[linux.git] / mm / userfaultfd.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 *  mm/userfaultfd.c
 *
 *  Copyright (C) 2015  Red Hat, Inc.
 */

#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;
	pgoff_t offset, max_off;
	struct inode *inode;

	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 = -ENOENT;
			*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));

	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
	if (dst_vma->vm_file) {
		/* the shmem MAP_PRIVATE case requires checking the i_size */
		inode = dst_vma->vm_file->f_inode;
		offset = linear_page_index(dst_vma, dst_addr);
		max_off = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
		ret = -EFAULT;
		if (unlikely(offset >= max_off))
			goto out_release_uncharge_unlock;
	}
	ret = -EEXIST;
	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;
	pgoff_t offset, max_off;
	struct inode *inode;

	_dst_pte = pte_mkspecial(pfn_pte(my_zero_pfn(dst_addr),
					 dst_vma->vm_page_prot));
	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
	if (dst_vma->vm_file) {
		/* the shmem MAP_PRIVATE case requires checking the i_size */
		inode = dst_vma->vm_file->f_inode;
		offset = linear_page_index(dst_vma, dst_addr);
		max_off = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
		ret = -EFAULT;
		if (unlikely(offset >= max_off))
			goto out_unlock;
	}
	ret = -EEXIST;
	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;
		/*
		 * Check the vma is registered in uffd, this is
		 * required to enforce the VM_MAYWRITE check done at
		 * uffd registration time.
		 */
		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, mapping, idx);
		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 == -ENOENT)) {
			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;

	/*
	 * The normal page fault path for a shmem will invoke the
	 * fault, fill the hole in the file and COW it right away. The
	 * result generates plain anonymous memory. So when we are
	 * asked to fill an hole in a MAP_PRIVATE shmem mapping, we'll
	 * generate anonymous memory directly without actually filling
	 * the hole. For the MAP_PRIVATE case the robustness check
	 * only happens in the pagetable (to verify it's still none)
	 * and not in the radix tree.
	 */
	if (!(dst_vma->vm_flags & VM_SHARED)) {
		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;
	/*
	 * Check the vma is registered in uffd, this is required to
	 * enforce the VM_MAYWRITE check done at uffd registration
	 * time.
	 */
	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 (!(dst_vma->vm_flags & VM_SHARED) &&
	    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))) {
			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 == -ENOENT)) {
			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
20c8ccb1	1	// SPDX-License-Identifier: GPL-2.0-only
c1a4de99 AA	2	/*
	3	* mm/userfaultfd.c
	4	*
	5	* Copyright (C) 2015 Red Hat, Inc.
c1a4de99 AA	6	*/
	7
	8	#include <linux/mm.h>
174cd4b1	9	#include <linux/sched/signal.h>
c1a4de99 AA	10	#include <linux/pagemap.h>
	11	#include <linux/rmap.h>
	12	#include <linux/swap.h>
	13	#include <linux/swapops.h>
	14	#include <linux/userfaultfd_k.h>
	15	#include <linux/mmu_notifier.h>
60d4d2d2	16	#include <linux/hugetlb.h>
26071ced	17	#include <linux/shmem_fs.h>
c1a4de99 AA	18	#include <asm/tlbflush.h>
	19	#include "internal.h"
	20
	21	static int mcopy_atomic_pte(struct mm_struct *dst_mm,
	22	pmd_t *dst_pmd,
	23	struct vm_area_struct *dst_vma,
	24	unsigned long dst_addr,
