[linux.git] / mm / mremap.c

/*
 *	mm/mremap.c
 *
 *	(C) Copyright 1996 Linus Torvalds
 *
 *	Address space accounting code	<[email protected]>
 *	(C) Copyright 2002 Red Hat Inc, All Rights Reserved
 */

#include <linux/mm.h>
#include <linux/hugetlb.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"

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);
	split_huge_page_pmd(mm, pmd);
	if (pmd_none_or_clear_bad(pmd))
		return NULL;

	return pmd;
}

static pmd_t *alloc_new_pmd(struct mm_struct *mm, struct vm_area_struct *vma,
			    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;

	VM_BUG_ON(pmd_trans_huge(*pmd));
	if (pmd_none(*pmd) && __pte_alloc(mm, vma, 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;
		mutex_lock(&mapping->i_mmap_mutex);
	}

	/*
	 * 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(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(new_pte - 1);
	pte_unmap_unlock(old_pte - 1, old_ptl);
	if (mapping)
		mutex_unlock(&mapping->i_mmap_mutex);
	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, vma, 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;

	/* Need to be careful about a growing mapping */
	if (new_len > old_len) {
		unsigned long pgoff;

		if (vma->vm_flags & (VM_DONTEXPAND | VM_PFNMAP))
			goto Efault;
		pgoff = (addr - vma->vm_start) >> PAGE_SHIFT;
		pgoff += vma->vm_pgoff;
		if (pgoff + (new_len >> PAGE_SHIFT) < pgoff)
			goto Einval;
	}

	if (vma->vm_flags & VM_LOCKED) {
		unsigned long locked, lock_limit;
		locked = mm->locked_vm << PAGE_SHIFT;
		lock_limit = rlimit(RLIMIT_MEMLOCK);
		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;
	unsigned long map_flags;

	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;
	}

	map_flags = MAP_FIXED;
	if (vma->vm_flags & VM_MAYSHARE)
		map_flags |= MAP_SHARED;

	ret = get_unmapped_area(vma->vm_file, new_addr, new_len, vma->vm_pgoff +
				((addr - vma->vm_start) >> PAGE_SHIFT),
				map_flags);
	if (ret & ~PAGE_MASK)
		goto out1;

	ret = move_vma(vma, addr, old_len, new_len, new_addr);
	if (!(ret & ~PAGE_MASK))
		goto out;
out1:
	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;
	if (end < vma->vm_end) /* overflow */
		return 0;
	if (vma->vm_next && vma->vm_next->vm_start < end) /* intersection */
		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;

			if (vma_adjust(vma, vma->vm_start, addr + new_len,
				       vma->vm_pgoff, NULL)) {
				ret = -ENOMEM;
				goto out;
			}

