Re: [PATCH v5] userfaultfd/shmem: fix MCOPY_ATOMIC_CONTINUE behavior

On Tue, Apr 6, 2021 at 4:49 PM Peter Xu peterx@redhat.com wrote:

On Mon, Apr 05, 2021 at 10:19:17AM -0700, Axel Rasmussen wrote:

...

Previously, the continue implementation in shmem_mcopy_atomic_pte was incorrect for two main reasons:

• It didn't correctly skip some sections of code which make sense for newly allocated pages, but absolutely don't make sense for pre-existing page cache pages.

• Because shmem_mcopy_continue_pte is called only if VM_SHARED is detected in mm/userfaultd.c, we were incorrectly not supporting the case where a tmpfs file had been mmap()-ed with MAP_PRIVATE.

So, this patch does the following:

In mm/userfaultfd.c, break the logic to install PTEs out of mcopy_atomic_pte, into a separate helper function.

In mfill_atomic_pte, for the MCOPY_ATOMIC_CONTINUE case, simply look up the existing page in the page cache, and then use the PTE installation helper to setup the mapping. This addresses the two issues noted above.

The previous code's bugs manifested clearly in the error handling path. So, to detect this kind of issue in the future, modify the selftest to exercise the error handling path as well.

Note that this patch is based on linux-next/akpm; the "fixes" line refers to a SHA1 in that branch.

Changes since v4:

• Added back the userfaultfd.c selftest changes from v3; this file was mistakenly reverted in v4.

Changes since v3:

• Significantly refactored the patch. Continue handling now happens in mm/userfaultfd.c, via a PTE installation helper. Most of the mm/shmem.c changes from the patch being fixed [1] are reverted.

Changes since v2:

• Drop the ClearPageDirty() entirely, instead of trying to remember the old value.
• Modify both pgoff / max_off checks to use pgoff. It's equivalent to offset, but offset wasn't initialized until the first check (which we're skipping).
• Keep the second pgoff / max_off check in the continue case.

Changes since v1:

• Refactor to skip ahead with goto, instead of adding several more "if (!is_continue)".
• Fix unconditional ClearPageDirty().
• Don't pte_mkwrite() when is_continue && !VM_SHARED.

[1] https://lore.kernel.org/patchwork/patch/1392464/

Fixes: 00da60b9d0a0 ("userfaultfd: support minor fault handling for shmem") Signed-off-by: Axel Rasmussen axelrasmussen@google.com

---

mm/shmem.c | 56 +++---- mm/userfaultfd.c | 183 ++++++++++++++++------- tools/testing/selftests/vm/userfaultfd.c | 12 ++ 3 files changed, 168 insertions(+), 83 deletions(-)

diff --git a/mm/shmem.c b/mm/shmem.c index 5cfd2fb6e52b..9d9a9f254f33 100644 --- a/mm/shmem.c +++ b/mm/shmem.c @@ -2366,7 +2366,6 @@ int shmem_mcopy_atomic_pte(struct mm_struct *dst_mm, pmd_t *dst_pmd, unsigned long dst_addr, unsigned long src_addr, enum mcopy_atomic_mode mode, struct page **pagep) {

• bool is_continue = (mode == MCOPY_ATOMIC_CONTINUE);
  struct inode *inode = file_inode(dst_vma->vm_file);
  struct shmem_inode_info *info = SHMEM_I(inode);
  struct address_space *mapping = inode->i_mapping;
  

@@ -2377,18 +2376,17 @@ int shmem_mcopy_atomic_pte(struct mm_struct *dst_mm, pmd_t *dst_pmd, struct page *page; pte_t _dst_pte, *dst_pte; int ret;

• pgoff_t offset, max_off;
  

• pgoff_t max_off;
  

• /* Handled by mcontinue_atomic_pte instead. */
  

• if (WARN_ON_ONCE(mode == MCOPY_ATOMIC_CONTINUE))
  

•         return -EINVAL;
  

(It would be ideal if this patch can be squashed into original one, since ideally shmem_mcopy_atomic_pte could be mostly untouched to support uffd-minor then we can avoid a lot of code churns; it's just that we noticed it too late.. then this warn_on_one will not be needed too since previously it was a bool)

...

  ret = -ENOMEM;
  if (!shmem_inode_acct_block(inode, 1))
          goto out;

• if (is_continue) {
  

•         ret = -EFAULT;
  

•         page = find_lock_page(mapping, pgoff);
  

•         if (!page)
  

•                 goto out_unacct_blocks;
  

• } else if (!*pagep) {
  

• if (!*pagep) {
          page = shmem_alloc_page(gfp, info, pgoff);
          if (!page)
                  goto out_unacct_blocks;
  