b6ebaedb AA	25	unsigned long src_addr,
b6ebaedb AA	26	struct page **pagep)
c1a4de99 AA	27	{
	28	struct mem_cgroup *memcg;
	29	pte_t _dst_pte, *dst_pte;
	30	spinlock_t *ptl;
c1a4de99 AA	31	void *page_kaddr;
c1a4de99 AA	32	int ret;
b6ebaedb	33	struct page *page;
e2a50c1f AA	34	pgoff_t offset, max_off;
e2a50c1f AA	35	struct inode *inode;
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)) {
9e368259	51	ret = -ENOENT;
b6ebaedb AA	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
c1a4de99	76	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
e2a50c1f AA	77	if (dst_vma->vm_file) {
	78	/* the shmem MAP_PRIVATE case requires checking the i_size */
	79	inode = dst_vma->vm_file->f_inode;
	80	offset = linear_page_index(dst_vma, dst_addr);
	81	max_off = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
	82	ret = -EFAULT;
	83	if (unlikely(offset >= max_off))
	84	goto out_release_uncharge_unlock;
	85	}
	86	ret = -EEXIST;
c1a4de99 AA	87	if (!pte_none(*dst_pte))
	88	goto out_release_uncharge_unlock;
	89
	90	inc_mm_counter(dst_mm, MM_ANONPAGES);
d281ee61	91	page_add_new_anon_rmap(page, dst_vma, dst_addr, false);
f627c2f5	92	mem_cgroup_commit_charge(page, memcg, false, false);
c1a4de99 AA	93	lru_cache_add_active_or_unevictable(page, dst_vma);
	94
	95	set_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);
	96
	97	/* No need to invalidate - it was non-present before */
	98	update_mmu_cache(dst_vma, dst_addr, dst_pte);
	99
	100	pte_unmap_unlock(dst_pte, ptl);
	101	ret = 0;
	102	out:
	103	return ret;
	104	out_release_uncharge_unlock:
	105	pte_unmap_unlock(dst_pte, ptl);
f627c2f5	106	mem_cgroup_cancel_charge(page, memcg, false);
c1a4de99	107	out_release:
09cbfeaf	108	put_page(page);
c1a4de99	109	goto out;
c1a4de99 AA	110	}
	111
	112	static int mfill_zeropage_pte(struct mm_struct *dst_mm,
	113	pmd_t *dst_pmd,
	114	struct vm_area_struct *dst_vma,
	115	unsigned long dst_addr)
	116	{
	117	pte_t _dst_pte, *dst_pte;
	118	spinlock_t *ptl;
	119	int ret;
e2a50c1f AA	120	pgoff_t offset, max_off;
e2a50c1f AA	121	struct inode *inode;
c1a4de99 AA	122
	123	_dst_pte = pte_mkspecial(pfn_pte(my_zero_pfn(dst_addr),
	124	dst_vma->vm_page_prot));
c1a4de99	125	dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
e2a50c1f AA	126	if (dst_vma->vm_file) {
	127	/* the shmem MAP_PRIVATE case requires checking the i_size */
	128	inode = dst_vma->vm_file->f_inode;
	129	offset = linear_page_index(dst_vma, dst_addr);
	130	max_off = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
	131	ret = -EFAULT;
	132	if (unlikely(offset >= max_off))
	133	goto out_unlock;
	134	}
	135	ret = -EEXIST;
c1a4de99 AA	136	if (!pte_none(*dst_pte))
	137	goto out_unlock;
	138	set_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);
	139	/* No need to invalidate - it was non-present before */
	140	update_mmu_cache(dst_vma, dst_addr, dst_pte);
	141	ret = 0;
	142	out_unlock:
	143	pte_unmap_unlock(dst_pte, ptl);
	144	return ret;
	145	}
	146
	147	static pmd_t mm_alloc_pmd(struct mm_struct mm, unsigned long address)
	148	{
	149	pgd_t *pgd;
c2febafc	150	p4d_t *p4d;
c1a4de99	151	pud_t *pud;
c1a4de99 AA	152
c1a4de99 AA	153	pgd = pgd_offset(mm, address);
c2febafc KS	154	p4d = p4d_alloc(mm, pgd, address);
	155	if (!p4d)
	156	return NULL;
	157	pud = pud_alloc(mm, p4d, address);
	158	if (!pud)
	159	return NULL;
	160	/*
	161	* Note that we didn't run this because the pmd was
	162	* missing, the *pmd may be already established and in
	163	* turn it may also be a trans_huge_pmd.
