[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/sched/signal.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 <linux/hugetlb.h>
#include <linux/shmem_fs.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, false))
		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, false);
	mem_cgroup_commit_charge(page, memcg, false, 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, false);
out_release:
	put_page(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;
	p4d_t *p4d;
	pud_t *pud;

	pgd = pgd_offset(mm, address);
	p4d = p4d_alloc(mm, pgd, address);
	if (!p4d)
		return NULL;
	pud = pud_alloc(mm, p4d, address);
	if (!pud)
		return NULL;
	/*
	 * 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.
	 */
	return pmd_alloc(mm, pud, address);
}

#ifdef CONFIG_HUGETLB_PAGE
/*
 * __mcopy_atomic processing for HUGETLB vmas.  Note that this routine is
 * called with mmap_sem held, it will release mmap_sem before returning.
 */
static __always_inline ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
					      struct vm_area_struct *dst_vma,
					      unsigned long dst_start,
					      unsigned long src_start,
					      unsigned long len,
					      bool zeropage)
{
	int vm_alloc_shared = dst_vma->vm_flags & VM_SHARED;
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
	ssize_t err;
	pte_t *dst_pte;
	unsigned long src_addr, dst_addr;
	long copied;
	struct page *page;
	struct hstate *h;
	unsigned long vma_hpagesize;
	pgoff_t idx;
	u32 hash;
	struct address_space *mapping;

	/*
	 * There is no default zero huge page for all huge page sizes as
	 * supported by hugetlb.  A PMD_SIZE huge pages may exist as used
	 * by THP.  Since we can not reliably insert a zero page, this
	 * feature is not supported.
	 */
	if (zeropage) {
		up_read(&dst_mm->mmap_sem);
		return -EINVAL;
	}

	src_addr = src_start;
	dst_addr = dst_start;
	copied = 0;
	page = NULL;
	vma_hpagesize = vma_kernel_pagesize(dst_vma);

	/*
	 * Validate alignment based on huge page size
	 */
	err = -EINVAL;
	if (dst_start & (vma_hpagesize - 1) || len & (vma_hpagesize - 1))
		goto out_unlock;

retry:
	/*
	 * On routine entry dst_vma is set.  If we had to drop mmap_sem and
	 * retry, dst_vma will be set to NULL and we must lookup again.
	 */
	if (!dst_vma) {
		err = -ENOENT;
		dst_vma = find_vma(dst_mm, dst_start);
		if (!dst_vma || !is_vm_hugetlb_page(dst_vma))
			goto out_unlock;
		/*
		 * Only allow __mcopy_atomic_hugetlb on userfaultfd
		 * registered ranges.
		 */
		if (!dst_vma->vm_userfaultfd_ctx.ctx)
			goto out_unlock;

		if (dst_start < dst_vma->vm_start ||
		    dst_start + len > dst_vma->vm_end)
			goto out_unlock;

		err = -EINVAL;
		if (vma_hpagesize != vma_kernel_pagesize(dst_vma))
			goto out_unlock;

		vm_shared = dst_vma->vm_flags & VM_SHARED;
	}

	if (WARN_ON(dst_addr & (vma_hpagesize - 1) ||
		    (len - copied) & (vma_hpagesize - 1)))
		goto out_unlock;

	/*
	 * If not shared, ensure the dst_vma has a anon_vma.
	 */
	err = -ENOMEM;
	if (!vm_shared) {
		if (unlikely(anon_vma_prepare(dst_vma)))
			goto out_unlock;
	}

	h = hstate_vma(dst_vma);

	while (src_addr < src_start + len) {
		pte_t dst_pteval;

		BUG_ON(dst_addr >= dst_start + len);
		VM_BUG_ON(dst_addr & ~huge_page_mask(h));

