[OpenPower-Firmware] [POWER8] OCC Firdata over IPMI

[Andrew Jeffery]

On Wed, 29 May 2019, at 17:42, Artem Senichev wrote:
> On Wed, May 29, 2019 at 09:59:54AM +0930, Andrew Jeffery wrote:
> > On Wed, 29 May 2019, at 07:03, Douglas Gilbert wrote:
> > > Everything I have observed so far has the correct ecc calculation. 
> > > Another thing that could cause the symptom being see is that the pnor 
> > > as not been erased. If the existing value of a pnor 8 byte field is ff 
> > > ff ff ff ff ff ff ff 00 and new data is written w/o an "erase" then the 
> > > new data would look fine, except the ecc byte would not be changed from 
> > > 00.
> > > When does the erase happen?
> > > 
> > 
> > mboxd performs the erase when flushing changed regions to flash.
> > 
> > Artem: It might be interesting to capture the DBus traffic when the OCC FIR data
> > write occurs. It might help determine whether it's the protocol parameters or
> > the data that is broken. This can be done with:
> > 
> > # busctl capture > /tmp/dbus.pcap
> > 
> > Separately, was there a github issue open for this? I think it's probably time
> > we tracked the problem somewhere more convenient than on a mailing list.
> 
> Guys, I think we did not understand one another :)
> 
> OCC doesn't use IPMI to send FIRDATA, despite it has support for this.
> This is the problem. 

*record scratch*

Oh.

Doug: Did the patches get backported to whatever branch we ship for P8?

> Using SFC for writing causes ECC errors.
> 
> Other components of OpenPOWER firmware work with PNOR flash over IPMI,
> only OCC cannot do it this way.
> 
> I created issue on github:
> https://github.com/open-power/occ/issues/24

Thanks for that.

Andrew

> 
> -- 
> Regards,
> Artem Senichev
> Software Engineer, YADRO.
>