[coreboot] Serial Console only starts after reset

[Joseph Smith]

On 06/15/2010 11:54 AM, Joseph Smith wrote:
>
>
>
> On Tue, 15 Jun 2010 10:28:51 -0400, Joseph Smith<joe at settoplinux.org>
> wrote:
>>
>>
>>
>> On Tue, 15 Jun 2010 16:16:46 +0200, Stefan Reinauer
> <stepan at coresystems.de>
>> wrote:
>>> On 6/15/10 4:07 PM, Joseph Smith wrote:
>>>> When
>>>> I cold boot it, if I hit the reset button before Linux starts, it
>>> restarts
>>>> with coreboot serial console just fine.
>>>>
>>>
>>> Hm...
>>>
>>> Maybe 0x2e/0x4e are mapped to LPC after the SuperIO is configured (and
>>> that mapping survives a reset)... that would be a southbridge init
> issue.
>>>
>> Hmm, This is interesting. The first time I booted it, I had copied
> romstage
>> from another board that had:
>>
>> #define SERIAL_DEV PNP_DEV(0x4e, SMSCSUPERIO_SP1)
>>
>> And I got coreboot serial console (early_serial) just fine but when it
> came
>> to detecting and inializing the superio, I got all kinds of errors from
>> coreboot about device not found.
>>
>> So I changed it to 0x2e as described in the datasheet and detected with
>> superiotool(vendor bios), and that is when this problem started. But now
>> coreboot is happy, detects the device and the resource allocator is
> happy,
>> I just do not get early_serial on cold boot.
>>
>> This is weird!
>>
>>> Or, the hardware has to stabilize after power on before you can
>>> configure the SuperIO.
>>>
>> Maybe?
>>
> Hmm from page 142 in the datasheet, I need to figure out of there is a
> pull-down resistor connected to the SYSOPT pin to use 0x2e or if there is a
> pull-up resistor connected to the SYSOPT pin to use 0x4e. And once powered
> up the configuration port base address can be changed through CR26 and
> CR27.
>
> I bet you that there is a pull-up resistor connected to the SYSOPT pin and
> 0x4e is supposed to be used. And the vendor bios changes this to 0x2e at
> some point. That is what is throwing me off!
>
> So I will try:
>
> #define SERIAL_DEV PNP_DEV(0x4e, SMSCSUPERIO_SP1)
>
> and change all the devices in devicetree.cb to 4e.
>
> My question is does it need to get changed to 0x2e at some point? Or can I
> just leave it at 0x4e? Will it matter to OS?
>
> Thanks for all the brainstorming help Stefan and Ron.
>
Yup that did it :-)
Looks like SYSOPT had a pull-up resistor so 0x4e works great.

But there is one other small issue. Looks like the resource allocater is 
not allocating any space for device 43.a the Runtime Registers.

My devicetree.cb looks like:
           device pnp 4e.a on            # Runtime registers
             io 0x60 = 0x800
           end

But my bootlog (attached) shows:
PNP: 004e.a resource base 800 size 0 align 0 gran 0 limit 0 flags 
c0000100 index 60

skipping PNP: 004e.a at 60 fixed resource, size=0!

PNP: 004e.a 60 <- [0x0000000800 - 0x00000007ff] size 0x00000000 gran 0x00 io

     PNP: 004e.a resource base 800 size 0 align 0 gran 0 limit 0 flags 
e0000100 index 60


Is it supposed to look like that?

-- 
Thanks,
Joseph Smith
Set-Top-Linux
www.settoplinux.org
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: i810_boot_test11.txt
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20100615/c5662790/attachment.txt>