[RFC -mm] mm, userfaultfd, THP: Avoid waiting when PMD under THP migration

[Huang, Ying]

From: Huang Ying <ying.huang@xxxxxxxxx>

If THP migration is enabled, the following situation is possible,

- A THP is mapped at source address
- Migration is started to move the THP to another node
- Page fault occurs
- The PMD (migration entry) is copied to the destination address in mremap

That is, it is possible for handle_userfault() encounter a PMD entry
which has been handled but !pmd_present().  In the current
implementation, we will wait for such PMD entries, which may cause
unnecessary waiting, and potential soft lockup.

This is fixed via avoiding to wait when !pmd_present(), only wait when
pmd_none().

Question:

I found userfaultfd_must_wait() is always called when PMD or PTE is
none, and with mm->mmap_sem read-lock held.  mremap() will write-lock
mm->mmap_sem.  And UFFDIO_COPY don't support to copy THP mapping.  So
the situation described above couldn't happen in practice?

Signed-off-by: "Huang, Ying" <ying.huang@xxxxxxxxx>
Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
Cc: Mike Kravetz <mike.kravetz@xxxxxxxxxx>
Cc: Mike Rapoport <rppt@xxxxxxxxxxxxxxxxxx>
Cc: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>
Cc: Alexander Viro <viro@xxxxxxxxxxxxxxxxxx>
Cc: Zi Yan <zi.yan@xxxxxxxxxxxxxx>
Cc: Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx>
---
 fs/userfaultfd.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
index b5a0193e1960..0fcf66c3e439 100644
--- a/fs/userfaultfd.c
+++ b/fs/userfaultfd.c
@@ -294,10 +294,13 @@ static inline bool userfaultfd_must_wait(struct userfaultfd_ctx *ctx,
 	 * pmd_trans_unstable) of the pmd.
 	 */
 	_pmd = READ_ONCE(*pmd);
-	if (!pmd_present(_pmd))
+	if (pmd_none(_pmd))
 		goto out;
 
 	ret = false;
+	if (!pmd_present(_pmd))
+		goto out;
+
 	if (pmd_trans_huge(_pmd))
 		goto out;
 
-- 
2.14.2