In article <040822021708.M0125288@mirage.ceres.dti.ne.jp>
I wrote:

> I rewrite your patch a bit:
 :

> Well, now I wonder if it's really better to attach the timer in GT64111.
> 
> It may be simpler to map whole region of GT64111 and to initialize
> all registers in gt_attach() rather than to have separate gttimer
> device attachment. I'll rethink how this should be implemented.

I've committed these changes which fix the "hardclock(9) is called
before cpu_initclocks(9)" problem.

If you find any problem, please report here, or via send-pr(1).
---
izumi Tsutsui
tsutsui@ceres.dti.ne.jp