Re: [coreboot] AMD Serengeti Cheetah devices

[ron minnich]

it's working now, here's patch.

it even probes for options roms etc. We can try vga now!

Need an ack.

ron

AMD Serengeti Cheetah devices

[Peter Stuge]

ron minnich wrote:
> This gets us to etherboot again, but this time devices are set 
> up correctly on bus 1 --- i.e., the scan of the 8111 bridge works. It even 
> tries to find the vga rom to run it, which we did not get before.
>
> Signed-off-by: Ronald G. Minnich

If you explain the below, it's:

Acked-by: Peter Stuge
> +++ southbridge/amd/amd8111/usb.c	(working copy)
> .phase4_enable_disable           = amd8111_enable,

Ok, enable USB.
> +++ mainboard/amd/serengeti/dts	(working copy)
>    	pci0@18,0 {
>    		/config/("northbridge/amd/k8/pci");
>    		pci@0,0 {
> 		/config/("southbridge/amd/amd8111/amd8111.dts");
> 		/config/("southbridge/amd/amd8111/pci.dts");
> 		pci@5,0{
> 			/config/("southbridge/amd/amd8111/nic.dts");
> 		};
> };
> pci@4,0 {
> 	/config/("southbridge/amd/amd8111/ide.dts");
> };
> 	pci@5,0 {
> 		/config/("southbridge/amd/amd8111/nic.dts");
> 	};
> };

Why treat nic and ide differently here? And what about that USB? Is
it automatically added - and ide+nic needs to be explicitly listed?
> +++ arch/x86/Makefile	(working copy)
> 	pci_ops_conf1.c resourcemap.c 
> 	pci_ops_conf1.c 

Just cleanup? Isn't resourcemap needed because of the change or was
it never needed?

//Peter

AMD Serengeti Cheetah devices

[Myles Watson]

> Why treat nic and ide differently here?

I can answer this one.  They are on different busses.  The nic hangs off the
pci controller in the 8111, and IDE doesn't.  You can reference the picture
from the data sheet in this thread.
> > +++ arch/x86/Makefile	(working copy)
> > 	pci_ops_conf1.c resourcemap.c
> > 	pci_ops_conf1.c
>
> Just cleanup? Isn't resourcemap needed because of the change or was
> it never needed?

It just got moved to a different Makefile.

Thanks,
Myles

AMD Serengeti Cheetah devices

[ron minnich]

OK, here you go, I cleaned the dts up a bit more and get:
<8>dynamic PCI: 00:00.0(PCI: 00:06.0): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:07.0(PCI: 00:07.0): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:07.1(PCI: 00:07.1): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:07.2(PCI: 00:07.2): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:07.3(PCI: 00:07.3): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:07.5(PCI: 00:07.5): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:07.6(PCI: 00:07.6): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:07.7(PCI: 00:07.7): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:0a.1(PCI: 00:0a.1): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 00:0b.0(PCI: 00:0b.0): enabled 1 have_resources 1 initialized 0
<8>dynamic PCI: 00:0b.1(PCI: 00:0b.1): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 01:00.0(PCI: 01:00.0): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 01:00.1(PCI: 01:00.1): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 01:00.2(PCI: 01:00.2): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 01:01.0(PCI: 01:01.0): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 01:04.0(PCI: 01:04.0): enabled 1 have_resources 1 initialized 1
<8>dynamic PCI: 01:05.0(PCI: 01:05.0): enabled 1 have_resources 1 initialized 1

which exactly matches simnow.

On Thu, Oct 9, 2008 at 6:41 AM, Peter Stuge  wrote:
> Ok, enable USB.

I just added it in to dts and it finds both usb and usb2.
> Why treat nic and ide differently here? And what about that USB? Is
> it automatically added - and ide+nic needs to be explicitly listed?

