missing RELOC()s in prom.c?

[Paul Mackerras]

Jimi Xenidis writes:

> Fellow coder (cc'd) found the following issue
> -JX

Hmmm, good catch, but it's not obvious to me why smt_setup() has to be
done at prom_init time.  I don't see anything in there that has to be
done then - we are just looking at the command line and the device
tree and setting some fields in the naca.  That could be done more
easily (without the RELOCs) once we have set up the MMU, surely?
We use _naca->smt_state in prom_hold_cpus, it is true, but that only
affects the setting of cpu_available_map and cpu_present_at_boot, and
the values of those bitmaps doesn't affect any calls to OF.

Does anyone have a reason why smt_setup has to be called from
prom_init?

Paul.

** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/