Running `mcelog` on an AMD processor

Question

When I run mcelog (version 154), I get the following output.

mcelog: ERROR: AMD Processor family 23: mcelog does not support this processor.  Please use the edac_mce_amd module instead.
CPU is unsupported

This to me feels like a category error, because mcelog is an application and edac_mce_amd is a kernel module.

Moreover, I do have this module loaded. lsmod returns edac_mce_amd 28672 0.

So what is this error message really saying? Is there an alternative binary that I should be running to get MCEs from my AMD processor, or is there a way to make mcelog compatible?

The background for this is that I am getting sudden crashes about once a day with no incriminating messages in journalctl, so I am trying to see if there are MCEs that might be highlighting the problem.

Solution

mcelog is deprecated, you should use rasdaemon instead (also available on GitHub, and probably in your distribution of choice).

If packaged versions of rasdaemon don’t appear to work with your CPU, try building a newer version; in particular, support for your CPU family requires at least version 0.6.3 of rasdaemon. (But even older CPU families might require recent versions of rasdaemon.)