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

/*
 *	mm/mremap.c
 *
 *	(C) Copyright 1996 Linus Torvalds
 *
 *	Address space accounting code	<alan@lxorguk.ukuu.org.uk>
 *	(C) Copyright 2002 Red Hat Inc, All Rights Reserved
 */

#include <linux/mm.h>
#include <linux/hugetlb.h>
#include <linux/slab.h>
#include <linux/shm.h>
#include <linux/ksm.h>
#include <linux/mman.h>
#include <linux/swap.h>
#include <linux/capability.h>
#include <linux/fs.h>
#include <linux/highmem.h>
#include <linux/security.h>
#include <linux/syscalls.h>
#include <linux/mmu_notifier.h>

#include <asm/uaccess.h>
#include <asm/cacheflush.h>
#include <asm/tlbflush.h>

#include "internal.h"

#ifndef arch_mmap_check
#define arch_mmap_check(addr, len, flags)	(0)
#endif

static pmd_t *get_old_pmd(struct mm_struct *mm, unsigned long addr)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;

	pgd = pgd_offset(mm, addr);
	if (pgd_none_or_clear_bad(pgd))
		return NULL;

	pud = pud_offset(pgd, addr);
	if (pud_none_or_clear_bad(pud))
		return NULL;

	pmd = pmd_offset(pud, addr);
	if (pmd_none_or_clear_bad(pmd))
		return NULL;

	return pmd;
}

static pmd_t *alloc_new_pmd(struct mm_struct *mm, unsigned long addr)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;

	pgd = pgd_offset(mm, addr);
	pud = pud_alloc(mm, pgd, addr);
	if (!pud)
		return NULL;

	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return NULL;

	if (!pmd_present(*pmd) && __pte_alloc(mm, pmd, addr))
		return NULL;

	return pmd;
}

static void move_ptes(struct vm_area_struct *vma, pmd_t *old_pmd,
		unsigned long old_addr, unsigned long old_end,
		struct vm_area_struct *new_vma, pmd_t *new_pmd,
		unsigned long new_addr)
{
	struct address_space *mapping = NULL;
	struct mm_struct *mm = vma->vm_mm;
	pte_t *old_pte, *new_pte, pte;
	spinlock_t *old_ptl, *new_ptl;
	unsigned long old_start;

	old_start = old_addr;
	mmu_notifier_invalidate_range_start(vma->vm_mm,
					    old_start, old_end);
	if (vma->vm_file) {
		/*
		 * Subtle point from Rajesh Venkatasubramanian: before
		 * moving file-based ptes, we must lock truncate_pagecache
		 * out, since it might clean the dst vma before the src vma,
		 * and we propagate stale pages into the dst afterward.
		 */
		mapping = vma->vm_file->f_mapping;
		spin_lock(&mapping->i_mmap_lock);
		if (new_vma->vm_truncate_count &&
		    new_vma->vm_truncate_count != vma->vm_truncate_count)
			new_vma->vm_truncate_count = 0;
	}

	/*
	 * We don't have to worry about the ordering of src and dst
	 * pte locks because exclusive mmap_sem prevents deadlock.
	 */
	old_pte = pte_offset_map_lock(mm, old_pmd, old_addr, &old_ptl);
 	new_pte = pte_offset_map_nested(new_pmd, new_addr);
	new_ptl = pte_lockptr(mm, new_pmd);
	if (new_ptl != old_ptl)
		spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING);
	arch_enter_lazy_mmu_mode();

	for (; old_addr < old_end; old_pte++, old_addr += PAGE_SIZE,
				   new_pte++, new_addr += PAGE_SIZE) {
		if (pte_none(*old_pte))
			continue;
		pte = ptep_clear_flush(vma, old_addr, old_pte);
		pte = move_pte(pte, new_vma->vm_page_prot, old_addr, new_addr);
		set_pte_at(mm, new_addr, new_pte, pte);
	}

	arch_leave_lazy_mmu_mode();
	if (new_ptl != old_ptl)
		spin_unlock(new_ptl);
	pte_unmap_nested(new_pte - 1);
	pte_unmap_unlock(old_pte - 1, old_ptl);
	if (mapping)
		spin_unlock(&mapping->i_mmap_lock);
	mmu_notifier_invalidate_range_end(vma->vm_mm, old_start, old_end);
}

