[deliverable/linux.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 {
		unsigned long pfn;
		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;
		pfn = pte_pfn(*pte);
		if (unlikely(!pfn_valid(pfn))) {
			print_bad_pte(vma, *pte, addr);
			continue;
		}
		page = pfn_to_page(pfn);

		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().
	 * Can't do anything with VM_RESERVED regions either.
	 */
	if (vma->vm_flags & (VM_HUGETLB|VM_RESERVED))
		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_sem 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 LT	29	do {
	30	unsigned long pfn;
	31	struct page *page;
	32
0c942a45 HD	33	if (progress >= 64) {
0c942a45 HD	34	progress = 0;
705e87c0	35	if (need_resched() \|\| need_lockbreak(ptl))
0c942a45 HD	36	break;
	37	}
	38	progress++;
1da177e4 LT	39	if (!pte_present(*pte))
1da177e4 LT	40	continue;
b4955ce3 AK	41	if (!pte_maybe_dirty(*pte))
b4955ce3 AK	42	continue;
1da177e4	43	pfn = pte_pfn(*pte);
b5810039 NP	44	if (unlikely(!pfn_valid(pfn))) {
b5810039 NP	45	print_bad_pte(vma, *pte, addr);
1da177e4	46	continue;
b5810039	47	}
1da177e4	48	page = pfn_to_page(pfn);
1da177e4 LT	49
	50	if (ptep_clear_flush_dirty(vma, addr, pte) \|\|
	51	page_test_and_clear_dirty(page))
	52	set_page_dirty(page);
0c942a45	53	progress += 3;
1da177e4	54	} while (pte++, addr += PAGE_SIZE, addr != end);
705e87c0 HD	55	pte_unmap_unlock(pte - 1, ptl);
705e87c0 HD	56	cond_resched();
0c942a45 HD	57	if (addr != end)
0c942a45 HD	58	goto again;
1da177e4 LT	59	}
1da177e4 LT	60
b57b98d1	61	static inline void msync_pmd_range(struct vm_area_struct vma, pud_t pud,
1da177e4 LT	62	unsigned long addr, unsigned long end)
	63	{
	64	pmd_t *pmd;
	65	unsigned long next;
	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;
b57b98d1	72	msync_pte_range(vma, pmd, addr, next);
1da177e4 LT	73	} while (pmd++, addr = next, addr != end);
	74	}
	75
b57b98d1	76	static inline void msync_pud_range(struct vm_area_struct vma, pgd_t pgd,
1da177e4 LT	77	unsigned long addr, unsigned long end)
	78	{
	79	pud_t *pud;
	80	unsigned long next;
	81
	82	pud = pud_offset(pgd, addr);
	83	do {
	84	next = pud_addr_end(addr, end);
	85	if (pud_none_or_clear_bad(pud))
	86	continue;
b57b98d1	87	msync_pmd_range(vma, pud, addr, next);
1da177e4 LT	88	} while (pud++, addr = next, addr != end);
	89	}
	90
b57b98d1	91	static void msync_page_range(struct vm_area_struct *vma,
1da177e4 LT	92	unsigned long addr, unsigned long end)
1da177e4 LT	93	{
1da177e4 LT	94	pgd_t *pgd;
	95	unsigned long next;
	96
	97	/* For hugepages we can't go walking the page table normally,
	98	* but that's ok, hugetlbfs is memory based, so we don't need
b5810039 NP	99	* to do anything more on an msync().
	100	* Can't do anything with VM_RESERVED regions either.
	101	*/
	102	if (vma->vm_flags & (VM_HUGETLB\|VM_RESERVED))
1da177e4 LT	103	return;
	104
	105	BUG_ON(addr >= end);
705e87c0	106	pgd = pgd_offset(vma->vm_mm, addr);
1da177e4	107	flush_cache_range(vma, addr, end);
1da177e4 LT	108	do {
	109	next = pgd_addr_end(addr, end);
	110	if (pgd_none_or_clear_bad(pgd))
	111	continue;
b57b98d1	112	msync_pud_range(vma, pgd, addr, next);
1da177e4	113	} while (pgd++, addr = next, addr != end);
1da177e4 LT	114	}
1da177e4 LT	115
1da177e4 LT	116	/*
	117	* MS_SYNC syncs the entire file - including mappings.