	164	*/
	165	return pmd_alloc(mm, pud, address);
c1a4de99 AA	166	}
c1a4de99 AA	167
60d4d2d2 MK	168	#ifdef CONFIG_HUGETLB_PAGE
	169	/*
	170	* __mcopy_atomic processing for HUGETLB vmas. Note that this routine is
	171	* called with mmap_sem held, it will release mmap_sem before returning.
	172	*/
	173	static __always_inline ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
	174	struct vm_area_struct *dst_vma,
	175	unsigned long dst_start,
	176	unsigned long src_start,
	177	unsigned long len,
	178	bool zeropage)
	179	{
1c9e8def MK	180	int vm_alloc_shared = dst_vma->vm_flags & VM_SHARED;
1c9e8def MK	181	int vm_shared = dst_vma->vm_flags & VM_SHARED;
60d4d2d2 MK	182	ssize_t err;
	183	pte_t *dst_pte;
	184	unsigned long src_addr, dst_addr;
	185	long copied;
	186	struct page *page;
	187	struct hstate *h;
	188	unsigned long vma_hpagesize;
	189	pgoff_t idx;
	190	u32 hash;
	191	struct address_space *mapping;
	192
	193	/*
	194	* There is no default zero huge page for all huge page sizes as
	195	* supported by hugetlb. A PMD_SIZE huge pages may exist as used
	196	* by THP. Since we can not reliably insert a zero page, this
	197	* feature is not supported.
	198	*/
	199	if (zeropage) {
	200	up_read(&dst_mm->mmap_sem);
	201	return -EINVAL;
	202	}
	203
	204	src_addr = src_start;
	205	dst_addr = dst_start;
	206	copied = 0;
	207	page = NULL;
	208	vma_hpagesize = vma_kernel_pagesize(dst_vma);
	209
	210	/*
	211	* Validate alignment based on huge page size
	212	*/
	213	err = -EINVAL;
	214	if (dst_start & (vma_hpagesize - 1) \|\| len & (vma_hpagesize - 1))
	215	goto out_unlock;
	216
	217	retry:
	218	/*
	219	* On routine entry dst_vma is set. If we had to drop mmap_sem and
	220	* retry, dst_vma will be set to NULL and we must lookup again.
	221	*/
	222	if (!dst_vma) {
27d02568	223	err = -ENOENT;
60d4d2d2 MK	224	dst_vma = find_vma(dst_mm, dst_start);
	225	if (!dst_vma \|\| !is_vm_hugetlb_page(dst_vma))
	226	goto out_unlock;
60d4d2d2	227	/*
29ec9066 AA	228	* Check the vma is registered in uffd, this is
	229	* required to enforce the VM_MAYWRITE check done at
	230	* uffd registration time.
60d4d2d2	231	*/
27d02568 MR	232	if (!dst_vma->vm_userfaultfd_ctx.ctx)
	233	goto out_unlock;
	234
60d4d2d2 MK	235	if (dst_start < dst_vma->vm_start \|\|
	236	dst_start + len > dst_vma->vm_end)
	237	goto out_unlock;
1c9e8def	238
27d02568 MR	239	err = -EINVAL;
	240	if (vma_hpagesize != vma_kernel_pagesize(dst_vma))
	241	goto out_unlock;
	242
1c9e8def	243	vm_shared = dst_vma->vm_flags & VM_SHARED;
60d4d2d2 MK	244	}
	245
	246	if (WARN_ON(dst_addr & (vma_hpagesize - 1) \|\|
	247	(len - copied) & (vma_hpagesize - 1)))
	248	goto out_unlock;
	249
60d4d2d2	250	/*
1c9e8def	251	* If not shared, ensure the dst_vma has a anon_vma.
