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

/*
 *	linux/mm/madvise.c
 *
 * Copyright (C) 1999  Linus Torvalds
 * Copyright (C) 2002  Christoph Hellwig
 */

#include <linux/mman.h>
#include <linux/pagemap.h>
#include <linux/syscalls.h>
#include <linux/mempolicy.h>
#include <linux/hugetlb.h>

/*
 * We can potentially split a vm area into separate
 * areas, each area with its own behavior.
 */
static long madvise_behavior(struct vm_area_struct * vma,
		     struct vm_area_struct **prev,
		     unsigned long start, unsigned long end, int behavior)
{
	struct mm_struct * mm = vma->vm_mm;
	int error = 0;
	pgoff_t pgoff;
	int new_flags = vma->vm_flags & ~VM_READHINTMASK;

	switch (behavior) {
	case MADV_SEQUENTIAL:
		new_flags |= VM_SEQ_READ;
		break;
	case MADV_RANDOM:
		new_flags |= VM_RAND_READ;
		break;
	default:
		break;
	}

	if (new_flags == vma->vm_flags) {
		*prev = vma;
		goto success;
	}

	pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT);
	*prev = vma_merge(mm, *prev, start, end, new_flags, vma->anon_vma,
				vma->vm_file, pgoff, vma_policy(vma));
	if (*prev) {
		vma = *prev;
		goto success;
	}

	*prev = vma;

	if (start != vma->vm_start) {
		error = split_vma(mm, vma, start, 1);
		if (error)
			goto out;
	}

	if (end != vma->vm_end) {
		error = split_vma(mm, vma, end, 0);
		if (error)
			goto out;
	}

	/*
	 * vm_flags is protected by the mmap_sem held in write mode.
	 */
	vma->vm_flags = new_flags;

out:
	if (error == -ENOMEM)
		error = -EAGAIN;
success:
	return error;
}

/*
 * Schedule all required I/O operations.  Do not wait for completion.
 */
static long madvise_willneed(struct vm_area_struct * vma,
			     struct vm_area_struct ** prev,
			     unsigned long start, unsigned long end)
{
	struct file *file = vma->vm_file;

	if (!file)
		return -EBADF;

	*prev = vma;
	start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	if (end > vma->vm_end)
		end = vma->vm_end;
	end = ((end - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;

	force_page_cache_readahead(file->f_mapping,
			file, start, max_sane_readahead(end - start));
	return 0;
}

/*
 * Application no longer needs these pages.  If the pages are dirty,
 * it's OK to just throw them away.  The app will be more careful about
 * data it wants to keep.  Be sure to free swap resources too.  The
 * zap_page_range call sets things up for refill_inactive to actually free
 * these pages later if no one else has touched them in the meantime,
 * although we could add these pages to a global reuse list for
 * refill_inactive to pick up before reclaiming other pages.
 *
 * NB: This interface discards data rather than pushes it out to swap,
 * as some implementations do.  This has performance implications for
 * applications like large transactional databases which want to discard
 * pages in anonymous maps after committing to backing store the data
 * that was kept in them.  There is no reason to write this data out to
 * the swap area if the application is discarding it.
 *
 * An interface that causes the system to free clean pages and flush
 * dirty pages is already available as msync(MS_INVALIDATE).
 */
static long madvise_dontneed(struct vm_area_struct * vma,
			     struct vm_area_struct ** prev,
			     unsigned long start, unsigned long end)
{
	*prev = vma;
	if ((vma->vm_flags & VM_LOCKED) || is_vm_hugetlb_page(vma))
		return -EINVAL;

	if (unlikely(vma->vm_flags & VM_NONLINEAR)) {
		struct zap_details details = {
			.nonlinear_vma = vma,
			.last_index = ULONG_MAX,
		};
		zap_page_range(vma, start, end - start, &details);
	} else
		zap_page_range(vma, start, end - start, NULL);
	return 0;
}

static long madvise_vma(struct vm_area_struct *vma, struct vm_area_struct **prev,
			unsigned long start, unsigned long end, int behavior)
{
	long error = -EBADF;

	switch (behavior) {
	case MADV_NORMAL:
	case MADV_SEQUENTIAL:
	case MADV_RANDOM:
		error = madvise_behavior(vma, prev, start, end, behavior);
		break;

	case MADV_WILLNEED:
		error = madvise_willneed(vma, prev, start, end);
		break;

	case MADV_DONTNEED:
		error = madvise_dontneed(vma, prev, start, end);
		break;

