[lkml]   [2002]   [Aug]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: MM patches against 2.5.31
Christian Ehrhardt wrote:
> 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):

So we do. It's a hugely improbable race, so there's no huge rush
to fix it. Looks like the same race is present in -ac kernels,
actually, if add_to_swap() fails. Also perhaps 2.4 is exposed if
swap_writepage() is synchronous, and page reclaim races with
zap_page_range. ho-hum.

What I'm inclined to do there is to change __page_cache_release()
to not attempt to free the page at all. Just let it sit on the
LRU until page reclaim encounters it. With the anon-free-via-pagevec
patch, very, very, very few pages actually get their final release in
__page_cache_release() - zero on uniprocessor, I expect.

And change pagevec_release() to take the LRU lock before dropping
the refcount on the pages.

That means there will need to be two flavours of pagevec_release():
one which expects the pages to become freeable (and takes the LRU
lock in anticipation of this). And one which doesn't expect the
pages to become freeable. The latter will leave the occasional
zero-count page on the LRU, as above.

Sound sane?
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

 \ /
  Last update: 2005-03-22 13:28    [W:0.232 / U:0.496 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site