Dear Maintainer,

the kexec-tools package in trixie has lost its SysV scripts, because

Is it possible to take over at least /etc/init.d/kexec in o-s-s?

Without that script, the kexec mechanism is unusable under sysvinit, because

# kexec -l /vmlinuz --initrd=/initrd.img --reuse-cmdline
# init 6

results in a standard reboot, instead of a fast reboot the user expects.

Best regards,
	g.b.

Hi,
needs kexec package, but why then remove support for for users who
still rely on it?
Perhaps kexec maintainer can be convinced to restore the sysvinit
script?

Regards,
Lorenzo

Hi,
needs kexec package, but why then remove support for for users who
still rely on it?
Perhaps kexec maintainer can be convinced to restore the sysvinit
script?

Regards,
Lorenzo

lifting is done by /sbin/kexec.

If I understand correctly, what changed in 2023 (see #1039248) is the
addition of a proper systemd unit for a kexec reboot, instead of the
auto-generated stub which used to invoke /etc/init.d/kexec (i.e. a
wrapper around "kexec -e") at the right time in the shutdown sequence.
Therefore, /etc/init.d/kexec is now superfluous and can be dropped:
sysvinit users don't matter.

Perhaps, but similar requests for e.g. rsyslog and nftables have been
summarily discarded, so I'm a bit pessimistic.  Anyway, it might be too
late for trixie.

Regards,
	g.

lifting is done by /sbin/kexec.

If I understand correctly, what changed in 2023 (see #1039248) is the
addition of a proper systemd unit for a kexec reboot, instead of the
auto-generated stub which used to invoke /etc/init.d/kexec (i.e. a
wrapper around "kexec -e") at the right time in the shutdown sequence.
Therefore, /etc/init.d/kexec is now superfluous and can be dropped:
sysvinit users don't matter.

Perhaps, but similar requests for e.g. rsyslog and nftables have been
summarily discarded, so I'm a bit pessimistic.  Anyway, it might be too
late for trixie.

Regards,
	g.

[...]

It *is* too late for trixie but often package maintainers simply don't
realise there remains a constituency of users for the initscript and
will accept them back into the package, so I think it's worth a try!

One thing I like to do is to trim unnecessary complexity that is no
longer relevant, doing convoluted things with legacy environment
variables which is contrary to initscript best practice. A slimmer
script may be more palatable to an uncertain maintainer.

Often one can go further and write a fresh initscript using
init-d-script(5) with only a few lines - that might help change package
maintainers' preconceptions about initscripts!

Andrew

Hi,

A similar request was made to Devuan a while ago[1].

My reading is that an initscript is no longer required. Based on
the changelog entry[2], the non-systemd way of doing this is now

 kexec-load-kernel; reboot

Mark


[1]  https://bugs.devuan.org/871

[2]  https://sources.debian.org/src/kexec-tools/1%3A2.0.29-2/debian/changelog#L60

submitted #1111802

Not sure about Devuan, but definitely it is required on debian.

I tried "kexec-load-kernel; reboot" on my trixie system, and it did
not work as expected: just a standard reboot through the boot loader.

A kexec reboot requires two steps:

1. load kernel, initrd, and command line with "kexec -l": this is done
   by kexec-load-kernel (or manually)

2. in the /etc/rc6.d sequence, just after the root filesystem is unmounted,
   invoke "kexec -e" instead of the usual "reboot -f -d".

Don't know about systemd: perhaps it detects that a kernel has been loaded
(peeking at /sys/kernel/kexec_loaded) and does the right thing.

But for SysV, /etc/init.d/kexec is the only place where "kexec -e"
is invoked.  If it's missing, then after /etc/init.d/umountroot the
sequence proceeds to /etc/init.d/reboot, which just does "reboot -d -f".

Regards,
	g.b.

Thanks.

Just following up the latest (non-maintainer) comment in #1111802, even if the
maintainer of kexec-tools refuses to reinstate the LSB initscript, I am not sure
o-s-s is the right place for it. IIUC, o-s-s triggers on the systemd unit but
kexec-tools has no unit.

I wonder if bin:initscripts might be a better option.

Matthew, what do you think?

But I hope this is moot and the kexec-tools maintainer is more enlightened.

Mark

Hi,

o-s-s does require a service file to trigger on (I think any path under
/lib/systemd/system/ would do at a pinch), which kexec-tools doesn't
have. So I think initscripts would fit better if it's necessary (it's
not clear to me that an init script is still needed, and/or whether
shutdown or equivalent should do the necessary instead now).

Regards,

Matthew

Hi,

I see that #1111802 has been fixed, and it looks like kexec-tools have
init scripts again, so I'm closing this bug.

Thanks,

Matthew