#define LATENCY_LIMIT	(64 * PAGE_SIZE)

unsigned long move_page_tables(struct vm_area_struct *vma,
		unsigned long old_addr, struct vm_area_struct *new_vma,
		unsigned long new_addr, unsigned long len)
{
	unsigned long extent, next, old_end;
	pmd_t *old_pmd, *new_pmd;

	old_end = old_addr + len;
	flush_cache_range(vma, old_addr, old_end);

	for (; old_addr < old_end; old_addr += extent, new_addr += extent) {
		cond_resched();
		next = (old_addr + PMD_SIZE) & PMD_MASK;
		if (next - 1 > old_end)
			next = old_end;
		extent = next - old_addr;
		old_pmd = get_old_pmd(vma->vm_mm, old_addr);
		if (!old_pmd)
			continue;
		new_pmd = alloc_new_pmd(vma->vm_mm, new_addr);
		if (!new_pmd)
			break;
		next = (new_addr + PMD_SIZE) & PMD_MASK;
		if (extent > next - new_addr)
			extent = next - new_addr;
		if (extent > LATENCY_LIMIT)
			extent = LATENCY_LIMIT;
		move_ptes(vma, old_pmd, old_addr, old_addr + extent,
				new_vma, new_pmd, new_addr);
	}

	return len + old_addr - old_end;	/* how much done */
}

static unsigned long move_vma(struct vm_area_struct *vma,
		unsigned long old_addr, unsigned long old_len,
		unsigned long new_len, unsigned long new_addr)
{
	struct mm_struct *mm = vma->vm_mm;
	struct vm_area_struct *new_vma;
	unsigned long vm_flags = vma->vm_flags;
	unsigned long new_pgoff;
	unsigned long moved_len;
	unsigned long excess = 0;
	unsigned long hiwater_vm;
	int split = 0;
	int err;

	/*
	 * We'd prefer to avoid failure later on in do_munmap:
	 * which may split one vma into three before unmapping.
	 */
	if (mm->map_count >= sysctl_max_map_count - 3)
		return -ENOMEM;

	/*
	 * Advise KSM to break any KSM pages in the area to be moved:
	 * it would be confusing if they were to turn up at the new
	 * location, where they happen to coincide with different KSM
	 * pages recently unmapped.  But leave vma->vm_flags as it was,
	 * so KSM can come around to merge on vma and new_vma afterwards.
	 */
	err = ksm_madvise(vma, old_addr, old_addr + old_len,
						MADV_UNMERGEABLE, &vm_flags);
	if (err)
		return err;

	new_pgoff = vma->vm_pgoff + ((old_addr - vma->vm_start) >> PAGE_SHIFT);
	new_vma = copy_vma(&vma, new_addr, new_len, new_pgoff);
	if (!new_vma)
		return -ENOMEM;

	moved_len = move_page_tables(vma, old_addr, new_vma, new_addr, old_len);
	if (moved_len < old_len) {
		/*
		 * On error, move entries back from new area to old,
		 * which will succeed since page tables still there,
		 * and then proceed to unmap new area instead of old.
		 */
		move_page_tables(new_vma, new_addr, vma, old_addr, moved_len);
		vma = new_vma;
		old_len = new_len;
		old_addr = new_addr;
		new_addr = -ENOMEM;
	}

	/* Conceal VM_ACCOUNT so old reservation is not undone */
	if (vm_flags & VM_ACCOUNT) {
		vma->vm_flags &= ~VM_ACCOUNT;
		excess = vma->vm_end - vma->vm_start - old_len;
		if (old_addr > vma->vm_start &&
		    old_addr + old_len < vma->vm_end)
			split = 1;
	}

	/*
	 * If we failed to move page tables we still do total_vm increment
	 * since do_munmap() will decrement it by old_len == new_len.
	 *
	 * Since total_vm is about to be raised artificially high for a
	 * moment, we need to restore high watermark afterwards: if stats
	 * are taken meanwhile, total_vm and hiwater_vm appear too high.
	 * If this were a serious issue, we'd add a flag to do_munmap().
	 */
	hiwater_vm = mm->hiwater_vm;
	mm->total_vm += new_len >> PAGE_SHIFT;
	vm_stat_account(mm, vma->vm_flags, vma->vm_file, new_len>>PAGE_SHIFT);

	if (do_munmap(mm, old_addr, old_len) < 0) {
		/* OOM: unable to split vma, just get accounts right */
		vm_unacct_memory(excess >> PAGE_SHIFT);
		excess = 0;
	}
	mm->hiwater_vm = hiwater_vm;

	/* Restore VM_ACCOUNT if one or two pieces of vma left */
	if (excess) {
		vma->vm_flags |= VM_ACCOUNT;
		if (split)
			vma->vm_next->vm_flags |= VM_ACCOUNT;
	}

	if (vm_flags & VM_LOCKED) {
		mm->locked_vm += new_len >> PAGE_SHIFT;
		if (new_len > old_len)
			mlock_vma_pages_range(new_vma, new_addr + old_len,
						       new_addr + new_len);
	}

	return new_addr;
}

static struct vm_area_struct *vma_to_resize(unsigned long addr,
	unsigned long old_len, unsigned long new_len, unsigned long *p)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma = find_vma(mm, addr);

	if (!vma || vma->vm_start > addr)
		goto Efault;

	if (is_vm_hugetlb_page(vma))
		goto Einval;

