[Skiboot] [PATCH] hw/phb3: Update capi initialization sequence

Frederic Barrat fbarrat at linux.vnet.ibm.com
Mon Aug 1 19:15:32 AEST 2016


Le 01/08/2016 à 07:23, Stewart Smith a écrit :
> I don't seem to have access to that bug, plus, I really don't like us
> referencing internal bug numbers in external repositories.
>
> so, what is this change? Why? Where is it documented? If it's internal
> documentation, please point me at it.


Apologies. I had seen a few references to defects in the logs, but that 
was old, at a time where it may have been more acceptable.

The issue described in the defect/circumvention document is actually 
getting old (March 2014). A "link down" error detected by part of the 
PHB logic was converted from Fatal to Endpoint Recoverable. It affects 
the initialization sequence of some of the phb3 registers, but that part 
had already been taken care of in skiboot from the initial release, as 
far as I can tell. I'll send the internal document I got from the 
hardware folks, but it shouldn't be news (from what they said and what I 
see in the skiboot code).

However, there was also an impact on CAPI, which somehow was dropped. 
The 2 registers modified by the patch are meant to have the CAPP react 
to the "link down" error. It was noticed on a least a couple of defects 
recently when hunting EEH recovery issues for cxlflash.

   Fred



More information about the Skiboot mailing list