		/*
		 * Serialize via hugetlb_fault_mutex
		 */
		idx = linear_page_index(dst_vma, dst_addr);
		mapping = dst_vma->vm_file->f_mapping;
		hash = hugetlb_fault_mutex_hash(h, dst_mm, dst_vma, mapping,
								idx, dst_addr);
		mutex_lock(&hugetlb_fault_mutex_table[hash]);

		err = -ENOMEM;
		dst_pte = huge_pte_alloc(dst_mm, dst_addr, huge_page_size(h));
		if (!dst_pte) {
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			goto out_unlock;
		}

		err = -EEXIST;
		dst_pteval = huge_ptep_get(dst_pte);
		if (!huge_pte_none(dst_pteval)) {
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			goto out_unlock;
		}

		err = hugetlb_mcopy_atomic_pte(dst_mm, dst_pte, dst_vma,
						dst_addr, src_addr, &page);

		mutex_unlock(&hugetlb_fault_mutex_table[hash]);
		vm_alloc_shared = vm_shared;

		cond_resched();

		if (unlikely(err == -EFAULT)) {
			up_read(&dst_mm->mmap_sem);
			BUG_ON(!page);

			err = copy_huge_page_from_user(page,
						(const void __user *)src_addr,
						pages_per_huge_page(h), true);
			if (unlikely(err)) {
				err = -EFAULT;
				goto out;
			}
			down_read(&dst_mm->mmap_sem);

			dst_vma = NULL;
			goto retry;
		} else
			BUG_ON(page);

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

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

out_unlock:
	up_read(&dst_mm->mmap_sem);
out:
	if (page) {
		/*
		 * We encountered an error and are about to free a newly
		 * allocated huge page.
		 *
		 * Reservation handling is very subtle, and is different for
		 * private and shared mappings.  See the routine
		 * restore_reserve_on_error for details.  Unfortunately, we
		 * can not call restore_reserve_on_error now as it would
		 * require holding mmap_sem.
		 *
		 * If a reservation for the page existed in the reservation
		 * map of a private mapping, the map was modified to indicate
		 * the reservation was consumed when the page was allocated.
		 * We clear the PagePrivate flag now so that the global
		 * reserve count will not be incremented in free_huge_page.
		 * The reservation map will still indicate the reservation
		 * was consumed and possibly prevent later page allocation.
		 * This is better than leaking a global reservation.  If no
		 * reservation existed, it is still safe to clear PagePrivate
		 * as no adjustments to reservation counts were made during
		 * allocation.
		 *
		 * The reservation map for shared mappings indicates which
		 * pages have reservations.  When a huge page is allocated
		 * for an address with a reservation, no change is made to
		 * the reserve map.  In this case PagePrivate will be set
		 * to indicate that the global reservation count should be
		 * incremented when the page is freed.  This is the desired
		 * behavior.  However, when a huge page is allocated for an
		 * address without a reservation a reservation entry is added
		 * to the reservation map, and PagePrivate will not be set.
		 * When the page is freed, the global reserve count will NOT
		 * be incremented and it will appear as though we have leaked
		 * reserved page.  In this case, set PagePrivate so that the
		 * global reserve count will be incremented to match the
		 * reservation map entry which was created.
		 *
		 * Note that vm_alloc_shared is based on the flags of the vma
		 * for which the page was originally allocated.  dst_vma could
		 * be different or NULL on error.
		 */
		if (vm_alloc_shared)
			SetPagePrivate(page);
		else
			ClearPagePrivate(page);
		put_page(page);
	}
	BUG_ON(copied < 0);
	BUG_ON(err > 0);
	BUG_ON(!copied && !err);
	return copied ? copied : err;
}
#else /* !CONFIG_HUGETLB_PAGE */
/* fail at build time if gcc attempts to use this */
extern ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
				      struct vm_area_struct *dst_vma,
				      unsigned long dst_start,
				      unsigned long src_start,
				      unsigned long len,
				      bool zeropage);
#endif /* CONFIG_HUGETLB_PAGE */

static __always_inline ssize_t mfill_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 **page,
						bool zeropage)
{
	ssize_t err;

	if (vma_is_anonymous(dst_vma)) {
		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);
	} else {
		if (!zeropage)
			err = shmem_mcopy_atomic_pte(dst_mm, dst_pmd,
						     dst_vma, dst_addr,
						     src_addr, page);
		else
			err = shmem_mfill_zeropage_pte(dst_mm, dst_pmd,
						       dst_vma, dst_addr);
	}

	return err;
}

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,
					      bool *mmap_changing)
{
	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);

	/*
	 * If memory mappings are changing because of non-cooperative
	 * operation (e.g. mremap) running in parallel, bail out and
	 * request the user to retry later
	 */
	err = -EAGAIN;
	if (mmap_changing && READ_ONCE(*mmap_changing))
		goto out_unlock;