@@ -2415,27 +2413,21 @@ int shmem_mcopy_atomic_pte(struct mm_struct *dst_mm, pmd_t *dst_pmd, *pagep = NULL; }

• if (!is_continue) {
  

•         VM_BUG_ON(PageSwapBacked(page));
  

•         VM_BUG_ON(PageLocked(page));
  

•         __SetPageLocked(page);
  

•         __SetPageSwapBacked(page);
  

•         __SetPageUptodate(page);
  

• }
  

• VM_BUG_ON(PageSwapBacked(page));
  

• VM_BUG_ON(PageLocked(page));
  

• __SetPageLocked(page);
  

• __SetPageSwapBacked(page);
  

• __SetPageUptodate(page);
  
  ret = -EFAULT;
  

• offset = linear_page_index(dst_vma, dst_addr);
  max_off = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
  

• if (unlikely(offset >= max_off))
  

• if (unlikely(pgoff >= max_off))
          goto out_release;
  
  

• /* If page wasn't already in the page cache, add it. */
  

• if (!is_continue) {
  

•         ret = shmem_add_to_page_cache(page, mapping, pgoff, NULL,
  

•                                       gfp & GFP_RECLAIM_MASK, dst_mm);
  

•         if (ret)
  

•                 goto out_release;
  

• }
  

• ret = shmem_add_to_page_cache(page, mapping, pgoff, NULL,
  

•                               gfp & GFP_RECLAIM_MASK, dst_mm);
  

• if (ret)
  

•         goto out_release;
  
  _dst_pte = mk_pte(page, dst_vma->vm_page_prot);
  if (dst_vma->vm_flags & VM_WRITE)
  

@@ -2455,22 +2447,20 @@ int shmem_mcopy_atomic_pte(struct mm_struct *dst_mm, pmd_t *dst_pmd,

  ret = -EFAULT;
  max_off = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);

• if (unlikely(offset >= max_off))
  

• if (unlikely(pgoff >= max_off))
          goto out_release_unlock;
  
  ret = -EEXIST;
  if (!pte_none(*dst_pte))
          goto out_release_unlock;
  
  

• if (!is_continue) {
  

•         lru_cache_add(page);
  

• lru_cache_add(page);
  
  

•         spin_lock_irq(&info->lock);
  

•         info->alloced++;
  

•         inode->i_blocks += BLOCKS_PER_PAGE;
  

•         shmem_recalc_inode(inode);
  

•         spin_unlock_irq(&info->lock);
  

• }
  

• spin_lock_irq(&info->lock);
  

• info->alloced++;
  

• inode->i_blocks += BLOCKS_PER_PAGE;
  

• shmem_recalc_inode(inode);
  

• spin_unlock_irq(&info->lock);
  
  inc_mm_counter(dst_mm, mm_counter_file(page));
  page_add_file_rmap(page, false);
  

diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c index cbb7c8d79a4d..286d0657fbe2 100644 --- a/mm/userfaultfd.c +++ b/mm/userfaultfd.c @@ -48,21 +48,103 @@ struct vm_area_struct *find_dst_vma(struct mm_struct *dst_mm, return dst_vma; }

+/*

• • Install PTEs, to map dst_addr (within dst_vma) to page.
• • This function handles MCOPY_ATOMIC_CONTINUE (which is always file-backed),
• • whether or not dst_vma is VM_SHARED. It also handles the more general
• • MCOPY_ATOMIC_NORMAL case, when dst_vma is *not* VM_SHARED (it may be file
• • backed, or not).
• • Note that MCOPY_ATOMIC_NORMAL for a VM_SHARED dst_vma is handled by
• • shmem_mcopy_atomic_pte instead.
• */

+static int mcopy_atomic_install_ptes(struct mm_struct *dst_mm, pmd_t *dst_pmd,

•                              struct vm_area_struct *dst_vma,
  

•                              unsigned long dst_addr, struct page *page,
  

•                              enum mcopy_atomic_mode mode, bool wp_copy)
  

+{

• int ret;
  

• pte_t _dst_pte, *dst_pte;
  

• bool is_continue = mode == MCOPY_ATOMIC_CONTINUE;
  

Nit: brackets?

...

• int writable;
  

• bool vm_shared = dst_vma->vm_flags & VM_SHARED;
  

• bool is_file_backed = dst_vma->vm_file;
  

Nit: Maybe "!vma_is_anonymous(dst_vma)" is better?

...

• spinlock_t *ptl;
  

• struct inode *inode;
  

• pgoff_t offset, max_off;
  

• _dst_pte = mk_pte(page, dst_vma->vm_page_prot);
  

• writable = dst_vma->vm_flags & VM_WRITE;
  

• /* For CONTINUE on a non-shared VMA, don't pte_mkwrite for CoW. */
  

• if (is_continue && !vm_shared)
  

Curious whether we can drop "mode" in this function.

