[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;

	switch (behavior) {
	case MADV_NORMAL:
		new_flags = new_flags & ~VM_RAND_READ & ~VM_SEQ_READ;
		break;
	case MADV_SEQUENTIAL:
		new_flags = (new_flags & ~VM_RAND_READ) | VM_SEQ_READ;
		break;
	case MADV_RANDOM:
		new_flags = (new_flags & ~VM_SEQ_READ) | VM_RAND_READ;
		break;
	case MADV_DONTFORK:
		new_flags |= VM_DONTCOPY;
		break;
	case MADV_DOFORK:
		new_flags &= ~VM_DONTCOPY;
		break;
	}

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

	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;
	}

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

out:
	if (error == -ENOMEM)
		error = -EAGAIN;
	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;

	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|VM_HUGETLB|VM_PFNMAP))
		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;
}

/*
 * Application wants to free up the pages and associated backing store.
 * This is effectively punching a hole into the middle of a file.
 *
 * NOTE: Currently, only shmfs/tmpfs is supported for this operation.
 * Other filesystems return -ENOSYS.
 */
static long madvise_remove(struct vm_area_struct *vma,
				struct vm_area_struct **prev,
				unsigned long start, unsigned long end)
{
	struct address_space *mapping;
        loff_t offset, endoff;

	*prev = vma;

	if (vma->vm_flags & (VM_LOCKED|VM_NONLINEAR|VM_HUGETLB))
		return -EINVAL;

	if (!vma->vm_file || !vma->vm_file->f_mapping
		|| !vma->vm_file->f_mapping->host) {
			return -EINVAL;
	}

	if ((vma->vm_flags & (VM_SHARED|VM_WRITE)) != (VM_SHARED|VM_WRITE))
		return -EACCES;

	mapping = vma->vm_file->f_mapping;

	offset = (loff_t)(start - vma->vm_start)
			+ ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
	endoff = (loff_t)(end - vma->vm_start - 1)
			+ ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
	return  vmtruncate_range(mapping->host, offset, endoff);
}

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

	switch (behavior) {
	case MADV_DOFORK:
		if (vma->vm_flags & VM_IO) {
			error = -EINVAL;
			break;
		}
	case MADV_DONTFORK:
	case MADV_NORMAL:
	case MADV_SEQUENTIAL:
	case MADV_RANDOM:
		error = madvise_behavior(vma, prev, start, end, behavior);
		break;
	case MADV_REMOVE:
		error = madvise_remove(vma, prev, start, end);
		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.
 *  MADV_REMOVE - the application wants to free up the given range of
 *		pages and associated backing store.
 *
 * 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 && start > vma->vm_start)
		prev = vma;

