yenta_socket: no PCI interrupts after resume if intel-agp loaded on HP 1105

yenta_socket: no PCI interrupts after resume if intel-agp loaded on HP 1105

[Matthew Garrett]

If I boot a kernel without any modules loaded and then load intel-agp, I
can successfully suspend and resume this machine via ACPI. However, the
b44 module is then unable to determine its MAC address correctly, and
ipw2100 claims that it can't find a device. yenta_socket gives 

Yenta : Enabling burst memory read transactions
Yenta: Using CSCINT to route CSC interrupts to PCI
Yenta: Routing CardBus interrupts to PCI
Yenta TI: socket 0000:02:06.0, mfunc 0x00001000, devctl 0x66
Yenta TI: socket 0000:02:06.0 probing PCI interrupt failed, trying to fix
Yenta TI: socket 0000:02:06.0 falling back to parallel PCI interrupts
Yenta TI: socket 0000:02:06.0 no PCI interrupts. Fish. Please report.

The only output from the AGP driver is:

agpgart: Detected an Intel 855PM Chipset.
agpgart: AGP aperture is 32M @ 0xe2000000

If I have all of these drivers loaded at suspend, resume fails. Without
intel-agp, I can suspend and resume happily (well, X won't start after
resume, but that's potentially an entirely separate issue)
-- 
Matthew Garrett | mjg59@srcf.ucam.org

Re: yenta_socket: no PCI interrupts after resume if intel-agp loaded on HP 1105

[Matthew Garrett]

> If I have all of these drivers loaded at suspend, resume fails. Without
> intel-agp, I can suspend and resume happily (well, X won't start after
> resume, but that's potentially an entirely separate issue)

I've tracked this down some further. If I comment out these lines in
intel_845_configure:

        /* agpm */
        pci_read_config_byte(agp_bridge->dev, INTEL_I845_AGPM, &temp2);
        pci_write_config_byte(agp_bridge->dev, INTEL_I845_AGPM, temp2 |
(1 << 1));

then I can load intel-agp, suspend and resume and still get correct
interrupts. If I leave then uncommented, then attempting to load
yenta-socket after a suspend/resume with intel-agp loaded results in the
"no PCI interrupts. Fish" error. This is an i855PM device using nvidia
graphics. 

-- 
Matthew Garrett | mjg59@srcf.ucam.org