	/*
	 * Make sure the vma is not shared, that the dst range is
	 * both valid and fully within a single existing vma.
	 */
	err = -ENOENT;
	dst_vma = find_vma(dst_mm, dst_start);
	if (!dst_vma)
		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;

	if (dst_start < dst_vma->vm_start ||
	    dst_start + len > dst_vma->vm_end)
		goto out_unlock;

	err = -EINVAL;
	/*
	 * shmem_zero_setup is invoked in mmap for MAP_ANONYMOUS|MAP_SHARED but
	 * it will overwrite vm_ops, so vma_is_anonymous must return false.
	 */
	if (WARN_ON_ONCE(vma_is_anonymous(dst_vma) &&
	    dst_vma->vm_flags & VM_SHARED))
		goto out_unlock;

	/*
	 * If this is a HUGETLB vma, pass off to appropriate routine
	 */
	if (is_vm_hugetlb_page(dst_vma))
		return  __mcopy_atomic_hugetlb(dst_mm, dst_vma, dst_start,
						src_start, len, zeropage);

	if (!vma_is_anonymous(dst_vma) && !vma_is_shmem(dst_vma))
		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 (vma_is_anonymous(dst_vma) && 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_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));

		err = mfill_atomic_pte(dst_mm, dst_pmd, dst_vma, dst_addr,
				       src_addr, &page, zeropage);
		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)
		put_page(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,
		     bool *mmap_changing)
{
	return __mcopy_atomic(dst_mm, dst_start, src_start, len, false,
			      mmap_changing);
}