	default:
		error = -EINVAL;
		break;
	}
		
	return error;
}

/*
 * The madvise(2) system call.
 *
 * Applications can use madvise() to advise the kernel how it should
 * handle paging I/O in this VM area.  The idea is to help the kernel
 * use appropriate read-ahead and caching techniques.  The information
 * provided is advisory only, and can be safely disregarded by the
 * kernel without affecting the correct operation of the application.
 *
 * behavior values:
 *  MADV_NORMAL - the default behavior is to read clusters.  This
 *		results in some read-ahead and read-behind.
 *  MADV_RANDOM - the system should read the minimum amount of data
 *		on any access, since it is unlikely that the appli-
 *		cation will need more than what it asks for.
 *  MADV_SEQUENTIAL - pages in the given range will probably be accessed
 *		once, so they can be aggressively read ahead, and
 *		can be freed soon after they are accessed.
 *  MADV_WILLNEED - the application is notifying the system to read
 *		some pages ahead.
 *  MADV_DONTNEED - the application is finished with the given range,
 *		so the kernel can free resources associated with it.
 *
 * return values:
 *  zero    - success
 *  -EINVAL - start + len < 0, start is not page-aligned,
 *		"behavior" is not a valid value, or application
 *		is attempting to release locked or shared pages.
 *  -ENOMEM - addresses in the specified range are not currently
 *		mapped, or are outside the AS of the process.
 *  -EIO    - an I/O error occurred while paging in data.
 *  -EBADF  - map exists, but area maps something that isn't a file.
 *  -EAGAIN - a kernel resource was temporarily unavailable.
 */
asmlinkage long sys_madvise(unsigned long start, size_t len_in, int behavior)
{
	unsigned long end, tmp;
	struct vm_area_struct * vma, *prev;
	int unmapped_error = 0;
	int error = -EINVAL;
	size_t len;

	down_write(&current->mm->mmap_sem);

	if (start & ~PAGE_MASK)
		goto out;
	len = (len_in + ~PAGE_MASK) & PAGE_MASK;

	/* Check to see whether len was rounded up from small -ve to zero */
	if (len_in && !len)
		goto out;

	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.
	 * - different from the way of handling in mlock etc.
	 */
	vma = find_vma_prev(current->mm, start, &prev);
	if (!vma && prev)
		vma = prev->vm_next;
	for (;;) {
		/* Still start < end. */
		error = -ENOMEM;
		if (!vma)
			goto out;

		/* Here start < (end|vma->vm_end). */
		if (start < vma->vm_start) {
			unmapped_error = -ENOMEM;
			start = vma->vm_start;
			if (start >= end)
				goto out;
		}

		/* Here vma->vm_start <= start < (end|vma->vm_end) */
		tmp = vma->vm_end;
		if (end < tmp)
			tmp = end;

