[linux.git] / mm / userfaultfd.c

/*
 *  mm/userfaultfd.c
 *
 *  Copyright (C) 2015  Red Hat, Inc.
 *
 *  This work is licensed under the terms of the GNU GPL, version 2. See
 *  the COPYING file in the top-level directory.
 */

#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/rmap.h>
#include <linux/swap.h>
#include <linux/swapops.h>
#include <linux/userfaultfd_k.h>
#include <linux/mmu_notifier.h>
#include <asm/tlbflush.h>
#include "internal.h"

static int mcopy_atomic_pte(struct mm_struct *dst_mm,
			    pmd_t *dst_pmd,
			    struct vm_area_struct *dst_vma,
			    unsigned long dst_addr,
			    unsigned long src_addr,
			    struct page **pagep)
{
	struct mem_cgroup *memcg;
	pte_t _dst_pte, *dst_pte;
	spinlock_t *ptl;
	void *page_kaddr;
	int ret;
	struct page *page;

	if (!*pagep) {
		ret = -ENOMEM;
		page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, dst_vma, dst_addr);
		if (!page)
			goto out;

		page_kaddr = kmap_atomic(page);
		ret = copy_from_user(page_kaddr,
				     (const void __user *) src_addr,
				     PAGE_SIZE);
		kunmap_atomic(page_kaddr);

		/* fallback to copy_from_user outside mmap_sem */
		if (unlikely(ret)) {
			ret = -EFAULT;
			*pagep = page;
			/* don't free the page */
			goto out;
		}
	} else {
		page = *pagep;
		*pagep = NULL;
	}

	/*
	 * The memory barrier inside __SetPageUptodate makes sure that
	 * preceeding stores to the page contents become visible before
	 * the set_pte_at() write.
	 */
	__SetPageUptodate(page);

	ret = -ENOMEM;
	if (mem_cgroup_try_charge(page, dst_mm, GFP_KERNEL, &memcg))
		goto out_release;

	_dst_pte = mk_pte(page, dst_vma->vm_page_prot);
	if (dst_vma->vm_flags & VM_WRITE)
		_dst_pte = pte_mkwrite(pte_mkdirty(_dst_pte));

	ret = -EEXIST;
	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
	if (!pte_none(*dst_pte))
		goto out_release_uncharge_unlock;

	inc_mm_counter(dst_mm, MM_ANONPAGES);
	page_add_new_anon_rmap(page, dst_vma, dst_addr);
	mem_cgroup_commit_charge(page, memcg, false);
	lru_cache_add_active_or_unevictable(page, dst_vma);

	set_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);

	/* No need to invalidate - it was non-present before */
	update_mmu_cache(dst_vma, dst_addr, dst_pte);

	pte_unmap_unlock(dst_pte, ptl);
	ret = 0;
out:
	return ret;
out_release_uncharge_unlock:
	pte_unmap_unlock(dst_pte, ptl);
	mem_cgroup_cancel_charge(page, memcg);
out_release:
	page_cache_release(page);
	goto out;
}

static int mfill_zeropage_pte(struct mm_struct *dst_mm,
			      pmd_t *dst_pmd,
			      struct vm_area_struct *dst_vma,
			      unsigned long dst_addr)
{
	pte_t _dst_pte, *dst_pte;
	spinlock_t *ptl;
	int ret;

	_dst_pte = pte_mkspecial(pfn_pte(my_zero_pfn(dst_addr),
					 dst_vma->vm_page_prot));
	ret = -EEXIST;
	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
	if (!pte_none(*dst_pte))
		goto out_unlock;
	set_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);
	/* No need to invalidate - it was non-present before */
	update_mmu_cache(dst_vma, dst_addr, dst_pte);
	ret = 0;
out_unlock:
	pte_unmap_unlock(dst_pte, ptl);
	return ret;
}

static pmd_t *mm_alloc_pmd(struct mm_struct *mm, unsigned long address)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;

	pgd = pgd_offset(mm, address);
	pud = pud_alloc(mm, pgd, address);
	if (pud)
		/*
		 * Note that we didn't run this because the pmd was
		 * missing, the *pmd may be already established and in
		 * turn it may also be a trans_huge_pmd.
		 */
		pmd = pmd_alloc(mm, pud, address);
	return pmd;
}

static __always_inline ssize_t __mcopy_atomic(struct mm_struct *dst_mm,
					      unsigned long dst_start,
					      unsigned long src_start,
					      unsigned long len,
					      bool zeropage)
{
	struct vm_area_struct *dst_vma;
	ssize_t err;
	pmd_t *dst_pmd;
	unsigned long src_addr, dst_addr;
	long copied;
	struct page *page;