60d4d2d2 MK	252	*/
60d4d2d2 MK	253	err = -ENOMEM;
1c9e8def MK	254	if (!vm_shared) {
	255	if (unlikely(anon_vma_prepare(dst_vma)))
	256	goto out_unlock;
	257	}
60d4d2d2 MK	258
	259	h = hstate_vma(dst_vma);
	260
	261	while (src_addr < src_start + len) {
	262	pte_t dst_pteval;
	263
	264	BUG_ON(dst_addr >= dst_start + len);
	265	VM_BUG_ON(dst_addr & ~huge_page_mask(h));
	266
	267	/*
ddeaab32	268	* Serialize via hugetlb_fault_mutex
60d4d2d2	269	*/
b43a9990	270	idx = linear_page_index(dst_vma, dst_addr);
ddeaab32	271	mapping = dst_vma->vm_file->f_mapping;
55254636	272	hash = hugetlb_fault_mutex_hash(h, mapping, idx);
60d4d2d2 MK	273	mutex_lock(&hugetlb_fault_mutex_table[hash]);
	274
	275	err = -ENOMEM;
	276	dst_pte = huge_pte_alloc(dst_mm, dst_addr, huge_page_size(h));
	277	if (!dst_pte) {
	278	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
	279	goto out_unlock;
	280	}
	281
	282	err = -EEXIST;
	283	dst_pteval = huge_ptep_get(dst_pte);
	284	if (!huge_pte_none(dst_pteval)) {
	285	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
	286	goto out_unlock;
	287	}
	288
	289	err = hugetlb_mcopy_atomic_pte(dst_mm, dst_pte, dst_vma,
	290	dst_addr, src_addr, &page);
	291
	292	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
1c9e8def	293	vm_alloc_shared = vm_shared;
60d4d2d2 MK	294
	295	cond_resched();
	296
9e368259	297	if (unlikely(err == -ENOENT)) {
60d4d2d2 MK	298	up_read(&dst_mm->mmap_sem);
	299	BUG_ON(!page);
	300
	301	err = copy_huge_page_from_user(page,
	302	(const void __user *)src_addr,
810a56b9	303	pages_per_huge_page(h), true);
60d4d2d2 MK	304	if (unlikely(err)) {
	305	err = -EFAULT;
	306	goto out;
	307	}
	308	down_read(&dst_mm->mmap_sem);
	309
	310	dst_vma = NULL;
	311	goto retry;
	312	} else
	313	BUG_ON(page);
	314
	315	if (!err) {
	316	dst_addr += vma_hpagesize;
	317	src_addr += vma_hpagesize;
	318	copied += vma_hpagesize;
	319
	320	if (fatal_signal_pending(current))
	321	err = -EINTR;
	322	}
	323	if (err)
	324	break;
	325	}
	326
	327	out_unlock:
	328	up_read(&dst_mm->mmap_sem);
	329	out:
21205bf8 MK	330	if (page) {
	331	/*
	332	* We encountered an error and are about to free a newly
1c9e8def MK	333	* allocated huge page.
	334	*
	335	* Reservation handling is very subtle, and is different for
	336	* private and shared mappings. See the routine
	337	* restore_reserve_on_error for details. Unfortunately, we
	338	* can not call restore_reserve_on_error now as it would
	339	* require holding mmap_sem.
	340	*
	341	* If a reservation for the page existed in the reservation
	342	* map of a private mapping, the map was modified to indicate
	343	* the reservation was consumed when the page was allocated.
	344	* We clear the PagePrivate flag now so that the global
21205bf8 MK	345	* reserve count will not be incremented in free_huge_page.
	346	* The reservation map will still indicate the reservation
	347	* was consumed and possibly prevent later page allocation.
