Skip to content
Porthole organization iconPorthole

pstore has never produced a record on 7.2.2, even though ramoops registers and enables a console

scope: device:google-taimen · severity: trap · confidence: proven · subsystem: debug

Symptom — everything about pstore looks healthy, and there is never a record. On 7.2.2 the reserved region is carved, the backend registers, and the kernel even routes a console at it:

OF: reserved mem: 0x00000000b0000000..0x00000000b01fffff (2048 KiB) ramoops@b0000000
pstore: Registered ramoops as persistent store backend
ramoops: using 0x200000@0xb0000000, ecc: 0
printk: legacy console [ramoops-1] enabled

Then a real watchdog reset happens and /sys/fs/pstore is empty, with systemd’s own note for it:

Platform Persistent Storage Archival skipped, unmet condition
ConditionDirectoryNotEmpty=/sys/fs/pstore

Why it matters more than it looks — this is the instrument you reach for the moment the phone hangs where no console can follow: a resume hang, a suspend-side SoC stall, a panic before userspace. On 2026-09-20 an overnight resume hang left nothing at all to read, and the reflex “check pstore, then fix pstore” is a whole session with no output. There is no log across a reset on this kernel. Plan the investigation around that, not through it.

The older note taimen-pstore-ecc records that the ramoops region can be read out from TWRP and warns never to set ecc-size. That is about recovering the region with another OS, and it is not the same thing as pstore working from Linux – which, on 7.2.2, it does not.

Instead, for a hang with no console:

  • /sys/power/pm_test (freezer / devices / platform / processors / core) with pm_print_times=1. It runs the suspend and resume machinery and returns without entering the real sleep state, so the machine never goes down and dmesg survives. It is the only way to get a named, timed device list out of this failure class here. Its limit: nothing that requires the SoC to genuinely sleep – the IPA interconnect stall, the RPM handshake, clock gating – can reproduce under it.
  • Binary bisect by runtime toggle, taking “did it come back” as the whole oracle. Slower, but it does not need a log.
  • CONFIG_PM_TRACE_RTC, the x86 trick of hashing the last-resumed device into RTC scratch, is depends on X86. Not an option on arm64.

Do not spend the session making pstore work first.