On Fri, Jan 7, 2022 at 4:56 PM Adam Nielsen <a.nielsen@xxxxxxxxxxx> wrote:
Hi all,
If I have asked systemd to make use of a hardware watchdog by setting
RuntimeWatchdogSec in systemd.conf, how can I work out whether systemd
is actually using the hardware watchdog or not?
There is a non-systemd command "wdctl" which queries /dev/watchdog0 and
tells you things like what the current timeout is set to, however
on some of my machines this works, but on others it gives an error
saying /dev/watchdog0 is unavailable.
Some reading suggests only one program can access /dev/watchdog0 at a
time, so I am not sure whether this means systemd is already using the
watchdog and that's why wdctl can't access it, or does it mean the
watchdog hardware isn't working properly? On the machines where wdctl
does print details about /dev/watchdog0, does this mean systemd has not
taken ownership of it, or does that device allow multiple instances of
the watchdog?
Recent versions of util-linux will obtain the information through /sys if they're unable to open the watchdog device for whatever reason. Older versions just give up.
Generally, if systemd or something else is actively using the watchdog, then wdctl will report a "Timeleft" value lower than the total "Timeout" as it is actively counting down. (If it does report identical values, that *could* mean systemd just pinged the watchdog, so wait 1-2 seconds and check again.)
Mantas Mikulėnas