	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;
f8225661	25	int new_flags = vma->vm_flags;
e798c6e8 PM	26
e798c6e8 PM	27	switch (behavior) {
f8225661 MT	28	case MADV_NORMAL:
	29	new_flags = new_flags & ~VM_RAND_READ & ~VM_SEQ_READ;
	30	break;
e798c6e8	31	case MADV_SEQUENTIAL:
f8225661	32	new_flags = (new_flags & ~VM_RAND_READ) \| VM_SEQ_READ;
e798c6e8 PM	33	break;
e798c6e8 PM	34	case MADV_RANDOM:
f8225661	35	new_flags = (new_flags & ~VM_SEQ_READ) \| VM_RAND_READ;
e798c6e8	36	break;
f8225661 MT	37	case MADV_DONTFORK:
	38	new_flags \|= VM_DONTCOPY;
	39	break;
	40	case MADV_DOFORK:
	41	new_flags &= ~VM_DONTCOPY;
e798c6e8 PM	42	break;
	43	}
	44
05b74384 PM	45	if (new_flags == vma->vm_flags) {
05b74384 PM	46	*prev = vma;
836d5ffd	47	goto out;
05b74384 PM	48	}
	49
	50	pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT);
	51	prev = vma_merge(mm, prev, start, end, new_flags, vma->anon_vma,
	52	vma->vm_file, pgoff, vma_policy(vma));
	53	if (*prev) {
	54	vma = *prev;
	55	goto success;
	56	}
	57
	58	*prev = vma;
1da177e4 LT	59
	60	if (start != vma->vm_start) {
	61	error = split_vma(mm, vma, start, 1);
	62	if (error)
	63	goto out;
	64	}
	65
	66	if (end != vma->vm_end) {
	67	error = split_vma(mm, vma, end, 0);
	68	if (error)
	69	goto out;
	70	}
	71
836d5ffd	72	success:
1da177e4 LT	73	/*
	74	* vm_flags is protected by the mmap_sem held in write mode.
	75	*/
e798c6e8	76	vma->vm_flags = new_flags;
1da177e4 LT	77
	78	out:
	79	if (error == -ENOMEM)
	80	error = -EAGAIN;
	81	return error;
	82	}
	83
	84	/*
	85	* Schedule all required I/O operations. Do not wait for completion.
	86	*/
	87	static long madvise_willneed(struct vm_area_struct * vma,
05b74384	88	struct vm_area_struct ** prev,
1da177e4 LT	89	unsigned long start, unsigned long end)
	90	{
	91	struct file *file = vma->vm_file;
	92
1bef4003 S	93	if (!file)
	94	return -EBADF;
	95
fe77ba6f CO	96	if (file->f_mapping->a_ops->get_xip_page) {
	97	/* no bad return value, but ignore advice */
	98	return 0;
	99	}
	100
05b74384	101	*prev = vma;
1da177e4 LT	102	start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	103	if (end > vma->vm_end)
	104	end = vma->vm_end;
	105	end = ((end - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	106
	107	force_page_cache_readahead(file->f_mapping,
	108	file, start, max_sane_readahead(end - start));
	109	return 0;
	110	}
	111
	112	/*
	113	* Application no longer needs these pages. If the pages are dirty,
	114	* it's OK to just throw them away. The app will be more careful about
	115	* data it wants to keep. Be sure to free swap resources too. The
	116	* zap_page_range call sets things up for refill_inactive to actually free
	117	* these pages later if no one else has touched them in the meantime,
	118	* although we could add these pages to a global reuse list for
	119	* refill_inactive to pick up before reclaiming other pages.
	120	*
	121	* NB: This interface discards data rather than pushes it out to swap,
	122	* as some implementations do. This has performance implications for
	123	* applications like large transactional databases which want to discard
	124	* pages in anonymous maps after committing to backing store the data
	125	* that was kept in them. There is no reason to write this data out to
	126	* the swap area if the application is discarding it.
	127	*
	128	* An interface that causes the system to free clean pages and flush
	129	* dirty pages is already available as msync(MS_INVALIDATE).
	130	*/
	131	static long madvise_dontneed(struct vm_area_struct * vma,
05b74384	132	struct vm_area_struct ** prev,
1da177e4 LT	133	unsigned long start, unsigned long end)
1da177e4 LT	134	{
05b74384	135	*prev = vma;
6aab341e	136	if (vma->vm_flags & (VM_LOCKED\|VM_HUGETLB\|VM_PFNMAP))
1da177e4 LT	137	return -EINVAL;
	138
	139	if (unlikely(vma->vm_flags & VM_NONLINEAR)) {
	140	struct zap_details details = {
	141	.nonlinear_vma = vma,
	142	.last_index = ULONG_MAX,
	143	};
	144	zap_page_range(vma, start, end - start, &details);
	145	} else
	146	zap_page_range(vma, start, end - start, NULL);
	147	return 0;
	148	}
	149
f6b3ec23 BP	150	/*
	151	* Application wants to free up the pages and associated backing store.
	152	* This is effectively punching a hole into the middle of a file.
	153	*
	154	* NOTE: Currently, only shmfs/tmpfs is supported for this operation.
	155	* Other filesystems return -ENOSYS.
	156	*/
	157	static long madvise_remove(struct vm_area_struct *vma,
00e9fa2d	158	struct vm_area_struct **prev,
f6b3ec23 BP	159	unsigned long start, unsigned long end)
	160	{
	161	struct address_space *mapping;
	162	loff_t offset, endoff;
	163
00e9fa2d NP	164	*prev = vma;
00e9fa2d NP	165
f6b3ec23 BP	166	if (vma->vm_flags & (VM_LOCKED\|VM_NONLINEAR\|VM_HUGETLB))
	167	return -EINVAL;
	168
	169	if (!vma->vm_file \|\| !vma->vm_file->f_mapping
	170	\|\| !vma->vm_file->f_mapping->host) {
	171	return -EINVAL;
	172	}
	173
69cf0fac HD	174	if ((vma->vm_flags & (VM_SHARED\|VM_WRITE)) != (VM_SHARED\|VM_WRITE))
	175	return -EACCES;
	176
f6b3ec23 BP	177	mapping = vma->vm_file->f_mapping;
	178
	179	offset = (loff_t)(start - vma->vm_start)