ssize_t mfill_zeropage(struct mm_struct *dst_mm, unsigned long start,
		       unsigned long len, bool *mmap_changing)
{
	return __mcopy_atomic(dst_mm, start, 0, len, true, mmap_changing);
}
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>
174cd4b1	11	#include <linux/sched/signal.h>
c1a4de99 AA	12	#include <linux/pagemap.h>
	13	#include <linux/rmap.h>
	14	#include <linux/swap.h>
	15	#include <linux/swapops.h>
	16	#include <linux/userfaultfd_k.h>
	17	#include <linux/mmu_notifier.h>
60d4d2d2	18	#include <linux/hugetlb.h>
26071ced	19	#include <linux/shmem_fs.h>
c1a4de99 AA	20	#include <asm/tlbflush.h>
	21	#include "internal.h"
	22
	23	static int mcopy_atomic_pte(struct mm_struct *dst_mm,
	24	pmd_t *dst_pmd,
	25	struct vm_area_struct *dst_vma,
	26	unsigned long dst_addr,
b6ebaedb AA	27	unsigned long src_addr,
b6ebaedb AA	28	struct page **pagep)
c1a4de99 AA	29	{
	30	struct mem_cgroup *memcg;
	31	pte_t _dst_pte, *dst_pte;
	32	spinlock_t *ptl;
c1a4de99 AA	33	void *page_kaddr;
c1a4de99 AA	34	int ret;
b6ebaedb	35	struct page *page;
c1a4de99	36
b6ebaedb AA	37	if (!*pagep) {
	38	ret = -ENOMEM;
	39	page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, dst_vma, dst_addr);
	40	if (!page)
	41	goto out;
	42
	43	page_kaddr = kmap_atomic(page);
	44	ret = copy_from_user(page_kaddr,
	45	(const void __user *) src_addr,
	46	PAGE_SIZE);
	47	kunmap_atomic(page_kaddr);
	48
	49	/* fallback to copy_from_user outside mmap_sem */
	50	if (unlikely(ret)) {
	51	ret = -EFAULT;
	52	*pagep = page;
	53	/* don't free the page */
	54	goto out;
	55	}
	56	} else {
	57	page = *pagep;
	58	*pagep = NULL;
	59	}
c1a4de99 AA	60
	61	/*
	62	* The memory barrier inside __SetPageUptodate makes sure that
	63	* preceeding stores to the page contents become visible before
	64	* the set_pte_at() write.
	65	*/
	66	__SetPageUptodate(page);
	67
	68	ret = -ENOMEM;
f627c2f5	69	if (mem_cgroup_try_charge(page, dst_mm, GFP_KERNEL, &memcg, false))
c1a4de99 AA	70	goto out_release;
	71
	72	_dst_pte = mk_pte(page, dst_vma->vm_page_prot);
	73	if (dst_vma->vm_flags & VM_WRITE)
	74	_dst_pte = pte_mkwrite(pte_mkdirty(_dst_pte));
	75
	76	ret = -EEXIST;
	77	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
	78	if (!pte_none(*dst_pte))
	79	goto out_release_uncharge_unlock;
	80
	81	inc_mm_counter(dst_mm, MM_ANONPAGES);
d281ee61	82	page_add_new_anon_rmap(page, dst_vma, dst_addr, false);
f627c2f5	83	mem_cgroup_commit_charge(page, memcg, false, false);
c1a4de99 AA	84	lru_cache_add_active_or_unevictable(page, dst_vma);
	85
	86	set_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);
	87
	88	/* No need to invalidate - it was non-present before */
	89	update_mmu_cache(dst_vma, dst_addr, dst_pte);
	90
	91	pte_unmap_unlock(dst_pte, ptl);
	92	ret = 0;
	93	out:
	94	return ret;
	95	out_release_uncharge_unlock:
	96	pte_unmap_unlock(dst_pte, ptl);
f627c2f5	97	mem_cgroup_cancel_charge(page, memcg, false);
c1a4de99	98	out_release:
09cbfeaf	99	put_page(page);
c1a4de99	100	goto out;
c1a4de99 AA	101	}
	102
	103	static int mfill_zeropage_pte(struct mm_struct *dst_mm,
	104	pmd_t *dst_pmd,
	105	struct vm_area_struct *dst_vma,
	106	unsigned long dst_addr)
	107	{
	108	pte_t _dst_pte, *dst_pte;
	109	spinlock_t *ptl;
	110	int ret;
	111
	112	_dst_pte = pte_mkspecial(pfn_pte(my_zero_pfn(dst_addr),
	113	dst_vma->vm_page_prot));
	114	ret = -EEXIST;
	115	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
	116	if (!pte_none(*dst_pte))
	117	goto out_unlock;
	118	set_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);
	119	/* No need to invalidate - it was non-present before */
	120	update_mmu_cache(dst_vma, dst_addr, dst_pte);
	121	ret = 0;
	122	out_unlock:
