mm: fix locking order in mm_take_all_locks()

Dmitry Vyukov has reported[1] possible deadlock (triggered by his
syzkaller fuzzer):

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&hugetlbfs_i_mmap_rwsem_key);
                               lock(&mapping->i_mmap_rwsem);
                               lock(&hugetlbfs_i_mmap_rwsem_key);
  lock(&mapping->i_mmap_rwsem);

Both traces points to mm_take_all_locks() as a source of the problem.
It doesn't take care about ordering or hugetlbfs_i_mmap_rwsem_key (aka
mapping->i_mmap_rwsem for hugetlb mapping) vs.  i_mmap_rwsem.

huge_pmd_share() does memory allocation under hugetlbfs_i_mmap_rwsem_key
and allocator can take i_mmap_rwsem if it hit reclaim.  So we need to
take i_mmap_rwsem from all hugetlb VMAs before taking i_mmap_rwsem from
rest of VMAs.

The patch also documents locking order for hugetlbfs_i_mmap_rwsem_key.

[1] http://lkml.kernel.org/r/CACT4Y+Zu95tBs-0EvdiAKzUOsb4tczRRfCRTpLr4bg_OP9HuVg@mail.gmail.com

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Michal Hocko <mhocko@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/hugetlbfs/inode.c

@@ -708,7 +708,7 @@ static struct inode *hugetlbfs_get_root(struct super_block *sb,
 /*
  * Hugetlbfs is not reclaimable; therefore its i_mmap_rwsem will never
  * be taken from reclaim -- unlike regular filesystems. This needs an
- * annotation because huge_pmd_share() does an allocation under
+ * annotation because huge_pmd_share() does an allocation under hugetlb's
  * i_mmap_rwsem.
  */
 static struct lock_class_key hugetlbfs_i_mmap_rwsem_key;

mm/mmap.c

@@ -3184,10 +3184,16 @@ static void vm_lock_mapping(struct mm_struct *mm, struct address_space *mapping)
  * mapping->flags avoid to take the same lock twice, if more than one
  * vma in this mm is backed by the same anon_vma or address_space.
  *
- * We can take all the locks in random order because the VM code
- * taking i_mmap_rwsem or anon_vma->rwsem outside the mmap_sem never
- * takes more than one of them in a row. Secondly we're protected
- * against a concurrent mm_take_all_locks() by the mm_all_locks_mutex.
+ * We take locks in following order, accordingly to comment at beginning
+ * of mm/rmap.c:
+ *   - all hugetlbfs_i_mmap_rwsem_key locks (aka mapping->i_mmap_rwsem for
+ *     hugetlb mapping);
+ *   - all i_mmap_rwsem locks;
+ *   - all anon_vma->rwseml
+ *
+ * We can take all locks within these types randomly because the VM code
+ * doesn't nest them and we protected from parallel mm_take_all_locks() by
+ * mm_all_locks_mutex.
  *
  * mm_take_all_locks() and mm_drop_all_locks are expensive operations
  * that may have to take thousand of locks.
@@ -3206,7 +3212,16 @@ int mm_take_all_locks(struct mm_struct *mm)
 	for (vma = mm->mmap; vma; vma = vma->vm_next) {
 		if (signal_pending(current))
 			goto out_unlock;
-		if (vma->vm_file && vma->vm_file->f_mapping)
+		if (vma->vm_file && vma->vm_file->f_mapping &&
+				is_vm_hugetlb_page(vma))
+			vm_lock_mapping(mm, vma->vm_file->f_mapping);
+	}
+
+	for (vma = mm->mmap; vma; vma = vma->vm_next) {
+		if (signal_pending(current))
+			goto out_unlock;
+		if (vma->vm_file && vma->vm_file->f_mapping &&
+				!is_vm_hugetlb_page(vma))
 			vm_lock_mapping(mm, vma->vm_file->f_mapping);
 	}
 

mm/rmap.c

@@ -23,6 +23,7 @@
  * inode->i_mutex	(while writing or truncating, not reading or faulting)
  *   mm->mmap_sem
  *     page->flags PG_locked (lock_page)
+ *       hugetlbfs_i_mmap_rwsem_key (in huge_pmd_share)
  *         mapping->i_mmap_rwsem
  *           anon_vma->rwsem
  *             mm->page_table_lock or pte_lock