On Mon, 2 Oct 2000, Daniel Phillips wrote:
> Rik van Riel wrote:
> >
> > Local mechanisms simply CANNOT make page replacement work
> > well on a system-wide level.
>
> I think you mean 'local mechanisms alone'. The question is not
> *whether* the subsystems will work together, but *how*. I have
> a nagging feeling we can do a little better than the mechanism
> currently on the table.
That's up to the filesystems ;)
If the VM does a page->mapping->flush() callback and the
filesystem flushes (on average) 20 pages as a result of
such a callback, we'll only have to do the callback for
5% of the pages ...
Also, by speeding up / slowing down bdflush (kflushd) as
appropriate, you can avoid having to do the flush callback
at all most of the time. (for "normal" access patterns
and under "normal" loads)
The ->flush() callback is mostly there for those situations
in which we have a big discrepancy between the filesystem
writeout order and the VM page reclaim order...
regards,
Rik
--
"What you're running that piece of shit Gnome?!?!"
-- Miguel de Icaza, UKUUG 2000
http://www.conectiva.com/ http://www.surriel.com/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Oct 07 2000 - 21:00:09 EST