	/* We can't remap across vm area boundaries */
	if (old_len > vma->vm_end - addr)
		goto Efault;

	if (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP)) {
		if (new_len > old_len)
			goto Efault;
	}

	if (vma->vm_flags & VM_LOCKED) {
		unsigned long locked, lock_limit;
		locked = mm->locked_vm << PAGE_SHIFT;
		lock_limit = current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur;
		locked += new_len - old_len;
		if (locked > lock_limit && !capable(CAP_IPC_LOCK))
			goto Eagain;
	}

	if (!may_expand_vm(mm, (new_len - old_len) >> PAGE_SHIFT))
		goto Enomem;

	if (vma->vm_flags & VM_ACCOUNT) {
		unsigned long charged = (new_len - old_len) >> PAGE_SHIFT;
		if (security_vm_enough_memory(charged))
			goto Efault;
		*p = charged;
	}

	return vma;

Efault:	/* very odd choice for most of the cases, but... */
	return ERR_PTR(-EFAULT);
Einval:
	return ERR_PTR(-EINVAL);
Enomem:
	return ERR_PTR(-ENOMEM);
Eagain:
	return ERR_PTR(-EAGAIN);
}

static unsigned long mremap_to(unsigned long addr,
	unsigned long old_len, unsigned long new_addr,
	unsigned long new_len)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma;
	unsigned long ret = -EINVAL;
	unsigned long charged = 0;

	if (new_addr & ~PAGE_MASK)
		goto out;

	if (new_len > TASK_SIZE || new_addr > TASK_SIZE - new_len)
		goto out;

	/* Check if the location we're moving into overlaps the
	 * old location at all, and fail if it does.
	 */
	if ((new_addr <= addr) && (new_addr+new_len) > addr)
		goto out;

	if ((addr <= new_addr) && (addr+old_len) > new_addr)
		goto out;

	ret = security_file_mmap(NULL, 0, 0, 0, new_addr, 1);
	if (ret)
		goto out;

	ret = do_munmap(mm, new_addr, new_len);
	if (ret)
		goto out;

	if (old_len >= new_len) {
		ret = do_munmap(mm, addr+new_len, old_len - new_len);
		if (ret && old_len != new_len)
			goto out;
		old_len = new_len;
	}

	vma = vma_to_resize(addr, old_len, new_len, &charged);
	if (IS_ERR(vma)) {
		ret = PTR_ERR(vma);
		goto out;
	}

	ret = move_vma(vma, addr, old_len, new_len, new_addr);
	if (ret & ~PAGE_MASK)
		vm_unacct_memory(charged);

out:
	return ret;
}

static int vma_expandable(struct vm_area_struct *vma, unsigned long delta)
{
	unsigned long end = vma->vm_end + delta;
	unsigned long max_addr = TASK_SIZE;
	if (vma->vm_next)
		max_addr = vma->vm_next->vm_start;
	if (max_addr < end || end < vma->vm_end)
		return 0;
	if (arch_mmap_check(vma->vm_start, end - vma->vm_start, MAP_FIXED))
		return 0;
	if (get_unmapped_area(NULL, vma->vm_start, end - vma->vm_start,
			      0, MAP_FIXED) & ~PAGE_MASK)
		return 0;
	return 1;
}

/*
 * Expand (or shrink) an existing mapping, potentially moving it at the
 * same time (controlled by the MREMAP_MAYMOVE flag and available VM space)
 *
 * MREMAP_FIXED option added 5-Dec-1999 by Benjamin LaHaise
 * This option implies MREMAP_MAYMOVE.
 */
unsigned long do_mremap(unsigned long addr,
	unsigned long old_len, unsigned long new_len,
	unsigned long flags, unsigned long new_addr)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma;
	unsigned long ret = -EINVAL;
	unsigned long charged = 0;

	if (flags & ~(MREMAP_FIXED | MREMAP_MAYMOVE))
		goto out;

	if (addr & ~PAGE_MASK)
		goto out;

	old_len = PAGE_ALIGN(old_len);
	new_len = PAGE_ALIGN(new_len);

	/*
	 * We allow a zero old-len as a special case
	 * for DOS-emu "duplicate shm area" thing. But
	 * a zero new-len is nonsensical.
	 */
	if (!new_len)
		goto out;

	if (flags & MREMAP_FIXED) {
		if (flags & MREMAP_MAYMOVE)
			ret = mremap_to(addr, old_len, new_addr, new_len);
		goto out;
	}

	/*
	 * Always allow a shrinking remap: that just unmaps
	 * the unnecessary pages..
	 * do_munmap does all the needed commit accounting
	 */
	if (old_len >= new_len) {
		ret = do_munmap(mm, addr+new_len, old_len - new_len);
		if (ret && old_len != new_len)
			goto out;
		ret = addr;
		goto out;
	}

	/*
	 * Ok, we need to grow..
	 */
	vma = vma_to_resize(addr, old_len, new_len, &charged);
	if (IS_ERR(vma)) {
		ret = PTR_ERR(vma);
		goto out;
	}

