[linux.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/page-isolation.h>
#include <linux/hugetlb.h>
#include <linux/sched.h>
#include <linux/ksm.h>

/*
 * Any behaviour which results in changes to the vma->vm_flags needs to
 * take mmap_sem for writing. Others, which simply traverse vmas, need
 * to only take it for reading.
 */
static int madvise_need_mmap_write(int behavior)
{
	switch (behavior) {
	case MADV_REMOVE:
	case MADV_WILLNEED:
	case MADV_DONTNEED:
		return 0;
	default:
		/* be safe, default to 1. list exceptions explicitly */
		return 1;
	}
}

/*
 * 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;
	unsigned long 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:
		if (vma->vm_flags & VM_IO) {
			error = -EINVAL;
			goto out;
		}
		new_flags &= ~VM_DONTCOPY;
		break;
	case MADV_MERGEABLE:
	case MADV_UNMERGEABLE:
		error = ksm_madvise(vma, start, end, behavior, &new_flags);
		if (error)
			goto out;
		break;
	case MADV_HUGEPAGE:
	case MADV_NOHUGEPAGE:
		error = hugepage_madvise(vma, &new_flags, behavior);
		if (error)
			goto out;
		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_mem) {
		/* 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, 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 shrink_active_list 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
 * shrink_active_list 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;
	int error;

	*prev = NULL;	/* tell sys_madvise we drop mmap_sem */

	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);

	/* vmtruncate_range needs to take i_mutex */
	up_read(&current->mm->mmap_sem);
	error = vmtruncate_range(mapping->host, offset, endoff);
	down_read(&current->mm->mmap_sem);
	return error;
}

#ifdef CONFIG_MEMORY_FAILURE
/*
 * Error injection support for memory error handling.
 */
static int madvise_hwpoison(int bhv, unsigned long start, unsigned long end)
{
	int ret = 0;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	for (; start < end; start += PAGE_SIZE) {
		struct page *p;
		int ret = get_user_pages_fast(start, 1, 0, &p);
		if (ret != 1)
			return ret;
		if (bhv == MADV_SOFT_OFFLINE) {
			printk(KERN_INFO "Soft offlining page %lx at %lx\n",
				page_to_pfn(p), start);
			ret = soft_offline_page(p, MF_COUNT_INCREASED);
			if (ret)
				break;
			continue;
		}
		printk(KERN_INFO "Injecting memory failure for page %lx at %lx\n",
		       page_to_pfn(p), start);
		/* Ignore return value for now */
		__memory_failure(page_to_pfn(p), 0, MF_COUNT_INCREASED);
	}
	return ret;
}
#endif

static long
madvise_vma(struct vm_area_struct *vma, struct vm_area_struct **prev,
		unsigned long start, unsigned long end, int behavior)
{
	switch (behavior) {
	case MADV_REMOVE:
		return madvise_remove(vma, prev, start, end);
	case MADV_WILLNEED:
		return madvise_willneed(vma, prev, start, end);
	case MADV_DONTNEED:
		return madvise_dontneed(vma, prev, start, end);
	default:
		return madvise_behavior(vma, prev, start, end, behavior);
	}
}

static int
madvise_behavior_valid(int behavior)
{
	switch (behavior) {
	case MADV_DOFORK:
	case MADV_DONTFORK:
	case MADV_NORMAL:
	case MADV_SEQUENTIAL:
	case MADV_RANDOM:
	case MADV_REMOVE:
	case MADV_WILLNEED:
	case MADV_DONTNEED:
#ifdef CONFIG_KSM
	case MADV_MERGEABLE:
	case MADV_UNMERGEABLE:
#endif
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	case MADV_HUGEPAGE:
	case MADV_NOHUGEPAGE:
#endif
		return 1;

	default:
		return 0;
	}
}

/*
 * 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.
 *  MADV_DONTFORK - omit this area from child's address space when forking:
 *		typically, to avoid COWing pages pinned by get_user_pages().
 *  MADV_DOFORK - cancel MADV_DONTFORK: no longer omit this area when forking.
 *  MADV_MERGEABLE - the application recommends that KSM try to merge pages in
 *		this area with pages of identical content from other such areas.
 *  MADV_UNMERGEABLE- cancel MADV_MERGEABLE: no longer merge pages with others.
 *
 * 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.
 */
SYSCALL_DEFINE3(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;
	int write;
	size_t len;

