Re: Report of install for a power6 with debian-squeeze-di-beta1-powerpc-netinst.iso
Hi,
> i've also experienced the "two passes behaviour" occasionally, but
> up to now haven't been able to figure out a pattern. It's definately
> Linux-specific, since we don't see that behaviour with our AIX LPARs.
> Not sure though if its yaboot or Kernel related.
It's not specific to Linux, however it may be more obvious because we
have an extra step with yaboot.
The kernel communicates to firmware via the
ibm,client-architecture-support call. This call allows us to tell
firmware about all the various features we support. Unfortunately when
we ask for something different than the previous boot firmware is
often past the point of no return and has to reboot and reconfigure.
As an example, I recently added support for a more detailed NUMA
topology feature:
http://patchwork.ozlabs.org/patch/49682/
Since this feature changes the meaning of some parts of the Open
Firmware device tree, when firmware see this field change between boots
it has no option but to reboot and reconfigure the device tree it
presents.
yaboot has code to recognise this reboot and to retry the previous
kernel automatically and without delay. So bottom line is we should
expect double booting when going between AIX and Linux or sometimes
between versions of Linux (eg a kernel before and after my NUMA change).
If we are seeing the partition continually go around in a loop due to
the client architecture then that is a bug.
Anton
Reply to: