[OpenPower-Firmware] Error logging on POWER9

[Stewart Smith]

Hi all,

On the OpenPOWER Foundation System Software workgroup this morning, we
started a discussion on error logging for POWER9 and how we may want to
differ from how things are done on POWER8.

In the call, we covered a bit of history of how we got to now (tl;dr: we
had to ship, so we borrowed the data structures and the like from FSP
land), what Google did with P8 (tl;dr: replaced a bunch of code and
unified things into one format) and how we may want to approach thinking
of what to do for POWER9.

The result of that was that we should start at what we want the end user
experience to be and work backwards into data formats (plain text string
is a data format too).

I also said I'd start a discussion here.... so here we go.

My 2c:
- the specification for whatever formats we decide should be open
- It should be human readable and understandable to someone with a
  decent knowledge of computers, but isn't an expert in POWER.
- SRC codes are incomprehensible to all but a few people on the planet
- Our current data structure for logs: the Platform Error Log (PEL) is
  both POWER specific and requires special tools to turn into text, let
  alone decode what they mean.
- There's been issues in publishing the complete PEL spec
- PEL is a custom binary format that means we have to maintain a lot of
  POWER specific code, which isn't ideal.
- We'll unlikely *completely* get rid of the PEL for P9 as we still are
  going to have IBM FSP systems producing them, but we could deprecate
  and shift to something better on pure OpenPOWER systems.

-- 
Stewart Smith
OPAL Architect, IBM.