	/* old_len exactly to the end of the area..
	 */
	if (old_len == vma->vm_end - addr) {
		/* can we just expand the current mapping? */
		if (vma_expandable(vma, new_len - old_len)) {
			int pages = (new_len - old_len) >> PAGE_SHIFT;

			vma_adjust(vma, vma->vm_start,
				addr + new_len, vma->vm_pgoff, NULL);

			mm->total_vm += pages;
			vm_stat_account(mm, vma->vm_flags, vma->vm_file, pages);
			if (vma->vm_flags & VM_LOCKED) {
				mm->locked_vm += pages;
				mlock_vma_pages_range(vma, addr + old_len,
						   addr + new_len);
			}
			ret = addr;
			goto out;
		}
	}

	/*
	 * We weren't able to just expand or shrink the area,
	 * we need to create a new one and move it..
	 */
	ret = -ENOMEM;
	if (flags & MREMAP_MAYMOVE) {
		unsigned long map_flags = 0;
		if (vma->vm_flags & VM_MAYSHARE)
			map_flags |= MAP_SHARED;

		new_addr = get_unmapped_area(vma->vm_file, 0, new_len,
					vma->vm_pgoff, map_flags);
		if (new_addr & ~PAGE_MASK) {
			ret = new_addr;
			goto out;
		}

		ret = security_file_mmap(NULL, 0, 0, 0, new_addr, 1);
		if (ret)
			goto out;
		ret = move_vma(vma, addr, old_len, new_len, new_addr);
	}
out:
	if (ret & ~PAGE_MASK)
		vm_unacct_memory(charged);
	return ret;
}

SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len,
		unsigned long, new_len, unsigned long, flags,
		unsigned long, new_addr)
{
	unsigned long ret;