			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 +
					((addr - vma->vm_start) >> PAGE_SHIFT),
					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 <[email protected]>
1da177e4 LT	7	* (C) Copyright 2002 Red Hat Inc, All Rights Reserved
	8	*/
	9
	10	#include <linux/mm.h>
	11	#include <linux/hugetlb.h>
1da177e4	12	#include <linux/shm.h>
1ff82995	13	#include <linux/ksm.h>
1da177e4 LT	14	#include <linux/mman.h>
1da177e4 LT	15	#include <linux/swap.h>
c59ede7b	16	#include <linux/capability.h>
1da177e4 LT	17	#include <linux/fs.h>
	18	#include <linux/highmem.h>
	19	#include <linux/security.h>
	20	#include <linux/syscalls.h>
cddb8a5c	21	#include <linux/mmu_notifier.h>
1da177e4 LT	22
	23	#include <asm/uaccess.h>
	24	#include <asm/cacheflush.h>
	25	#include <asm/tlbflush.h>
	26
ba470de4 RR	27	#include "internal.h"
ba470de4 RR	28
7be7a546	29	static pmd_t get_old_pmd(struct mm_struct mm, unsigned long addr)
1da177e4 LT	30	{
	31	pgd_t *pgd;
	32	pud_t *pud;
	33	pmd_t *pmd;
	34
	35	pgd = pgd_offset(mm, addr);
	36	if (pgd_none_or_clear_bad(pgd))
	37	return NULL;
	38
	39	pud = pud_offset(pgd, addr);
	40	if (pud_none_or_clear_bad(pud))
	41	return NULL;
	42
	43	pmd = pmd_offset(pud, addr);
bae9c19b	44	split_huge_page_pmd(mm, pmd);
1da177e4 LT	45	if (pmd_none_or_clear_bad(pmd))
	46	return NULL;
	47
7be7a546	48	return pmd;
1da177e4 LT	49	}
1da177e4 LT	50
8ac1f832 AA	51	static pmd_t alloc_new_pmd(struct mm_struct mm, struct vm_area_struct *vma,
8ac1f832 AA	52	unsigned long addr)
1da177e4 LT	53	{
	54	pgd_t *pgd;
	55	pud_t *pud;
c74df32c	56	pmd_t *pmd;
1da177e4 LT	57
1da177e4 LT	58	pgd = pgd_offset(mm, addr);
1da177e4 LT	59	pud = pud_alloc(mm, pgd, addr);
1da177e4 LT	60	if (!pud)
c74df32c	61	return NULL;
7be7a546	62
1da177e4	63	pmd = pmd_alloc(mm, pud, addr);
7be7a546	64	if (!pmd)
c74df32c	65	return NULL;
7be7a546	66
8ac1f832 AA	67	VM_BUG_ON(pmd_trans_huge(*pmd));
8ac1f832 AA	68	if (pmd_none(*pmd) && __pte_alloc(mm, vma, pmd, addr))
c74df32c HD	69	return NULL;
c74df32c HD	70
7be7a546	71	return pmd;
1da177e4 LT	72	}
1da177e4 LT	73
7be7a546 HD	74	static void move_ptes(struct vm_area_struct vma, pmd_t old_pmd,
	75	unsigned long old_addr, unsigned long old_end,
	76	struct vm_area_struct new_vma, pmd_t new_pmd,
	77	unsigned long new_addr)
1da177e4 LT	78	{
	79	struct address_space *mapping = NULL;
	80	struct mm_struct *mm = vma->vm_mm;
7be7a546	81	pte_t old_pte, new_pte, pte;
4c21e2f2	82	spinlock_t old_ptl, new_ptl;
cddb8a5c	83	unsigned long old_start;
1da177e4	84
cddb8a5c AA	85	old_start = old_addr;
	86	mmu_notifier_invalidate_range_start(vma->vm_mm,
	87	old_start, old_end);
1da177e4 LT	88	if (vma->vm_file) {
	89	/*
	90	* Subtle point from Rajesh Venkatasubramanian: before
25d9e2d1 NP	91	* moving file-based ptes, we must lock truncate_pagecache
25d9e2d1 NP	92	* out, since it might clean the dst vma before the src vma,
1da177e4 LT	93	* and we propagate stale pages into the dst afterward.
	94	*/
	95	mapping = vma->vm_file->f_mapping;
3d48ae45	96	mutex_lock(&mapping->i_mmap_mutex);
1da177e4	97	}
1da177e4	98
4c21e2f2 HD	99	/*
	100	* We don't have to worry about the ordering of src and dst
	101	* pte locks because exclusive mmap_sem prevents deadlock.
	102	*/
c74df32c	103	old_pte = pte_offset_map_lock(mm, old_pmd, old_addr, &old_ptl);
ece0e2b6	104	new_pte = pte_offset_map(new_pmd, new_addr);
4c21e2f2 HD	105	new_ptl = pte_lockptr(mm, new_pmd);
4c21e2f2 HD	106	if (new_ptl != old_ptl)