	180	+ ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
	181	endoff = (loff_t)(end - vma->vm_start - 1)
	182	+ ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
	183	return vmtruncate_range(mapping->host, offset, endoff);
	184	}
	185
165cd402	186	static long
	187	madvise_vma(struct vm_area_struct vma, struct vm_area_struct *prev,
	188	unsigned long start, unsigned long end, int behavior)
1da177e4	189	{
1bef4003	190	long error;
165cd402	191
1da177e4	192	switch (behavior) {
f8225661 MT	193	case MADV_DOFORK:
	194	if (vma->vm_flags & VM_IO) {
	195	error = -EINVAL;
	196	break;
	197	}
	198	case MADV_DONTFORK:
1da177e4 LT	199	case MADV_NORMAL:
	200	case MADV_SEQUENTIAL:
	201	case MADV_RANDOM:
05b74384	202	error = madvise_behavior(vma, prev, start, end, behavior);
1da177e4	203	break;
f6b3ec23	204	case MADV_REMOVE:
00e9fa2d	205	error = madvise_remove(vma, prev, start, end);
f6b3ec23	206	break;
1da177e4 LT	207
1da177e4 LT	208	case MADV_WILLNEED:
05b74384	209	error = madvise_willneed(vma, prev, start, end);
1da177e4 LT	210	break;
	211
	212	case MADV_DONTNEED:
05b74384	213	error = madvise_dontneed(vma, prev, start, end);
1da177e4 LT	214	break;
	215
	216	default:
	217	error = -EINVAL;
	218	break;
	219	}
1da177e4 LT	220	return error;
	221	}
	222
	223	/*
	224	* The madvise(2) system call.
	225	*
	226	* Applications can use madvise() to advise the kernel how it should
	227	* handle paging I/O in this VM area. The idea is to help the kernel
	228	* use appropriate read-ahead and caching techniques. The information
	229	* provided is advisory only, and can be safely disregarded by the
	230	* kernel without affecting the correct operation of the application.
	231	*
	232	* behavior values:
	233	* MADV_NORMAL - the default behavior is to read clusters. This
	234	* results in some read-ahead and read-behind.
	235	* MADV_RANDOM - the system should read the minimum amount of data
	236	* on any access, since it is unlikely that the appli-
	237	* cation will need more than what it asks for.
	238	* MADV_SEQUENTIAL - pages in the given range will probably be accessed
	239	* once, so they can be aggressively read ahead, and
	240	* can be freed soon after they are accessed.
	241	* MADV_WILLNEED - the application is notifying the system to read
	242	* some pages ahead.
	243	* MADV_DONTNEED - the application is finished with the given range,
	244	* so the kernel can free resources associated with it.
f6b3ec23 BP	245	* MADV_REMOVE - the application wants to free up the given range of
f6b3ec23 BP	246	* pages and associated backing store.
1da177e4 LT	247	*
	248	* return values:
	249	* zero - success
	250	* -EINVAL - start + len < 0, start is not page-aligned,
	251	* "behavior" is not a valid value, or application
	252	* is attempting to release locked or shared pages.
	253	* -ENOMEM - addresses in the specified range are not currently
	254	* mapped, or are outside the AS of the process.
	255	* -EIO - an I/O error occurred while paging in data.
	256	* -EBADF - map exists, but area maps something that isn't a file.
	257	* -EAGAIN - a kernel resource was temporarily unavailable.
	258	*/
	259	asmlinkage long sys_madvise(unsigned long start, size_t len_in, int behavior)
	260	{
05b74384 PM	261	unsigned long end, tmp;
05b74384 PM	262	struct vm_area_struct * vma, *prev;
1da177e4 LT	263	int unmapped_error = 0;
	264	int error = -EINVAL;
	265	size_t len;
	266
	267	down_write(&current->mm->mmap_sem);
	268
	269	if (start & ~PAGE_MASK)
	270	goto out;
	271	len = (len_in + ~PAGE_MASK) & PAGE_MASK;
	272
	273	/* Check to see whether len was rounded up from small -ve to zero */
	274	if (len_in && !len)
	275	goto out;
	276
	277	end = start + len;
	278	if (end < start)
	279	goto out;
	280
	281	error = 0;
	282	if (end == start)
	283	goto out;
	284
	285	/*
	286	* If the interval [start,end) covers some unmapped address
	287	* ranges, just ignore them, but return -ENOMEM at the end.
05b74384	288	* - different from the way of handling in mlock etc.
1da177e4	289	*/
05b74384	290	vma = find_vma_prev(current->mm, start, &prev);
836d5ffd HD	291	if (vma && start > vma->vm_start)
	292	prev = vma;
	293
1da177e4 LT	294	for (;;) {
	295	/* Still start < end. */
	296	error = -ENOMEM;
	297	if (!vma)
	298	goto out;
	299
05b74384	300	/* Here start < (end\|vma->vm_end). */
1da177e4 LT	301	if (start < vma->vm_start) {
	302	unmapped_error = -ENOMEM;
	303	start = vma->vm_start;
05b74384 PM	304	if (start >= end)
05b74384 PM	305	goto out;
1da177e4 LT	306	}
1da177e4 LT	307
05b74384 PM	308	/* Here vma->vm_start <= start < (end\|vma->vm_end) */
	309	tmp = vma->vm_end;
	310	if (end < tmp)
	311	tmp = end;
1da177e4	312
05b74384 PM	313	/* Here vma->vm_start <= start < tmp <= (end\|vma->vm_end). */
05b74384 PM	314	error = madvise_vma(vma, &prev, start, tmp, behavior);
1da177e4 LT	315	if (error)
1da177e4 LT	316	goto out;
05b74384 PM	317	start = tmp;
	318	if (start < prev->vm_end)
	319	start = prev->vm_end;
	320	error = unmapped_error;
	321	if (start >= end)
	322	goto out;
	323	vma = prev->vm_next;
1da177e4	324	}
1da177e4 LT	325	out:
	326	up_write(&current->mm->mmap_sem);
	327	return error;
	328	}