Frank, On Sat, Nov 18, 2017 at 2:01 PM, Frank Scheiner <frank.schei...@web.de> wrote: > Hi Mathieu, > > On 11/18/2017 01:17 PM, Mathieu Malaterre wrote: >> >> Hi Frank, >> >> On Sat, Nov 18, 2017 at 11:52 AM, Frank Scheiner <frank.schei...@web.de> >> wrote: >>> >>> This patch requires the ofpath tool which is part of the yaboot package, >>> as both ofpathname (part of powerpc-ibm-utils package) and >>> grub-ofpathname (not available packaged but part of the GRUB source >>> code) do not translate a given device node into a working OF path on >>> type 11,2 and 7,3 G5 Power Macs. >> >> >> Sorry if this sound silly, but are you sure you want to use a tool >> from the yaboot package, while the original intent is to get rid of >> the yaboot package ? > > > I originally didn't knew that the ofpath tool was part of the yaboot > package. It was available in-target when starting the GRUB installation in > expert mode (i.e. despite skipping the yaboot installation) hence I assumed > it might be part of some DEB package installed by default on ppc64 and in > addition it also was the only device-node-to-OF-path-translation-tool that > worked "correctly" for the tested G5 Power Macs, so I just used it. > > E.g. for /dev/sda ofpathname returned > "/ht@0,f2000000/pci@9/k2-sata-root@c/scsi@0/sd@0,0" while devalias in OF > says it's "/ht/pci@9/k2-sata-root/k2-sata@0/disk@0". > > And for /dev/sda2: > > * I unfortunately don't have a result for ofpathname currently > > * grub-ofpathname returned "/ht@0,f2000000/pci@9/k2-sata-root@c/disk@0:b" > which didn't work > > * ofpath returned "/ht@0,f2000000/pci@9/k2-sata-root@c/@0/@0:2" which worked > > It's of course not ideal, and we should maybe try to fix either ofpathname > or grub-ofpathname or include ofpath (which btw is just a shell script) in > another package (e.g. grub-installer already has an additional tool for PReP > boot devices or integration in powerpc-ibm-utils (for ppc64/newworld alone) > or powerpc-utils (for both powerpc/ and ppc64/newworld). > > Up until now yaboot was the "only" boot loader for powerpc and ppc64, so it > made some sense to include ofpath there (also because it came from the same > developer). With an additional boot loader, it would IMHO make more sense to > include ofpath in another DEB package to be able to share it's > functionality.
Well that's surprising. Thanks for the detailed explanation. I've filled #882076 and forwarded upstream. I fail to understand how people would install grub on ppc64 if grub-ofpathname does not return the correct path. -M
