[PATCH 01/23] shmem/userfaultfd: Take care of UFFDIO_COPY_MODE_WP

[Peter Xu <peterx@redhat.com>]

Firstly, pass wp_copy into shmem_mfill_atomic_pte() through the stack.
Then apply the UFFD_WP bit properly when the UFFDIO_COPY on shmem is with
UFFDIO_COPY_MODE_WP.

One thing to mention is that shmem_mfill_atomic_pte() needs to set the dirty
bit in pte even if UFFDIO_COPY_MODE_WP is set.  The reason is similar to
dcf7fe9d8976 ("userfaultfd: shmem: UFFDIO_COPY: set the page dirty if VM_WRITE
is not set") where we need to set page as dirty even if VM_WRITE is no there.
It's just that shmem can drop the pte any time later, and if it's not dirty the
data will be dropped.  For uffd-wp, that could lead to data loss if without the
dirty bit set.

Note that shmem_mfill_zeropage_pte() will always call shmem_mfill_atomic_pte()
with wp_copy==false because UFFDIO_ZEROCOPY does not support UFFDIO_COPY_MODE_WP.

Signed-off-by: Peter Xu <peterx@redhat.com>
---
 include/linux/shmem_fs.h |  5 +++--
 mm/shmem.c               | 18 ++++++++++++++----
 mm/userfaultfd.c         |  2 +-
 3 files changed, 18 insertions(+), 7 deletions(-)

diff --git a/include/linux/shmem_fs.h b/include/linux/shmem_fs.h
index f0919c3722e7..dfd0369657d8 100644
--- a/include/linux/shmem_fs.h
+++ b/include/linux/shmem_fs.h
@@ -128,10 +128,11 @@ extern void shmem_uncharge(struct inode *inode, long pages);
 int shmem_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,
-			   enum mcopy_atomic_mode mode, struct page **pagep);
+			   enum mcopy_atomic_mode mode, struct page **pagep,
+			   bool wp_copy);
 #else /* !CONFIG_SHMEM */
 #define shmem_mcopy_atomic_pte(dst_mm, dst_pmd, dst_vma, dst_addr, \
-			       src_addr, mode, pagep)        ({ BUG(); 0; })
+			       src_addr, mode, pagep, wp_copy) ({ BUG(); 0; })
 #endif /* CONFIG_SHMEM */
 #endif /* CONFIG_USERFAULTFD */
 
diff --git a/mm/shmem.c b/mm/shmem.c
index 5cfd2fb6e52b..e88aaabaeb27 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -2364,7 +2364,8 @@ static struct inode *shmem_get_inode(struct super_block *sb, const struct inode
 int shmem_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,
-			   enum mcopy_atomic_mode mode, struct page **pagep)
+			   enum mcopy_atomic_mode mode, struct page **pagep,
+			   bool wp_copy)
 {
 	bool is_continue = (mode == MCOPY_ATOMIC_CONTINUE);
 	struct inode *inode = file_inode(dst_vma->vm_file);
@@ -2438,9 +2439,18 @@ int shmem_mcopy_atomic_pte(struct mm_struct *dst_mm, pmd_t *dst_pmd,
 	}
 
 	_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));
-	else {
+	if (dst_vma->vm_flags & VM_WRITE) {
+		if (wp_copy)
+			_dst_pte = pte_mkuffd_wp(pte_wrprotect(_dst_pte));
+		else
+			_dst_pte = pte_mkwrite(_dst_pte);
+		/*
+		 * Similar reason to set_page_dirty(), that we need to mark the
+		 * pte dirty even if wp_copy==true here, otherwise the pte and
+		 * its page could be dropped at anytime when e.g. swapped out.
+		 */
+		_dst_pte = pte_mkdirty(_dst_pte);
+	} else {
 		/*
 		 * We don't set the pte dirty if the vma has no
 		 * VM_WRITE permission, so mark the page dirty or it
diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index cbb7c8d79a4d..0963e0d9ed20 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -448,7 +448,7 @@ static __always_inline ssize_t mfill_atomic_pte(struct mm_struct *dst_mm,
 	} else {
 		VM_WARN_ON_ONCE(wp_copy);
 		err = shmem_mcopy_atomic_pte(dst_mm, dst_pmd, dst_vma, dst_addr,
-					     src_addr, mode, page);
+					     src_addr, mode, page, wp_copy);
 	}
 
 	return err;
-- 
2.26.2