For this one, can it be replaced with "!vma_is_anonymous() && !vm_shared"?

...

•         writable = 0;
  

• if (writable) {
  

•         _dst_pte = pte_mkdirty(_dst_pte);
  

This was unconditional previously at least for anonymous, see [1] below. I think we need to keep that behavior..

Probably you saw that shmem code has the dirty bit conditionally set, however I'm thinking maybe it's easier to always set it (I'll do it in uffd-wp shmem series anyways), then we can even drop the set_page_dirty() below, afaict.

...

•         if (wp_copy)
  

•                 _dst_pte = pte_mkuffd_wp(_dst_pte);
  

•         else
  

•                 _dst_pte = pte_mkwrite(_dst_pte);
  

• } else if (vm_shared) {
  

•         /*
  

•          * Since we didn't pte_mkdirty(), mark the page dirty or it
  

•          * could be freed from under us. We could do this
  

•          * unconditionally, but doing it only if !writable is faster.
  

•          */
  

•         set_page_dirty(page);
  

• }
  

• dst_pte = pte_offset_map_lock(dst_mm, dst_pmd, dst_addr, &ptl);
  

• if (is_file_backed) {
  

•         /* 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;
  

Frankly I don't really know why this must be put into pgtable lock.. Since if not required then it can be moved into UFFDIO_COPY path, as CONTINUE doesn't need it iiuc. Just raise it up as a pure question.

It's not clear to me either. shmem_getpage_gfp() does check this twice kinda like we're doing, but it doesn't ever touch the PTL. What it seems to be worried about is, what happens if a concurrent FALLOC_FL_PUNCH_HOLE happens somewhere in the middle of whatever manipulation we're doing? From looking at shmem_fallocate(), I think the basic point is that truncation happens while "inode_lock(inode)" is held, but neither shmem_mcopy_atomic_pte() or the new mcopy_atomic_install_ptes() take that lock.

I'm a bit hesitant to just remove it, run some tests, and then declare victory, because it seems plausible it's there to catch some semi-hard-to-induce race. I'm not sure how to prove that *isn't* needed, so my inclination is to just keep it?

I'll send a series addressing the feedback so far this afternoon, and I'll leave this alone for now - at least, it doesn't seem to hurt anything. Maybe Hugh or someone else has some more advice about it. If so, I'm happy to remove it in a follow-up.

...

• }
  

• ret = -EEXIST;
  

• if (!pte_none(*dst_pte))
  

•         goto out_unlock;
  

• inc_mm_counter(dst_mm, mm_counter(page));
  

• if (is_file_backed)
  

•         page_add_file_rmap(page, false);
  

• else
  

•         page_add_new_anon_rmap(page, dst_vma, dst_addr, false);
  

• if (!is_continue)
  

•         lru_cache_add_inactive_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:

This label seems not needed any more, we can "return 0" here and "return ret" below so as to drop the last reference.

...

• return ret;
  

+out_unlock:

• pte_unmap_unlock(dst_pte, ptl);
  

• goto out;
  

+}

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,

•                     enum mcopy_atomic_mode mode,
  

When reach this function, mode must be MCOPY_ATOMIC_NORMAL, right? Then I think we can drop it and use MCOPY_ATOMIC_NORMAL directly below. Moreover, if you read above and agree "mode" can be dropped in mcopy_atomic_install_ptes(), then it's even cleaner since we just drop all "mode" parameters.

...

                      bool wp_copy)

{

• 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;
  

@@ -99,43 +181,12 @@ static int mcopy_atomic_pte(struct mm_struct *dst_mm, if (mem_cgroup_charge(page, dst_mm, GFP_KERNEL)) goto out_release;

• _dst_pte = pte_mkdirty(mk_pte(page, dst_vma->vm_page_prot));
  

[1]

...

• if (dst_vma->vm_flags & VM_WRITE) {
  

•         if (wp_copy)
  

•                 _dst_pte = pte_mkuffd_wp(_dst_pte);
  

•         else
  

•                 _dst_pte = pte_mkwrite(_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);
  

• lru_cache_add_inactive_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;
  

• ret = mcopy_atomic_install_ptes(dst_mm, dst_pmd, dst_vma, dst_addr,
  

•                                 page, mode, wp_copy);
  

• if (ret)
  

•         goto out_release;
  

out: return ret; -out_release_uncharge_unlock:

• pte_unmap_unlock(dst_pte, ptl);
  

out_release: put_page(page); goto out; @@ -176,6 +227,38 @@ static int mfill_zeropage_pte(struct mm_struct *dst_mm, return ret; }