		/* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */
		error = madvise_vma(vma, &prev, start, tmp, behavior);
		if (error)
			goto out;
		start = tmp;
		if (start < prev->vm_end)
			start = prev->vm_end;
		error = unmapped_error;
		if (start >= end)
			goto out;
		vma = prev->vm_next;
	}
out:
	up_write(&current->mm->mmap_sem);
	return error;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/mm/madvise.c
	3	*
	4	* Copyright (C) 1999 Linus Torvalds
	5	* Copyright (C) 2002 Christoph Hellwig
	6	*/
	7
	8	#include <linux/mman.h>
	9	#include <linux/pagemap.h>
	10	#include <linux/syscalls.h>
05b74384	11	#include <linux/mempolicy.h>
1da177e4 LT	12	#include <linux/hugetlb.h>
	13
	14	/*
	15	* We can potentially split a vm area into separate
	16	* areas, each area with its own behavior.
	17	*/
05b74384 PM	18	static long madvise_behavior(struct vm_area_struct * vma,
	19	struct vm_area_struct **prev,
	20	unsigned long start, unsigned long end, int behavior)
1da177e4 LT	21	{
	22	struct mm_struct * mm = vma->vm_mm;
	23	int error = 0;
05b74384	24	pgoff_t pgoff;
e798c6e8 PM	25	int new_flags = vma->vm_flags & ~VM_READHINTMASK;
	26
	27	switch (behavior) {
	28	case MADV_SEQUENTIAL:
	29	new_flags \|= VM_SEQ_READ;
	30	break;
	31	case MADV_RANDOM:
	32	new_flags \|= VM_RAND_READ;
	33	break;
	34	default:
	35	break;
	36	}
	37
05b74384 PM	38	if (new_flags == vma->vm_flags) {
	39	*prev = vma;
	40	goto success;
	41	}
	42
	43	pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT);
	44	prev = vma_merge(mm, prev, start, end, new_flags, vma->anon_vma,
	45	vma->vm_file, pgoff, vma_policy(vma));
	46	if (*prev) {
	47	vma = *prev;
	48	goto success;
	49	}
	50
	51	*prev = vma;
1da177e4 LT	52
	53	if (start != vma->vm_start) {
	54	error = split_vma(mm, vma, start, 1);
	55	if (error)
	56	goto out;
	57	}
	58
	59	if (end != vma->vm_end) {
	60	error = split_vma(mm, vma, end, 0);
	61	if (error)
	62	goto out;
	63	}
	64
	65	/*
	66	* vm_flags is protected by the mmap_sem held in write mode.
	67	*/
e798c6e8	68	vma->vm_flags = new_flags;
1da177e4 LT	69
	70	out:
	71	if (error == -ENOMEM)
	72	error = -EAGAIN;
05b74384	73	success:
1da177e4 LT	74	return error;
	75	}
	76
	77	/*
	78	* Schedule all required I/O operations. Do not wait for completion.
	79	*/
	80	static long madvise_willneed(struct vm_area_struct * vma,
05b74384	81	struct vm_area_struct ** prev,
1da177e4 LT	82	unsigned long start, unsigned long end)
	83	{
	84	struct file *file = vma->vm_file;
	85
	86	if (!file)
	87	return -EBADF;
	88
05b74384	89	*prev = vma;
1da177e4 LT	90	start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	91	if (end > vma->vm_end)
	92	end = vma->vm_end;
	93	end = ((end - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	94
	95	force_page_cache_readahead(file->f_mapping,
	96	file, start, max_sane_readahead(end - start));
	97	return 0;
	98	}
	99
	100	/*
	101	* Application no longer needs these pages. If the pages are dirty,
	102	* it's OK to just throw them away. The app will be more careful about
	103	* data it wants to keep. Be sure to free swap resources too. The
	104	* zap_page_range call sets things up for refill_inactive to actually free
	105	* these pages later if no one else has touched them in the meantime,
	106	* although we could add these pages to a global reuse list for
	107	* refill_inactive to pick up before reclaiming other pages.
	108	*
	109	* NB: This interface discards data rather than pushes it out to swap,
	110	* as some implementations do. This has performance implications for
	111	* applications like large transactional databases which want to discard
	112	* pages in anonymous maps after committing to backing store the data
	113	* that was kept in them. There is no reason to write this data out to
	114	* the swap area if the application is discarding it.
	115	*
	116	* An interface that causes the system to free clean pages and flush
	117	* dirty pages is already available as msync(MS_INVALIDATE).
	118	*/
	119	static long madvise_dontneed(struct vm_area_struct * vma,
05b74384	120	struct vm_area_struct ** prev,
1da177e4 LT	121	unsigned long start, unsigned long end)
1da177e4 LT	122	{
05b74384	123	*prev = vma;
1da177e4 LT	124	if ((vma->vm_flags & VM_LOCKED) \|\| is_vm_hugetlb_page(vma))
	125	return -EINVAL;
	126
	127	if (unlikely(vma->vm_flags & VM_NONLINEAR)) {
	128	struct zap_details details = {
	129	.nonlinear_vma = vma,
	130	.last_index = ULONG_MAX,
	131	};
	132	zap_page_range(vma, start, end - start, &details);
	133	} else
