[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/writeback.h>
#include <linux/file.h>
#include <linux/syscalls.h>

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

static unsigned long 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;
	unsigned long ret = 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))
			ret += 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;
	return ret;
}

static inline unsigned long msync_pmd_range(struct vm_area_struct *vma,
			pud_t *pud, unsigned long addr, unsigned long end)
{
	pmd_t *pmd;
	unsigned long next;
	unsigned long ret = 0;

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

static inline unsigned long msync_pud_range(struct vm_area_struct *vma,
			pgd_t *pgd, unsigned long addr, unsigned long end)
{
	pud_t *pud;
	unsigned long next;
	unsigned long ret = 0;

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

static unsigned long msync_page_range(struct vm_area_struct *vma,
				unsigned long addr, unsigned long end)
{
	pgd_t *pgd;
	unsigned long next;
	unsigned long ret = 0;

	/* 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 0;

	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;
		ret += msync_pud_range(vma, pgd, addr, next);
	} while (pgd++, addr = next, addr != end);
	return ret;
}

/*
 * 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,
			unsigned long *nr_pages_dirtied)
{
	struct file *file = vma->vm_file;

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

	if (file && (vma->vm_flags & VM_SHARED))
		*nr_pages_dirtied = msync_page_range(vma, addr, end);
	return 0;
}

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;
	int done = 0;

	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.
	 */
	down_read(&current->mm->mmap_sem);
	if (flags & MS_SYNC)
		current->flags |= PF_SYNCWRITE;
	vma = find_vma(current->mm, start);
	unmapped_error = 0;
	do {
		unsigned long nr_pages_dirtied = 0;
		struct file *file;

		/* Still start < end. */
		error = -ENOMEM;
		if (!vma)
			goto out_unlock;
		/* 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,
							&nr_pages_dirtied);
				if (error)
					goto out_unlock;
			}
			error = unmapped_error;
			done = 1;
		} else {
			/* Here vma->vm_start <= start < vma->vm_end < end. */
			error = msync_interval(vma, start, vma->vm_end, flags,
						&nr_pages_dirtied);
			if (error)
				goto out_unlock;
		}
		file = vma->vm_file;
		start = vma->vm_end;
		if ((flags & MS_ASYNC) && file && nr_pages_dirtied) {
			get_file(file);
			up_read(&current->mm->mmap_sem);
			balance_dirty_pages_ratelimited_nr(file->f_mapping,
							nr_pages_dirtied);
			fput(file);
			down_read(&current->mm->mmap_sem);
			vma = find_vma(current->mm, start);
		} else if ((flags & MS_SYNC) && file &&
				(vma->vm_flags & VM_SHARED)) {
			struct address_space *mapping;
			int err;

