>On a Gateway Handbook 486 running Linux 1.1.62 with "noblink" patch, I'm
>having trouble with disk power management. Specifically, I've set
>update/bdflush to a fairly long interval to supposedly cut down on disk
>access to save on batteries.
Don't do that. If anything, you want bdflush to run more often so that that
dirty pages hit the disk ASAP rather than being slowly dribbled out.
Quote:>However, when the disk shuts down, or tries to
>do so, it immediately starts up again. The behavior is as if hd.c receives
>the interrupt, classes it as "unknown but ok", but someone is telling
>update/bdflush to run because of the interrupt (which came from the apm
>bios).
Been there, done that. Check /usr/adm/messages and you'll see the problem
is that the kernel is logging the power down interrupt as an unknown disk
event. A solution is to change /etc/syslog.conf to not log those messages.
Alternatively you can get a new 1.1.6* kernel that ignores the disk spin
down interrupt.
Quote:>I'm not running the apmd daemon because it locks up the machine trying to do
>an 'outb' to what I suppose should be the apm bios.
I was never able to find a laptop where the apmd daemon would work.
I use my own APM interface code, but it's to grungy to release. The APM
spec says that you have to call the APM BIOS in real mode to establish the
protected mode interface, and thereafter call it with both 16 and 32 bit
code segments. Uhhhggg. That means patches to boot/setup.S and
init/main.c, as well as hooks in a bunch of other files. After all of this
the only thing I'm certain works correctly is the battery charge report in
/proc/apm.
--
USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
Code 930.5, Goddard Space Flight Center, Greenbelt, MD. 20771
301-286-0882 http://cesdis.gsfc.nasa.gov/pub/people/becker/whoiam.html
by