I now list them all.
> +++ arch/x86/Makefile (working copy)
>                 pci_ops_conf1.c resourcemap.c
>                 pci_ops_conf1.c
> Just cleanup? Isn't resourcemap needed because of the change or was
> it never needed?

It turns out many newer targets will not need this done in stage0, and
space is limited. So mainboards that need it in stage0 will have to
list it.

Committed revision 910.

And etherboot is working :-)

ron

AMD Serengeti Cheetah devices

[Myles Watson]

> > Why treat nic and ide differently here? And what about that USB? Is
> > it automatically added - and ide+nic needs to be explicitly listed?
>
> I now list them all.

Very nice.  Now how do we disable the 8111 nic and EHCI (like v2 does)?  I'm
still curious how that works.

Thanks,
Myles

AMD Serengeti Cheetah devices

[ron minnich]

On Thu, Oct 9, 2008 at 8:07 AM, Myles Watson  wrote:
> > > Why treat nic and ide differently here? And what about that USB? Is
> > > it automatically added - and ide+nic needs to be explicitly listed?
> >
> > I now list them all.
>
> Very nice.  Now how do we disable the 8111 nic and EHCI (like v2 does)?  I'm
> still curious how that works.

Disable for scanning or disable once it is all scanned and address
space allocated?

ron

AMD Serengeti Cheetah devices

[Myles Watson]

> > Very nice.  Now how do we disable the 8111 nic and EHCI (like v2 does)?
>
> I'm
> > still curious how that works.
>
> Disable for scanning or disable once it is all scanned and address
> space allocated?

Good question.  What did the v2 disable mean?

Thanks,
Myles

AMD Serengeti Cheetah devices

[ron minnich]

On Thu, Oct 9, 2008 at 8:30 AM, Myles Watson  wrote:
> > > Very nice.  Now how do we disable the 8111 nic and EHCI (like v2 does)?
> >
> > I'm
> > > still curious how that works.
> >
> > Disable for scanning or disable once it is all scanned and address
> > space allocated?
>
> Good question.  What did the v2 disable mean?

so here are the cases. You might not want to scan the device, or you
might want the device to actually not be visible.

make it not scanned, put enabled = "0" in dts.
*possibly* make it disappear, put enabled = "0" in dts but also have a
phase4_enable_disable that will look at dev->enabled and make the
magic happen.

For now, you can try
enabled = "0"; in the dts entry for the nic.

ron

AMD Serengeti Cheetah devices

[Carl-Daniel Hailfinger]

On 09.10.2008 17:40, ron minnich wrote:
> On Thu, Oct 9, 2008 at 8:30 AM, Myles Watson  wrote:
> > > > Very nice.  Now how do we disable the 8111 nic and EHCI (like v2 does)?
> > >
> > > I'm
> > > > still curious how that works.
> > >
> > > Disable for scanning or disable once it is all scanned and address
> > > space allocated?
> >
> > Good question.  What did the v2 disable mean?
>
> so here are the cases. You might not want to scan the device, or you
> might want the device to actually not be visible.
>
> make it not scanned, put enabled = "0" in dts.
> *possibly* make it disappear, put enabled = "0" in dts but also have a
> phase4_enable_disable that will look at dev->enabled and make the
> magic happen.

That disappearance trick only works only with geode right now and needs
disable_vpci or somesuch set in the dts. I had a patch which allowed you
to use "hidden" as a property in the dts. I can dig it up if you want.
> For now, you can try
> enabled = "0"; in the dts entry for the nic.

I thought it was
disabled;
but I'm not so sure anymore.

