On Mon, 2003-09-22 at 15:47, Karol Kozimor wrote:
> Thus wrote Steve Lord:
> > > Please do search for laptop-mode patches published on lkml several times.
> > > They have a sysctl setting that, when turned on, dumps the processes or
> > > threads doing I/O on screen. Using that patch, I was able to successfully
> > > spin my disk down to about 1 minute (the maximum sysctl setting for
> > > xfssyncd or the other).
> > Using a current CVS kernel, and redhat 9.0 user space, I can get
> > my laptop to sit there with a suspended ide drive for 10 minutes
> > before something pops off and writes to a log file. I mounted all
> > filesystems with noatime and stopped crond.
> When doing no work on the machine itself? Right, if I leave mine alone, the
> drive is spun down for several minutes (never actually bothered to count
> them) -- not that it saves much power, though. However, I believe we would
> want to actually let people do some work, and as far as I understand it
> now, whenever the pages get dirty or whatever write gets pending, the disk
> is powered on after a minute or so. Is the there a specific reason for
> which the sysctls max out at 60 seconds?
> Best regards,
Ahh, I went and looked at bdflush's maximums, and yes, for folks who
want to live in ram, we can tweak a few maximums.
Steve Lord voice: +1-651-683-3511
Principal Engineer, Filesystem Software email: lord@xxxxxxx