[mirror_ubuntu-artful-kernel.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->f_mapping->a_ops->get_xip_page) {
		/* no bad return value, but ignore advice */
		return 0;
	}

	*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)
{
	struct file *filp = vma->vm_file;
	long error = -EBADF;

	if (!filp)
		goto  out;

	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;
	}
		
out:
	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
fe77ba6f CO	86	if (file->f_mapping->a_ops->get_xip_page) {
	87	/* no bad return value, but ignore advice */
	88	return 0;
	89	}
	90
05b74384	91	*prev = vma;
1da177e4 LT	92	start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	93	if (end > vma->vm_end)
	94	end = vma->vm_end;
	95	end = ((end - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	96
	97	force_page_cache_readahead(file->f_mapping,
	98	file, start, max_sane_readahead(end - start));
	99	return 0;
	100	}
	101
	102	/*
	103	* Application no longer needs these pages. If the pages are dirty,
	104	* it's OK to just throw them away. The app will be more careful about
	105	* data it wants to keep. Be sure to free swap resources too. The
	106	* zap_page_range call sets things up for refill_inactive to actually free
	107	* these pages later if no one else has touched them in the meantime,
	108	* although we could add these pages to a global reuse list for
	109	* refill_inactive to pick up before reclaiming other pages.
	110	*
	111	* NB: This interface discards data rather than pushes it out to swap,
	112	* as some implementations do. This has performance implications for
	113	* applications like large transactional databases which want to discard
	114	* pages in anonymous maps after committing to backing store the data
	115	* that was kept in them. There is no reason to write this data out to
	116	* the swap area if the application is discarding it.
	117	*
	118	* An interface that causes the system to free clean pages and flush
	119	* dirty pages is already available as msync(MS_INVALIDATE).
	120	*/
	121	static long madvise_dontneed(struct vm_area_struct * vma,
05b74384	122	struct vm_area_struct ** prev,
1da177e4 LT	123	unsigned long start, unsigned long end)
1da177e4 LT	124	{
05b74384	125	*prev = vma;
1da177e4 LT	126	if ((vma->vm_flags & VM_LOCKED) \|\| is_vm_hugetlb_page(vma))
	127	return -EINVAL;
	128
	129	if (unlikely(vma->vm_flags & VM_NONLINEAR)) {
	130	struct zap_details details = {
	131	.nonlinear_vma = vma,
	132	.last_index = ULONG_MAX,
	133	};
	134	zap_page_range(vma, start, end - start, &details);
	135	} else
	136	zap_page_range(vma, start, end - start, NULL);
	137	return 0;
	138	}
	139
165cd402	140	static long
	141	madvise_vma(struct vm_area_struct vma, struct vm_area_struct *prev,
	142	unsigned long start, unsigned long end, int behavior)
1da177e4	143	{
165cd402	144	struct file *filp = vma->vm_file;
1da177e4 LT	145	long error = -EBADF;
1da177e4 LT	146
165cd402	147	if (!filp)
	148	goto out;
	149
1da177e4 LT	150	switch (behavior) {
	151	case MADV_NORMAL:
	152	case MADV_SEQUENTIAL:
	153	case MADV_RANDOM:
05b74384	154	error = madvise_behavior(vma, prev, start, end, behavior);
1da177e4 LT	155	break;
	156
	157	case MADV_WILLNEED:
05b74384	158	error = madvise_willneed(vma, prev, start, end);
1da177e4 LT	159	break;
	160
	161	case MADV_DONTNEED:
05b74384	162	error = madvise_dontneed(vma, prev, start, end);
1da177e4 LT	163	break;
	164
	165	default:
	166	error = -EINVAL;
	167	break;
	168	}
	169
165cd402	170	out:
1da177e4 LT	171	return error;
	172	}
	173
	174	/*
	175	* The madvise(2) system call.
	176	*
