Re: INTEL 845G Chipset IDE Quandry

[Martin Dalecki <dalecki@evision-ventures.com>]

Vojtech Pavlik wrote:
> On Mon, Jun 03, 2002 at 06:46:24AM +0200, Martin Dalecki wrote:
> 
>>Vojtech Pavlik wrote:
>>
>>>On Sun, Jun 02, 2002 at 09:36:35PM +0200, Martin Dalecki wrote:
>>>
>>>
>>>>Anthony Spinillo wrote:
>>>>
>>>>
>>>>>Back to my original problem, will there be a fix before 2010? ;)
>>>>
>>>>Well since you have already tyred yourself to poke at it.
>>>>Well please just go ahead and atd an entry to the table
>>>>at the end of piix.c which encompasses the device.
>>>>Do it by copying over the next familiar one and I would
>>>>be really geald if you could just test whatever this
>>>>worked. If yes well please send me just the patch and
>>>>I will include it.
>>>
>>>
>>>Note it works with 2.5 already. We have the device there.
>>
>>Yes after looking it up I realized it's already there.
> 
> 
> But as Alan pointer out, in 2.4 the missing PCI ID isn't the problem -
> it would work with no tuning without it, but the fact the on-board BIOS
> incorrectly assigns io-ranges to the PCI device is a problem we may have
> on 2.5 as well.


Well I don't know that much about the ever changing PCI/ACPI support
in kernel - the only thing I could imagine
would be that we sanitize the handling of it at the generic
"chipset quirk handling" there. Right during the "bios table
scan" time... (I mean drivers/pci/quirks.c)

The following function there looks like the right tool for this
purpose:

static void __init quirk_io_region(struct pci_dev *dev, unsigned region, 
unsigned size, int nr)

Well after looking closer I'm convinced that this is
the right place... will you have a look at this plase...
I'm more then busy enbough with other things right now.