Regards,
Carl-Daniel

AMD Serengeti Cheetah devices

[ron minnich]

On Thu, Oct 9, 2008 at 3:29 PM, Carl-Daniel Hailfinger
 wrote:
> That disappearance trick only works only with geode right now and needs
> disable_vpci or somesuch set in the dts. I had a patch which allowed you
> to use "hidden" as a property in the dts. I can dig it up if you want.

there has been hardware in the past -- in v1 days in particular --
wherein a bit set in function 0 made another function disappear -- not
visible in config space. It really happens ...

ron

AMD Serengeti Cheetah devices

[Carl-Daniel Hailfinger]

On 10.10.2008 00:33, ron minnich wrote:
> On Thu, Oct 9, 2008 at 3:29 PM, Carl-Daniel Hailfinger
>  wrote:
> > That disappearance trick only works only with geode right now and needs
> > disable_vpci or somesuch set in the dts. I had a patch which allowed you
> > to use "hidden" as a property in the dts. I can dig it up if you want.
>
> there has been hardware in the past -- in v1 days in particular --
> wherein a bit set in function 0 made another function disappear -- not
> visible in config space. It really happens ...

Yes, the "works only in geode right now" was about us having only
geode-specific code for disabling devices.

My ICH4 laptop can hide the SMBus device as well. Later ICH can probably
do the same.

Regards,
Carl-Daniel

AMD Serengeti Cheetah devices

[Marc Jones]

ron minnich wrote:
> On Thu, Oct 9, 2008 at 3:29 PM, Carl-Daniel Hailfinger
>  wrote:
> > That disappearance trick only works only with geode right now and needs
> > disable_vpci or somesuch set in the dts. I had a patch which allowed you
> > to use "hidden" as a property in the dts. I can dig it up if you want.
>
> there has been hardware in the past -- in v1 days in particular --
> wherein a bit set in function 0 made another function disappear -- not
> visible in config space. It really happens ...

Many chipsets liek the 690.600 can do this as well.
Marc

AMD Serengeti Cheetah devices

[Corey Osgood]

On Thu, Oct 9, 2008 at 6:58 PM, Marc Jones  wrote:
> ron minnich wrote:
> > On Thu, Oct 9, 2008 at 3:29 PM, Carl-Daniel Hailfinger
> >  wrote:
> > > That disappearance trick only works only with geode right now and needs
> > > disable_vpci or somesuch set in the dts. I had a patch which allowed you
> > > to use "hidden" as a property in the dts. I can dig it up if you want.
> >
> > there has been hardware in the past -- in v1 days in particular --
> > wherein a bit set in function 0 made another function disappear -- not
> > visible in config space. It really happens ...
>
> Many chipsets liek the 690.600 can do this as well.
> Marc

On CN700, everything but d0f0 is hidden by default, and needs to be enabled
by setting a bit.

-Corey

AMD Serengeti Cheetah devices

[Carl-Daniel Hailfinger]

On 10.10.2008 01:02, Corey Osgood wrote:
> On Thu, Oct 9, 2008 at 6:58 PM, Marc Jones  wrote
> > ron minnich wrote
> > > On Thu, Oct 9, 2008 at 3:29 PM, Carl-Daniel Hailfinger wrote:
> > > > That disappearance trick only works only with geode right now and needs
> > > > disable_vpci or somesuch set in the dts. I had a patch which allowed you
> > > > to use "hidden" as a property in the dts. I can dig it up if you want.
> > >
> > > there has been hardware in the past -- in v1 days in particular --
> > > wherein a bit set in function 0 made another function disappear -- not
> > > visible in config space. It really happens ...
> >
> > Many chipsets liek the 690.600 can do this as well.
> > Marc
>
> On CN700, everything but d0f0 is hidden by default, and needs to be enabled
> by setting a bit.

OK, you all get your "disabled" dts property. Ping me after I submit the
C7 CAR for v3.

Regards,
Carl-Daniel

AMD Serengeti Cheetah devices

[ron minnich]

On Thu, Oct 9, 2008 at 3:29 PM, Carl-Daniel Hailfinger
 wrote:
> I thought it was
> disabled;
> but I'm not so sure anymore.

if (streq(prop->name, "enabled")){
    		enabled = 1;
    	}
    	if (streq(prop->name, "disabled")){
    		enabled = 0;
    	}

sorry, you are right.

ron