	177	* Applications can use madvise() to advise the kernel how it should
	178	* handle paging I/O in this VM area. The idea is to help the kernel
	179	* use appropriate read-ahead and caching techniques. The information
	180	* provided is advisory only, and can be safely disregarded by the
	181	* kernel without affecting the correct operation of the application.
	182	*
	183	* behavior values:
	184	* MADV_NORMAL - the default behavior is to read clusters. This
	185	* results in some read-ahead and read-behind.
	186	* MADV_RANDOM - the system should read the minimum amount of data
	187	* on any access, since it is unlikely that the appli-
	188	* cation will need more than what it asks for.
	189	* MADV_SEQUENTIAL - pages in the given range will probably be accessed
	190	* once, so they can be aggressively read ahead, and
	191	* can be freed soon after they are accessed.
	192	* MADV_WILLNEED - the application is notifying the system to read
	193	* some pages ahead.
	194	* MADV_DONTNEED - the application is finished with the given range,
	195	* so the kernel can free resources associated with it.
	196	*
	197	* return values:
	198	* zero - success
	199	* -EINVAL - start + len < 0, start is not page-aligned,
	200	* "behavior" is not a valid value, or application
	201	* is attempting to release locked or shared pages.
	202	* -ENOMEM - addresses in the specified range are not currently
	203	* mapped, or are outside the AS of the process.
	204	* -EIO - an I/O error occurred while paging in data.
	205	* -EBADF - map exists, but area maps something that isn't a file.
	206	* -EAGAIN - a kernel resource was temporarily unavailable.
	207	*/
	208	asmlinkage long sys_madvise(unsigned long start, size_t len_in, int behavior)
	209	{
05b74384 PM	210	unsigned long end, tmp;
05b74384 PM	211	struct vm_area_struct * vma, *prev;
1da177e4 LT	212	int unmapped_error = 0;
	213	int error = -EINVAL;
	214	size_t len;
	215
	216	down_write(&current->mm->mmap_sem);
	217
	218	if (start & ~PAGE_MASK)
	219	goto out;
	220	len = (len_in + ~PAGE_MASK) & PAGE_MASK;
	221
	222	/* Check to see whether len was rounded up from small -ve to zero */
	223	if (len_in && !len)
	224	goto out;
	225
	226	end = start + len;
	227	if (end < start)
	228	goto out;
	229
	230	error = 0;
	231	if (end == start)
	232	goto out;
	233
	234	/*
	235	* If the interval [start,end) covers some unmapped address
	236	* ranges, just ignore them, but return -ENOMEM at the end.
05b74384	237	* - different from the way of handling in mlock etc.
1da177e4	238	*/
05b74384 PM	239	vma = find_vma_prev(current->mm, start, &prev);
	240	if (!vma && prev)
	241	vma = prev->vm_next;
1da177e4 LT	242	for (;;) {
	243	/* Still start < end. */
	244	error = -ENOMEM;
	245	if (!vma)
	246	goto out;
	247
05b74384	248	/* Here start < (end\|vma->vm_end). */
1da177e4 LT	249	if (start < vma->vm_start) {
	250	unmapped_error = -ENOMEM;
	251	start = vma->vm_start;
05b74384 PM	252	if (start >= end)
05b74384 PM	253	goto out;
1da177e4 LT	254	}
1da177e4 LT	255
05b74384 PM	256	/* Here vma->vm_start <= start < (end\|vma->vm_end) */
	257	tmp = vma->vm_end;
	258	if (end < tmp)
	259	tmp = end;
1da177e4	260
05b74384 PM	261	/* Here vma->vm_start <= start < tmp <= (end\|vma->vm_end). */
05b74384 PM	262	error = madvise_vma(vma, &prev, start, tmp, behavior);
1da177e4 LT	263	if (error)
1da177e4 LT	264	goto out;
05b74384 PM	265	start = tmp;
	266	if (start < prev->vm_end)
	267	start = prev->vm_end;
	268	error = unmapped_error;
	269	if (start >= end)
	270	goto out;
	271	vma = prev->vm_next;
1da177e4	272	}
1da177e4 LT	273	out:
	274	up_write(&current->mm->mmap_sem);
	275	return error;
	276	}