Re: [RFC: 2.6 patch] m68knommu: remove the broken COMEMPCI code

[Greg Ungerer <gerg@snapgear.com>]

Hi Adrian,

Adrian Bunk wrote:
> On Fri, Aug 29, 2008 at 02:13:24PM +1000, Greg Ungerer wrote:
>> Hi Adrian,
>>
>> Adrian Bunk wrote:
>>> On Mon, Aug 25, 2008 at 04:16:12PM +1000, Greg Ungerer wrote:
>>>> Hi Adrian,
>>>>
>>>> Adrian Bunk wrote:
>>>>> There exists blackfin hardware with PCI support, but the support
>>>>> currently in the kernel fails to build starting with:
>>>>>
>>>>> <--  snip  -->
>>>>>
>>>>> ...
>>>>>   CC      init/main.o
>>>>> In file included from dma-mapping.h:5,
>>>>>                  from dma-mapping.h:52,
>>>>>                  from dmaengine.h:29,
>>>>>                  from skbuff.h:29,
>>>>>                  from netlink.h:155,
>>>>>                  from genetlink.h:4,
>>>>>                  from genetlink.h:4,
>>>>>                  from taskstats_kern.h:12,
>>>>>                  from main.c:46:
>>>>> dma-mapping.h: In function 'dma_supported': ma-mapping.h:24: error: 
>>>>> implicit declaration of function 'pci_dma_supported'
>>>>> ...
>>>>> make[2]: *** [init/main.o] Error 1
>>>> I think we should just remove the underlying comempci access
>>>> code which is the only user of this (arch/m68knommu/kernel/comempci.c).
>>>> Nobody has used it in a very long time. And the hardware itself is
>>>> old, and completely brain-damaged by design.
>>> A patch for removing the comempci code is below.
>>>
>>> It's a bit suspicious that asm/elia.h is now removed since comempci.c  
>>> was the only user. What's the sttus of the eLIA platform?
>> Only ever a handful where made, and that was in 1999.
>> I doubt anyone could ever use the PCI interface on it
>> in any useful way.
> 
> So I can send a patch to remove the platform?

Yeah, I guess we can add in back in of anyone complains.
I don't know of any users of it anymore.


>>> But COMEMPCI fails with a different error - this init/main.o build 
>>> error is present for CONFIG_PCI=y, CONFIG_COMEMPCI=n, so the dependency 
>>> of PCI on BROKEN is still required (unless it gets fixed).
>> Unless I am mistaken there is now no need for the CONFIG_PCI option
>> if selecting m68knommu. The only m68knommu platforms that had PCI
>> as far as I know where those that use the comempci part.
> 
> That implies I can also kill the m68k PCI code since m68knommu will 
> never use it?

Yep.

Regards
Greg


>> The patch looks ok to me too (acked below).
> 
> Thanks.
> 
>> Regards
>> Greg
>> ...
> 
> cu
> Adrian
> 


-- 
------------------------------------------------------------------------
Greg Ungerer  --  Chief Software Dude       EMAIL:     gerg@snapgear.com
SnapGear -- a Secure Computing Company      PHONE:       +61 7 3435 2888
825 Stanley St,                             FAX:         +61 7 3891 3630
Woolloongabba, QLD, 4102, Australia         WEB: http://www.SnapGear.com