petersulyok/smfc v5.4.0

on GitHub

New

• AMD GPU support: gpu_type=amd enables temperature monitoring via rocm-smi - based on PR #112 by @GCoffland
  • New amd_temp_sensor= parameter selects the temperature sensor (0-junction, 1-edge, 2-memory, default=0)
  • New rocm_smi_path= parameter specifies the path to the rocm-smi command
  • GPU type validation added to dependency checker in service startup
  • New AMD GPU docker image (petersulyok/smfc:5.4.0-amd / latest-amd) added
  • Unit test and smoke tests updated
• Man page and documentation (README.md) updated
• Dynamic temperature generation in Smoke tests for all fan controllers
• Extended DEBUG level logging across the codebase for better internal state monitoring:
  • Fan controller: temperature smoothing details (raw vs smoothed, window fill), sensitivity check results, calculated fan level, level-unchanged confirmation, polling skipped with remaining time
  • Per-device temperatures logged in multi-device setups (min/avg/max aggregation)
  • IPMI: raw ipmitool command execution and response tracing, fan mode and fan level changes
  • Shared IPMI zone arbitration: desired levels logged on change, zone ownership map at startup
  • HD fan controller: smartctl fallback path, standby guard state
  • CONST fan controller: current vs expected fan level per zone
• Command line help text added to /etc/default/smfc configuration file
• Feature list added to README.md "How does it work?" section

Changed

• Docker: many files refactored from -gpu to -nvidia and -amd naming
• Docker: docker-build.sh and docker-push.sh updated to build and push all three image variants in a single call
• Installation script (install.sh) now preserves /etc/default/smfc when --keep-config is set and the file already exists
• Shared IPMI zone arbitration log ("Arbitration desired levels") now only fires when desired levels change, reducing log noise in steady state
• Improved docstrings with better test descriptions
• Pylint warnings corrected