[OpenPower-Firmware] How to flash openPOWER on a Habanero system (8348-21C)

[Dan Horák]

Hi Nick,

On Fri, 5 Feb 2021 13:22:20 +0100
Nick Singer <nsinger at suse.de> wrote:

> Hey all,
> 
> Am 02.02.21 um 19:36 schrieb Klaus Heinrich Kiwi:
> > Didn't realize Nick wasn't cc'ed in Dan's email, so adding him now.
> 
> Thanks! I've followed the archives so I found it anyway :)
> 
> > On 2/2/2021 3:35 PM, Klaus Heinrich Kiwi wrote:
> >>> have you tried flashing the PNOR from the BMC? IIRC that's how I have
> >>> updated mine.
> >>>
> >>
> >> Nick,
> >>
> >>   any updates here? Have you tried to update through the BMC?
> 
> I didn't consider it as
> https://github.com/open-power/op-build/issues/848#issuecomment-276980698
> states that the HPM format is not build by op-build. The second option I
> have in the BMC is "AMI" which seems to be proprietary and signed. At
> least when I tried to upload my pnor file it complained about unsigned code.

were you using the BMC web UI or shell/command line? I am pretty sure I
used "pflash" from the BMC shell (ssh into the BMC) to write our
firmware (http://fedora.danny.cz/ppc/) build into the PNOR. Not sure if
we have updated the BMC firmware too before updating PNOR.

Jakub (CCed), do you recall the details?


		Dan

> >> What skiboot version are you using? There was a fix on 5.4.0 that
> >> could be related:
> >> https://github.com/open-power/skiboot/issues/44
> >>
> >> (it might mean you need to pair the pflash utility with the skiboot
> >> version
> >> you are currently using)
> 
> After you mentioning this I tried several versions of pflash. Starting
> from a very old release (early 5.x, as I couldn't build 4.x) up to the
> latest release and even tried master. Some of the older versions
> couldn't even retrieve the information (./pflash -i) while the later
> where perfectly able to do this. Also reading out the pnor works
> flawlessly (I took a "backup" with pflash as well as with dd from
> /dev/mtd0).
> 
> >> Thanks,
> >>
> >>   -Klaus
> >>
> > 
> 
> I also read that flash devices might be locked and one can unlock them
> with "flash_unlock" from mtd-utils but this just yielded: "error 95
> (Operation not supported)" so I guess it does not apply to power8 chips.
> 
> Any other ideas what I could try out?