	down_write(&current->mm->mmap_sem);
	ret = do_mremap(addr, old_len, new_len, flags, new_addr);
	up_write(&current->mm->mmap_sem);
	return ret;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* mm/mremap.c
	3	*
	4	* (C) Copyright 1996 Linus Torvalds
	5	*
046c6884	6	* Address space accounting code <alan@lxorguk.ukuu.org.uk>
1da177e4 LT	7	* (C) Copyright 2002 Red Hat Inc, All Rights Reserved
	8	*/
	9
	10	#include <linux/mm.h>
	11	#include <linux/hugetlb.h>
	12	#include <linux/slab.h>
	13	#include <linux/shm.h>
1ff82995	14	#include <linux/ksm.h>
1da177e4 LT	15	#include <linux/mman.h>
1da177e4 LT	16	#include <linux/swap.h>
c59ede7b	17	#include <linux/capability.h>
1da177e4 LT	18	#include <linux/fs.h>
	19	#include <linux/highmem.h>
	20	#include <linux/security.h>
	21	#include <linux/syscalls.h>
cddb8a5c	22	#include <linux/mmu_notifier.h>
1da177e4 LT	23
	24	#include <asm/uaccess.h>
	25	#include <asm/cacheflush.h>
	26	#include <asm/tlbflush.h>
	27
ba470de4 RR	28	#include "internal.h"
ba470de4 RR	29
f106af4e AV	30	#ifndef arch_mmap_check
	31	#define arch_mmap_check(addr, len, flags) (0)
	32	#endif
	33
7be7a546	34	static pmd_t get_old_pmd(struct mm_struct mm, unsigned long addr)
1da177e4 LT	35	{
	36	pgd_t *pgd;
	37	pud_t *pud;
	38	pmd_t *pmd;
	39
	40	pgd = pgd_offset(mm, addr);
	41	if (pgd_none_or_clear_bad(pgd))
	42	return NULL;
	43
	44	pud = pud_offset(pgd, addr);
	45	if (pud_none_or_clear_bad(pud))
	46	return NULL;
	47
	48	pmd = pmd_offset(pud, addr);
	49	if (pmd_none_or_clear_bad(pmd))
	50	return NULL;
	51
7be7a546	52	return pmd;
1da177e4 LT	53	}
1da177e4 LT	54
7be7a546	55	static pmd_t alloc_new_pmd(struct mm_struct mm, unsigned long addr)
1da177e4 LT	56	{
	57	pgd_t *pgd;
	58	pud_t *pud;
c74df32c	59	pmd_t *pmd;
1da177e4 LT	60
1da177e4 LT	61	pgd = pgd_offset(mm, addr);
1da177e4 LT	62	pud = pud_alloc(mm, pgd, addr);
1da177e4 LT	63	if (!pud)
c74df32c	64	return NULL;
7be7a546	65
1da177e4	66	pmd = pmd_alloc(mm, pud, addr);
7be7a546	67	if (!pmd)
c74df32c	68	return NULL;
7be7a546	69
1bb3630e	70	if (!pmd_present(*pmd) && __pte_alloc(mm, pmd, addr))
c74df32c HD	71	return NULL;
c74df32c HD	72
7be7a546	73	return pmd;
1da177e4 LT	74	}
1da177e4 LT	75
7be7a546 HD	76	static void move_ptes(struct vm_area_struct vma, pmd_t old_pmd,
	77	unsigned long old_addr, unsigned long old_end,
	78	struct vm_area_struct new_vma, pmd_t new_pmd,
	79	unsigned long new_addr)
1da177e4 LT	80	{
	81	struct address_space *mapping = NULL;
	82	struct mm_struct *mm = vma->vm_mm;
7be7a546	83	pte_t old_pte, new_pte, pte;
4c21e2f2	84	spinlock_t old_ptl, new_ptl;
cddb8a5c	85	unsigned long old_start;
1da177e4	86
cddb8a5c AA	87	old_start = old_addr;
	88	mmu_notifier_invalidate_range_start(vma->vm_mm,
	89	old_start, old_end);
1da177e4 LT	90	if (vma->vm_file) {
	91	/*
	92	* Subtle point from Rajesh Venkatasubramanian: before
25d9e2d1	93	* moving file-based ptes, we must lock truncate_pagecache
25d9e2d1	94	* out, since it might clean the dst vma before the src vma,
1da177e4 LT	95	* and we propagate stale pages into the dst afterward.
	96	*/
	97	mapping = vma->vm_file->f_mapping;
	98	spin_lock(&mapping->i_mmap_lock);
	99	if (new_vma->vm_truncate_count &&
	100	new_vma->vm_truncate_count != vma->vm_truncate_count)
	101	new_vma->vm_truncate_count = 0;
	102	}
1da177e4	103
4c21e2f2 HD	104	/*
	105	* We don't have to worry about the ordering of src and dst
	106	* pte locks because exclusive mmap_sem prevents deadlock.
	107	*/
c74df32c HD	108	old_pte = pte_offset_map_lock(mm, old_pmd, old_addr, &old_ptl);
c74df32c HD	109	new_pte = pte_offset_map_nested(new_pmd, new_addr);
4c21e2f2 HD	110	new_ptl = pte_lockptr(mm, new_pmd);
4c21e2f2 HD	111	if (new_ptl != old_ptl)
f20dc5f7	112	spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING);
6606c3e0	113	arch_enter_lazy_mmu_mode();
7be7a546 HD	114
	115	for (; old_addr < old_end; old_pte++, old_addr += PAGE_SIZE,
	116	new_pte++, new_addr += PAGE_SIZE) {
	117	if (pte_none(*old_pte))
	118	continue;
	119	pte = ptep_clear_flush(vma, old_addr, old_pte);
7be7a546 HD	120	pte = move_pte(pte, new_vma->vm_page_prot, old_addr, new_addr);
7be7a546 HD	121	set_pte_at(mm, new_addr, new_pte, pte);