	/*
	 * Sanitize the command parameters:
	 */
	BUG_ON(dst_start & ~PAGE_MASK);
	BUG_ON(len & ~PAGE_MASK);

	/* Does the address range wrap, or is the span zero-sized? */
	BUG_ON(src_start + len <= src_start);
	BUG_ON(dst_start + len <= dst_start);

	src_addr = src_start;
	dst_addr = dst_start;
	copied = 0;
	page = NULL;
retry:
	down_read(&dst_mm->mmap_sem);

	/*
	 * Make sure the vma is not shared, that the dst range is
	 * both valid and fully within a single existing vma.
	 */
	err = -EINVAL;
	dst_vma = find_vma(dst_mm, dst_start);
	if (!dst_vma || (dst_vma->vm_flags & VM_SHARED))
		goto out_unlock;
	if (dst_start < dst_vma->vm_start ||
	    dst_start + len > dst_vma->vm_end)
		goto out_unlock;

	/*
	 * Be strict and only allow __mcopy_atomic on userfaultfd
	 * registered ranges to prevent userland errors going
	 * unnoticed. As far as the VM consistency is concerned, it
	 * would be perfectly safe to remove this check, but there's
	 * no useful usage for __mcopy_atomic ouside of userfaultfd
	 * registered ranges. This is after all why these are ioctls
	 * belonging to the userfaultfd and not syscalls.
	 */
	if (!dst_vma->vm_userfaultfd_ctx.ctx)
		goto out_unlock;

	/*
	 * FIXME: only allow copying on anonymous vmas, tmpfs should
	 * be added.
	 */
	if (dst_vma->vm_ops)
		goto out_unlock;

	/*
	 * Ensure the dst_vma has a anon_vma or this page
	 * would get a NULL anon_vma when moved in the
	 * dst_vma.
	 */
	err = -ENOMEM;
	if (unlikely(anon_vma_prepare(dst_vma)))
		goto out_unlock;

	while (src_addr < src_start + len) {
		pmd_t dst_pmdval;

		BUG_ON(dst_addr >= dst_start + len);

		dst_pmd = mm_alloc_pmd(dst_mm, dst_addr);
		if (unlikely(!dst_pmd)) {
			err = -ENOMEM;
			break;
		}

		dst_pmdval = pmd_read_atomic(dst_pmd);
		/*
		 * If the dst_pmd is mapped as THP don't
		 * override it and just be strict.
		 */
		if (unlikely(pmd_trans_huge(dst_pmdval))) {
			err = -EEXIST;
			break;
		}
		if (unlikely(pmd_none(dst_pmdval)) &&
		    unlikely(__pte_alloc(dst_mm, dst_vma, dst_pmd,
					 dst_addr))) {
			err = -ENOMEM;
			break;
		}
		/* If an huge pmd materialized from under us fail */
		if (unlikely(pmd_trans_huge(*dst_pmd))) {
			err = -EFAULT;
			break;
		}

		BUG_ON(pmd_none(*dst_pmd));
		BUG_ON(pmd_trans_huge(*dst_pmd));

		if (!zeropage)
			err = mcopy_atomic_pte(dst_mm, dst_pmd, dst_vma,
					       dst_addr, src_addr, &page);
		else
			err = mfill_zeropage_pte(dst_mm, dst_pmd, dst_vma,
						 dst_addr);

		cond_resched();

		if (unlikely(err == -EFAULT)) {
			void *page_kaddr;

			up_read(&dst_mm->mmap_sem);
			BUG_ON(!page);

			page_kaddr = kmap(page);
			err = copy_from_user(page_kaddr,
					     (const void __user *) src_addr,
					     PAGE_SIZE);
			kunmap(page);
			if (unlikely(err)) {
				err = -EFAULT;
				goto out;
			}
			goto retry;
		} else
			BUG_ON(page);

		if (!err) {
			dst_addr += PAGE_SIZE;
			src_addr += PAGE_SIZE;
			copied += PAGE_SIZE;

