Description:
The base of my devbox was built around Supermicro devices, one of them - X8DTH-6F - motherboard with w83795 hardware monitoring chip. A couple of weeks ago I started to load the hardware with heavy tasks and that is why I decided to add FANs speed to monitoring (in addition to RAID status and CPUs temperature).
I executed 'sensors-detect' as usual, added necessary metrics to the monitoring and continued my work. In ~12 hours I received a message from the monitoring - the server was unavailable. What were in logs:
May 5 00:22:07 okd-master2 kernel: [39313.223132] i801_smbus 0000:00:1f.3: SMBus is busy, can't use it!
May 5 00:22:07 okd-master2 kernel: [39313.223799] w83795 10-002f: Failed to write to register 0x040, err -16
<= freeze
After some tests I put w83795, w83627ehf, hwmon_vid kernel modules to the blacklist and after that my server works as expected (i.e. without problems).
BTW: Intrusion sensor changed its value every several seconds without any interventions :)