2.6 agpgart and acpi standby/resume

2.6 agpgart and acpi standby/resume

[Terence Ripperda, tripperda]

when working on supporting the 2.6 kernel's acpi support, I ran into some issues on a Dell Inspiron 4100 notebook (w/ Intel Corp. 82830 830 Chipset Host Bridge) when resuming from standby. the basic problem was that agp configuration was lost in the host bridge's pci config space, causing dma problems with our chip.

I checked through the agpgart code (specifically the intel-agp.c driver) and saw that the driver did support acpi resume and attempted to reconfigure the chipset when resuming. but the problem is that the driver only does this if the bridge is using the generic or 845 driver. in my case, the 830mp driver was being used. I hacked in support for my driver and everything worked fine.

I'm curious why support was only added for 2 cases, instead of reconfiguring the chipset in every case. Is this because there were problems with some drivers, or is support added only on an "as-needed" basis?

(in this case, I happen to be testing with 2.6.0, but I see the same problem in the 2.6.1 code)

Thanks,
Terence

this fixed my problem:

--- intel-agp.c.orig    2004-01-22 12:51:31.000000000 -0600
+++ intel-agp.c 2004-01-22 12:51:20.000000000 -0600
@@ -1432,6 +1432,8 @@
                intel_configure();
        else if (bridge->driver == &intel_845_driver)
                intel_845_configure();
+       else if (bridge->driver == &intel_830mp_driver)
+               intel_830mp_configure();
 
        return 0;
 }

Re: 2.6 agpgart and acpi standby/resume

[Dave Jones]

On Thu, Jan 22, 2004 at 12:58:07PM -0600, Terence Ripperda wrote:

 > I'm curious why support was only added for 2 cases, instead of reconfiguring
 > the chipset in every case. Is this because there were problems with some
 > drivers, or is support added only on an "as-needed" basis?

The latter, though more an "as-tested" basis.

		Dave

Re: 2.6 agpgart and acpi standby/resume

[Terence Ripperda, tripperda]

ok thanks. can you consider the patch I attached to the first email as submitted and verified? I'll send more as I run across them.

Thanks,
Terence

On Thu, Jan 22, 2004 at 07:44:04PM +0000, davej@redhat.com wrote:
> On Thu, Jan 22, 2004 at 12:58:07PM -0600, Terence Ripperda wrote:
> 
>  > I'm curious why support was only added for 2 cases, instead of reconfiguring
>  > the chipset in every case. Is this because there were problems with some
>  > drivers, or is support added only on an "as-needed" basis?
> 
> The latter, though more an "as-tested" basis.
> 
> 		Dave

Re: 2.6 agpgart and acpi standby/resume

[Dave Jones]

On Thu, Jan 22, 2004 at 01:53:38PM -0600, Terence Ripperda wrote:
 > 
 > ok thanks. can you consider the patch I attached to the first email as
 > submitted and verified? I'll send more as I run across them.
 
Will do, (sorry for not making that obvious).

Thanks,

		Dave