			if (fatal_signal_pending(current))
				err = -EINTR;
		}
		if (err)
			break;
	}

out_unlock:
	up_read(&dst_mm->mmap_sem);
out:
	if (page)
		page_cache_release(page);
	BUG_ON(copied < 0);
	BUG_ON(err > 0);
	BUG_ON(!copied && !err);
	return copied ? copied : err;
}

ssize_t mcopy_atomic(struct mm_struct *dst_mm, unsigned long dst_start,
		     unsigned long src_start, unsigned long len)
{
	return __mcopy_atomic(dst_mm, dst_start, src_start, len, false);
}

ssize_t mfill_zeropage(struct mm_struct *dst_mm, unsigned long start,
		       unsigned long len)
{
	return __mcopy_atomic(dst_mm, start, 0, len, true);
}
Commit	Line	Data
c1a4de99 AA	1	/*
	2	* mm/userfaultfd.c
	3	*
	4	* Copyright (C) 2015 Red Hat, Inc.
	5	*
	6	* This work is licensed under the terms of the GNU GPL, version 2. See
	7	* the COPYING file in the top-level directory.
	8	*/
	9
	10	#include <linux/mm.h>
	11	#include <linux/pagemap.h>
	12	#include <linux/rmap.h>
	13	#include <linux/swap.h>
	14	#include <linux/swapops.h>
	15	#include <linux/userfaultfd_k.h>
	16	#include <linux/mmu_notifier.h>
	17	#include <asm/tlbflush.h>
	18	#include "internal.h"
	19
	20	static int mcopy_atomic_pte(struct mm_struct *dst_mm,
	21	pmd_t *dst_pmd,
	22	struct vm_area_struct *dst_vma,
	23	unsigned long dst_addr,
b6ebaedb AA	24	unsigned long src_addr,
b6ebaedb AA	25	struct page **pagep)
c1a4de99 AA	26	{
	27	struct mem_cgroup *memcg;
	28	pte_t _dst_pte, *dst_pte;
	29	spinlock_t *ptl;
c1a4de99 AA	30	void *page_kaddr;
c1a4de99 AA	31	int ret;
b6ebaedb	32	struct page *page;
c1a4de99	33
b6ebaedb AA	34	if (!*pagep) {
	35	ret = -ENOMEM;
	36	page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, dst_vma, dst_addr);
	37	if (!page)
	38	goto out;
	39
	40	page_kaddr = kmap_atomic(page);
	41	ret = copy_from_user(page_kaddr,
	42	(const void __user *) src_addr,
	43	PAGE_SIZE);
	44	kunmap_atomic(page_kaddr);
	45
	46	/* fallback to copy_from_user outside mmap_sem */
	47	if (unlikely(ret)) {
	48	ret = -EFAULT;
	49	*pagep = page;
	50	/* don't free the page */
	51	goto out;
	52	}
	53	} else {
	54	page = *pagep;
	55	*pagep = NULL;
	56	}
c1a4de99 AA	57
	58	/*
	59	* The memory barrier inside __SetPageUptodate makes sure that
	60	* preceeding stores to the page contents become visible before
	61	* the set_pte_at() write.
	62	*/
	63	__SetPageUptodate(page);
	64
	65	ret = -ENOMEM;
	66	if (mem_cgroup_try_charge(page, dst_mm, GFP_KERNEL, &memcg))
	67	goto out_release;
	68
	69	_dst_pte = mk_pte(page, dst_vma->vm_page_prot);
	70	if (dst_vma->vm_flags & VM_WRITE)
	71	_dst_pte = pte_mkwrite(pte_mkdirty(_dst_pte));
	72
	73	ret = -EEXIST;
	74	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
	75	if (!pte_none(*dst_pte))
	76	goto out_release_uncharge_unlock;
	77
	78	inc_mm_counter(dst_mm, MM_ANONPAGES);
	79	page_add_new_anon_rmap(page, dst_vma, dst_addr);
	80	mem_cgroup_commit_charge(page, memcg, false);
	81	lru_cache_add_active_or_unevictable(page, dst_vma);
	82
	83	set_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);
	84
	85	/* No need to invalidate - it was non-present before */
