Re: [PATCH] kdump-lib-initramfs.sh: use -f option for failure_action halt

[Kairui Song]

On Wed, Oct 30, 2019 at 9:02 AM d.hatayama(a)fujitsu.com
<d.hatayama(a)fujitsu.com&gt; wrote:
>  > -----Original Message-----
>  > From: Kairui Song [mailto:[e-mail]]
>  > Sent: Monday, October 28, 2019 5:29 PM
>  > To: Hatayama, Daisuke/畑山 大輔 <d.hatayama(a)fujitsu.com&gt;
>  > Subject: Re: [PATCH] kdump-lib-initramfs.sh: use -f option for failure_action
>  > halt
>  >
>  > On Mon, Oct 28, 2019 at 11:29 AM d.hatayama(a)fujitsu.com
>  > <d.hatayama(a)fujitsu.com&gt; wrote:
>  > >
>  > > failure_action halt doesn't work well now. This appears to be similar
>  > > to the previous issues such as
>  > > b2534348199ad5010c86890a068653d970c9933d and
>  > > 89565289c64aa5323278cf5722c725a58b369af8.
>  >
>  > Hi,
>  >
>  > Thanks for the patch, it looks good to me in general, but can you add
>  > more detail about the problem you are trying to solve?
>  > Eg, add some word about the system behavior or some error logs. Like
>  > in the two commits you mentioned.
>
>  I see. I'll reflect your comment in the next version.
>
>  I would write that there is a message indicating halt command is executed
>  but actually system is reboot, not halted together with log message.
>
>  Also I a little more looked into the behavior using debug message
>  with systemd.log-leve=debug. It looks that systemd-halt.service
>  is once pulled-in but it is somehow not executed and finally system
>  is rebooted.
>
>  systemd-udevd-control.socket: Changed listening -> dead
>  systemd-udevd-control.socket: Job 100 systemd-udevd-control.socket/stop finished,
> result=done
>
>  [^[[0;32m  OK  ^[[0m] Closed ^[[0;1;39mudev Control Socket^[[0m.
>  Kdump: Executing failure action halt
>
>    ^-- failure_action halt is executed
>
>  Bus private-bus-connection: changing state UNSET → OPENING
>  Bus private-bus-connection: changing state OPENING → AUTHENTICATING
>  Accepted new private connection.
>  Bus private-bus-connection: changing state AUTHENTICATING → RUNNING
>  Got message type=method_call sender=n/a destination=org.freedesktop.systemd1
> path=/org/freedesktop/systemd1 interface=org.freedesktop.systemd1.Manager member=StartUnit
> cookie=1 reply_cookie=0 signature=ss error-name=n/a error-message=n/\
>  a
>  halt.target: Trying to enqueue job halt.target/start/replace-irreversibly
>
>     ^-- halt.target gets started being pulled-in here
>
>  Added job halt.target/start to transaction.
>  Pulling in systemd-halt.service/start from halt.target/start
>  Added job systemd-halt.service/start to transaction.
>  Pulling in shutdown.target/start from systemd-halt.service/start
>  Added job shutdown.target/start to transaction.
>  Pulling in dracut-cmdline.service/stop from shutdown.target/start
>  Added job dracut-cmdline.service/stop to transaction.
>  Pulling in cryptsetup.target/stop from shutdown.target/start
>  Added job cryptsetup.target/stop to transaction.
>  ...<snip>...
>  squash-root.mount: squash-root.mount lost dependency After=dev-loop0.device
>  squash-root.mount: squash-root.mount lost dependency References=dev-loop0.device
>  squash-root.mount: Collecting.
>  Child 400 (umount) died (code=exited, status=0/SUCCESS)
>  etc.mount: Child 400 belongs to etc.mount.
>  etc.mount: Mount process exited, code=exited, status=0/SUCCESS
>  etc.mount: Succeeded.
>  etc.mount: Changed unmounting -> dead
>  etc.mount: Job 181 etc.mount/stop finished, result=done
>  [^[[0;32m  OK  ^[[0m] Unmounted ^[[0;1;39m/etc^[[0m.
>  etc.mount: Collecting.
>  Received SIGCHLD from PID 400 (n/a).
>  inotify event for /etc/localtime
>  Failed to stat /etc/localtime, ignoring: No such file or directory
>  /etc/localtime doesn't exist yet, watching /etc instead.
>  Timezone has been changed (now: UTC).
>  Bus private-bus-connection: changing state AUTHENTICATING → RUNNING
>  Got message type=method_call sender=n/a destination=org.freedesktop.systemd1
> path=/org/freedesktop/systemd1 interface=org.freedesktop.systemd1.Manager member=Reboot
> cookie=1 reply_cookie=0 signature=n/a error-name=n/a error-message=n/aO\
>  perating on architecture: x86-64
>
>  Sent message type=method_return sender=org.freedesktop.systemd1 destination=n/a path=n/a
> interface=n/a member=n/a cookie=1 reply_cookie=1 signature=n/a error-name=n/a
> error-message=n/a
>  Shutting down.
>  Bus private-bus-connection: changing state RUNNING → CLOSING
>  dev-vda.device: Failed to send unit remove signal for dev-vda.device: Connection reset by
> peer
>  systemd-journald.service: Executing: /usr/lib/systemd/systemd-journald
>  Bus private-bus-connection: changing state CLOSING → CLOSED
>  [   15.956995] Unregister pv shared memory for cpu 0
>  [   15.967559] reboot: Restarting system
>  [   15.970216] reboot: machine restart
>
>    ^---- but finally system gets restarted.
>
>  This behavior doesn't change even if I changed ExecStart=/usr/bin/systemctl --force
> halt
>  in systemd-halt.service into ExecStart=/usr/bin/systemctl --force --force halt.
>  systemctl halt with --force twice means systemctl calls immediately directly reboot
>  system call. Thus, I think systemd somehow reboot system before entering
> systemd-halt.service.
>  
Thanks, I remember Pingfan have encountered similar problem before,
that force shutdown turns into reboot and that seems to be a systemd
bug. CC Pingfan see if he has any idea about this.

But it's very strange that systemctl --force --force halt doesn't work
either. I can reproduce this locally with Fedora.

Also add the missing kexec(a)lists.fedoraproject.org in CC list.


--
Best Regards,
Kairui Song