androidboot.bootreason says watchdog on every boot of taimen, including clean ones -- it is not a reset-reason oracle
scope: device:google-taimen · severity: trap · confidence: proven · subsystem: boot
Do not read androidboot.bootreason out of /proc/cmdline as the reason
the phone went down. On taimen it says watchdog every time, including
after a clean reboot typed over ssh.
This is a tempting instrument precisely because there is no other one: ram-does-not-survive-a-reset-here means pstore comes back empty from every reset, so a hard reset normally leaves a journal that stops mid-line and nothing else. A cmdline field that survives – because the bootloader writes it – looks like the answer. It is a constant.
The mechanism is in the same cmdline, two fields earlier:
qcom_wdt.arm_on_probe=1 watchdog.open_timeout=300 reboot=wreboot=w puts the kernel in watchdog reboot mode: an ordinary shutdown ends
by letting the qcom watchdog bite, so the bootloader records watchdog for a
deliberate reboot exactly as it does for a real hard lockup. The field
describes how the SoC was reset, not why.
Instead: separate a crash from a reboot with things that do differ –
uptime on the far side, whether the previous boot’s journal ends with a
shutdown sequence or stops mid-line
(journalctl -b -1 -n 40), and host-side liveness during the event the way
tools/ph-hang-matrix.sh does it. For a hang you intend to catch, arm a
witness before it happens; there is no post-hoc one here.