mm/hugetlb: fix uffd wr-protection for CoW optimization path

commit 60d5b473d61be61ac315e544fcd6a8234a79500e upstream.

This patch fixes an issue that a hugetlb uffd-wr-protected mapping can be
writable even with uffd-wp bit set.  It only happens with hugetlb private
mappings, when someone firstly wr-protects a missing pte (which will
install a pte marker), then a write to the same page without any prior
access to the page.

Userfaultfd-wp trap for hugetlb was implemented in hugetlb_fault() before
reaching hugetlb_wp() to avoid taking more locks that userfault won't
need.  However there's one CoW optimization path that can trigger
hugetlb_wp() inside hugetlb_no_page(), which will bypass the trap.

This patch skips hugetlb_wp() for CoW and retries the fault if uffd-wp bit
is detected.  The new path will only trigger in the CoW optimization path
because generic hugetlb_fault() (e.g.  when a present pte was
wr-protected) will resolve the uffd-wp bit already.  Also make sure
anonymous UNSHARE won't be affected and can still be resolved, IOW only
skip CoW not CoR.

This patch will be needed for v5.19+ hence copy stable.

[peterx@redhat.com: v2]
Link: https://lkml.kernel.org/r/ZBzOqwF2wrHgBVZb@x1n
[peterx@redhat.com: v3]
Link: https://lkml.kernel.org/r/20230324142620.2344140-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20230321191840.1897940-1-peterx@redhat.com
Fixes: 166f3ecc0daf ("mm/hugetlb: hook page faults for uffd write protection")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reported-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Tested-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Acked-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 52475c4262e45ef8ea06b27873cb7befc6092925..d3ffa0fd49e57b479f9896f7752fb90fed735eed 100644 (file)
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -5469,7 +5469,7 @@ static vm_fault_t hugetlb_wp(struct mm_struct *mm, struct vm_area_struct *vma,
                       struct page *pagecache_page, spinlock_t *ptl)
 {
        const bool unshare = flags & FAULT_FLAG_UNSHARE;
-       pte_t pte;
+       pte_t pte = huge_ptep_get(ptep);
        struct hstate *h = hstate_vma(vma);
        struct page *old_page, *new_page;
        int outside_reserve = 0;
@@ -5480,6 +5480,17 @@ static vm_fault_t hugetlb_wp(struct mm_struct *mm, struct vm_area_struct *vma,
        VM_BUG_ON(unshare && (flags & FOLL_WRITE));
        VM_BUG_ON(!unshare && !(flags & FOLL_WRITE));
 
+       /*
+        * Never handle CoW for uffd-wp protected pages.  It should be only
+        * handled when the uffd-wp protection is removed.
+        *
+        * Note that only the CoW optimization path (in hugetlb_no_page())
+        * can trigger this, because hugetlb_fault() will always resolve
+        * uffd-wp bit first.
+        */
+       if (!unshare && huge_pte_uffd_wp(pte))
+               return 0;
+
        /*
         * hugetlb does not support FOLL_FORCE-style write faults that keep the
         * PTE mapped R/O such as maybe_mkwrite() would do.
@@ -5495,7 +5506,6 @@ static vm_fault_t hugetlb_wp(struct mm_struct *mm, struct vm_area_struct *vma,
                return 0;
        }
 
-       pte = huge_ptep_get(ptep);
        old_page = pte_page(pte);
 
        delayacct_wpcopy_start();