Re: [PATCH] x86: adjust GFP mask handling for coherent allocations

["Jan Beulich" <JBeulich@novell.com>]

>>> Ingo Molnar <mingo@elte.hu> 26.10.09 21:19 >>>
>
>* Jan Beulich <JBeulich@novell.com> wrote:
>
>> >>> Ingo Molnar <mingo@elte.hu> 26.10.09 16:22 >>>
>> >* Jan Beulich <JBeulich@novell.com> wrote:
>> >> And any attempt to eliminate the conditional another way would just 
>> >> introduce a very similar conditional elsewhere; with this having a 
>> >> single user (and foreseeably not ever a second one) I would think this 
>> >> would just make the code less readable.
>> >
>> >There's 3 other current uses of DMA_BIT_MASK(24) in arch/x86 - couldnt 
>> >those use ISA_DMA_BIT_MASK too?
>> 
>> Oh, so you didn't mean me to eliminate the conditional in pci-dma.c, 
>> but just to replace the DMA_BIT_MASK(24) here an elsewhere. Sure, I'm 
>> fine with adding this to the patch.
>
>Well, can ISA_BIT_MASK fall back to DMA_BIT_MASK(32) on !CONFIG_ISA? If 
>we have ISA support disabled we might as well pretend the whole world is 
>PCI, right?
>
>That way we'd get rid of that #ifdef in the .c code too.

I can certainly code it that way, but then we can't use it to replace any
instance of DMA_BIT_MASK(24) used to derive the need for GFP_DMA.
So I'm not sure elimination of which of the instances of DMA_BIT_MASK(24)
is more desirable...

Jan