant pages dirty. The application may now run fsync() to
	115	* write out the dirty pages and wait on the writeout and check the result.
	116	* Or the application may run fadvise(FADV_DONTNEED) against the fd to start
	117	* async writeout immediately.
	118	* So my _not_ starting I/O in MS_ASYNC we provide complete flexibility to
	119	* applications.
	120	*/
	121	static int msync_interval(struct vm_area_struct *vma,
	122	unsigned long addr, unsigned long end, int flags)
	123	{
	124	int ret = 0;
	125	struct file *file = vma->vm_file;
	126
	127	if ((flags & MS_INVALIDATE) && (vma->vm_flags & VM_LOCKED))
	128	return -EBUSY;
	129
	130	if (file && (vma->vm_flags & VM_SHARED)) {
0c942a45	131	msync_page_range(vma, addr, end);
1da177e4 LT	132
	133	if (flags & MS_SYNC) {
	134	struct address_space *mapping = file->f_mapping;
	135	int err;
	136
	137	ret = filemap_fdatawrite(mapping);
	138	if (file->f_op && file->f_op->fsync) {
	139	/*
1b1dcc1b	140	* We don't take i_mutex here because mmap_sem
1da177e4 LT	141	* is already held.
	142	*/
	143	err = file->f_op->fsync(file,file->f_dentry,1);
	144	if (err && !ret)
	145	ret = err;
	146	}
	147	err = filemap_fdatawait(mapping);
	148	if (!ret)
	149	ret = err;
	150	}
	151	}
	152	return ret;
	153	}
	154
	155	asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
	156	{
	157	unsigned long end;
	158	struct vm_area_struct *vma;
	159	int unmapped_error, error = -EINVAL;
	160
	161	if (flags & MS_SYNC)
	162	current->flags \|= PF_SYNCWRITE;
	163
	164	down_read(&current->mm->mmap_sem);
	165	if (flags & ~(MS_ASYNC \| MS_INVALIDATE \| MS_SYNC))
	166	goto out;
	167	if (start & ~PAGE_MASK)
	168	goto out;
	169	if ((flags & MS_ASYNC) && (flags & MS_SYNC))
	170	goto out;
	171	error = -ENOMEM;
	172	len = (len + ~PAGE_MASK) & PAGE_MASK;
	173	end = start + len;
	174	if (end < start)
	175	goto out;
	176	error = 0;
	177	if (end == start)
	178	goto out;
	179	/*
	180	* If the interval [start,end) covers some unmapped address ranges,
	181	* just ignore them, but return -ENOMEM at the end.
	182	*/
	183	vma = find_vma(current->mm, start);
	184	unmapped_error = 0;
	185	for (;;) {
	186	/* Still start < end. */
	187	error = -ENOMEM;
	188	if (!vma)
	189	goto out;
	190	/* Here start < vma->vm_end. */
	191	if (start < vma->vm_start) {
	192	unmapped_error = -ENOMEM;
	193	start = vma->vm_start;
	194	}
	195	/* Here vma->vm_start <= start < vma->vm_end. */
	196	if (end <= vma->vm_end) {
	197	if (start < end) {
	198	error = msync_interval(vma, start, end, flags);
	199	if (error)
	200	goto out;
	201	}
	202	error = unmapped_error;
	203	goto out;
	204	}
205	/* Here vma->vm_start <= start < vma->vm_end < end. */
206	error = msync_interval(vma, start, vma->vm_end, flags);
207	if (error)
208	goto out;
209	start = vma->vm_end;
210	vma = vma->vm_next;
211	}
212	out:
213	up_read(&current->mm->mmap_sem);
214	current->flags &= ~PF_SYNCWRITE;
215	return error;
216	}