f20dc5f7	107	spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING);
6606c3e0	108	arch_enter_lazy_mmu_mode();
7be7a546 HD	109
	110	for (; old_addr < old_end; old_pte++, old_addr += PAGE_SIZE,
	111	new_pte++, new_addr += PAGE_SIZE) {
	112	if (pte_none(*old_pte))
	113	continue;
	114	pte = ptep_clear_flush(vma, old_addr, old_pte);
7be7a546 HD	115	pte = move_pte(pte, new_vma->vm_page_prot, old_addr, new_addr);
7be7a546 HD	116	set_pte_at(mm, new_addr, new_pte, pte);
1da177e4	117	}
7be7a546	118
6606c3e0	119	arch_leave_lazy_mmu_mode();
4c21e2f2 HD	120	if (new_ptl != old_ptl)
4c21e2f2 HD	121	spin_unlock(new_ptl);
ece0e2b6	122	pte_unmap(new_pte - 1);
c74df32c	123	pte_unmap_unlock(old_pte - 1, old_ptl);
1da177e4	124	if (mapping)
3d48ae45	125	mutex_unlock(&mapping->i_mmap_mutex);
cddb8a5c	126	mmu_notifier_invalidate_range_end(vma->vm_mm, old_start, old_end);
1da177e4 LT	127	}
1da177e4 LT	128
7be7a546 HD	129	#define LATENCY_LIMIT (64 * PAGE_SIZE)
7be7a546 HD	130
b6a2fea3	131	unsigned long move_page_tables(struct vm_area_struct *vma,
1da177e4 LT	132	unsigned long old_addr, struct vm_area_struct *new_vma,
	133	unsigned long new_addr, unsigned long len)
	134	{
7be7a546 HD	135	unsigned long extent, next, old_end;
7be7a546 HD	136	pmd_t old_pmd, new_pmd;
1da177e4	137
7be7a546 HD	138	old_end = old_addr + len;
7be7a546 HD	139	flush_cache_range(vma, old_addr, old_end);
1da177e4	140
7be7a546	141	for (; old_addr < old_end; old_addr += extent, new_addr += extent) {
1da177e4	142	cond_resched();
7be7a546 HD	143	next = (old_addr + PMD_SIZE) & PMD_MASK;
	144	if (next - 1 > old_end)
	145	next = old_end;
	146	extent = next - old_addr;
	147	old_pmd = get_old_pmd(vma->vm_mm, old_addr);
	148	if (!old_pmd)
	149	continue;
8ac1f832	150	new_pmd = alloc_new_pmd(vma->vm_mm, vma, new_addr);
7be7a546 HD	151	if (!new_pmd)
	152	break;
	153	next = (new_addr + PMD_SIZE) & PMD_MASK;
	154	if (extent > next - new_addr)
	155	extent = next - new_addr;
	156	if (extent > LATENCY_LIMIT)
	157	extent = LATENCY_LIMIT;
	158	move_ptes(vma, old_pmd, old_addr, old_addr + extent,
	159	new_vma, new_pmd, new_addr);
1da177e4	160	}
7be7a546 HD	161
7be7a546 HD	162	return len + old_addr - old_end; /* how much done */
1da177e4 LT	163	}
	164
	165	static unsigned long move_vma(struct vm_area_struct *vma,
	166	unsigned long old_addr, unsigned long old_len,
	167	unsigned long new_len, unsigned long new_addr)
	168	{
	169	struct mm_struct *mm = vma->vm_mm;
	170	struct vm_area_struct *new_vma;
	171	unsigned long vm_flags = vma->vm_flags;
	172	unsigned long new_pgoff;
	173	unsigned long moved_len;
	174	unsigned long excess = 0;
365e9c87	175	unsigned long hiwater_vm;
1da177e4	176	int split = 0;
7103ad32	177	int err;
1da177e4 LT	178
	179	/*
	180	* We'd prefer to avoid failure later on in do_munmap:
	181	* which may split one vma into three before unmapping.
	182	*/
	183	if (mm->map_count >= sysctl_max_map_count - 3)
	184	return -ENOMEM;
	185
1ff82995 HD	186	/*
	187	* Advise KSM to break any KSM pages in the area to be moved:
	188	* it would be confusing if they were to turn up at the new
	189	* location, where they happen to coincide with different KSM
	190	* pages recently unmapped. But leave vma->vm_flags as it was,
	191	* so KSM can come around to merge on vma and new_vma afterwards.
	192	*/
7103ad32 HD	193	err = ksm_madvise(vma, old_addr, old_addr + old_len,
	194	MADV_UNMERGEABLE, &vm_flags);