#ifdef CONFIG_MEMORY_FAILURE
	if (behavior == MADV_HWPOISON || behavior == MADV_SOFT_OFFLINE)
		return madvise_hwpoison(behavior, start, start+len_in);
#endif
	if (!madvise_behavior_valid(behavior))
		return error;

	write = madvise_need_mmap_write(behavior);
	if (write)
		down_write(&current->mm->mmap_sem);
	else
		down_read(&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 (prev && start < prev->vm_end)
			start = prev->vm_end;
		error = unmapped_error;
		if (start >= end)
			goto out;
		if (prev)
			vma = prev->vm_next;
		else	/* madvise_remove dropped mmap_sem */
			vma = find_vma(current->mm, start);
	}
out:
	if (write)
		up_write(&current->mm->mmap_sem);
	else
		up_read(&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>
afcf938e	12	#include <linux/page-isolation.h>
1da177e4	13	#include <linux/hugetlb.h>
e8edc6e0	14	#include <linux/sched.h>
f8af4da3	15	#include <linux/ksm.h>
1da177e4	16
0a27a14a NP	17	/*
	18	* Any behaviour which results in changes to the vma->vm_flags needs to
	19	* take mmap_sem for writing. Others, which simply traverse vmas, need
	20	* to only take it for reading.
	21	*/
	22	static int madvise_need_mmap_write(int behavior)
	23	{
	24	switch (behavior) {
	25	case MADV_REMOVE:
	26	case MADV_WILLNEED:
	27	case MADV_DONTNEED:
	28	return 0;
	29	default:
	30	/* be safe, default to 1. list exceptions explicitly */
	31	return 1;
	32	}
	33	}
	34
1da177e4 LT	35	/*
	36	* We can potentially split a vm area into separate
	37	* areas, each area with its own behavior.
	38	*/
05b74384 PM	39	static long madvise_behavior(struct vm_area_struct * vma,
	40	struct vm_area_struct **prev,
	41	unsigned long start, unsigned long end, int behavior)
1da177e4 LT	42	{
	43	struct mm_struct * mm = vma->vm_mm;
	44	int error = 0;
05b74384	45	pgoff_t pgoff;
3866ea90	46	unsigned long new_flags = vma->vm_flags;
e798c6e8 PM	47
e798c6e8 PM	48	switch (behavior) {
f8225661 MT	49	case MADV_NORMAL:
	50	new_flags = new_flags & ~VM_RAND_READ & ~VM_SEQ_READ;
	51	break;
e798c6e8	52	case MADV_SEQUENTIAL:
f8225661	53	new_flags = (new_flags & ~VM_RAND_READ) \| VM_SEQ_READ;
e798c6e8 PM	54	break;
e798c6e8 PM	55	case MADV_RANDOM:
f8225661	56	new_flags = (new_flags & ~VM_SEQ_READ) \| VM_RAND_READ;
e798c6e8	57	break;
f8225661 MT	58	case MADV_DONTFORK:
	59	new_flags \|= VM_DONTCOPY;
	60	break;
	61	case MADV_DOFORK:
3866ea90 HD	62	if (vma->vm_flags & VM_IO) {
	63	error = -EINVAL;
	64	goto out;
	65	}
f8225661	66	new_flags &= ~VM_DONTCOPY;
e798c6e8	67	break;
f8af4da3 HD	68	case MADV_MERGEABLE:
	69	case MADV_UNMERGEABLE:
	70	error = ksm_madvise(vma, start, end, behavior, &new_flags);
	71	if (error)
	72	goto out;
	73	break;
0af4e98b	74	case MADV_HUGEPAGE:
a664b2d8	75	case MADV_NOHUGEPAGE:
60ab3244	76	error = hugepage_madvise(vma, &new_flags, behavior);
0af4e98b AA	77	if (error)
	78	goto out;
	79	break;
e798c6e8 PM	80	}
e798c6e8 PM	81
05b74384 PM	82	if (new_flags == vma->vm_flags) {
05b74384 PM	83	*prev = vma;
836d5ffd	84	goto out;
05b74384 PM	85	}
	86
	87	pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT);
