Re: [PATCH 3/5] Deforest the function pointer jungle in the time code.

["Franck Bui-Huu" <vagabon.xyz@gmail.com>]

Hi,

On 6/14/07, Thomas Bogendoerfer <tsbogend@alpha.franken.de> wrote:
> On Thu, Jun 14, 2007 at 12:19:59PM +0200, Franck Bui-Huu wrote:
> >  arch/mips/sni/a20r.c                      |    1 -
> >  arch/mips/sni/ds1216.c                    |    4 +-
> >  arch/mips/sni/pcimt.c                     |    3 -
> >  arch/mips/sni/pcit.c                      |    3 -
> >  arch/mips/sni/rm200.c                     |    2 -
> >  arch/mips/sni/time.c                      |    2 +-
>
> the SNI part is broken and can't work that way.
I don't get you there. Are you talking about patch 3/5 (you're
replying on this one) or the patch 5/5 ?

patch #3 is only clean up, so it shouldn't break anything....
patch #5 does not migrate any platforms, so it's actually broken for
all current platforms.

> 1. SNI used two different RTC chips (ds126 and mc146818) and it's no big
>    deal support them in just one kernel with the current framework
That's the point now, if the current implementation can fit well with
all platforms.

> 2. One line of SNI machines (a20r) can't use the cp0 counter, so it's not
>    a really good idea to calibrate it
The current code doesn't automatically calibrate any hpt. It was
really hard to guess which ones need that so now if you need to
calibrate your hpt, then you have to call calibrate_hpt().

Thanks
--
              Franck