On Thu, 13 Jan 2000, Kanoj Sarcar wrote:
[snip Linus' winning idea]
> Yes, that's what everyone seems to be pointing at. As I mentioned, I am
> looking into this as I type. The only thing is, as Andrea points out,
> 2.3 bh/irq handlers do not request HIGHMEM pages, so shouldn't the
> 2.3 kswapd do something more like:
>
> more_work = 0;
> for (i = 0; i < MAX_NR_ZONES; i++) {
> if (i != ZONE_HIGHMEM)
> more_work |= balance_zone(zone+i)
> }
> if (!more_work)
> sleep()
Nope. We want to do page aging and reclamation in ZONE_HIGHMEM
too, otherwise all memory `rotation' is going to happen in the
other zones and the system can thrash in the remaining 1G of
memory while there's 3G of unused data in ZONE_HIGHMEM...
But I agree, we probably don't have to reclaim that many pages
in ZONE_HIGHMEM, something like freepages.min should be enough.
regards,
Rik
-- The Internet is not a network of computers. It is a network of people. That is its real strength.- 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 Jan 15 2000 - 21:00:25 EST