	88	prev = vma_merge(mm, prev, start, end, new_flags, vma->anon_vma,
	89	vma->vm_file, pgoff, vma_policy(vma));
	90	if (*prev) {
	91	vma = *prev;
	92	goto success;
	93	}
	94
	95	*prev = vma;
1da177e4 LT	96
	97	if (start != vma->vm_start) {
	98	error = split_vma(mm, vma, start, 1);
	99	if (error)
	100	goto out;
	101	}
	102
	103	if (end != vma->vm_end) {
	104	error = split_vma(mm, vma, end, 0);
	105	if (error)
	106	goto out;
	107	}
	108
836d5ffd	109	success:
1da177e4 LT	110	/*
	111	* vm_flags is protected by the mmap_sem held in write mode.
	112	*/
e798c6e8	113	vma->vm_flags = new_flags;
1da177e4 LT	114
	115	out:
	116	if (error == -ENOMEM)
	117	error = -EAGAIN;
	118	return error;
	119	}
	120
	121	/*
	122	* Schedule all required I/O operations. Do not wait for completion.
	123	*/
	124	static long madvise_willneed(struct vm_area_struct * vma,
05b74384	125	struct vm_area_struct ** prev,
1da177e4 LT	126	unsigned long start, unsigned long end)
	127	{
	128	struct file *file = vma->vm_file;
	129
1bef4003 S	130	if (!file)
	131	return -EBADF;
	132
70688e4d	133	if (file->f_mapping->a_ops->get_xip_mem) {
fe77ba6f CO	134	/* no bad return value, but ignore advice */
	135	return 0;
	136	}
	137
05b74384	138	*prev = vma;
1da177e4 LT	139	start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	140	if (end > vma->vm_end)
	141	end = vma->vm_end;
	142	end = ((end - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
	143
f7e839dd	144	force_page_cache_readahead(file->f_mapping, file, start, end - start);
1da177e4 LT	145	return 0;
	146	}
	147
	148	/*
	149	* Application no longer needs these pages. If the pages are dirty,
	150	* it's OK to just throw them away. The app will be more careful about
	151	* data it wants to keep. Be sure to free swap resources too. The
7e6cbea3	152	* zap_page_range call sets things up for shrink_active_list to actually free
1da177e4 LT	153	* these pages later if no one else has touched them in the meantime,
1da177e4 LT	154	* although we could add these pages to a global reuse list for
7e6cbea3	155	* shrink_active_list to pick up before reclaiming other pages.
1da177e4 LT	156	*
	157	* NB: This interface discards data rather than pushes it out to swap,
	158	* as some implementations do. This has performance implications for
	159	* applications like large transactional databases which want to discard
	160	* pages in anonymous maps after committing to backing store the data
	161	* that was kept in them. There is no reason to write this data out to
	162	* the swap area if the application is discarding it.
	163	*
	164	* An interface that causes the system to free clean pages and flush
	165	* dirty pages is already available as msync(MS_INVALIDATE).
	166	*/
	167	static long madvise_dontneed(struct vm_area_struct * vma,
05b74384	168	struct vm_area_struct ** prev,
1da177e4 LT	169	unsigned long start, unsigned long end)
1da177e4 LT	170	{
05b74384	171	*prev = vma;
6aab341e	172	if (vma->vm_flags & (VM_LOCKED\|VM_HUGETLB\|VM_PFNMAP))
1da177e4 LT	173	return -EINVAL;
	174
	175	if (unlikely(vma->vm_flags & VM_NONLINEAR)) {
