FS#30380 - [kexec-tools] Implement wether we want to reboot with kexec with a configuration option
Attached to Project:
Arch Linux
Opened by Olivier (olive) - Wednesday, 20 June 2012, 15:21 GMT
Last edited by Tobias Powalowski (tpowa) - Thursday, 28 June 2012, 09:38 GMT
Opened by Olivier (olive) - Wednesday, 20 June 2012, 15:21 GMT
Last edited by Tobias Powalowski (tpowa) - Thursday, 28 June 2012, 09:38 GMT
|
Details
Description:
The way the system decides whether kexec should be used for reboot seems odd. We enable a "service" at boot time and kexec is used if this service is "running". That seems odd because there are in reality no process running. Moreover how this feature is implementing is very confusing with /etc/rc.d/kexec supporting unstandard option load/unload and the stop option not doing at all the same thing in runlevel 6 (loading the kernel to make the kexec) or the normal runlevel (not using kexec on the next reboot). This reflects that there are in reality no services to stop. I think it would make much more sense to enable this via a configuration option. I propose the following change: 1) Add an option USE_KEXEC=yes/no in /etc/conf.d/kexec 2) In case USE_KEXEC is yes; Load the initrd and the kernel somewhere in /etc/rc.shutdown (before the pivot_root) and perform the reboot via kexec at the end of the script. Probably add sanity checks and enable a normal BIOS reboot if anything fails. Package version kexec-tools 2.0.2-5 |
This task depends upon
Closed by Tobias Powalowski (tpowa)
Thursday, 28 June 2012, 09:38 GMT
Reason for closing: Won't implement
Thursday, 28 June 2012, 09:38 GMT
Reason for closing: Won't implement
Comment by
Dave Reisner (falconindy) -
Saturday, 23 June 2012, 01:05 GMT
Comment by Tom Gundersen (tomegun) -
Saturday, 23 June 2012, 01:34 GMT
Comment by
Dave Reisner (falconindy) -
Saturday, 23 June 2012, 01:35 GMT
Comment by
Tobias Powalowski (tpowa) - Saturday,
23 June 2012, 07:59 GMT
Sure, would you like to see alsa converted to USE_ALSA=y as well,
since it isn't really a daemon? i don't really understand the
point of this. The end result is the same. The logic to determine
whether to kexec or reboot is already in place. It's just a matter
of ensuring _something_ loads the jump kernel if desired. The
/etc/rc.d script is that something.
Seems a bit odd to make kexec a static configuration (either as a
variable somewhere or as a DAEMON). Wouldn't the admin just load
the kernel when s/he wants to reboot into a new kernel?
Most distros automate this somehow, and to some degree. To point
out the other extreme, ubuntu will always use kexec if its
installed.
Systemd has also a daemon for it.