On Fri, 14 Jul 2000, Andrea Arcangeli wrote:
> On Thu, 13 Jul 2000, Alan Cox wrote:
>
> >6 weeks to fix the VM is too short. But then I think we may have to accept
> >that we dont get the VM perfect until 2.5. We just need to get it back working
> >as well as 2.0 did
>
> IMHO the fact the memory balancing doesn't know about the
> _class_zone where the user want to allocate memory is one major
> problem that we have in 2.4.x.
*sigh* I don't think I will go over this again since you don't
seem to read your email anyway...
> Without that chance Juan's kernel will keep trying to take some
> ISA DMA memory free even if he doesn't use floppy or ISA
> soundcard and it won't provide LRU behaviour within the
> classzone.
"You're not using your soundcard right now. No reason to keep
ISA DMA memory free now."
Now imagine what happens when Juan wants to play some music
and the sound driver wants to allocate a DMA buffer...
(also, the "classzone doesn't try to free DMA memory" contradicts
an earlier email by you)
> What I mean is that if shrink_mmap doesn't get a zone_t pointer
> as parameter, the VM will always end doing something wrong.
> There's no way to fix that except reinserting the parameter that
> gives the information about the allocation as I did in
> classzone.
That _you_ can't see any other way doesn't mean there is
no better way to do these things. The solution is to have
a scavenge and inactive list.
Basically this means that:
1) we don't throw away the data of pages we want to free,
meaning that we can "free" any page we want ... because
the data is remembered there's no harm in doing that
2) we can see which zone has the biggest percentage of old
pages, so we can do allocations from that zone
3) we don't mess up page aging because we skip writing out
dirty pages, etc...
> I'll be glad to extract such part of classzone as a clean patch
> against latest Linus's tree if somebody is interested.
Don't think so. Stephen Tweedie, Juan Quintela, Alan Cox and
me have been thinking long and hard about a new VM design that
will work better than any modification to the current system.
We'll talk about it next week at OLS.
(the basic ideas are like the VM design I posted to linux-mm
about 5 weeks ago, we have just changed some of the details)
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.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Jul 15 2000 - 21:00:21 EST