1c9e8def MK	348	* This is better than leaking a global reservation. If no
	349	* reservation existed, it is still safe to clear PagePrivate
	350	* as no adjustments to reservation counts were made during
	351	* allocation.
	352	*
	353	* The reservation map for shared mappings indicates which
	354	* pages have reservations. When a huge page is allocated
	355	* for an address with a reservation, no change is made to
	356	* the reserve map. In this case PagePrivate will be set
	357	* to indicate that the global reservation count should be
	358	* incremented when the page is freed. This is the desired
	359	* behavior. However, when a huge page is allocated for an
	360	* address without a reservation a reservation entry is added
	361	* to the reservation map, and PagePrivate will not be set.
	362	* When the page is freed, the global reserve count will NOT
	363	* be incremented and it will appear as though we have leaked
	364	* reserved page. In this case, set PagePrivate so that the
	365	* global reserve count will be incremented to match the
	366	* reservation map entry which was created.
	367	*
	368	* Note that vm_alloc_shared is based on the flags of the vma
	369	* for which the page was originally allocated. dst_vma could
	370	* be different or NULL on error.
21205bf8	371	*/
1c9e8def MK	372	if (vm_alloc_shared)
	373	SetPagePrivate(page);
	374	else
	375	ClearPagePrivate(page);
60d4d2d2	376	put_page(page);
21205bf8	377	}
60d4d2d2 MK	378	BUG_ON(copied < 0);
	379	BUG_ON(err > 0);
	380	BUG_ON(!copied && !err);
	381	return copied ? copied : err;
	382	}
	383	#else /* !CONFIG_HUGETLB_PAGE */
	384	/* fail at build time if gcc attempts to use this */
	385	extern ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
	386	struct vm_area_struct *dst_vma,
	387	unsigned long dst_start,
	388	unsigned long src_start,
	389	unsigned long len,
	390	bool zeropage);
	391	#endif /* CONFIG_HUGETLB_PAGE */
	392
3217d3c7 MR	393	static __always_inline ssize_t mfill_atomic_pte(struct mm_struct *dst_mm,
	394	pmd_t *dst_pmd,
	395	struct vm_area_struct *dst_vma,
	396	unsigned long dst_addr,
	397	unsigned long src_addr,
	398	struct page **page,
	399	bool zeropage)
	400	{
	401	ssize_t err;
	402
5b51072e AA	403	/*
	404	* The normal page fault path for a shmem will invoke the
	405	* fault, fill the hole in the file and COW it right away. The
	406	* result generates plain anonymous memory. So when we are
	407	* asked to fill an hole in a MAP_PRIVATE shmem mapping, we'll
	408	* generate anonymous memory directly without actually filling
	409	* the hole. For the MAP_PRIVATE case the robustness check
	410	* only happens in the pagetable (to verify it's still none)
	411	* and not in the radix tree.
