Re: [PATCH] restore libata build on frv

[Jeff Garzik <jgarzik@pobox.com>]

David Woodhouse wrote:
> On Mon, 2006-09-25 at 16:46 +0100, Alan Cox wrote:
>> Ar Llu, 2006-09-25 am 15:39 +0100, ysgrifennodd David Howells:
>>> Why does the arch have to supply those numbers?  What's wrong with my
>>> suggested patch?  According to code in libata, these are _legacy_ access
>>> methods, and on FRV they aren't currently required, so why can't I dispense
>> "legacy, legacy, legacy" "wont wont wont"
>>
>> The ports in question are PCI values. They come from the PCI
>> specifications and apply to any device with PCI bus, unless it has
>> special mappings. The same logic you are whining about is already partly
>> handled in the generic pci quirks code, and in time will end up with the
>> I/O port value fixups there anyway.
>>
>> See quirk_ide_bases in drivers/pci/quirks.c
> 
> If we can do that with PCI quirks, why the need to hard-code it in the
> IDE driver too?
> 
> And IRQ zero isn't particularly helpful suggestion -- using an invalid
> IRQ number would be better. Like NO_IRQ or IDE_NO_IRQ, which should be
> -1.
> 
> Don't make me dig out the board where the PCI slots all get IRQ 0 :)

The irq is a special case no matter how we try to prettyify it.  We need 
two irqs, and PCI only gives us one per device.

	Jeff