Re: one more ACPI Error (utglobal-0125): Unknown exception code:0xFFFFFFEA [Re: 2.6.18-rc4-mm3]

[Shaohua Li <shaohua.li@intel.com>]

On Thu, 2006-09-07 at 09:25 -0600, Bjorn Helgaas wrote:
> On Wednesday 06 September 2006 20:03, Shaohua Li wrote:
> > On Thu, 2006-09-07 at 04:04 +0800, keith mannthey wrote:
> > > On Wed, 2006-09-06 at 11:59 -0700, Moore, Robert wrote: 
> > > > From one of the ACPI guys: 
> > > >  
> > > > > Get hid 
> > > > > Look for driver 
> > > > > If you find a match, load it 
> > > > > If no match, get CID 
> > > > > Look for driver 
> > > > > If you find a match, load it 
> > > > > If you did not find an hid or cid match, punt
> > > 
> > > I think this is what my patch is doing.
> > > 
> > > when looking for a driver: (acpi_bus_find_driver) 
> > > I check against the HID  
> > > return if found  
> > > Then I check against the CID 
> > > return if found 
> > > else 
> > > punt 
> > > 
> > > Any objections to pushing this into -mm and dropping the motherboard 
> > > change?
> 
> > I'd prefer not take this way. The ACPI driver model is already mess
> > enough, let's don't make it worse. We are converting the ACPI driver
> > model to Linux driver model, this will make the attempt difficult.
> 
> I see that driver_bind() and driver_probe_device() don't mesh well
> with the idea that multiple drivers might be able to claim a device,
> because there doesn't seem to be a way to prioritize one driver
> over another.  Is that the problem you're referring to?
Yes.

> If we decide that "try HID first, then try CID" is the right thing,
> I think we should figure out how to make that work.  Maybe that
> means extending the driver model somehow.
Don't think it's easy, especially no other bus needs it I guess.

> > We can let the motherboard driver not bind to your device (say we didn't
> > register the motherboard driver, but just reserve the resource of the
> > deivce). Is it ok to you? (I remember Bjorn said he wants to reserve the
> > mem region of the device too).
> 
> My point was that ACPI tells us what resources the device uses,
> and we should reserve all of them so we accurately model the system.
> 
> Reserving resources without registering the driver sounds like a hack
> to work around broken behavior elsewhere, so I don't think it's a
> good idea.
Do we really need the memory hotplug device returns pnp0c01/pnp0c02?
What's the purpose?

Thanks,
Shaohua