Subject: Re: hardclock(9) for cobalt
To: None <port-cobalt@netbsd.org>
From: Izumi Tsutsui <tsutsui@ceres.dti.ne.jp>
List: port-cobalt
Date: 08/23/2004 23:40:19
In article <20040823.222423.74749864.kiyohara@kk.iij4u.or.jp>
kiyohara@kk.iij4u.or.jp wrote:

> I looked at the source code of Linux. Respectively original
> initialization is performed. We can perform processing which
> corresponds after PCI_NETBSD_CONFIGURE.

What's the "original initialzation" on Linux?
According to linux-2.0.34C53_SK/arch/mips/cobalt/pci.c,
fixup for I/O space and mem space could be done by
your PCI_NETBSD_CONFIGURE patch.
Maybe it's worth to fixup interrupt lines, but I have
no idea about the latency timer and cache line size.

BTW, pci.c of Linux also says in its comment:
	/* Fixup base addresses, we only support I/O at the moment. */
so I wonder if cobalt pci really supports mem space...
(Does Linux 53c8xx driver use memory mapped access?)
---
Izumi Tsutsui
tsutsui@ceres.dti.ne.jp