	195	if (err)
	196	return err;
1ff82995	197
1da177e4 LT	198	new_pgoff = vma->vm_pgoff + ((old_addr - vma->vm_start) >> PAGE_SHIFT);
	199	new_vma = copy_vma(&vma, new_addr, new_len, new_pgoff);
	200	if (!new_vma)
	201	return -ENOMEM;
	202
	203	moved_len = move_page_tables(vma, old_addr, new_vma, new_addr, old_len);
	204	if (moved_len < old_len) {
	205	/*
	206	* On error, move entries back from new area to old,
	207	* which will succeed since page tables still there,
	208	* and then proceed to unmap new area instead of old.
	209	*/
	210	move_page_tables(new_vma, new_addr, vma, old_addr, moved_len);
	211	vma = new_vma;
	212	old_len = new_len;
	213	old_addr = new_addr;
	214	new_addr = -ENOMEM;
	215	}
	216
	217	/* Conceal VM_ACCOUNT so old reservation is not undone */
	218	if (vm_flags & VM_ACCOUNT) {
	219	vma->vm_flags &= ~VM_ACCOUNT;
	220	excess = vma->vm_end - vma->vm_start - old_len;
	221	if (old_addr > vma->vm_start &&
	222	old_addr + old_len < vma->vm_end)
	223	split = 1;
	224	}
	225
71799062	226	/*
365e9c87 HD	227	* If we failed to move page tables we still do total_vm increment
	228	* since do_munmap() will decrement it by old_len == new_len.
	229	*
	230	* Since total_vm is about to be raised artificially high for a
	231	* moment, we need to restore high watermark afterwards: if stats
	232	* are taken meanwhile, total_vm and hiwater_vm appear too high.
	233	* If this were a serious issue, we'd add a flag to do_munmap().
71799062	234	*/
365e9c87	235	hiwater_vm = mm->hiwater_vm;
71799062	236	mm->total_vm += new_len >> PAGE_SHIFT;
ab50b8ed	237	vm_stat_account(mm, vma->vm_flags, vma->vm_file, new_len>>PAGE_SHIFT);
71799062	238
1da177e4 LT	239	if (do_munmap(mm, old_addr, old_len) < 0) {
	240	/* OOM: unable to split vma, just get accounts right */
	241	vm_unacct_memory(excess >> PAGE_SHIFT);
	242	excess = 0;
	243	}
365e9c87	244	mm->hiwater_vm = hiwater_vm;
1da177e4 LT	245
	246	/* Restore VM_ACCOUNT if one or two pieces of vma left */
	247	if (excess) {
	248	vma->vm_flags \|= VM_ACCOUNT;
	249	if (split)
	250	vma->vm_next->vm_flags \|= VM_ACCOUNT;
	251	}
	252
1da177e4 LT	253	if (vm_flags & VM_LOCKED) {
	254	mm->locked_vm += new_len >> PAGE_SHIFT;
	255	if (new_len > old_len)
ba470de4 RR	256	mlock_vma_pages_range(new_vma, new_addr + old_len,
ba470de4 RR	257	new_addr + new_len);
1da177e4 LT	258	}
	259
	260	return new_addr;
	261	}
	262
54f5de70 AV	263	static struct vm_area_struct *vma_to_resize(unsigned long addr,
	264	unsigned long old_len, unsigned long new_len, unsigned long *p)
	265	{
	266	struct mm_struct *mm = current->mm;
	267	struct vm_area_struct *vma = find_vma(mm, addr);
	268
	269	if (!vma \|\| vma->vm_start > addr)
	270	goto Efault;
	271
	272	if (is_vm_hugetlb_page(vma))
	273	goto Einval;
	274
	275	/* We can't remap across vm area boundaries */
	276	if (old_len > vma->vm_end - addr)
	277	goto Efault;
	278
982134ba LT	279	/* Need to be careful about a growing mapping */
	280	if (new_len > old_len) {
	281	unsigned long pgoff;
	282
	283	if (vma->vm_flags & (VM_DONTEXPAND \| VM_PFNMAP))
54f5de70	284	goto Efault;
982134ba LT	285	pgoff = (addr - vma->vm_start) >> PAGE_SHIFT;
	286	pgoff += vma->vm_pgoff;
	287	if (pgoff + (new_len >> PAGE_SHIFT) < pgoff)
	288	goto Einval;
54f5de70 AV	289	}
	290
	291	if (vma->vm_flags & VM_LOCKED) {
	292	unsigned long locked, lock_limit;
	293	locked = mm->locked_vm << PAGE_SHIFT;
59e99e5b	294	lock_limit = rlimit(RLIMIT_MEMLOCK);
