Re(2): Perfect buffer/swapping in pre-patch-2.0.31-2 + a few changes

James B. MacLean (
Sun, 13 Jul 1997 10:57:21 -0300 (ADT)

Looks good here too ;),

Just to report that this appears to be the smoothest running kernel for me
also. All machines are pre-patch-2.0.31-2 + perfect-patches. My FireWall
box has not shown a 'free pages' report in it's 2 days of up time (not
unusual though). I tried a make -j on my 16meg 486DX/2 66 and watched it
use up all of my 32meg swap and spit out some:

kernel: hdb: read_intr: status=0x59 { DriveReady SeekComplete DataRequest Error }
kernel: hdb: read_intr: error=0x10 {SectorIdNotFound }, CHS=649/6/17, sector=51600

which I expect are caused from my failing drive that I should have fixed
before now. It then failed to complete this make because virtual memory
was exausted. So I set it to make -j10 which it seemed happy to run with a
load of 12 or so and I could still use/do other things.

On an 8meg 486 on a UMSDOS partition, not really stressed, it certainly
runs fine.

On a 4meg 386sx25 UMSDOS, I could not see any difference though between
this and the original 'Werner?' patches, but again, this is too

On Sat, 12 Jul 1997, Erik Andersen wrote:

> I just tried your patch, and it seems to work VERY well. I just did a
> make -j kernel compile while running netscape, and reading the linux-kernel
> mailing list. My system load got so high that xload was completely black,
> but everything worked, and mutt (my mail program) was only a bit sluggish.
> This is on an AMD K6 200, 32 Megs ram, 64 Megs swap, with the perfect buffer
> cache patch on top of pre-patch-2.0.31-2 and some IDE patches from Gadi Oxman.
> This may really be the perfect buffer cache patch. No more thrashing
> for swap when the system load goes balistic. Good work!
> -Erik
> --
> Erik B. Andersen Web:
> email:


James B. MacLean
Department of Education
Nova Scotia, Canada
B3M 4B2