Re: classzone-VM + mapped pages out of lru_cache

From: Craig Kulesa (
Date: Thu May 04 2000 - 04:54:14 EST

In light of my previous message about Andrea's patch against pre7-3,
here's a few comments about pre7-4:

It would appear that kswapd isn't eating up the CPU as badly as pre6, but
interactive response under I/O load sure hasn't gotten any better for me.

A token example might be the following on a 64 MB laptop:

'dd if=/dev/zero of=dummy.dat bs=1024k count=200'

In pre5, this moves along smoothly, with no obvious impact upon the
user. With Andrea's classzone patch, the load is slightly higher
(probably within measurement error), with a little mouse hesitation. But
in pre7-4, the X session and mouse freeze for many seconds at a time (it
can't even update vmstat). System load goes to 2-3, versus less than 1 in
pre5 or with classzone-18. kswapd doesn't claim responsibility; it claims
perhaps 1 second of CPU time.

That system is running WindowMaker/X11, and an xterm, pretty much nothing
else. I haven't been able to get that particular machine to lockup with
classzone-18, but there's no mail spool file there either... Hmmm.

The same general trends in behavior are seen on two other boxes, a 32 MB
PCI-less 486 and a 128 MB PIII box.


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

This archive was generated by hypermail 2b29 : Sun May 07 2000 - 21:00:14 EST