	412	*/
	413	if (!(dst_vma->vm_flags & VM_SHARED)) {
3217d3c7 MR	414	if (!zeropage)
	415	err = mcopy_atomic_pte(dst_mm, dst_pmd, dst_vma,
	416	dst_addr, src_addr, page);
	417	else
	418	err = mfill_zeropage_pte(dst_mm, dst_pmd,
	419	dst_vma, dst_addr);
	420	} else {
8fb44e54	421	if (!zeropage)
3217d3c7 MR	422	err = shmem_mcopy_atomic_pte(dst_mm, dst_pmd,
	423	dst_vma, dst_addr,
	424	src_addr, page);
8fb44e54 MR	425	else
	426	err = shmem_mfill_zeropage_pte(dst_mm, dst_pmd,
	427	dst_vma, dst_addr);
3217d3c7 MR	428	}
	429
	430	return err;
	431	}
	432
c1a4de99 AA	433	static __always_inline ssize_t __mcopy_atomic(struct mm_struct *dst_mm,
	434	unsigned long dst_start,
	435	unsigned long src_start,
	436	unsigned long len,
df2cc96e MR	437	bool zeropage,
df2cc96e MR	438	bool *mmap_changing)
c1a4de99 AA	439	{
	440	struct vm_area_struct *dst_vma;
	441	ssize_t err;
	442	pmd_t *dst_pmd;
	443	unsigned long src_addr, dst_addr;
b6ebaedb AA	444	long copied;
b6ebaedb AA	445	struct page *page;
c1a4de99 AA	446
	447	/*
	448	* Sanitize the command parameters:
	449	*/
	450	BUG_ON(dst_start & ~PAGE_MASK);
	451	BUG_ON(len & ~PAGE_MASK);
	452
	453	/* Does the address range wrap, or is the span zero-sized? */
	454	BUG_ON(src_start + len <= src_start);
	455	BUG_ON(dst_start + len <= dst_start);
	456
b6ebaedb AA	457	src_addr = src_start;
	458	dst_addr = dst_start;
	459	copied = 0;
	460	page = NULL;
	461	retry:
c1a4de99 AA	462	down_read(&dst_mm->mmap_sem);
c1a4de99 AA	463
df2cc96e MR	464	/*
	465	* If memory mappings are changing because of non-cooperative
	466	* operation (e.g. mremap) running in parallel, bail out and
	467	* request the user to retry later
	468	*/
	469	err = -EAGAIN;
	470	if (mmap_changing && READ_ONCE(*mmap_changing))
	471	goto out_unlock;
	472
c1a4de99 AA	473	/*
	474	* Make sure the vma is not shared, that the dst range is
	475	* both valid and fully within a single existing vma.
	476	*/
27d02568	477	err = -ENOENT;
c1a4de99	478	dst_vma = find_vma(dst_mm, dst_start);
26071ced MR	479	if (!dst_vma)
26071ced MR	480	goto out_unlock;
1c9e8def	481	/*
29ec9066 AA	482	* Check the vma is registered in uffd, this is required to
	483	* enforce the VM_MAYWRITE check done at uffd registration
	484	* time.
1c9e8def	485	*/
27d02568	486	if (!dst_vma->vm_userfaultfd_ctx.ctx)
b6ebaedb	487	goto out_unlock;
1c9e8def	488
c1a4de99 AA	489	if (dst_start < dst_vma->vm_start \|\|
c1a4de99 AA	490	dst_start + len > dst_vma->vm_end)
b6ebaedb	491	goto out_unlock;
c1a4de99	492
27d02568 MR	493	err = -EINVAL;
	494	/*
	495	* shmem_zero_setup is invoked in mmap for MAP_ANONYMOUS\|MAP_SHARED but
	496	* it will overwrite vm_ops, so vma_is_anonymous must return false.
	497	*/
	498	if (WARN_ON_ONCE(vma_is_anonymous(dst_vma) &&
	499	dst_vma->vm_flags & VM_SHARED))
	500	goto out_unlock;
	501
60d4d2d2 MK	502	/*
	503	* If this is a HUGETLB vma, pass off to appropriate routine
	504	*/
	505	if (is_vm_hugetlb_page(dst_vma))
	506	return __mcopy_atomic_hugetlb(dst_mm, dst_vma, dst_start,
	507	src_start, len, zeropage);
	508
26071ced	509	if (!vma_is_anonymous(dst_vma) && !vma_is_shmem(dst_vma))
b6ebaedb	510	goto out_unlock;
c1a4de99 AA	511
	512	/*
	513	* Ensure the dst_vma has a anon_vma or this page
	514	* would get a NULL anon_vma when moved in the
	515	* dst_vma.