	176	struct zap_details details = {
	177	.nonlinear_vma = vma,
	178	.last_index = ULONG_MAX,
	179	};
	180	zap_page_range(vma, start, end - start, &details);
	181	} else
	182	zap_page_range(vma, start, end - start, NULL);
	183	return 0;
	184	}
	185
f6b3ec23 BP	186	/*
	187	* Application wants to free up the pages and associated backing store.
	188	* This is effectively punching a hole into the middle of a file.
	189	*
	190	* NOTE: Currently, only shmfs/tmpfs is supported for this operation.
	191	* Other filesystems return -ENOSYS.
	192	*/
	193	static long madvise_remove(struct vm_area_struct *vma,
00e9fa2d	194	struct vm_area_struct **prev,
f6b3ec23 BP	195	unsigned long start, unsigned long end)
	196	{
	197	struct address_space *mapping;
90ed52eb HD	198	loff_t offset, endoff;
90ed52eb HD	199	int error;
f6b3ec23	200
90ed52eb	201	prev = NULL; / tell sys_madvise we drop mmap_sem */
00e9fa2d	202
f6b3ec23 BP	203	if (vma->vm_flags & (VM_LOCKED\|VM_NONLINEAR\|VM_HUGETLB))
	204	return -EINVAL;
	205
	206	if (!vma->vm_file \|\| !vma->vm_file->f_mapping
	207	\|\| !vma->vm_file->f_mapping->host) {
	208	return -EINVAL;
	209	}
	210
69cf0fac HD	211	if ((vma->vm_flags & (VM_SHARED\|VM_WRITE)) != (VM_SHARED\|VM_WRITE))
	212	return -EACCES;
	213
f6b3ec23 BP	214	mapping = vma->vm_file->f_mapping;
	215
	216	offset = (loff_t)(start - vma->vm_start)
	217	+ ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
	218	endoff = (loff_t)(end - vma->vm_start - 1)
	219	+ ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
90ed52eb	220
bd5fe6c5	221	/* vmtruncate_range needs to take i_mutex */
0a27a14a	222	up_read(&current->mm->mmap_sem);
90ed52eb	223	error = vmtruncate_range(mapping->host, offset, endoff);
0a27a14a	224	down_read(&current->mm->mmap_sem);
90ed52eb	225	return error;
f6b3ec23 BP	226	}
f6b3ec23 BP	227
9893e49d AK	228	#ifdef CONFIG_MEMORY_FAILURE
	229	/*
	230	* Error injection support for memory error handling.
	231	*/
afcf938e	232	static int madvise_hwpoison(int bhv, unsigned long start, unsigned long end)
9893e49d AK	233	{
	234	int ret = 0;
	235
	236	if (!capable(CAP_SYS_ADMIN))
	237	return -EPERM;
	238	for (; start < end; start += PAGE_SIZE) {
	239	struct page *p;
d15f107d	240	int ret = get_user_pages_fast(start, 1, 0, &p);
9893e49d AK	241	if (ret != 1)
9893e49d AK	242	return ret;
afcf938e AK	243	if (bhv == MADV_SOFT_OFFLINE) {
	244	printk(KERN_INFO "Soft offlining page %lx at %lx\n",
	245	page_to_pfn(p), start);
	246	ret = soft_offline_page(p, MF_COUNT_INCREASED);
	247	if (ret)
	248	break;
	249	continue;
	250	}
9893e49d AK	251	printk(KERN_INFO "Injecting memory failure for page %lx at %lx\n",
	252	page_to_pfn(p), start);
	253	/* Ignore return value for now */
82ba011b	254	__memory_failure(page_to_pfn(p), 0, MF_COUNT_INCREASED);
9893e49d AK	255	}
	256	return ret;
	257	}
	258	#endif
	259
165cd402	260	static long
	261	madvise_vma(struct vm_area_struct vma, struct vm_area_struct *prev,
	262	unsigned long start, unsigned long end, int behavior)
1da177e4	263	{
1da177e4	264	switch (behavior) {
f6b3ec23	265	case MADV_REMOVE:
