Re: [PATCH] vmscan: do not evict inactive pages when skipping an active list scan

[Rik van Riel <riel@redhat.com>]

On 11/25/2009 03:35 PM, Johannes Weiner wrote:
> Hello all,
>
> On Wed, Nov 25, 2009 at 01:37:52PM -0500, Rik van Riel wrote:
>    
>> In AIM7 runs, recent kernels start swapping out anonymous pages
>> well before they should.  This is due to shrink_list falling
>> through to shrink_inactive_list if !inactive_anon_is_low(zone, sc),
>> when all we really wanted to do is pre-age some anonymous pages to
>> give them extra time to be referenced while on the inactive list.
>>      
> I do not quite understand what changed 'recently'.
>
> That fall-through logic to keep eating inactives when the ratio is off
> came in a year ago with the second-chance-for-anon-pages patch..?
>
>    
The confusion comes from my use of the word
"recently" here.  Larry started testing with
RHEL 5 as the baseline.

And yeah - I believe the code may well have
been wrong ever since it was merged. The
indentation just looked so pretty that noone
spotted the bug.

> Acked-by: Johannes Weiner<hannes@cmpxchg.org>
>
>    
Thank you.