54f5de70 AV	295	locked += new_len - old_len;
	296	if (locked > lock_limit && !capable(CAP_IPC_LOCK))
	297	goto Eagain;
	298	}
	299
	300	if (!may_expand_vm(mm, (new_len - old_len) >> PAGE_SHIFT))
	301	goto Enomem;
	302
	303	if (vma->vm_flags & VM_ACCOUNT) {
	304	unsigned long charged = (new_len - old_len) >> PAGE_SHIFT;
	305	if (security_vm_enough_memory(charged))
	306	goto Efault;
	307	*p = charged;
	308	}
	309
	310	return vma;
	311
	312	Efault: /* very odd choice for most of the cases, but... */
	313	return ERR_PTR(-EFAULT);
	314	Einval:
	315	return ERR_PTR(-EINVAL);
	316	Enomem:
	317	return ERR_PTR(-ENOMEM);
	318	Eagain:
	319	return ERR_PTR(-EAGAIN);
	320	}
	321
ecc1a899 AV	322	static unsigned long mremap_to(unsigned long addr,
	323	unsigned long old_len, unsigned long new_addr,
	324	unsigned long new_len)
	325	{
	326	struct mm_struct *mm = current->mm;
	327	struct vm_area_struct *vma;
	328	unsigned long ret = -EINVAL;
	329	unsigned long charged = 0;
097eed10	330	unsigned long map_flags;
ecc1a899 AV	331
	332	if (new_addr & ~PAGE_MASK)
	333	goto out;
	334
	335	if (new_len > TASK_SIZE \|\| new_addr > TASK_SIZE - new_len)
	336	goto out;
	337
	338	/* Check if the location we're moving into overlaps the
	339	* old location at all, and fail if it does.
	340	*/
	341	if ((new_addr <= addr) && (new_addr+new_len) > addr)
	342	goto out;
	343
	344	if ((addr <= new_addr) && (addr+old_len) > new_addr)
	345	goto out;
	346
	347	ret = security_file_mmap(NULL, 0, 0, 0, new_addr, 1);
	348	if (ret)
	349	goto out;
	350
	351	ret = do_munmap(mm, new_addr, new_len);
	352	if (ret)
	353	goto out;
	354
	355	if (old_len >= new_len) {
	356	ret = do_munmap(mm, addr+new_len, old_len - new_len);
	357	if (ret && old_len != new_len)
	358	goto out;
	359	old_len = new_len;
	360	}
	361
	362	vma = vma_to_resize(addr, old_len, new_len, &charged);
	363	if (IS_ERR(vma)) {
	364	ret = PTR_ERR(vma);
	365	goto out;
	366	}
	367
097eed10 AV	368	map_flags = MAP_FIXED;
	369	if (vma->vm_flags & VM_MAYSHARE)
	370	map_flags \|= MAP_SHARED;
9206de95	371
097eed10 AV	372	ret = get_unmapped_area(vma->vm_file, new_addr, new_len, vma->vm_pgoff +
	373	((addr - vma->vm_start) >> PAGE_SHIFT),
	374	map_flags);
ecc1a899	375	if (ret & ~PAGE_MASK)
097eed10 AV	376	goto out1;
	377
	378	ret = move_vma(vma, addr, old_len, new_len, new_addr);
	379	if (!(ret & ~PAGE_MASK))
	380	goto out;
	381	out1:
	382	vm_unacct_memory(charged);
ecc1a899 AV	383
	384	out:
	385	return ret;
	386	}
	387
1a0ef85f AV	388	static int vma_expandable(struct vm_area_struct *vma, unsigned long delta)
1a0ef85f AV	389	{
f106af4e	390	unsigned long end = vma->vm_end + delta;
9206de95	391	if (end < vma->vm_end) /* overflow */
f106af4e	392	return 0;
9206de95	393	if (vma->vm_next && vma->vm_next->vm_start < end) /* intersection */
f106af4e AV	394	return 0;
	395	if (get_unmapped_area(NULL, vma->vm_start, end - vma->vm_start,
	396	0, MAP_FIXED) & ~PAGE_MASK)
1a0ef85f	397	return 0;
1a0ef85f AV	398	return 1;
	399	}
	400
1da177e4 LT	401	/*
	402	* Expand (or shrink) an existing mapping, potentially moving it at the
	403	* same time (controlled by the MREMAP_MAYMOVE flag and available VM space)
	404	*
	405	* MREMAP_FIXED option added 5-Dec-1999 by Benjamin LaHaise
	406	* This option implies MREMAP_MAYMOVE.
	407	*/
	408	unsigned long do_mremap(unsigned long addr,
	409	unsigned long old_len, unsigned long new_len,
	410	unsigned long flags, unsigned long new_addr)