	123	pte_unmap_unlock(dst_pte, ptl);
	124	return ret;
	125	}
	126
	127	static pmd_t mm_alloc_pmd(struct mm_struct mm, unsigned long address)
	128	{
	129	pgd_t *pgd;
c2febafc	130	p4d_t *p4d;
c1a4de99	131	pud_t *pud;
c1a4de99 AA	132
c1a4de99 AA	133	pgd = pgd_offset(mm, address);
c2febafc KS	134	p4d = p4d_alloc(mm, pgd, address);
	135	if (!p4d)
	136	return NULL;
	137	pud = pud_alloc(mm, p4d, address);
	138	if (!pud)
	139	return NULL;
	140	/*
	141	* Note that we didn't run this because the pmd was
	142	* missing, the *pmd may be already established and in
	143	* turn it may also be a trans_huge_pmd.
	144	*/
	145	return pmd_alloc(mm, pud, address);
c1a4de99 AA	146	}
c1a4de99 AA	147
60d4d2d2 MK	148	#ifdef CONFIG_HUGETLB_PAGE
	149	/*
	150	* __mcopy_atomic processing for HUGETLB vmas. Note that this routine is
	151	* called with mmap_sem held, it will release mmap_sem before returning.
	152	*/
	153	static __always_inline ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
	154	struct vm_area_struct *dst_vma,
	155	unsigned long dst_start,
	156	unsigned long src_start,
	157	unsigned long len,
	158	bool zeropage)
	159	{
1c9e8def MK	160	int vm_alloc_shared = dst_vma->vm_flags & VM_SHARED;
1c9e8def MK	161	int vm_shared = dst_vma->vm_flags & VM_SHARED;
60d4d2d2 MK	162	ssize_t err;
	163	pte_t *dst_pte;
	164	unsigned long src_addr, dst_addr;
	165	long copied;
	166	struct page *page;
	167	struct hstate *h;
	168	unsigned long vma_hpagesize;
	169	pgoff_t idx;
	170	u32 hash;
	171	struct address_space *mapping;
	172
	173	/*
	174	* There is no default zero huge page for all huge page sizes as
	175	* supported by hugetlb. A PMD_SIZE huge pages may exist as used
	176	* by THP. Since we can not reliably insert a zero page, this
	177	* feature is not supported.
	178	*/
	179	if (zeropage) {
	180	up_read(&dst_mm->mmap_sem);
	181	return -EINVAL;
	182	}
	183
	184	src_addr = src_start;
	185	dst_addr = dst_start;
	186	copied = 0;
	187	page = NULL;
	188	vma_hpagesize = vma_kernel_pagesize(dst_vma);
	189
	190	/*
	191	* Validate alignment based on huge page size
	192	*/
	193	err = -EINVAL;
	194	if (dst_start & (vma_hpagesize - 1) \|\| len & (vma_hpagesize - 1))
	195	goto out_unlock;
	196
	197	retry:
	198	/*
	199	* On routine entry dst_vma is set. If we had to drop mmap_sem and
	200	* retry, dst_vma will be set to NULL and we must lookup again.
	201	*/
	202	if (!dst_vma) {
27d02568	203	err = -ENOENT;
60d4d2d2 MK	204	dst_vma = find_vma(dst_mm, dst_start);
	205	if (!dst_vma \|\| !is_vm_hugetlb_page(dst_vma))
	206	goto out_unlock;
60d4d2d2	207	/*
27d02568 MR	208	* Only allow __mcopy_atomic_hugetlb on userfaultfd
27d02568 MR	209	* registered ranges.
60d4d2d2	210	*/
27d02568 MR	211	if (!dst_vma->vm_userfaultfd_ctx.ctx)
	212	goto out_unlock;
	213
60d4d2d2 MK	214	if (dst_start < dst_vma->vm_start \|\|
	215	dst_start + len > dst_vma->vm_end)
	216	goto out_unlock;
1c9e8def	217
27d02568 MR	218	err = -EINVAL;
	219	if (vma_hpagesize != vma_kernel_pagesize(dst_vma))
	220	goto out_unlock;
	221
1c9e8def	222	vm_shared = dst_vma->vm_flags & VM_SHARED;
60d4d2d2 MK	223	}
	224
	225	if (WARN_ON(dst_addr & (vma_hpagesize - 1) \|\|
	226	(len - copied) & (vma_hpagesize - 1)))
	227	goto out_unlock;
	228
60d4d2d2	229	/*
1c9e8def	230	* If not shared, ensure the dst_vma has a anon_vma.
60d4d2d2 MK	231	*/
60d4d2d2 MK	232	err = -ENOMEM;
1c9e8def MK	233	if (!vm_shared) {
	234	if (unlikely(anon_vma_prepare(dst_vma)))
	235	goto out_unlock;
	236	}
60d4d2d2 MK	237
	238	h = hstate_vma(dst_vma);
	239
	240	while (src_addr < src_start + len) {
	241	pte_t dst_pteval;
	242
	243	BUG_ON(dst_addr >= dst_start + len);
	244	VM_BUG_ON(dst_addr & ~huge_page_mask(h));