3866ea90	266	return madvise_remove(vma, prev, start, end);
1da177e4	267	case MADV_WILLNEED:
3866ea90	268	return madvise_willneed(vma, prev, start, end);
1da177e4	269	case MADV_DONTNEED:
3866ea90	270	return madvise_dontneed(vma, prev, start, end);
1da177e4	271	default:
3866ea90	272	return madvise_behavior(vma, prev, start, end, behavior);
1da177e4	273	}
1da177e4 LT	274	}
1da177e4 LT	275
75927af8 NP	276	static int
	277	madvise_behavior_valid(int behavior)
	278	{
	279	switch (behavior) {
	280	case MADV_DOFORK:
	281	case MADV_DONTFORK:
	282	case MADV_NORMAL:
	283	case MADV_SEQUENTIAL:
	284	case MADV_RANDOM:
	285	case MADV_REMOVE:
	286	case MADV_WILLNEED:
	287	case MADV_DONTNEED:
f8af4da3 HD	288	#ifdef CONFIG_KSM
	289	case MADV_MERGEABLE:
	290	case MADV_UNMERGEABLE:
0af4e98b AA	291	#endif
	292	#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	293	case MADV_HUGEPAGE:
a664b2d8	294	case MADV_NOHUGEPAGE:
f8af4da3	295	#endif
75927af8 NP	296	return 1;
	297
	298	default:
	299	return 0;
	300	}
	301	}
3866ea90	302
1da177e4 LT	303	/*
	304	* The madvise(2) system call.
	305	*
	306	* Applications can use madvise() to advise the kernel how it should
	307	* handle paging I/O in this VM area. The idea is to help the kernel
	308	* use appropriate read-ahead and caching techniques. The information
	309	* provided is advisory only, and can be safely disregarded by the
	310	* kernel without affecting the correct operation of the application.
	311	*
	312	* behavior values:
	313	* MADV_NORMAL - the default behavior is to read clusters. This
	314	* results in some read-ahead and read-behind.
	315	* MADV_RANDOM - the system should read the minimum amount of data
	316	* on any access, since it is unlikely that the appli-
	317	* cation will need more than what it asks for.
	318	* MADV_SEQUENTIAL - pages in the given range will probably be accessed
	319	* once, so they can be aggressively read ahead, and
	320	* can be freed soon after they are accessed.
	321	* MADV_WILLNEED - the application is notifying the system to read
	322	* some pages ahead.
	323	* MADV_DONTNEED - the application is finished with the given range,
	324	* so the kernel can free resources associated with it.
f6b3ec23 BP	325	* MADV_REMOVE - the application wants to free up the given range of
f6b3ec23 BP	326	* pages and associated backing store.
3866ea90 HD	327	* MADV_DONTFORK - omit this area from child's address space when forking:
	328	* typically, to avoid COWing pages pinned by get_user_pages().
	329	* MADV_DOFORK - cancel MADV_DONTFORK: no longer omit this area when forking.
f8af4da3 HD	330	* MADV_MERGEABLE - the application recommends that KSM try to merge pages in
	331	* this area with pages of identical content from other such areas.
	332	* MADV_UNMERGEABLE- cancel MADV_MERGEABLE: no longer merge pages with others.
1da177e4 LT	333	*
	334	* return values:
	335	* zero - success
	336	* -EINVAL - start + len < 0, start is not page-aligned,
	337	* "behavior" is not a valid value, or application
	338	* is attempting to release locked or shared pages.
