Re: MM patches against 2.5.31

From: Christian Ehrhardt (
Date: Thu Aug 22 2002 - 06:28:05 EST

On Wed, Aug 21, 2002 at 07:29:04PM -0700, Andrew Morton wrote:
> I've uploaded a rollup of pending fixes and feature work
> against 2.5.31 to
> The rolled up patch there is suitable for ongoing testing and
> development. The individual patches are in the broken-out/
> directory and should all be documented.

Sorry, but we still have the page release race in multiple places.
Look at the following (page starts with page_count == 1):

Processor 1 Processor 2
refill_inactive: lines 378-395
   as page count == 1 we'll
   continue with line 401

                                     __pagevec_release: line 138
                                       calls release_pages
                                     release_pages: line 100-111
                                       put_page_test_zero brings the
                                       page count to 0 and we'll continue
                                       at line 114. Note that this may
                                       happen while another processor holds
                                       the lru lock, i.e. there is no
                                       point in checking for page count == 0
                                       with the lru lock held because
                                       the lru lock doesn't protect against
                                       decrements of page count after
                                       the check.
  line 401: page_cache_get
  resurrects the page, page
  count is now 1.
  lines 402-448.
  line 448 calls __pagevec_release

__pagevec_release: line 138
  calls release_pages
release_pages: lines 100-111
  put_page_test_zero brings the
  page count back to 0 (!!!)
  i.e. we continue at line 114:

  lines 114-123.
  The page count == 0 check in line
  123 is successful and the page
  is returned to the buddy allocator
                                       lines 114-123.
                                       The page count == 0 check in line
                                       123 is successful, i.e. the page
                                       is returned to the buddy allocator
                                       a second time. ===> BOOM

Neither the lru lock nor any of the page count == 0 checks can
prevent this from happening.

    regards Christian

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

This archive was generated by hypermail 2b29 : Fri Aug 23 2002 - 22:00:24 EST