	86	update_mmu_cache(dst_vma, dst_addr, dst_pte);
	87
	88	pte_unmap_unlock(dst_pte, ptl);
	89	ret = 0;
	90	out:
	91	return ret;
	92	out_release_uncharge_unlock:
	93	pte_unmap_unlock(dst_pte, ptl);
	94	mem_cgroup_cancel_charge(page, memcg);
	95	out_release:
	96	page_cache_release(page);
	97	goto out;
c1a4de99 AA	98	}
	99
	100	static int mfill_zeropage_pte(struct mm_struct *dst_mm,
	101	pmd_t *dst_pmd,
	102	struct vm_area_struct *dst_vma,
	103	unsigned long dst_addr)
	104	{
	105	pte_t _dst_pte, *dst_pte;
	106	spinlock_t *ptl;
	107	int ret;
	108
	109	_dst_pte = pte_mkspecial(pfn_pte(my_zero_pfn(dst_addr),
	110	dst_vma->vm_page_prot));
	111	ret = -EEXIST;
	112	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
	113	if (!pte_none(*dst_pte))
	114	goto out_unlock;
	115	set_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);
	116	/* No need to invalidate - it was non-present before */
	117	update_mmu_cache(dst_vma, dst_addr, dst_pte);
	118	ret = 0;
	119	out_unlock:
	120	pte_unmap_unlock(dst_pte, ptl);
	121	return ret;
	122	}
	123
	124	static pmd_t mm_alloc_pmd(struct mm_struct mm, unsigned long address)
	125	{
	126	pgd_t *pgd;
	127	pud_t *pud;
	128	pmd_t *pmd = NULL;
	129
	130	pgd = pgd_offset(mm, address);
	131	pud = pud_alloc(mm, pgd, address);
	132	if (pud)
	133	/*
	134	* Note that we didn't run this because the pmd was
	135	* missing, the *pmd may be already established and in
	136	* turn it may also be a trans_huge_pmd.
	137	*/
	138	pmd = pmd_alloc(mm, pud, address);
	139	return pmd;
	140	}
	141
	142	static __always_inline ssize_t __mcopy_atomic(struct mm_struct *dst_mm,
	143	unsigned long dst_start,
	144	unsigned long src_start,
	145	unsigned long len,
	146	bool zeropage)
	147	{
	148	struct vm_area_struct *dst_vma;
	149	ssize_t err;
	150	pmd_t *dst_pmd;
	151	unsigned long src_addr, dst_addr;
b6ebaedb AA	152	long copied;
b6ebaedb AA	153	struct page *page;
c1a4de99 AA	154
	155	/*
	156	* Sanitize the command parameters:
	157	*/
	158	BUG_ON(dst_start & ~PAGE_MASK);
	159	BUG_ON(len & ~PAGE_MASK);
	160
	161	/* Does the address range wrap, or is the span zero-sized? */
	162	BUG_ON(src_start + len <= src_start);
	163	BUG_ON(dst_start + len <= dst_start);
	164
b6ebaedb AA	165	src_addr = src_start;
	166	dst_addr = dst_start;
	167	copied = 0;
	168	page = NULL;
	169	retry:
c1a4de99 AA	170	down_read(&dst_mm->mmap_sem);
	171
	172	/*
	173	* Make sure the vma is not shared, that the dst range is
	174	* both valid and fully within a single existing vma.
	175	*/
	176	err = -EINVAL;
	177	dst_vma = find_vma(dst_mm, dst_start);
	178	if (!dst_vma \|\| (dst_vma->vm_flags & VM_SHARED))
b6ebaedb	179	goto out_unlock;
c1a4de99 AA	180	if (dst_start < dst_vma->vm_start \|\|
c1a4de99 AA	181	dst_start + len > dst_vma->vm_end)
b6ebaedb	182	goto out_unlock;
c1a4de99 AA	183
	184	/*
	185	* Be strict and only allow __mcopy_atomic on userfaultfd
	186	* registered ranges to prevent userland errors going
	187	* unnoticed. As far as the VM consistency is concerned, it
	188	* would be perfectly safe to remove this check, but there's