	411	{
d0de32d9	412	struct mm_struct *mm = current->mm;
1da177e4 LT	413	struct vm_area_struct *vma;
	414	unsigned long ret = -EINVAL;
	415	unsigned long charged = 0;
	416
	417	if (flags & ~(MREMAP_FIXED \| MREMAP_MAYMOVE))
	418	goto out;
	419
	420	if (addr & ~PAGE_MASK)
	421	goto out;
	422
	423	old_len = PAGE_ALIGN(old_len);
	424	new_len = PAGE_ALIGN(new_len);
	425
	426	/*
	427	* We allow a zero old-len as a special case
	428	* for DOS-emu "duplicate shm area" thing. But
	429	* a zero new-len is nonsensical.
	430	*/
	431	if (!new_len)
	432	goto out;
	433
1da177e4	434	if (flags & MREMAP_FIXED) {
ecc1a899 AV	435	if (flags & MREMAP_MAYMOVE)
	436	ret = mremap_to(addr, old_len, new_addr, new_len);
	437	goto out;
1da177e4 LT	438	}
	439
	440	/*
	441	* Always allow a shrinking remap: that just unmaps
	442	* the unnecessary pages..
	443	* do_munmap does all the needed commit accounting
	444	*/
	445	if (old_len >= new_len) {
d0de32d9	446	ret = do_munmap(mm, addr+new_len, old_len - new_len);
1da177e4 LT	447	if (ret && old_len != new_len)
	448	goto out;
	449	ret = addr;
ecc1a899	450	goto out;
1da177e4 LT	451	}
	452
	453	/*
ecc1a899	454	* Ok, we need to grow..
1da177e4	455	*/
54f5de70 AV	456	vma = vma_to_resize(addr, old_len, new_len, &charged);
	457	if (IS_ERR(vma)) {
	458	ret = PTR_ERR(vma);
1da177e4	459	goto out;
119f657c	460	}
1da177e4	461
1da177e4	462	/* old_len exactly to the end of the area..
1da177e4	463	*/
ecc1a899	464	if (old_len == vma->vm_end - addr) {
1da177e4	465	/* can we just expand the current mapping? */
1a0ef85f	466	if (vma_expandable(vma, new_len - old_len)) {
1da177e4 LT	467	int pages = (new_len - old_len) >> PAGE_SHIFT;
1da177e4 LT	468
5beb4930 RR	469	if (vma_adjust(vma, vma->vm_start, addr + new_len,
	470	vma->vm_pgoff, NULL)) {
	471	ret = -ENOMEM;
	472	goto out;
	473	}
1da177e4	474
d0de32d9 HD	475	mm->total_vm += pages;
d0de32d9 HD	476	vm_stat_account(mm, vma->vm_flags, vma->vm_file, pages);
1da177e4	477	if (vma->vm_flags & VM_LOCKED) {
d0de32d9	478	mm->locked_vm += pages;
ba470de4	479	mlock_vma_pages_range(vma, addr + old_len,
1da177e4 LT	480	addr + new_len);
	481	}
	482	ret = addr;
	483	goto out;
	484	}
	485	}
	486
	487	/*
	488	* We weren't able to just expand or shrink the area,
	489	* we need to create a new one and move it..
	490	*/
	491	ret = -ENOMEM;
	492	if (flags & MREMAP_MAYMOVE) {
ecc1a899 AV	493	unsigned long map_flags = 0;
	494	if (vma->vm_flags & VM_MAYSHARE)
	495	map_flags \|= MAP_SHARED;
	496
	497	new_addr = get_unmapped_area(vma->vm_file, 0, new_len,
93587414 AV	498	vma->vm_pgoff +
	499	((addr - vma->vm_start) >> PAGE_SHIFT),
	500	map_flags);
ecc1a899 AV	501	if (new_addr & ~PAGE_MASK) {
	502	ret = new_addr;
	503	goto out;
1da177e4	504	}
ecc1a899 AV	505
	506	ret = security_file_mmap(NULL, 0, 0, 0, new_addr, 1);
	507	if (ret)
	508	goto out;
1da177e4 LT	509	ret = move_vma(vma, addr, old_len, new_len, new_addr);
	510	}
	511	out:
	512	if (ret & ~PAGE_MASK)
	513	vm_unacct_memory(charged);
1da177e4 LT	514	return ret;
	515	}
	516
6a6160a7 HC	517	SYSCALL_DEFINE5(mremap, unsigned long, addr, unsigned long, old_len,
	518	unsigned long, new_len, unsigned long, flags,
	519	unsigned long, new_addr)
1da177e4 LT	520	{
	521	unsigned long ret;
	522
	523	down_write(&current->mm->mmap_sem);
	524	ret = do_mremap(addr, old_len, new_len, flags, new_addr);
	525	up_write(&current->mm->mmap_sem);
	526	return ret;
	527	}