On 2/28/08, shyam burkule <shyam.linux@xxxxxxxxx> wrote:
On 2/28/08, Mulyadi Santosa <mulyadi.santosa@xxxxxxxxx> wrote:Hi...
On Wed, Feb 27, 2008 at 6:02 PM, shyam burkule <shyam.linux@xxxxxxxxx> wrote:
> Yes , your right. But why , this logic is only applied to fault pages ?
because in the case of hard page fault, disk access is considered
expensive and should be avoided as many as possible.
> If
> they are really really going to refer again , then later (on second
> reference ) they may get moved to active list .(assuming they are moved to
> inactive list).
at that time, and assuming VM pressure is quite high...it would be too
late (reclaimed first instead of moved into active list).
If this reasoning still can't convince you, i suggest to wait for
answers from people like Rik van riel...yes , i still not agree...... because we cant say anything about access patern...
and two more function add page directly to active list, unuse_pte and unuse_mm . These function are called when swap area is to make off.
Plase tell me, why page fault page are get added to active list. I think, there is another way to deas with this , like we SET PG_referenced and then if same page is referenced within short period , it will get added to active list.
regards,
Mulyadi.