+static int mcontinue_atomic_pte(struct mm_struct *dst_mm,

•                         pmd_t *dst_pmd,
  

•                         struct vm_area_struct *dst_vma,
  

•                         unsigned long dst_addr,
  

•                         bool wp_copy)
  

+{

• struct inode *inode = file_inode(dst_vma->vm_file);
  

• struct address_space *mapping = inode->i_mapping;
  

• pgoff_t pgoff = linear_page_index(dst_vma, dst_addr);
  

• struct page *page;
  

• int ret;
  

• ret = -EFAULT;
  

• page = find_lock_page(mapping, pgoff);
  

• if (!page)
  

•         goto out;
  

• ret = mcopy_atomic_install_ptes(dst_mm, dst_pmd, dst_vma, dst_addr,
  

•                                 page, MCOPY_ATOMIC_CONTINUE, wp_copy);
  

• if (ret)
  

•         goto out_release;
  

• unlock_page(page);
  

• ret = 0;
  

+out:

• return ret;
  

+out_release:

• unlock_page(page);
  

• put_page(page);
  

• goto out;
  

+}

static pmd_t *mm_alloc_pmd(struct mm_struct *mm, unsigned long address) { pgd_t *pgd; @@ -418,7 +501,13 @@ static __always_inline ssize_t mfill_atomic_pte(struct mm_struct *dst_mm, enum mcopy_atomic_mode mode, bool wp_copy) {

• ssize_t err;
  

• ssize_t err = 0;
  

• if (mode == MCOPY_ATOMIC_CONTINUE) {
  

•         err = mcontinue_atomic_pte(dst_mm, dst_pmd, dst_vma, dst_addr,
  

•                                    wp_copy);
  

•         goto out;
  

Why not return directly? :)

Thanks,

...

• }
  
  /*
   * The normal page fault path for a shmem will invoke the
  

@@ -431,26 +520,20 @@ static __always_inline ssize_t mfill_atomic_pte(struct mm_struct *dst_mm, * and not in the radix tree. */ if (!(dst_vma->vm_flags & VM_SHARED)) {

•         switch (mode) {
  

•         case MCOPY_ATOMIC_NORMAL:
  

•         if (mode == MCOPY_ATOMIC_NORMAL)
                  err = mcopy_atomic_pte(dst_mm, dst_pmd, dst_vma,
                                         dst_addr, src_addr, page,
  

•                                        wp_copy);
  

•                 break;
  

•         case MCOPY_ATOMIC_ZEROPAGE:
  

•                                        mode, wp_copy);
  

•         else if (mode == MCOPY_ATOMIC_ZEROPAGE)
                  err = mfill_zeropage_pte(dst_mm, dst_pmd,
                                           dst_vma, dst_addr);
  

•                 break;
  

•         case MCOPY_ATOMIC_CONTINUE:
  

•                 err = -EINVAL;
  

•                 break;
  

•         }
  } else {
          VM_WARN_ON_ONCE(wp_copy);
          err = shmem_mcopy_atomic_pte(dst_mm, dst_pmd, dst_vma, dst_addr,
                                       src_addr, mode, page);
  }
  
  

+out: return err; }

diff --git a/tools/testing/selftests/vm/userfaultfd.c b/tools/testing/selftests/vm/userfaultfd.c index f6c86b036d0f..d8541a59dae5 100644 --- a/tools/testing/selftests/vm/userfaultfd.c +++ b/tools/testing/selftests/vm/userfaultfd.c @@ -485,6 +485,7 @@ static void wp_range(int ufd, __u64 start, __u64 len, bool wp) static void continue_range(int ufd, __u64 start, __u64 len) { struct uffdio_continue req;

• int ret;
  
  req.range.start = start;
  req.range.len = len;
  

@@ -493,6 +494,17 @@ static void continue_range(int ufd, __u64 start, __u64 len) if (ioctl(ufd, UFFDIO_CONTINUE, &req)) err("UFFDIO_CONTINUE failed for address 0x%" PRIx64, (uint64_t)start);

• /*
  

•  * Error handling within the kernel for continue is subtly different
  

•  * from copy or zeropage, so it may be a source of bugs. Trigger an
  

•  * error (-EEXIST) on purpose, to verify doing so doesn't cause a BUG.
  

•  */
  

• req.mapped = 0;
  

• ret = ioctl(ufd, UFFDIO_CONTINUE, &req);
  

• if (ret >= 0 || req.mapped != -EEXIST)
  

•         err("failed to exercise UFFDIO_CONTINUE error handling, ret=%d, mapped=%" PRId64,
  

•             ret, req.mapped);
  

}

static void *locking_thread(void *arg)

2.31.0.208.g409f899ff0-goog

-- Peter Xu