	245
	246	/*
	247	* Serialize via hugetlb_fault_mutex
	248	*/
	249	idx = linear_page_index(dst_vma, dst_addr);
	250	mapping = dst_vma->vm_file->f_mapping;
	251	hash = hugetlb_fault_mutex_hash(h, dst_mm, dst_vma, mapping,
	252	idx, dst_addr);
	253	mutex_lock(&hugetlb_fault_mutex_table[hash]);
	254
	255	err = -ENOMEM;
	256	dst_pte = huge_pte_alloc(dst_mm, dst_addr, huge_page_size(h));
	257	if (!dst_pte) {
	258	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
	259	goto out_unlock;
	260	}
	261
	262	err = -EEXIST;
	263	dst_pteval = huge_ptep_get(dst_pte);
	264	if (!huge_pte_none(dst_pteval)) {
	265	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
	266	goto out_unlock;
	267	}
	268
	269	err = hugetlb_mcopy_atomic_pte(dst_mm, dst_pte, dst_vma,
	270	dst_addr, src_addr, &page);
	271
	272	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
1c9e8def	273	vm_alloc_shared = vm_shared;
60d4d2d2 MK	274
	275	cond_resched();
	276
	277	if (unlikely(err == -EFAULT)) {
	278	up_read(&dst_mm->mmap_sem);
	279	BUG_ON(!page);
	280
	281	err = copy_huge_page_from_user(page,
	282	(const void __user *)src_addr,
810a56b9	283	pages_per_huge_page(h), true);
60d4d2d2 MK	284	if (unlikely(err)) {
	285	err = -EFAULT;
	286	goto out;
	287	}
	288	down_read(&dst_mm->mmap_sem);
	289
	290	dst_vma = NULL;
	291	goto retry;
	292	} else
	293	BUG_ON(page);
	294
	295	if (!err) {
	296	dst_addr += vma_hpagesize;
	297	src_addr += vma_hpagesize;
	298	copied += vma_hpagesize;
	299
	300	if (fatal_signal_pending(current))
	301	err = -EINTR;
	302	}
	303	if (err)
	304	break;
	305	}
	306
	307	out_unlock:
	308	up_read(&dst_mm->mmap_sem);
	309	out:
21205bf8 MK	310	if (page) {
	311	/*
	312	* We encountered an error and are about to free a newly
1c9e8def MK	313	* allocated huge page.
	314	*
	315	* Reservation handling is very subtle, and is different for
	316	* private and shared mappings. See the routine
	317	* restore_reserve_on_error for details. Unfortunately, we
	318	* can not call restore_reserve_on_error now as it would
	319	* require holding mmap_sem.
	320	*
	321	* If a reservation for the page existed in the reservation
	322	* map of a private mapping, the map was modified to indicate
	323	* the reservation was consumed when the page was allocated.
	324	* We clear the PagePrivate flag now so that the global
21205bf8 MK	325	* reserve count will not be incremented in free_huge_page.
	326	* The reservation map will still indicate the reservation
	327	* was consumed and possibly prevent later page allocation.
1c9e8def MK	328	* This is better than leaking a global reservation. If no
	329	* reservation existed, it is still safe to clear PagePrivate
	330	* as no adjustments to reservation counts were made during
	331	* allocation.
	332	*
	333	* The reservation map for shared mappings indicates which
	334	* pages have reservations. When a huge page is allocated
	335	* for an address with a reservation, no change is made to
	336	* the reserve map. In this case PagePrivate will be set
	337	* to indicate that the global reservation count should be
	338	* incremented when the page is freed. This is the desired
	339	* behavior. However, when a huge page is allocated for an
	340	* address without a reservation a reservation entry is added
	341	* to the reservation map, and PagePrivate will not be set.
	342	* When the page is freed, the global reserve count will NOT
	343	* be incremented and it will appear as though we have leaked
	344	* reserved page. In this case, set PagePrivate so that the
	345	* global reserve count will be incremented to match the
	346	* reservation map entry which was created.
	347	*