	134	zap_page_range(vma, start, end - start, NULL);
	135	return 0;
	136	}
	137
05b74384 PM	138	static long madvise_vma(struct vm_area_struct vma, struct vm_area_struct *prev,
05b74384 PM	139	unsigned long start, unsigned long end, int behavior)
1da177e4 LT	140	{
	141	long error = -EBADF;
	142
	143	switch (behavior) {
	144	case MADV_NORMAL:
	145	case MADV_SEQUENTIAL:
	146	case MADV_RANDOM:
05b74384	147	error = madvise_behavior(vma, prev, start, end, behavior);
1da177e4 LT	148	break;
	149
	150	case MADV_WILLNEED:
05b74384	151	error = madvise_willneed(vma, prev, start, end);
1da177e4 LT	152	break;
	153
	154	case MADV_DONTNEED:
05b74384	155	error = madvise_dontneed(vma, prev, start, end);
1da177e4 LT	156	break;
	157
	158	default:
	159	error = -EINVAL;
	160	break;
	161	}
	162
	163	return error;
	164	}
	165
	166	/*
	167	* The madvise(2) system call.
	168	*
	169	* Applications can use madvise() to advise the kernel how it should
	170	* handle paging I/O in this VM area. The idea is to help the kernel
	171	* use appropriate read-ahead and caching techniques. The information
	172	* provided is advisory only, and can be safely disregarded by the
	173	* kernel without affecting the correct operation of the application.
	174	*
	175	* behavior values:
	176	* MADV_NORMAL - the default behavior is to read clusters. This
	177	* results in some read-ahead and read-behind.
	178	* MADV_RANDOM - the system should read the minimum amount of data
	179	* on any access, since it is unlikely that the appli-
	180	* cation will need more than what it asks for.
	181	* MADV_SEQUENTIAL - pages in the given range will probably be accessed
	182	* once, so they can be aggressively read ahead, and
	183	* can be freed soon after they are accessed.
	184	* MADV_WILLNEED - the application is notifying the system to read
	185	* some pages ahead.
	186	* MADV_DONTNEED - the application is finished with the given range,
	187	* so the kernel can free resources associated with it.
	188	*
	189	* return values:
	190	* zero - success
	191	* -EINVAL - start + len < 0, start is not page-aligned,
	192	* "behavior" is not a valid value, or application
	193	* is attempting to release locked or shared pages.
	194	* -ENOMEM - addresses in the specified range are not currently
	195	* mapped, or are outside the AS of the process.
	196	* -EIO - an I/O error occurred while paging in data.
	197	* -EBADF - map exists, but area maps something that isn't a file.
	198	* -EAGAIN - a kernel resource was temporarily unavailable.
	199	*/
	200	asmlinkage long sys_madvise(unsigned long start, size_t len_in, int behavior)
	201	{
05b74384 PM	202	unsigned long end, tmp;
05b74384 PM	203	struct vm_area_struct * vma, *prev;
1da177e4 LT	204	int unmapped_error = 0;
	205	int error = -EINVAL;
	206	size_t len;
	207
	208	down_write(&current->mm->mmap_sem);
	209
	210	if (start & ~PAGE_MASK)
	211	goto out;
	212	len = (len_in + ~PAGE_MASK) & PAGE_MASK;
	213
	214	/* Check to see whether len was rounded up from small -ve to zero */
	215	if (len_in && !len)
	216	goto out;
	217
	218	end = start + len;
	219	if (end < start)
	220	goto out;
	221
	222	error = 0;
	223	if (end == start)
	224	goto out;
	225
	226	/*
	227	* If the interval [start,end) covers some unmapped address
	228	* ranges, just ignore them, but return -ENOMEM at the end.
05b74384	229	* - different from the way of handling in mlock etc.
1da177e4	230	*/
05b74384 PM	231	vma = find_vma_prev(current->mm, start, &prev);
	232	if (!vma && prev)
	233	vma = prev->vm_next;
1da177e4 LT	234	for (;;) {
	235	/* Still start < end. */
	236	error = -ENOMEM;
	237	if (!vma)
	238	goto out;
	239
05b74384	240	/* Here start < (end\|vma->vm_end). */
1da177e4 LT	241	if (start < vma->vm_start) {
	242	unmapped_error = -ENOMEM;
	243	start = vma->vm_start;
05b74384 PM	244	if (start >= end)
05b74384 PM	245	goto out;
1da177e4 LT	246	}
1da177e4 LT	247
05b74384 PM	248	/* Here vma->vm_start <= start < (end\|vma->vm_end) */
	249	tmp = vma->vm_end;
	250	if (end < tmp)
	251	tmp = end;
1da177e4	252
05b74384 PM	253	/* Here vma->vm_start <= start < tmp <= (end\|vma->vm_end). */
05b74384 PM	254	error = madvise_vma(vma, &prev, start, tmp, behavior);
1da177e4 LT	255	if (error)
1da177e4 LT	256	goto out;
05b74384 PM	257	start = tmp;
	258	if (start < prev->vm_end)
	259	start = prev->vm_end;
	260	error = unmapped_error;
	261	if (start >= end)
	262	goto out;
	263	vma = prev->vm_next;
1da177e4	264	}
1da177e4 LT	265	out:
	266	up_write(&current->mm->mmap_sem);
	267	return error;
	268	}