	516	*/
	517	err = -ENOMEM;
5b51072e AA	518	if (!(dst_vma->vm_flags & VM_SHARED) &&
5b51072e AA	519	unlikely(anon_vma_prepare(dst_vma)))
b6ebaedb	520	goto out_unlock;
c1a4de99	521
b6ebaedb	522	while (src_addr < src_start + len) {
c1a4de99	523	pmd_t dst_pmdval;
b6ebaedb	524
c1a4de99	525	BUG_ON(dst_addr >= dst_start + len);
b6ebaedb	526
c1a4de99 AA	527	dst_pmd = mm_alloc_pmd(dst_mm, dst_addr);
	528	if (unlikely(!dst_pmd)) {
	529	err = -ENOMEM;
	530	break;
	531	}
	532
	533	dst_pmdval = pmd_read_atomic(dst_pmd);
	534	/*
	535	* If the dst_pmd is mapped as THP don't
	536	* override it and just be strict.
	537	*/
	538	if (unlikely(pmd_trans_huge(dst_pmdval))) {
	539	err = -EEXIST;
	540	break;
	541	}
	542	if (unlikely(pmd_none(dst_pmdval)) &&
4cf58924	543	unlikely(__pte_alloc(dst_mm, dst_pmd))) {
c1a4de99 AA	544	err = -ENOMEM;
	545	break;
	546	}
	547	/* If an huge pmd materialized from under us fail */
	548	if (unlikely(pmd_trans_huge(*dst_pmd))) {
	549	err = -EFAULT;
	550	break;
	551	}
	552
	553	BUG_ON(pmd_none(*dst_pmd));
	554	BUG_ON(pmd_trans_huge(*dst_pmd));
	555
3217d3c7 MR	556	err = mfill_atomic_pte(dst_mm, dst_pmd, dst_vma, dst_addr,
3217d3c7 MR	557	src_addr, &page, zeropage);
c1a4de99 AA	558	cond_resched();
c1a4de99 AA	559
9e368259	560	if (unlikely(err == -ENOENT)) {
b6ebaedb AA	561	void *page_kaddr;
	562
	563	up_read(&dst_mm->mmap_sem);
	564	BUG_ON(!page);
	565
	566	page_kaddr = kmap(page);
	567	err = copy_from_user(page_kaddr,
	568	(const void __user *) src_addr,
	569	PAGE_SIZE);
	570	kunmap(page);
	571	if (unlikely(err)) {
	572	err = -EFAULT;
	573	goto out;
	574	}
	575	goto retry;
	576	} else
	577	BUG_ON(page);
	578
c1a4de99 AA	579	if (!err) {
	580	dst_addr += PAGE_SIZE;
	581	src_addr += PAGE_SIZE;
	582	copied += PAGE_SIZE;
	583
	584	if (fatal_signal_pending(current))
	585	err = -EINTR;
	586	}
	587	if (err)
	588	break;
	589	}
	590
b6ebaedb	591	out_unlock:
c1a4de99	592	up_read(&dst_mm->mmap_sem);
b6ebaedb AA	593	out:
b6ebaedb AA	594	if (page)
09cbfeaf	595	put_page(page);
c1a4de99 AA	596	BUG_ON(copied < 0);
	597	BUG_ON(err > 0);
	598	BUG_ON(!copied && !err);
	599	return copied ? copied : err;
	600	}
	601
	602	ssize_t mcopy_atomic(struct mm_struct *dst_mm, unsigned long dst_start,
df2cc96e MR	603	unsigned long src_start, unsigned long len,
df2cc96e MR	604	bool *mmap_changing)
c1a4de99	605	{
df2cc96e MR	606	return __mcopy_atomic(dst_mm, dst_start, src_start, len, false,
df2cc96e MR	607	mmap_changing);
c1a4de99 AA	608	}
	609
	610	ssize_t mfill_zeropage(struct mm_struct *dst_mm, unsigned long start,
df2cc96e	611	unsigned long len, bool *mmap_changing)
c1a4de99	612	{
df2cc96e	613	return __mcopy_atomic(dst_mm, start, 0, len, true, mmap_changing);
c1a4de99	614	}