			get_file(file);
			up_read(&current->mm->mmap_sem);
			mapping = file->f_mapping;
			error = filemap_fdatawrite(mapping);
			if (file->f_op && file->f_op->fsync) {
				mutex_lock(&mapping->host->i_mutex);
				err = file->f_op->fsync(file,file->f_dentry,1);
				mutex_unlock(&mapping->host->i_mutex);
				if (err && !error)
					error = err;
			}
			err = filemap_fdatawait(mapping);
			if (err && !error)
				error = err;
			fput(file);
			down_read(&current->mm->mmap_sem);
			if (error)
				goto out_unlock;
			vma = find_vma(current->mm, start);
		} else {
			vma = vma->vm_next;
		}
	} while (!done);
out_unlock:
	current->flags &= ~PF_SYNCWRITE;
	up_read(&current->mm->mmap_sem);
out:
	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>
9c50823e AM	15	#include <linux/writeback.h>
9c50823e AM	16	#include <linux/file.h>
1da177e4 LT	17	#include <linux/syscalls.h>
	18
	19	#include <asm/pgtable.h>
	20	#include <asm/tlbflush.h>
	21
9c50823e	22	static unsigned long msync_pte_range(struct vm_area_struct vma, pmd_t pmd,
1da177e4 LT	23	unsigned long addr, unsigned long end)
	24	{
	25	pte_t *pte;
705e87c0	26	spinlock_t *ptl;
0c942a45	27	int progress = 0;
9c50823e	28	unsigned long ret = 0;
1da177e4	29
0c942a45	30	again:
705e87c0	31	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
1da177e4	32	do {
1da177e4 LT	33	struct page *page;
1da177e4 LT	34
0c942a45 HD	35	if (progress >= 64) {
0c942a45 HD	36	progress = 0;
705e87c0	37	if (need_resched() \|\| need_lockbreak(ptl))
0c942a45 HD	38	break;
	39	}
	40	progress++;
1da177e4 LT	41	if (!pte_present(*pte))
1da177e4 LT	42	continue;
b4955ce3 AK	43	if (!pte_maybe_dirty(*pte))
b4955ce3 AK	44	continue;
6aab341e LT	45	page = vm_normal_page(vma, addr, *pte);
6aab341e LT	46	if (!page)
1da177e4	47	continue;
1da177e4	48	if (ptep_clear_flush_dirty(vma, addr, pte) \|\|
9c50823e AM	49	page_test_and_clear_dirty(page))
9c50823e AM	50	ret += set_page_dirty(page);
0c942a45	51	progress += 3;
1da177e4	52	} while (pte++, addr += PAGE_SIZE, addr != end);
705e87c0 HD	53	pte_unmap_unlock(pte - 1, ptl);
705e87c0 HD	54	cond_resched();
0c942a45 HD	55	if (addr != end)
0c942a45 HD	56	goto again;
9c50823e	57	return ret;
1da177e4 LT	58	}
1da177e4 LT	59
9c50823e AM	60	static inline unsigned long msync_pmd_range(struct vm_area_struct *vma,
9c50823e AM	61	pud_t *pud, unsigned long addr, unsigned long end)
1da177e4 LT	62	{
	63	pmd_t *pmd;
	64	unsigned long next;
9c50823e	65	unsigned long ret = 0;
1da177e4 LT	66
	67	pmd = pmd_offset(pud, addr);
	68	do {
	69	next = pmd_addr_end(addr, end);
	70	if (pmd_none_or_clear_bad(pmd))
	71	continue;
9c50823e	72	ret += msync_pte_range(vma, pmd, addr, next);
1da177e4	73	} while (pmd++, addr = next, addr != end);
9c50823e	74	return ret;
1da177e4 LT	75	}
1da177e4 LT	76
9c50823e AM	77	static inline unsigned long msync_pud_range(struct vm_area_struct *vma,
9c50823e AM	78	pgd_t *pgd, unsigned long addr, unsigned long end)
1da177e4 LT	79	{
	80	pud_t *pud;
	81	unsigned long next;
9c50823e	82	unsigned long ret = 0;
1da177e4 LT	83
	84	pud = pud_offset(pgd, addr);
	85	do {
	86	next = pud_addr_end(addr, end);
	87	if (pud_none_or_clear_bad(pud))
	88	continue;
9c50823e	89	ret += msync_pmd_range(vma, pud, addr, next);
1da177e4	90	} while (pud++, addr = next, addr != end);
9c50823e	91	return ret;
1da177e4 LT	92	}
1da177e4 LT	93
9c50823e	94	static unsigned long msync_page_range(struct vm_area_struct *vma,
1da177e4 LT	95	unsigned long addr, unsigned long end)
1da177e4 LT	96	{
1da177e4 LT	97	pgd_t *pgd;
1da177e4 LT	98	unsigned long next;
9c50823e	99	unsigned long ret = 0;
1da177e4 LT	100
	101	/* For hugepages we can't go walking the page table normally,
	102	* but that's ok, hugetlbfs is memory based, so we don't need
b5810039	103	* to do anything more on an msync().
b5810039	104	*/
6aab341e	105	if (vma->vm_flags & VM_HUGETLB)
9c50823e	106	return 0;
1da177e4 LT	107
1da177e4 LT	108	BUG_ON(addr >= end);
705e87c0	109	pgd = pgd_offset(vma->vm_mm, addr);
1da177e4	110	flush_cache_range(vma, addr, end);
1da177e4 LT	111	do {
	112	next = pgd_addr_end(addr, end);
	113	if (pgd_none_or_clear_bad(pgd))
	114	continue;
9c50823e	115	ret += msync_pud_range(vma, pgd, addr, next);
1da177e4	116	} while (pgd++, addr = next, addr != end);
9c50823e	117	return ret;
1da177e4 LT	118	}
