Re: CATS todclock reorganization (Re: CVS commit: syssrc/sys/arch)

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

In article <E16Ztpj-0005Uj-0K%tele-post-20.mail.demon.net@localhost>
chris%paradox.demon.co.uk@localhost wrote:

> > It seems todclock.c does not have footbridge-specific functions.
> > Anyway, I think we should reorganize todclock attachment on cats port.
> > My idea is:
 :
> Sounds fine to me, I was more interested in making sure we can still build 
> kernels and the files are in roughly the right place.

Yeah, I agree it is important to keep kernels (at least) compile.
Now it is fixed, so I'd like to suggest the next step..

> certainly todclock and 
> dsrtc shouldn't really be in footbridge, but they needed to go somewhere 
> shared and I think netwinder uses them.

I don't know about footbridge spec, but does it always require DS1687?
if not, the attachment can be shared as dev/isa/ds1687_isa.c.

In todclock.c case, I think it should be really MI in the perfect world.
(some of them are in clock_subr.c already)
todclock.c should have microtime(9), inittodr(9), resettodr(9) and
cpu_initclocks() etc. and all driver of RTC and interval timer
should pass device dependent functions to MI part.
But maybe it requires a log of work, so I think it is OK
to keep todclock.c port dependent for now.
---
Izumi Tsutsui
tsutsui%ceres.dti.ne.jp@localhost