[systemd-devel] Regarding systemd-boot counting.

[Lennart Poettering]

On Mo, 09.09.24 20:15, Srinivas Naik (nivasnaik at gmail.com) wrote:

> Hello Everyone,
> I am exploring systemd-boot counting for HLOS roll back. I am new to
> systemd-boot. As part of boot counting systemd-boot will decrement the
> "tries-left" counter of the selected configuration and tries to boot with
> that.
> *For example **4.14.11-300.fc27.x86_64+2-1.conf*.
> When the device successfully boots up, systemd-bless-boot.service is
> expected to mark this entry as good.
> *For example 4.14.11-300.fc27.x86_64.conf.*
> But if for some reason, the HLOS boot up fails, is it expected to reboot
> the device manually so that the next retry happens? Or is there a service
> similar to systemd-bless-boot.service which reboots the device when the
> boot up fails?

Well, what reboots the machine on failure really depends on the
subsystem you want to cover.

Many freezes cannot be caught at all. Kernel crashes might be
something you can catch via panic=5 (or similar) on the kernel
cmdline. If you want to put a timeout on the boot process, then add
something like this to
/etc/systemd/system/multi-user.target.d/50-job-timeout.conf:

[Unit]
JobTimeoutSec=5min
JobTimeoutAction=reboot-force

And then there is the hw watchdog you can use via RuntimeWatchdogSec=
in /etc/systemd/system.conf, which can catch some more failures.

If you do all three you should have quite OK coverage.

Lennart

--
Lennart Poettering, Berlin