1da177e4 LT	119
1da177e4 LT	120	/*
	121	* MS_SYNC syncs the entire file - including mappings.
	122	*
	123	* MS_ASYNC does not start I/O (it used to, up to 2.5.67). Instead, it just
	124	* marks the relevant pages dirty. The application may now run fsync() to
	125	* write out the dirty pages and wait on the writeout and check the result.
	126	* Or the application may run fadvise(FADV_DONTNEED) against the fd to start
	127	* async writeout immediately.
	128	* So my _not_ starting I/O in MS_ASYNC we provide complete flexibility to
	129	* applications.
	130	*/
9c50823e AM	131	static int msync_interval(struct vm_area_struct *vma, unsigned long addr,
	132	unsigned long end, int flags,
	133	unsigned long *nr_pages_dirtied)
1da177e4	134	{
1da177e4 LT	135	struct file *file = vma->vm_file;
	136
	137	if ((flags & MS_INVALIDATE) && (vma->vm_flags & VM_LOCKED))
	138	return -EBUSY;
	139
707c21c8	140	if (file && (vma->vm_flags & VM_SHARED))
9c50823e	141	*nr_pages_dirtied = msync_page_range(vma, addr, end);
707c21c8	142	return 0;
1da177e4 LT	143	}
	144
	145	asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
	146	{
	147	unsigned long end;
	148	struct vm_area_struct *vma;
	149	int unmapped_error, error = -EINVAL;
9c50823e	150	int done = 0;
1da177e4	151
1da177e4 LT	152	if (flags & ~(MS_ASYNC \| MS_INVALIDATE \| MS_SYNC))
	153	goto out;
	154	if (start & ~PAGE_MASK)
	155	goto out;
	156	if ((flags & MS_ASYNC) && (flags & MS_SYNC))
	157	goto out;
	158	error = -ENOMEM;
	159	len = (len + ~PAGE_MASK) & PAGE_MASK;
	160	end = start + len;
	161	if (end < start)
	162	goto out;
	163	error = 0;
	164	if (end == start)
	165	goto out;
	166	/*
	167	* If the interval [start,end) covers some unmapped address ranges,
	168	* just ignore them, but return -ENOMEM at the end.
	169	*/
9c50823e AM	170	down_read(&current->mm->mmap_sem);
	171	if (flags & MS_SYNC)
	172	current->flags \|= PF_SYNCWRITE;
1da177e4 LT	173	vma = find_vma(current->mm, start);
1da177e4 LT	174	unmapped_error = 0;
9c50823e AM	175	do {
	176	unsigned long nr_pages_dirtied = 0;
	177	struct file *file;
	178
1da177e4 LT	179	/* Still start < end. */
	180	error = -ENOMEM;
	181	if (!vma)
9c50823e	182	goto out_unlock;
1da177e4 LT	183	/* Here start < vma->vm_end. */
	184	if (start < vma->vm_start) {
	185	unmapped_error = -ENOMEM;
	186	start = vma->vm_start;
	187	}
	188	/* Here vma->vm_start <= start < vma->vm_end. */
	189	if (end <= vma->vm_end) {
	190	if (start < end) {
9c50823e AM	191	error = msync_interval(vma, start, end, flags,
9c50823e AM	192	&nr_pages_dirtied);
1da177e4	193	if (error)
9c50823e	194	goto out_unlock;
1da177e4 LT	195	}
1da177e4 LT	196	error = unmapped_error;
9c50823e AM	197	done = 1;
	198	} else {
	199	/* Here vma->vm_start <= start < vma->vm_end < end. */
	200	error = msync_interval(vma, start, vma->vm_end, flags,
	201	&nr_pages_dirtied);
	202	if (error)
	203	goto out_unlock;
1da177e4	204	}
9c50823e	205	file = vma->vm_file;
1da177e4	206	start = vma->vm_end;
9c50823e AM	207	if ((flags & MS_ASYNC) && file && nr_pages_dirtied) {
	208	get_file(file);
	209	up_read(&current->mm->mmap_sem);
	210	balance_dirty_pages_ratelimited_nr(file->f_mapping,
	211	nr_pages_dirtied);
	212	fput(file);
	213	down_read(&current->mm->mmap_sem);
	214	vma = find_vma(current->mm, start);
707c21c8 AM	215	} else if ((flags & MS_SYNC) && file &&
	216	(vma->vm_flags & VM_SHARED)) {
	217	struct address_space *mapping;
	218	int err;
	219
	220	get_file(file);
	221	up_read(&current->mm->mmap_sem);
	222	mapping = file->f_mapping;
	223	error = filemap_fdatawrite(mapping);
	224	if (file->f_op && file->f_op->fsync) {
	225	mutex_lock(&mapping->host->i_mutex);
	226	err = file->f_op->fsync(file,file->f_dentry,1);
	227	mutex_unlock(&mapping->host->i_mutex);
	228	if (err && !error)
	229	error = err;
	230	}
	231	err = filemap_fdatawait(mapping);
	232	if (err && !error)
	233	error = err;
	234	fput(file);
	235	down_read(&current->mm->mmap_sem);
	236	if (error)
	237	goto out_unlock;
	238	vma = find_vma(current->mm, start);
9c50823e AM	239	} else {
	240	vma = vma->vm_next;
	241	}
	242	} while (!done);
	243	out_unlock:
1da177e4	244	current->flags &= ~PF_SYNCWRITE;
9c50823e AM	245	up_read(&current->mm->mmap_sem);
9c50823e AM	246	out:
1da177e4 LT	247	return error;
1da177e4 LT	248	}