Re: first call softintr__init()

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

kiyohara%kk.iij4u.or.jp@localhost wrote:

> > - make sure which machines actually use MI ibm4xx_cpu_startup()
> >   (at least explora, virtex and walnut have their own cpu_startup())
> 
> hmm...
> I referred to oea.  softintr__init() is called from oea_startup() for
> oea.  softintr__init() actually initializes softintr_pool.  And,
> softintr__init() is called from intr_init() for ibm4xx.
> 
> Should oea also change?

Well, I don't say about OEA at all, but
only obs200 and obs405 use ibm4xx_machdep.c,
i.e. only evbppc/conf/files.obs{200,405} have
arch/powerpc/ibm4xx/ibm4xx_machdep.c.

So you shouldn't remove init_init() from
explora/machdep.c:cpu_startup(), and
you have to move intr_init() calls from
{virtex,walnut}/autoconf.c:cpu_configure() to
{virtex,walnut}/machdep.c:cpu_startup(),
not powerpc/ibm4xx/ibm4xx_machpdep.c:cpu_startup().
(I have no idea if we should make them also use MI ibm4xx_machdep.c)

Anyway, I think loginit() should be defered for now
as callout_startup2(), but we should ask Andrew first.
---
Izumi Tsutsui