Deliverable 4: Proposed boot process is available for review
pere at hungry.com
Mon Sep 11 19:37:09 UTC 2006
> I also got a reduction but just of 4 seconds: from 53 to 49. I just
> removed the scripts from rcS.d. Two seconds more than putting it on
> the background.
Right. And you removed both scripts? My test was after reordering
the boot using insserv and running the scripts in parallel. Perhaps
that made a difference. But I was using a very fast machine for my
> This still makes no difference in my machine :(. Still 53 seconds.
Yeah. But you seem to have a quick dhcp server, while the one I use
need three requests before it will answer. :/
> So I'll follow your lead and change "put hwclock in the background"
> for its removal for the deliverable 4.
Yeah. One way to disable it is by adding HWCLOCKACCESS=no in
/etc/default/rcS, but that will disable it also for shutdown, and I
suspect it is needed there to set the hardware clock based on the
current system time.
I guess a long term fix is to change util-linux to only run
hwclock*.sh if the hardware clock isn't using UTC and the platform
kernel is unable to read the hardware clock itself. And to also
change it the way Ubuntu have done it, to let udev call hwclock when
/dev/rtc become available, to make sure it isn't trying to run before
the device show up.
More information about the initscripts-ng-devel