Re: [coreboot] W83977EF vs. ASUS P2B-L vs. PS/2 keyboard

[Keith Hui]

> Message: 9
> Date: Tue, 23 Mar 2010 14:19:01 -0600
> From: Myles Watson 
> To: ron minnich 
> Cc: Mark Marshall , [e-mail]
> Subject: Re: [coreboot] W83977EF vs. ASUS P2B-L vs. PS/2 keyboard
> Message-ID:
>        
> Content-Type: text/plain; charset="iso-8859-1"
> > 155                     pnp_set_logical_device(dev);
> > > 156                     pnp_set_enable(dev, 1);
>
> This little snippet looks a lot like pnp_enable_resources, which is already
> getting called on all enabled devices.
>
> Maybe it's not getting called early enough?
>
> Thanks,
> Myles

Indeed. Any room for simplification?

http://tracker.coreboot.org/trac/coreboot/browser/trunk/src/devices/device.c

Line 801. This is very suspicious too. Shouldn't it be returning only
if it is ALREADY enabled?

My log points to only PCI devices getting enabled. Could this be the
root culprit?

I'll see what happens if I change this instead of applying the patch I
just floated.

Thanks
Keith

W83977EF vs. ASUS P2B-L vs. PS/2 keyboard

[Myles Watson]

> http://tracker.coreboot.org/trac/coreboot/browser/trunk/src/devices/device.c
>
> Line 801. This is very suspicious too. Shouldn't it be returning only
> if it is ALREADY enabled?

It returns if it isn't enabled.  It was disabled by pnp_enable if that was
needed.
> My log points to only PCI devices getting enabled. Could this be the
> root culprit?
>
> If you put in a debugging statement in pnp_enable_resources you'll see that

it gets called too.

Thanks,
Myles

W83977EF vs. ASUS P2B-L vs. PS/2 keyboard

[Keith Hui]

This is going to get a lot more interesting.

I switched back to my P2B-LS and tried the onboard SCSI now that I got
the keyboard going.

Its presence is known, but the linux module failed to probe it. lspci
shows that its I/O port range is disabled. No wonder. This is with a
BIOS built for P2B-F without the SCSI device listed in devicetree.cb
as I am investigating if the P2B-L/S/LS/F series can share the same
code as P2B-F. I thought coreboot enables all PCI devices it detects?

Same with the onboard Intel LAN. Its presence is also known. ifconfig
shows its MAC address, but assigning it a local IP fails.

That's it for tonight.

Cheers
Keith

On Tue, Mar 23, 2010 at 10:30 PM, Myles Watson  wrote:
> > http://tracker.coreboot.org/trac/coreboot/browser/trunk/src/devices/device.c
> >
> > Line 801. This is very suspicious too. Shouldn't it be returning only
> > if it is ALREADY enabled?
>
> It returns if it isn't enabled.  It was disabled by pnp_enable if that was
> needed.
> > My log points to only PCI devices getting enabled. Could this be the
> > root culprit?
>
> If you put in a debugging statement in pnp_enable_resources you'll see that
> it gets called too.
>
> Thanks,
> Myles

W83977EF vs. ASUS P2B-L vs. PS/2 keyboard

[Myles Watson]

> I switched back to my P2B-LS and tried the onboard SCSI now that I got
> the keyboard going.

So why did that change work if pnp_enable_resources was already doing the
same thing?
> Its presence is known, but the linux module failed to probe it. lspci
> shows that its I/O port range is disabled. No wonder. This is with a
> BIOS built for P2B-F without the SCSI device listed in devicetree.cb
> as I am investigating if the P2B-L/S/LS/F series can share the same
> code as P2B-F. I thought coreboot enables all PCI devices it detects?

It should.  You could send the log to the list.

I'd be interested in the log before and after the keyboard patch.  It seems
like something else is broken if you have to enable it twice.

Thanks,
Myles

W83977EF vs. ASUS P2B-L vs. PS/2 keyboard

[Stefan Reinauer]

On 3/24/10 2:04 PM, Myles Watson wrote:
> > I switched back to my P2B-LS and tried the onboard SCSI now that I got
> > the keyboard going.
>
> So why did that change work if pnp_enable_resources was already doing the
> same thing?

I seem to remember the southbridge driver had to contain some code to
make sure the pnp and i2c devices hanging off it are initialized
correctly. Maybe the device allocator does not try to call
pnp_enable_resources for the ASUS P2B?

Stefan

W83977EF vs. ASUS P2B-L vs. PS/2 keyboard

[Myles Watson]

On Wed, Mar 24, 2010 at 8:56 AM, Stefan Reinauer wrote:
> On 3/24/10 2:04 PM, Myles Watson wrote:
> > > I switched back to my P2B-LS and tried the onboard SCSI now that I got
> > > the keyboard going.
> >
> > So why did that change work if pnp_enable_resources was already doing the
> > same thing?
>
> I seem to remember the southbridge driver had to contain some code to
> make sure the pnp and i2c devices hanging off it are initialized
> correctly. Maybe the device allocator does not try to call
> pnp_enable_resources for the ASUS P2B?

That could be.  I just looked at the superio code, and I'm not sure it looks
right.  I'd like to see the log with this patch applied.  If it's correct,
it should be called for each enabled PNP device.

Index: svn/src/devices/pnp_device.c
===================================================================
--- svn.orig/src/devices/pnp_device.c
+++ svn/src/devices/pnp_device.c
@@ -145,6 +145,7 @@ void pnp_set_resources(device_t dev)

void pnp_enable_resources(device_t dev)
 {
+    printk(BIOS_DEBUG, "%s: %s\n", __func__, dev_path(dev));
     pnp_set_logical_device(dev);
     pnp_set_enable(dev, 1);
 }

Thanks,
Myles