	118	*
	119	* MS_ASYNC does not start I/O (it used to, up to 2.5.67). Instead, it just
	120	* marks the relevant pages dirty. The application may now run fsync() to
	121	* write out the dirty pages and wait on the writeout and check the result.
	122	* Or the application may run fadvise(FADV_DONTNEED) against the fd to start
	123	* async writeout immediately.
	124	* So my _not_ starting I/O in MS_ASYNC we provide complete flexibility to
	125	* applications.
	126	*/
	127	static int msync_interval(struct vm_area_struct *vma,
	128	unsigned long addr, unsigned long end, int flags)
	129	{
	130	int ret = 0;
	131	struct file *file = vma->vm_file;
	132
	133	if ((flags & MS_INVALIDATE) && (vma->vm_flags & VM_LOCKED))
	134	return -EBUSY;
	135
	136	if (file && (vma->vm_flags & VM_SHARED)) {
0c942a45	137	msync_page_range(vma, addr, end);
1da177e4 LT	138
	139	if (flags & MS_SYNC) {
	140	struct address_space *mapping = file->f_mapping;
	141	int err;
	142
	143	ret = filemap_fdatawrite(mapping);
	144	if (file->f_op && file->f_op->fsync) {
	145	/*
	146	* We don't take i_sem here because mmap_sem
	147	* is already held.
	148	*/
	149	err = file->f_op->fsync(file,file->f_dentry,1);
	150	if (err && !ret)
	151	ret = err;
	152	}
	153	err = filemap_fdatawait(mapping);
	154	if (!ret)
	155	ret = err;
	156	}
	157	}
	158	return ret;
	159	}
	160
	161	asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
	162	{
	163	unsigned long end;
	164	struct vm_area_struct *vma;
	165	int unmapped_error, error = -EINVAL;
	166
	167	if (flags & MS_SYNC)
	168	current->flags \|= PF_SYNCWRITE;
	169
	170	down_read(&current->mm->mmap_sem);
	171	if (flags & ~(MS_ASYNC \| MS_INVALIDATE \| MS_SYNC))
	172	goto out;
	173	if (start & ~PAGE_MASK)
	174	goto out;
	175	if ((flags & MS_ASYNC) && (flags & MS_SYNC))
	176	goto out;
	177	error = -ENOMEM;
	178	len = (len + ~PAGE_MASK) & PAGE_MASK;
	179	end = start + len;
	180	if (end < start)
	181	goto out;
	182	error = 0;
	183	if (end == start)
	184	goto out;
	185	/*
	186	* If the interval [start,end) covers some unmapped address ranges,
	187	* just ignore them, but return -ENOMEM at the end.
	188	*/
	189	vma = find_vma(current->mm, start);
	190	unmapped_error = 0;
	191	for (;;) {
	192	/* Still start < end. */
	193	error = -ENOMEM;
	194	if (!vma)
	195	goto out;
	196	/* Here start < vma->vm_end. */
	197	if (start < vma->vm_start) {
	198	unmapped_error = -ENOMEM;
	199	start = vma->vm_start;
	200	}
	201	/* Here vma->vm_start <= start < vma->vm_end. */
202	if (end <= vma->vm_end) {
203	if (start < end) {
204	error = msync_interval(vma, start, end, flags);
205	if (error)
206	goto out;
207	}
208	error = unmapped_error;
209	goto out;
210	}
211	/* Here vma->vm_start <= start < vma->vm_end < end. */
212	error = msync_interval(vma, start, vma->vm_end, flags);
213	if (error)
214	goto out;
215	start = vma->vm_end;
216	vma = vma->vm_next;
217	}
218	out:
219	up_read(&current->mm->mmap_sem);
220	current->flags &= ~PF_SYNCWRITE;
221	return error;
222	}