	339	* -ENOMEM - addresses in the specified range are not currently
	340	* mapped, or are outside the AS of the process.
	341	* -EIO - an I/O error occurred while paging in data.
	342	* -EBADF - map exists, but area maps something that isn't a file.
	343	* -EAGAIN - a kernel resource was temporarily unavailable.
	344	*/
3480b257	345	SYSCALL_DEFINE3(madvise, unsigned long, start, size_t, len_in, int, behavior)
1da177e4	346	{
05b74384 PM	347	unsigned long end, tmp;
05b74384 PM	348	struct vm_area_struct * vma, *prev;
1da177e4 LT	349	int unmapped_error = 0;
1da177e4 LT	350	int error = -EINVAL;
f7977793	351	int write;
1da177e4 LT	352	size_t len;
1da177e4 LT	353
9893e49d	354	#ifdef CONFIG_MEMORY_FAILURE
afcf938e AK	355	if (behavior == MADV_HWPOISON \|\| behavior == MADV_SOFT_OFFLINE)
afcf938e AK	356	return madvise_hwpoison(behavior, start, start+len_in);
9893e49d	357	#endif
75927af8 NP	358	if (!madvise_behavior_valid(behavior))
	359	return error;
	360
f7977793 JB	361	write = madvise_need_mmap_write(behavior);
f7977793 JB	362	if (write)
0a27a14a NP	363	down_write(&current->mm->mmap_sem);
	364	else
	365	down_read(&current->mm->mmap_sem);
1da177e4 LT	366
	367	if (start & ~PAGE_MASK)
	368	goto out;
	369	len = (len_in + ~PAGE_MASK) & PAGE_MASK;
	370
	371	/* Check to see whether len was rounded up from small -ve to zero */
	372	if (len_in && !len)
	373	goto out;
	374
	375	end = start + len;
	376	if (end < start)
	377	goto out;
	378
	379	error = 0;
	380	if (end == start)
	381	goto out;
	382
	383	/*
	384	* If the interval [start,end) covers some unmapped address
	385	* ranges, just ignore them, but return -ENOMEM at the end.
05b74384	386	* - different from the way of handling in mlock etc.
1da177e4	387	*/
05b74384	388	vma = find_vma_prev(current->mm, start, &prev);
836d5ffd HD	389	if (vma && start > vma->vm_start)
	390	prev = vma;
	391
1da177e4 LT	392	for (;;) {
	393	/* Still start < end. */
	394	error = -ENOMEM;
	395	if (!vma)
	396	goto out;
	397
05b74384	398	/* Here start < (end\|vma->vm_end). */
1da177e4 LT	399	if (start < vma->vm_start) {
	400	unmapped_error = -ENOMEM;
	401	start = vma->vm_start;
05b74384 PM	402	if (start >= end)
05b74384 PM	403	goto out;
1da177e4 LT	404	}
1da177e4 LT	405
05b74384 PM	406	/* Here vma->vm_start <= start < (end\|vma->vm_end) */
	407	tmp = vma->vm_end;
	408	if (end < tmp)
	409	tmp = end;
1da177e4	410
05b74384 PM	411	/* Here vma->vm_start <= start < tmp <= (end\|vma->vm_end). */
05b74384 PM	412	error = madvise_vma(vma, &prev, start, tmp, behavior);
1da177e4 LT	413	if (error)
1da177e4 LT	414	goto out;
05b74384	415	start = tmp;
90ed52eb	416	if (prev && start < prev->vm_end)
05b74384 PM	417	start = prev->vm_end;
	418	error = unmapped_error;
	419	if (start >= end)
	420	goto out;
90ed52eb HD	421	if (prev)
	422	vma = prev->vm_next;
	423	else /* madvise_remove dropped mmap_sem */
	424	vma = find_vma(current->mm, start);
1da177e4	425	}
1da177e4	426	out:
f7977793	427	if (write)
0a27a14a NP	428	up_write(&current->mm->mmap_sem);
	429	else
	430	up_read(&current->mm->mmap_sem);
	431
1da177e4 LT	432	return error;
1da177e4 LT	433	}