1da177e4	122	}
7be7a546	123
6606c3e0	124	arch_leave_lazy_mmu_mode();
4c21e2f2 HD	125	if (new_ptl != old_ptl)
4c21e2f2 HD	126	spin_unlock(new_ptl);
7be7a546	127	pte_unmap_nested(new_pte - 1);
c74df32c	128	pte_unmap_unlock(old_pte - 1, old_ptl);
1da177e4 LT	129	if (mapping)
1da177e4 LT	130	spin_unlock(&mapping->i_mmap_lock);
cddb8a5c	131	mmu_notifier_invalidate_range_end(vma->vm_mm, old_start, old_end);
1da177e4 LT	132	}
1da177e4 LT	133
7be7a546 HD	134	#define LATENCY_LIMIT (64 * PAGE_SIZE)
7be7a546 HD	135
b6a2fea3	136	unsigned long move_page_tables(struct vm_area_struct *vma,
1da177e4 LT	137	unsigned long old_addr, struct vm_area_struct *new_vma,
	138	unsigned long new_addr, unsigned long len)
	139	{
7be7a546 HD	140	unsigned long extent, next, old_end;
7be7a546 HD	141	pmd_t old_pmd, new_pmd;
1da177e4	142
7be7a546 HD	143	old_end = old_addr + len;
7be7a546 HD	144	flush_cache_range(vma, old_addr, old_end);
1da177e4	145
7be7a546	146	for (; old_addr < old_end; old_addr += extent, new_addr += extent) {
1da177e4	147	cond_resched();
7be7a546 HD	148	next = (old_addr + PMD_SIZE) & PMD_MASK;
	149	if (next - 1 > old_end)
	150	next = old_end;
	151	extent = next - old_addr;
	152	old_pmd = get_old_pmd(vma->vm_mm, old_addr);
	153	if (!old_pmd)
	154	continue;
	155	new_pmd = alloc_new_pmd(vma->vm_mm, new_addr);
	156	if (!new_pmd)
	157	break;
	158	next = (new_addr + PMD_SIZE) & PMD_MASK;
	159	if (extent > next - new_addr)
	160	extent = next - new_addr;
	161	if (extent > LATENCY_LIMIT)
	162	extent = LATENCY_LIMIT;
	163	move_ptes(vma, old_pmd, old_addr, old_addr + extent,
	164	new_vma, new_pmd, new_addr);
1da177e4	165	}
7be7a546 HD	166
7be7a546 HD	167	return len + old_addr - old_end; /* how much done */
1da177e4 LT	168	}
	169
	170	static unsigned long move_vma(struct vm_area_struct *vma,
	171	unsigned long old_addr, unsigned long old_len,
	172	unsigned long new_len, unsigned long new_addr)
	173	{
	174	struct mm_struct *mm = vma->vm_mm;
	175	struct vm_area_struct *new_vma;
	176	unsigned long vm_flags = vma->vm_flags;
	177	unsigned long new_pgoff;
	178	unsigned long moved_len;
	179	unsigned long excess = 0;
365e9c87	180	unsigned long hiwater_vm;
1da177e4	181	int split = 0;
7103ad32	182	int err;
1da177e4 LT	183
	184	/*
	185	* We'd prefer to avoid failure later on in do_munmap:
	186	* which may split one vma into three before unmapping.
	187	*/
	188	if (mm->map_count >= sysctl_max_map_count - 3)
	189	return -ENOMEM;
	190
1ff82995 HD	191	/*
	192	* Advise KSM to break any KSM pages in the area to be moved:
	193	* it would be confusing if they were to turn up at the new
	194	* location, where they happen to coincide with different KSM
	195	* pages recently unmapped. But leave vma->vm_flags as it was,
	196	* so KSM can come around to merge on vma and new_vma afterwards.
	197	*/
7103ad32 HD	198	err = ksm_madvise(vma, old_addr, old_addr + old_len,
	199	MADV_UNMERGEABLE, &vm_flags);
	200	if (err)
	201	return err;
1ff82995	202
1da177e4 LT	203	new_pgoff = vma->vm_pgoff + ((old_addr - vma->vm_start) >> PAGE_SHIFT);
	204	new_vma = copy_vma(&vma, new_addr, new_len, new_pgoff);
	205	if (!new_vma)
	206	return -ENOMEM;
	207
	208	moved_len = move_page_tables(vma, old_addr, new_vma, new_addr, old_len);
	209	if (moved_len < old_len) {
	210	/*
	211	* On error, move entries back from new area to old,
	212	* which will succeed since page tables still there,
	213	* and then proceed to unmap new area instead of old.
	214	*/
	215	move_page_tables(new_vma, new_addr, vma, old_addr, moved_len);
	216	vma = new_vma;
	217	old_len = new_len;
	218	old_addr = new_addr;
	219	new_addr = -ENOMEM;
	220	}
	221
	222	/* Conceal VM_ACCOUNT so old reservation is not undone */
	223	if (vm_flags & VM_ACCOUNT) {
	224	vma->vm_flags &= ~VM_ACCOUNT;
	225	excess = vma->vm_end - vma->vm_start - old_len;
	226	if (old_addr > vma->vm_start &&
	227	old_addr + old_len < vma->vm_end)
	228	split = 1;
	229	}
	230
71799062	231	/*
365e9c87 HD	232	* If we failed to move page tables we still do total_vm increment
	233	* since do_munmap() will decrement it by old_len == new_len.
	234	*
	235	* Since total_vm is about to be raised artificially high for a
