mm, dax: fix DAX deadlocks (COW fault)
authorYigal Korman <yigal@plexistor.com>
Mon, 16 Nov 2015 12:09:15 +0000 (14:09 +0200)
committerDan Williams <dan.j.williams@intel.com>
Thu, 19 Nov 2015 00:54:36 +0000 (16:54 -0800)
DAX handling of COW faults has wrong locking sequence:
dax_fault does i_mmap_lock_read
do_cow_fault does i_mmap_unlock_write

Ross's commit[1] missed a fix[2] that Kirill added to Matthew's
commit[3].

Original COW locking logic was introduced by Matthew here[4].

This should be applied to v4.3 as well.

[1] 0f90cc6609c7 mm, dax: fix DAX deadlocks
[2] 52a2b53ffde6 mm, dax: use i_mmap_unlock_write() in do_cow_fault()
[3] 843172978bb9 dax: fix race between simultaneous faults
[4] 2e4cdab0584f mm: allow page fault handlers to perform the COW

Cc: <stable@vger.kernel.org>
Cc: Boaz Harrosh <boaz@plexistor.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Jan Kara <jack@suse.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Acked-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Yigal Korman <yigal@plexistor.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
mm/memory.c

index deb679c..c387430 100644 (file)
@@ -3015,9 +3015,9 @@ static int do_cow_fault(struct mm_struct *mm, struct vm_area_struct *vma,
                } else {
                        /*
                         * The fault handler has no page to lock, so it holds
-                        * i_mmap_lock for write to protect against truncate.
+                        * i_mmap_lock for read to protect against truncate.
                         */
-                       i_mmap_unlock_write(vma->vm_file->f_mapping);
+                       i_mmap_unlock_read(vma->vm_file->f_mapping);
                }
                goto uncharge_out;
        }
@@ -3031,9 +3031,9 @@ static int do_cow_fault(struct mm_struct *mm, struct vm_area_struct *vma,
        } else {
                /*
                 * The fault handler has no page to lock, so it holds
-                * i_mmap_lock for write to protect against truncate.
+                * i_mmap_lock for read to protect against truncate.
                 */
-               i_mmap_unlock_write(vma->vm_file->f_mapping);
+               i_mmap_unlock_read(vma->vm_file->f_mapping);
        }
        return ret;
 uncharge_out: