On Tue, 20 Feb 2018, Ben Hutchings wrote:
> On Sun, 2018-02-18 at 22:31 +0100, Cristian Ionescu-Idbohrn wrote:
> > On Tue, 6 Feb 2018, Roger Lynn wrote:
> > > On 04/02/18 02:00, Jon DeVree wrote:
> > > > On Sun, Feb 04, 2018 at 01:03:59 +0100, Ben Hutchings wrote:
> > > > > This was unintentional, but I think it's correct. The
> > > > > Kconfig says to use rasdaemon which is already packaged and
> > > > > in stable.
> > > >
> > > > rasdaemon has a hard dependency on systemd, it isn't
> > > > installable on machines still running sysvinit.
> > >
> > > Yes it is installable. It depends on systemd, not systemd-sysv.
> > > systemd is co-installable with sysvinit-core. I have it
> > > installed on a Stretch sysvinit system (it then complains that
> > > it can't find a debugfs).
> >
> > Yes, and it should depend on blktrace (on non-systemd init) to
> > stop it complain about:
> >
> > rasdaemon: Can't find debugfs
> > rasdaemon: Can't locate a mounted debugfs
> [...]
>
> Why should blktrace be responsible for mounting debugfs? The init
> system should do that, same as for procfs and sysfs.
Right. But that's how it looks like right now. blktrace helps with
with mounting debugfs, on a non-systemd init based sysyem, AFAICT.
What's the other, sustainable, better way?
And by the way, this works better than previous (if you even reboot
the box, which I tend to avoid):
#!/bin/sh
### BEGIN INIT INFO
# Provides: rasdaemon
# Required-Start: mountdebugfs
# Required-Stop:
# Should-Start:
# Should-Stop:
# X-Start-Before:
# X-Stop-After:
# Default-Start: 2 3 4 5
# Default-Stop:
### END INIT INFO
SCRIPTNAME=/etc/init.d/rasdaemon
. /lib/lsb/init-functions
[ -x /usr/sbin/rasdaemon ] || exit 0
case $1 in
start)
log_action_begin_msg "Starting rasdaemon"
/usr/sbin/rasdaemon --record
log_action_end_msg "$?"
/usr/sbin/rasdaemon --enable
log_action_end_msg "$?"
/usr/sbin/ras-mc-ctl --register-labels
log_action_end_msg "$?"
;;
stop|restart|force-reload|status)
;;
*)
echo "Usage: $SCRIPTNAME start" >&2
exit 3
;;
esac
which can be better optimized to unduplicate paths.
--
Cristian