	236	* moment, we need to restore high watermark afterwards: if stats
	237	* are taken meanwhile, total_vm and hiwater_vm appear too high.
	238	* If this were a serious issue, we'd add a flag to do_munmap().
71799062	239	*/
365e9c87	240	hiwater_vm = mm->hiwater_vm;
71799062	241	mm->total_vm += new_len >> PAGE_SHIFT;
ab50b8ed	242	vm_stat_account(mm, vma->vm_flags, vma->vm_file, new_len>>PAGE_SHIFT);
71799062	243
1da177e4 LT	244	if (do_munmap(mm, old_addr, old_len) < 0) {
	245	/* OOM: unable to split vma, just get accounts right */
	246	vm_unacct_memory(excess >> PAGE_SHIFT);
	247	excess = 0;
	248	}
365e9c87	249	mm->hiwater_vm = hiwater_vm;
1da177e4 LT	250
	251	/* Restore VM_ACCOUNT if one or two pieces of vma left */
	252	if (excess) {
	253	vma->vm_flags \|= VM_ACCOUNT;
	254	if (split)
	255	vma->vm_next->vm_flags \|= VM_ACCOUNT;
	256	}
	257
1da177e4 LT	258	if (vm_flags & VM_LOCKED) {
	259	mm->locked_vm += new_len >> PAGE_SHIFT;
	260	if (new_len > old_len)
ba470de4 RR	261	mlock_vma_pages_range(new_vma, new_addr + old_len,
ba470de4 RR	262	new_addr + new_len);
1da177e4 LT	263	}
	264
	265	return new_addr;
	266	}
	267
54f5de70 AV	268	static struct vm_area_struct *vma_to_resize(unsigned long addr,
	269	unsigned long old_len, unsigned long new_len, unsigned long *p)
	270	{
	271	struct mm_struct *mm = current->mm;
	272	struct vm_area_struct *vma = find_vma(mm, addr);
	273
	274	if (!vma \|\| vma->vm_start > addr)
	275	goto Efault;
	276
	277	if (is_vm_hugetlb_page(vma))
	278	goto Einval;
	279
	280	/* We can't remap across vm area boundaries */
	281	if (old_len > vma->vm_end - addr)
	282	goto Efault;
	283
	284	if (vma->vm_flags & (VM_DONTEXPAND \| VM_PFNMAP)) {
	285	if (new_len > old_len)
	286	goto Efault;
	287	}
	288
	289	if (vma->vm_flags & VM_LOCKED) {
	290	unsigned long locked, lock_limit;
	291	locked = mm->locked_vm << PAGE_SHIFT;
	292	lock_limit = current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur;
	293	locked += new_len - old_len;
	294	if (locked > lock_limit && !capable(CAP_IPC_LOCK))
	295	goto Eagain;
	296	}
	297
	298	if (!may_expand_vm(mm, (new_len - old_len) >> PAGE_SHIFT))
	299	goto Enomem;
	300
	301	if (vma->vm_flags & VM_ACCOUNT) {
	302	unsigned long charged = (new_len - old_len) >> PAGE_SHIFT;
	303	if (security_vm_enough_memory(charged))
	304	goto Efault;
	305	*p = charged;
	306	}
	307
	308	return vma;
	309
	310	Efault: /* very odd choice for most of the cases, but... */
	311	return ERR_PTR(-EFAULT);
	312	Einval:
	313	return ERR_PTR(-EINVAL);
	314	Enomem:
	315	return ERR_PTR(-ENOMEM);
	316	Eagain:
	317	return ERR_PTR(-EAGAIN);
	318	}
	319
ecc1a899 AV	320	static unsigned long mremap_to(unsigned long addr,
	321	unsigned long old_len, unsigned long new_addr,
	322	unsigned long new_len)
	323	{
	324	struct mm_struct *mm = current->mm;
	325	struct vm_area_struct *vma;
	326	unsigned long ret = -EINVAL;
	327	unsigned long charged = 0;
	328
	329	if (new_addr & ~PAGE_MASK)
	330	goto out;
	331
	332	if (new_len > TASK_SIZE \|\| new_addr > TASK_SIZE - new_len)
	333	goto out;
	334
	335	/* Check if the location we're moving into overlaps the
	336	* old location at all, and fail if it does.
	337	*/
	338	if ((new_addr <= addr) && (new_addr+new_len) > addr)
	339	goto out;
	340
	341	if ((addr <= new_addr) && (addr+old_len) > new_addr)
	342	goto out;
	343
	344	ret = security_file_mmap(NULL, 0, 0, 0, new_addr, 1);
	345	if (ret)
	346	goto out;
	347
	348	ret = do_munmap(mm, new_addr, new_len);
	349	if (ret)
	350	goto out;
	351
	352	if (old_len >= new_len) {
	353	ret = do_munmap(mm, addr+new_len, old_len - new_len);
	354	if (ret && old_len != new_len)
	355	goto out;
	356	old_len = new_len;
	357	}
	358
	359	vma = vma_to_resize(addr, old_len, new_len, &charged);
	360	if (IS_ERR(vma)) {
	361	ret = PTR_ERR(vma);
	362	goto out;
	363	}
	364
	365	ret = move_vma(vma, addr, old_len, new_len, new_addr);
	366	if (ret & ~PAGE_MASK)
	367	vm_unacct_memory(charged);
	368
	369	out:
	370	return ret;
	371	}
	372
1a0ef85f AV	373	static int vma_expandable(struct vm_area_struct *vma, unsigned long delta)
1a0ef85f AV	374	{
f106af4e	375	unsigned long end = vma->vm_end + delta;