	348	* Note that vm_alloc_shared is based on the flags of the vma
	349	* for which the page was originally allocated. dst_vma could
	350	* be different or NULL on error.
21205bf8	351	*/
1c9e8def MK	352	if (vm_alloc_shared)
	353	SetPagePrivate(page);
	354	else
	355	ClearPagePrivate(page);
60d4d2d2	356	put_page(page);
21205bf8	357	}
60d4d2d2 MK	358	BUG_ON(copied < 0);
	359	BUG_ON(err > 0);
	360	BUG_ON(!copied && !err);
	361	return copied ? copied : err;
	362	}
	363	#else /* !CONFIG_HUGETLB_PAGE */
	364	/* fail at build time if gcc attempts to use this */
	365	extern ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
	366	struct vm_area_struct *dst_vma,
	367	unsigned long dst_start,
	368	unsigned long src_start,
	369	unsigned long len,
	370	bool zeropage);
	371	#endif /* CONFIG_HUGETLB_PAGE */
	372
3217d3c7 MR	373	static __always_inline ssize_t mfill_atomic_pte(struct mm_struct *dst_mm,
	374	pmd_t *dst_pmd,
	375	struct vm_area_struct *dst_vma,
	376	unsigned long dst_addr,
	377	unsigned long src_addr,
	378	struct page **page,
	379	bool zeropage)
	380	{
	381	ssize_t err;
	382
	383	if (vma_is_anonymous(dst_vma)) {
	384	if (!zeropage)
	385	err = mcopy_atomic_pte(dst_mm, dst_pmd, dst_vma,
	386	dst_addr, src_addr, page);
	387	else
	388	err = mfill_zeropage_pte(dst_mm, dst_pmd,
	389	dst_vma, dst_addr);
	390	} else {
8fb44e54	391	if (!zeropage)
3217d3c7 MR	392	err = shmem_mcopy_atomic_pte(dst_mm, dst_pmd,
	393	dst_vma, dst_addr,
	394	src_addr, page);
8fb44e54 MR	395	else
	396	err = shmem_mfill_zeropage_pte(dst_mm, dst_pmd,
	397	dst_vma, dst_addr);
3217d3c7 MR	398	}
	399
	400	return err;
	401	}
	402
c1a4de99 AA	403	static __always_inline ssize_t __mcopy_atomic(struct mm_struct *dst_mm,
	404	unsigned long dst_start,
	405	unsigned long src_start,
	406	unsigned long len,
df2cc96e MR	407	bool zeropage,
df2cc96e MR	408	bool *mmap_changing)
c1a4de99 AA	409	{
	410	struct vm_area_struct *dst_vma;
	411	ssize_t err;
	412	pmd_t *dst_pmd;
	413	unsigned long src_addr, dst_addr;
b6ebaedb AA	414	long copied;
b6ebaedb AA	415	struct page *page;
c1a4de99 AA	416
	417	/*
	418	* Sanitize the command parameters:
	419	*/
	420	BUG_ON(dst_start & ~PAGE_MASK);
	421	BUG_ON(len & ~PAGE_MASK);
	422
	423	/* Does the address range wrap, or is the span zero-sized? */
	424	BUG_ON(src_start + len <= src_start);
	425	BUG_ON(dst_start + len <= dst_start);
	426
b6ebaedb AA	427	src_addr = src_start;
	428	dst_addr = dst_start;
	429	copied = 0;
	430	page = NULL;
	431	retry:
c1a4de99 AA	432	down_read(&dst_mm->mmap_sem);
c1a4de99 AA	433
df2cc96e MR	434	/*
	435	* If memory mappings are changing because of non-cooperative
	436	* operation (e.g. mremap) running in parallel, bail out and
	437	* request the user to retry later
	438	*/
	439	err = -EAGAIN;
	440	if (mmap_changing && READ_ONCE(*mmap_changing))
	441	goto out_unlock;
	442
c1a4de99 AA	443	/*
	444	* Make sure the vma is not shared, that the dst range is
	445	* both valid and fully within a single existing vma.
	446	*/
27d02568	447	err = -ENOENT;
c1a4de99	448	dst_vma = find_vma(dst_mm, dst_start);
26071ced MR	449	if (!dst_vma)
26071ced MR	450	goto out_unlock;
1c9e8def	451	/*
27d02568 MR	452	* Be strict and only allow __mcopy_atomic on userfaultfd
	453	* registered ranges to prevent userland errors going
	454	* unnoticed. As far as the VM consistency is concerned, it
	455	* would be perfectly safe to remove this check, but there's
	456	* no useful usage for __mcopy_atomic ouside of userfaultfd
	457	* registered ranges. This is after all why these are ioctls
	458	* belonging to the userfaultfd and not syscalls.
1c9e8def	459	*/
27d02568	460	if (!dst_vma->vm_userfaultfd_ctx.ctx)
