Subject: Re: Debugging / SPL-levels / AT91RM9200 success
To: M. Warner Losh <>
From: Sami Kantoluoto <>
List: port-arm
Date: 03/06/2007 23:48:13
On Tue, Mar 06, 2007 at 02:03:48PM -0700, M. Warner Losh wrote:
> In message: <>
>             Sami Kantoluoto <> writes:
> : I'll take a look of FreeBSD's port.
> FreeBSD's port doesn't do CF cards right now.  It does a lot of other
> things.  Also, since FreeBSD doesn't have spl anymore, I'm not sure
> how helpful it would be.  That part of the systems have been diverging
> for a while now...

Ah, ok (I was thinking about spl stuff).

> : PS. When the port is somehow working, we would like to see it in NetBSD's
> : repository. Currently we've three arch-directories, arch/arm/at91,
> : arch/arm/at91rm9200 and arch/evbarm/mpcsa. I've separated at91 and at91rm9200
> : because the idea is to write AT91SAM926x support too someday. They (AT91RM9200
> : and AT91SAM926x) have quite a lot common but e.g. AT91SAM926x does not have
> : a system timer like AT91RM9200, ethernet controllers are different and so
> : on.
> Be careful here.  Many of the drivers are also applicable to the AVR32
> family as well.  The MCI controller is near enough to being the same
> that it could be shared between the two, but thankfully the
> AT91SAM926x processors don't share the byte order erratum with the
> AT91RM9200, so streaming performance isn't limited by how quickly you
> can byte swap...


> I'm also not convinced you need separate directories for each of the
> SOCs.  In FreeBSD all the files live in sys/arm/at91, so far quite
> happily.  I'm in the midst of bringing up an AT91SAM9260 eval board on

Well, that can be easily changed. I just thought this is the normal way
(each SOC in own directory) but I really don't know.

> FreeBSD, and so far it hasn't been too bad.  A lot easier than trying
> to bring up the AVR32 board I have...

Yes, writing whole architechture support must be much harder..