Re: MPC5200 support

[Robert Swindells <rjs%fdy2.demon.co.uk@localhost>]

Frank Wille wrote:
>>Robert Swindells wrote:
>>
>>>> Ok, makes sense. Do you have the impression that the interrupts are
>>>> working correctly?
>>> 
>>> They work for the PCI slot, I think they should work for other devices.
>
>Nice.
>I just added genfb* to my EFIKA config and the Radeon 9250 is recognized.

Other interrupts were not working, I have revised the interrupt controller
driver. The USB host controller seems to work now.

I will put a new snapshot onto ftp.n.o.

>The pci_configure_bus() doesn't work, but that doesn't seem to
>matter. Should we remove the option PCI_NETBSD_CONFIGURE for Efika, or
>do you plan to use a special efikapci device?

I wasn't planning on writing an efikapci device, how does pci_configure_bus()
fail ?

>> The disk controller needs to be in PIO mode until we get Bestcomm to
>> work.
>> 
>> Try this:
>> 
>> --- mpc5200_ide.c~      2009-06-02 14:35:09.000000000 +0100
>> +++ mpc5200_ide.c       2009-06-03 20:27:22.000000000 +0100
>
>Yes, works.
>
>BTW, there must be an endianess problem when reading the disk ID. OFW shows:
>
>  channel 0 unit 0 :   ata | TS64GSSD25-M                             | V0826
>
>But NetBSD reads:
>
>  wd0 at atabus0 drive 0: <ST46SGDS52M->
>
>I didn't notice, because I thought it must be a Seagate, but it's a
>Transcend SSD. :)

Looking at the other wdc attachments I think the bus_space functions
need to work in little endian mode.

>>> Unfortunately I had to boot the kernel directly, because I couldn't
>>> find out why ofwboot failed. If reports "root addr=0.0.0.0" and waits
>>> forever.
>> 
>> I just boot the kernel too, I boot the latest device tree supplement
>> first though.
>> 
>> I have put it in ftp.netbsd.org:/pub/NetBSD/misc/rjs/efika.forth

>Oh, I didn't know about it. Thanks.

>But is it really required? Or shouldn't it be our goal to run without
>such patches?

It is an official patch. There were plans to produce an installer to
put it into flash.

>The MPC5200 has known embedded devices at fixed addresses, so the
>alternative would be that they are defined by powerpc/mpc5200/,
>instead of taking them from OFW.

The builtin devices are at fixed addresses but they can be multiplexed
onto the external pins in different ways. The OFW device tree seems as
good a way as any to describe how the board designer has wired things
together.

>There might also be MPC5200 systems without OFW.

At my PPOE we made one. If I was designing it again though I would
have put OFW on it. Linux seems to require OFW for the MPC5200 drivers.

Robert Swindells