Parallelizing bootscripts [was: Make bootscripts more POSIX compliant]
ken at linuxfromscratch.org
Tue Feb 20 11:44:26 PST 2007
On Tue, Feb 20, 2007 at 12:46:04PM -0500, Bryan Kadzban wrote:
> On the topic of parallelizing the bootscripts, what do people think
> about doing this? DJ has added some easily-parallelizable scripts to
> the contrib/ directory in the bootscripts repo (basically, by making
> them LSB compliant, you make them easy to run in parallel). Should we
> look into making these scripts the default, perhaps for LFS 6.4 or 7?
> (And should we actually run them in parallel or not?)
If it causes no damage, and people think it's worth the time to
test it, yes to running in parallel. I'd better clarify that -
earlier this month I noted that the total time from power-on to a
login prompt on my desktops is dominated by the time to a boot
prompt, the time to get a dhcp lease, time for ntp to start up, and
on one by time for X to start.
For me, saving a couple of seconds in the bootscripts is neither
here nor there. If X can start while ntp is deciding whether or not
to get out of bed, that would be nice - but if ntp decides to call
in sick, I'd like to get the report.
As to testing, I'll mention that the via C7 I'm playing with for
possible lower-power (hah, 1 Watt less than my athlon64 when that is
at 1GHz with CnQ) seems to have an interesting race with the
bootscripts from December - when the console comes up with the
LatArCyrHeb font, one of the earlier messages gets rendered as if it
were mostly in cyrillic characters. I've only seen it on that box,
it's mostly harmless, and it's such a slow dog that I'm not motivated
to debug it ;) My point is that changing how the bootscripts are run
will need a *lot* of testing across different machines and
combinations of bootscripts.
das eine Mal als Tragödie, das andere Mal als Farce
More information about the lfs-dev