b6ebaedb	461	goto out_unlock;
1c9e8def	462
c1a4de99 AA	463	if (dst_start < dst_vma->vm_start \|\|
c1a4de99 AA	464	dst_start + len > dst_vma->vm_end)
b6ebaedb	465	goto out_unlock;
c1a4de99	466
27d02568 MR	467	err = -EINVAL;
	468	/*
	469	* shmem_zero_setup is invoked in mmap for MAP_ANONYMOUS\|MAP_SHARED but
	470	* it will overwrite vm_ops, so vma_is_anonymous must return false.
	471	*/
	472	if (WARN_ON_ONCE(vma_is_anonymous(dst_vma) &&
	473	dst_vma->vm_flags & VM_SHARED))
	474	goto out_unlock;
	475
60d4d2d2 MK	476	/*
	477	* If this is a HUGETLB vma, pass off to appropriate routine
	478	*/
	479	if (is_vm_hugetlb_page(dst_vma))
	480	return __mcopy_atomic_hugetlb(dst_mm, dst_vma, dst_start,
	481	src_start, len, zeropage);
	482
26071ced	483	if (!vma_is_anonymous(dst_vma) && !vma_is_shmem(dst_vma))
b6ebaedb	484	goto out_unlock;
c1a4de99 AA	485
	486	/*
	487	* Ensure the dst_vma has a anon_vma or this page
	488	* would get a NULL anon_vma when moved in the
	489	* dst_vma.
	490	*/
	491	err = -ENOMEM;
26071ced	492	if (vma_is_anonymous(dst_vma) && unlikely(anon_vma_prepare(dst_vma)))
b6ebaedb	493	goto out_unlock;
c1a4de99	494
b6ebaedb	495	while (src_addr < src_start + len) {
c1a4de99	496	pmd_t dst_pmdval;
b6ebaedb	497
c1a4de99	498	BUG_ON(dst_addr >= dst_start + len);
b6ebaedb	499
c1a4de99 AA	500	dst_pmd = mm_alloc_pmd(dst_mm, dst_addr);
	501	if (unlikely(!dst_pmd)) {
	502	err = -ENOMEM;
	503	break;
	504	}
	505
	506	dst_pmdval = pmd_read_atomic(dst_pmd);
	507	/*
	508	* If the dst_pmd is mapped as THP don't
	509	* override it and just be strict.
	510	*/
	511	if (unlikely(pmd_trans_huge(dst_pmdval))) {
	512	err = -EEXIST;
	513	break;
	514	}
	515	if (unlikely(pmd_none(dst_pmdval)) &&
3ed3a4f0	516	unlikely(__pte_alloc(dst_mm, dst_pmd, dst_addr))) {
c1a4de99 AA	517	err = -ENOMEM;
	518	break;
	519	}
	520	/* If an huge pmd materialized from under us fail */
	521	if (unlikely(pmd_trans_huge(*dst_pmd))) {
	522	err = -EFAULT;
	523	break;
	524	}
	525
	526	BUG_ON(pmd_none(*dst_pmd));
	527	BUG_ON(pmd_trans_huge(*dst_pmd));
	528
3217d3c7 MR	529	err = mfill_atomic_pte(dst_mm, dst_pmd, dst_vma, dst_addr,
3217d3c7 MR	530	src_addr, &page, zeropage);
c1a4de99 AA	531	cond_resched();
c1a4de99 AA	532
b6ebaedb AA	533	if (unlikely(err == -EFAULT)) {
	534	void *page_kaddr;
	535
	536	up_read(&dst_mm->mmap_sem);
	537	BUG_ON(!page);
	538
	539	page_kaddr = kmap(page);
	540	err = copy_from_user(page_kaddr,
	541	(const void __user *) src_addr,
	542	PAGE_SIZE);
	543	kunmap(page);
	544	if (unlikely(err)) {
	545	err = -EFAULT;
	546	goto out;
	547	}
	548	goto retry;
	549	} else
	550	BUG_ON(page);
	551
c1a4de99 AA	552	if (!err) {
	553	dst_addr += PAGE_SIZE;
	554	src_addr += PAGE_SIZE;
	555	copied += PAGE_SIZE;
	556
	557	if (fatal_signal_pending(current))
	558	err = -EINTR;
	559	}
	560	if (err)
	561	break;
	562	}
	563
b6ebaedb	564	out_unlock:
c1a4de99	565	up_read(&dst_mm->mmap_sem);
b6ebaedb AA	566	out:
b6ebaedb AA	567	if (page)
09cbfeaf	568	put_page(page);
c1a4de99 AA	569	BUG_ON(copied < 0);
	570	BUG_ON(err > 0);
	571	BUG_ON(!copied && !err);
	572	return copied ? copied : err;
	573	}
	574
	575	ssize_t mcopy_atomic(struct mm_struct *dst_mm, unsigned long dst_start,
df2cc96e MR	576	unsigned long src_start, unsigned long len,
df2cc96e MR	577	bool *mmap_changing)
c1a4de99	578	{
df2cc96e MR	579	return __mcopy_atomic(dst_mm, dst_start, src_start, len, false,
df2cc96e MR	580	mmap_changing);
c1a4de99 AA	581	}
	582
	583	ssize_t mfill_zeropage(struct mm_struct *dst_mm, unsigned long start,
df2cc96e	584	unsigned long len, bool *mmap_changing)
c1a4de99	585	{
df2cc96e	586	return __mcopy_atomic(dst_mm, start, 0, len, true, mmap_changing);
c1a4de99	587	}