	189	* no useful usage for __mcopy_atomic ouside of userfaultfd
	190	* registered ranges. This is after all why these are ioctls
	191	* belonging to the userfaultfd and not syscalls.
	192	*/
	193	if (!dst_vma->vm_userfaultfd_ctx.ctx)
b6ebaedb	194	goto out_unlock;
c1a4de99 AA	195
	196	/*
	197	* FIXME: only allow copying on anonymous vmas, tmpfs should
	198	* be added.
	199	*/
	200	if (dst_vma->vm_ops)
b6ebaedb	201	goto out_unlock;
c1a4de99 AA	202
	203	/*
	204	* Ensure the dst_vma has a anon_vma or this page
	205	* would get a NULL anon_vma when moved in the
	206	* dst_vma.
	207	*/
	208	err = -ENOMEM;
	209	if (unlikely(anon_vma_prepare(dst_vma)))
b6ebaedb	210	goto out_unlock;
c1a4de99	211
b6ebaedb	212	while (src_addr < src_start + len) {
c1a4de99	213	pmd_t dst_pmdval;
b6ebaedb	214
c1a4de99	215	BUG_ON(dst_addr >= dst_start + len);
b6ebaedb	216
c1a4de99 AA	217	dst_pmd = mm_alloc_pmd(dst_mm, dst_addr);
	218	if (unlikely(!dst_pmd)) {
	219	err = -ENOMEM;
	220	break;
	221	}
	222
	223	dst_pmdval = pmd_read_atomic(dst_pmd);
	224	/*
	225	* If the dst_pmd is mapped as THP don't
	226	* override it and just be strict.
	227	*/
	228	if (unlikely(pmd_trans_huge(dst_pmdval))) {
	229	err = -EEXIST;
	230	break;
	231	}
	232	if (unlikely(pmd_none(dst_pmdval)) &&
	233	unlikely(__pte_alloc(dst_mm, dst_vma, dst_pmd,
	234	dst_addr))) {
	235	err = -ENOMEM;
	236	break;
	237	}
	238	/* If an huge pmd materialized from under us fail */
	239	if (unlikely(pmd_trans_huge(*dst_pmd))) {
	240	err = -EFAULT;
	241	break;
	242	}
	243
	244	BUG_ON(pmd_none(*dst_pmd));
	245	BUG_ON(pmd_trans_huge(*dst_pmd));
	246
	247	if (!zeropage)
	248	err = mcopy_atomic_pte(dst_mm, dst_pmd, dst_vma,
b6ebaedb	249	dst_addr, src_addr, &page);
c1a4de99 AA	250	else
	251	err = mfill_zeropage_pte(dst_mm, dst_pmd, dst_vma,
	252	dst_addr);
	253
	254	cond_resched();
	255
b6ebaedb AA	256	if (unlikely(err == -EFAULT)) {
	257	void *page_kaddr;
	258
	259	up_read(&dst_mm->mmap_sem);
	260	BUG_ON(!page);
	261
	262	page_kaddr = kmap(page);
	263	err = copy_from_user(page_kaddr,
	264	(const void __user *) src_addr,
	265	PAGE_SIZE);
	266	kunmap(page);
	267	if (unlikely(err)) {
	268	err = -EFAULT;
	269	goto out;
	270	}
	271	goto retry;
	272	} else
	273	BUG_ON(page);
	274
c1a4de99 AA	275	if (!err) {
	276	dst_addr += PAGE_SIZE;
	277	src_addr += PAGE_SIZE;
	278	copied += PAGE_SIZE;
	279
	280	if (fatal_signal_pending(current))
	281	err = -EINTR;
	282	}
	283	if (err)
	284	break;
	285	}
	286
b6ebaedb	287	out_unlock:
c1a4de99	288	up_read(&dst_mm->mmap_sem);
b6ebaedb AA	289	out:
	290	if (page)
	291	page_cache_release(page);
c1a4de99 AA	292	BUG_ON(copied < 0);
	293	BUG_ON(err > 0);
	294	BUG_ON(!copied && !err);
	295	return copied ? copied : err;
	296	}
	297
	298	ssize_t mcopy_atomic(struct mm_struct *dst_mm, unsigned long dst_start,
	299	unsigned long src_start, unsigned long len)
	300	{
	301	return __mcopy_atomic(dst_mm, dst_start, src_start, len, false);
	302	}
	303
	304	ssize_t mfill_zeropage(struct mm_struct *dst_mm, unsigned long start,
	305	unsigned long len)
	306	{
	307	return __mcopy_atomic(dst_mm, start, 0, len, true);
	308	}