1a0ef85f AV	376	unsigned long max_addr = TASK_SIZE;
	377	if (vma->vm_next)
	378	max_addr = vma->vm_next->vm_start;
f106af4e AV	379	if (max_addr < end \|\| end < vma->vm_end)
	380	return 0;
	381	if (arch_mmap_check(vma->vm_start, end - vma->vm_start, MAP_FIXED))
	382	return 0;
	383	if (get_unmapped_area(NULL, vma->vm_start, end - vma->vm_start,
	384	0, MAP_FIXED) & ~PAGE_MASK)
1a0ef85f	385	return 0;
1a0ef85f AV	386	return 1;
	387	}
	388
1da177e4 LT	389	/*
	390	* Expand (or shrink) an existing mapping, potentially moving it at the
	391	* same time (controlled by the MREMAP_MAYMOVE flag and available VM space)
	392	*
	393	* MREMAP_FIXED option added 5-Dec-1999 by Benjamin LaHaise
	394	* This option implies MREMAP_MAYMOVE.
	395	*/
	396	unsigned long do_mremap(unsigned long addr,
	397	unsigned long old_len, unsigned long new_len,
	398	unsigned long flags, unsigned long new_addr)
	399	{
d0de32d9	400	struct mm_struct *mm = current->mm;
1da177e4 LT	401	struct vm_area_struct *vma;
	402	unsigned long ret = -EINVAL;
	403	unsigned long charged = 0;
	404
	405	if (flags & ~(MREMAP_FIXED \| MREMAP_MAYMOVE))
	406	goto out;
	407
	408	if (addr & ~PAGE_MASK)
	409	goto out;
	410
	411	old_len = PAGE_ALIGN(old_len);
	412	new_len = PAGE_ALIGN(new_len);
	413
	414	/*
	415	* We allow a zero old-len as a special case
	416	* for DOS-emu "duplicate shm area" thing. But
	417	* a zero new-len is nonsensical.
	418	*/
	419	if (!new_len)
	420	goto out;
	421
1da177e4	422	if (flags & MREMAP_FIXED) {
ecc1a899 AV	423	if (flags & MREMAP_MAYMOVE)
	424	ret = mremap_to(addr, old_len, new_addr, new_len);
	425	goto out;
1da177e4 LT	426	}
	427
	428	/*
	429	* Always allow a shrinking remap: that just unmaps
	430	* the unnecessary pages..
	431	* do_munmap does all the needed commit accounting
	432	*/
	433	if (old_len >= new_len) {
d0de32d9	434	ret = do_munmap(mm, addr+new_len, old_len - new_len);
1da177e4 LT	435	if (ret && old_len != new_len)
	436	goto out;
	437	ret = addr;
ecc1a899	438	goto out;
1da177e4 LT	439	}
	440
	441	/*
ecc1a899	442	* Ok, we need to grow..
1da177e4	443	*/
54f5de70 AV	444	vma = vma_to_resize(addr, old_len, new_len, &charged);
	445	if (IS_ERR(vma)) {
	446	ret = PTR_ERR(vma);
1da177e4	447	goto out;
119f657c	448	}
1da177e4	449
1da177e4	450	/* old_len exactly to the end of the area..
1da177e4	451	*/
ecc1a899	452	if (old_len == vma->vm_end - addr) {
1da177e4	453	/* can we just expand the current mapping? */
1a0ef85f	454	if (vma_expandable(vma, new_len - old_len)) {
1da177e4 LT	455	int pages = (new_len - old_len) >> PAGE_SHIFT;
	456
	457	vma_adjust(vma, vma->vm_start,
	458	addr + new_len, vma->vm_pgoff, NULL);
	459
d0de32d9 HD	460	mm->total_vm += pages;
d0de32d9 HD	461	vm_stat_account(mm, vma->vm_flags, vma->vm_file, pages);
1da177e4	462	if (vma->vm_flags & VM_LOCKED) {
d0de32d9	463	mm->locked_vm += pages;
ba470de4	464	mlock_vma_pages_range(vma, addr + old_len,
1da177e4 LT	465	addr + new_len);
	466	}
	467	ret = addr;
	468	goto out;
	469	}
	470	}
	471
	472	/*
	473	* We weren't able to just expand or shrink the area,
	474	* we need to create a new one and move it..
	475	*/
	476	ret = -ENOMEM;
	477	if (flags & MREMAP_MAYMOVE) {
ecc1a899 AV	478	unsigned long map_flags = 0;
	479	if (vma->vm_flags & VM_MAYSHARE)
	480	map_flags \|= MAP_SHARED;
	481
	482	new_addr = get_unmapped_area(vma->vm_file, 0, new_len,
	483	vma->vm_pgoff, map_flags);
	484	if (new_addr & ~PAGE_MASK) {
	485	ret = new_addr;
	486	goto out;
1da177e4	487	}
ecc1a899 AV	488
	489	ret = security_file_mmap(NULL, 0, 0, 0, new_addr, 1);
	490	if (ret)
	491	goto out;
1da177e4 LT	492	ret = move_vma(vma, addr, old_len, new_len, new_addr);
	493	}
	494	out:
	495	if (ret & ~PAGE_MASK)
	496	vm_unacct_memory(charged);
1da177e4 LT	497	return ret;
	498	}
	499
6a6160a7 HC	500	SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len,
	501	unsigned long, new_len, unsigned long, flags,
	502	unsigned long, new_addr)
1da177e4 LT	503	{
	504	unsigned long ret;
	505
	506	down_write(&current->mm->mmap_sem);
	507	ret = do_mremap(addr, old_len, new_len, flags, new_addr);
	508	up_write(&current->mm->mmap_sem);
	509	return ret;
	510	}