Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 05b74384 authored by Prasanna Meda's avatar Prasanna Meda Committed by Linus Torvalds
Browse files

[PATCH] madvise: merge the maps



This attempts to merge back the split maps.  This code is mostly copied
from Chrisw's mlock merging from post 2.6.11 trees.  The only difference is
in munmapped_error handling.  Also passed prev to willneed/dontneed,
eventhogh they do not handle it now, since I felt it will be cleaner,
instead of handling prev in madvise_vma in some cases and in subfunction in
some cases.

Signed-off-by: default avatarPrasanna Meda <pmeda@akamai.com>
Signed-off-by: default avatarAndrew Morton <akpm@osdl.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@osdl.org>
parent e798c6e8
Loading
Loading
Loading
Loading
+51 −29
Original line number Diff line number Diff line
@@ -8,17 +8,20 @@
#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, unsigned long start,
			     unsigned long end, int 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) {
@@ -32,8 +35,20 @@ static long madvise_behavior(struct vm_area_struct * vma, unsigned long start,
		break;
	}

	if (new_flags == vma->vm_flags)
		goto out;
	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);
@@ -56,6 +71,7 @@ static long madvise_behavior(struct vm_area_struct * vma, unsigned long start,
out:
	if (error == -ENOMEM)
		error = -EAGAIN;
success:
	return error;
}

@@ -63,6 +79,7 @@ static long madvise_behavior(struct vm_area_struct * vma, unsigned long start,
 * 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;
@@ -70,6 +87,7 @@ static long madvise_willneed(struct vm_area_struct * vma,
	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;
@@ -100,8 +118,10 @@ static long madvise_willneed(struct vm_area_struct * vma,
 * 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;

@@ -116,8 +136,8 @@ static long madvise_dontneed(struct vm_area_struct * vma,
	return 0;
}

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

@@ -125,15 +145,15 @@ static long madvise_vma(struct vm_area_struct * vma, unsigned long start,
	case MADV_NORMAL:
	case MADV_SEQUENTIAL:
	case MADV_RANDOM:
		error = madvise_behavior(vma, start, end, behavior);
		error = madvise_behavior(vma, prev, start, end, behavior);
		break;

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

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

	default:
@@ -180,8 +200,8 @@ static long madvise_vma(struct vm_area_struct * vma, unsigned long start,
 */
asmlinkage long sys_madvise(unsigned long start, size_t len_in, int behavior)
{
	unsigned long end;
	struct vm_area_struct * vma;
	unsigned long end, tmp;
	struct vm_area_struct * vma, *prev;
	int unmapped_error = 0;
	int error = -EINVAL;
	size_t len;
@@ -207,40 +227,42 @@ asmlinkage long sys_madvise(unsigned long start, size_t len_in, int behavior)
	/*
	 * 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(current->mm, start);
	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 < vma->vm_end. */
		/* 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 < vma->vm_end. */
		if (end <= vma->vm_end) {
			if (start < end) {
				error = madvise_vma(vma, start, end,
							behavior);
		/* 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;
	}

		/* Here vma->vm_start <= start < vma->vm_end < end. */
		error = madvise_vma(vma, start, vma->vm_end, behavior);
		if (error)
			goto out;
		start = vma->vm_end;
		vma = vma->vm_next;
	}

out:
	up_write(&current->mm->mmap_sem);
	return error;