RE: Re: boot NetBSD using u-boot

["Doug Fraser" <dwfraser%onebox.com@localhost>]

Robert,

Linux maybe, but on NetBSD port-powerpc (evbppc/marvell gt63260)
the machdep portion accepts the setup from the monitor.
If you set up no PCI/no chip selects, the kernel sets up
no PCI/no chip selects. At least that is how it is in evbppc for
NetBSD 2.0

We have a number of boards that run a common kernel.
The boot monitor configures that layer on a per board basis,
and the kernel happily accepts what we give it.
On some, we use all four CS, on others, three. On some we
use both PCI, on others only one.

So if U-Boot does not set those up now, it will have to be modified
to make NetBSD happy, or an intermediary will have to be written
that does. Perhaps that is why this all started with the discussion
of a pass 2 boot loader? Certainly, you could modify the kernel to
do this initialization. One of the advantages of having the source.
We just chose to do setup and POST in the monitor and leave the
kernel alone.

I am not very familiar with U-Boot. Does it support a user provided
callout to provide additional initialization, or do you just hack the
existing code directly?


Doug



-----Original Message-----
From:     Robert Swindells <rjs%fdy2.demon.co.uk@localhost>
Sent:     Thu,  6 Oct 2005 19:44:32 +0100 (BST)
To:       dwfraser%onebox.com@localhost
Cc:       port-powerpc%netbsd.org@localhost
Subject:  Re: boot NetBSD using u-boot


Doug Fraser wrote:
>Your statement about SDRAM is correct, I was just pointing out
>that his firmware, Das-U-Boot, or whatever, also has to set up
>the PCI mapping (if that applies) as well as any chip select
>mapping that is supported by the chip.

No it doesn't.

We are using u-boot at work and only configure the boot chip select
in firmware. All PCI mapping is done by the loaded operating system
(linux).

Robert Swindells