diff options
| author | Patrick NC Lin <patrick.nc.lin.wiwynn@gmail.com> | 2025-06-25 06:05:50 +0300 |
|---|---|---|
| committer | Patrick Williams <patrick@stwcx.xyz> | 2025-07-02 21:33:42 +0300 |
| commit | 12d8a2fe3bc1924fe714c18097910b08c5bd3af2 (patch) | |
| tree | 86c405c01c73752d8521ba622afe4238b417dd43 /meta-openembedded/meta-python/recipes-devtools/python/python3-dnspython | |
| parent | e030b774f8eab385f8e9f03ba6308d6e27233326 (diff) | |
| download | openbmc-12d8a2fe3bc1924fe714c18097910b08c5bd3af2.tar.xz | |
meta-facebook: yosemite4: Add fan missing detection mechanism
Summary:
Introduce two-stage fan missing handling to protect the system from
thermal risks due to insufficient cooling.
Description:
This commit adds mechanisms to detect and respond to fan rotor missing
events, covering both pre-boot and runtime conditions.
Case 1: Fan board missing before power-on
If the fan board is missing before power-on, its FRU will not be
detected. A new `monitor-fan-sensor.service` is introduced to run after
`multi-user.target` is reached. It checks the number of functional fan
rotors. If < 22 rotors are detected, it restarts the fan sensor once
and checks again. If there are still more than 2 missing rotors, an
event log is recorded and 8 blades are shut down via AC off. If the
rotor count recovers, no shutdown is triggered.
Case 2: Fan rotor missing at runtime
If fan rotors become missing during runtime, its DBus property will
show `"Value" = NaN`, `"Available" = true`, and `"Functional" = false`.
A new D-Bus monitor file named `fan-rotor-missing-poweroff.yaml`
is added to detect this condition. If more than 2 rotors are missing,
the system restarts the fan sensor and checks again. If more than 2
rotors are still missing, it logs an event and shuts down 8 blades via
AC off. If rotor count recovers, no shutdown occurs.
Additional fix:
While testing Case 1, it was found that if the system fails to detect
a fanboard FRU during boot-up, it hangs at `yosemite4-schematic-init`.
This blocks other services from proceeding. To resolve this, a timeout
is added when reading the fanboard FRU. If still missing after timeout,
the FSC source falls back to `"Unknown"`. This ensures the Watchdog
timeout uses a default value, allowing system boot to proceed.
Motivation:
Requested by the Thermal team. The goal is to ensure the system has
protection to avoid thermal damage when fans are missing.
Test Plan:
- Build and flash the firmware.
- Simulate power-on with fan board removed and confirm the mechanism
logs an event and shuts down 8 blades if more than 2 rotors remain
missing.
- Simulate fan rotor missing at runtime and confirm the system retries
once, logs event, and shuts down 8 blades if missing count persists.
Change-Id: I6e4c6da932660804efca5ec1b7b09d9380ea6d34
Signed-off-by: Patrick NC Lin <patrick.nc.lin.wiwynn@gmail.com>
Diffstat (limited to 'meta-openembedded/meta-python/recipes-devtools/python/python3-dnspython')
0 files changed, 0 insertions, 0 deletions