Re: AER/hotplug _OSC wierdness?

[Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>]

Robert Hancock wrote:
> I've been looking into why I get these messages on bootup on an Asus 
> P7P55D PRO motherboard:
> 
> Firmware did not grant requested _OSC control
> aer 0000:00:03.0:pcie02: AER service couldn't init device: no _OSC support
> pci_hotplug: PCI Hot Plug PCI Core version: 0.5
> Firmware did not grant requested _OSC control
> Firmware did not grant requested _OSC control
> Firmware did not grant requested _OSC control
> Firmware did not grant requested _OSC control
> 
> The AML code for _OSC in the DSDT (below) is almost identical to the 
> example code in the ACPI spec. From looking at the _OSC method, I don't 
> see why it should reject any requests for PCI Express hotplug or AER 
> control.
> 
> I can't quite make sense of the ACPI code that is supposed to handle 
> this. When the AER driver calls acpi_pci_osc_control_set, it first does 
> this:
> 
>     /* Need to query controls first before requesting them */
>     if (!root->osc_queried) {
>         status = acpi_pci_query_osc(root, root->osc_support_set);
>         if (ACPI_FAILURE(status))
>             goto out;
>     }


Just in case, does removing this 'if' statement fix the problem?


>     if ((root->osc_control_qry & control_req) != control_req) {
>         printk(KERN_DEBUG
>                "Firmware did not grant requested _OSC control\n");
>         status = AE_SUPPORT;
>         goto out;
>     }
> 
> Inside acpi_pci_query_osc:
> 
>     /* do _OSC query for all possible controls */
>     support_set = root->osc_support_set | (flags & OSC_SUPPORT_MASKS);
>     capbuf[OSC_QUERY_TYPE] = OSC_QUERY_ENABLE;
>     capbuf[OSC_SUPPORT_TYPE] = support_set;
>     capbuf[OSC_CONTROL_TYPE] = OSC_CONTROL_MASKS;
> 
>     status = acpi_pci_run_osc(root->device->handle, capbuf, &result);
>     if (ACPI_SUCCESS(status)) {
>         root->osc_support_set = support_set;
>         root->osc_control_qry = result;
>         root->osc_queried = 1;
>     }
> 
> The comment says "do _OSC query for all possible controls", but it 
> doesn't look like that's what the code is actually doing. The first time 
> this gets called, assuming osc_support_set is 0, it looks like 
> support_set passed into _OSC will also be 0. Then osc_queried would be 
> set to 1, _OSC would never get called again and the cached query result 
> is based on 0 support flags being passed into _OSC, which is wrong. Or 
> am I missing something?
> 

Though I might be missing something, I think _OSC is evaluated
and osc_control_qry is updated, whenever acpi_pci_osc_support()
gets called.

Thanks,
Kenji Kaneshige