Re: [BUG](-mm)pci_disable_device function clear bars_enabled element

[Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>]

Grant Grundler wrote:
> On Fri, Jun 02, 2006 at 11:57:36AM +0900, Kenji Kaneshige wrote:
> ...
> 
>>As Rajesh pointed out, there are many drivers which initialize the
>>device with the wrong order. They should be fixed.
> 
> 
> Then you also agree with the patch to pci.txt?

Yes. I agree with you.

>>I would like to
>>confirm the correct order to initialize the device again. Is the
>>following correct order?
>>
>>   (1) pci_request_regions()
>>   (2) pci_enable_device()
>>   (3) request_irq()
>>   (4) free_irq()
>>   (5) pci_disable_device()
>>   (6) pci_release_regions()
> 
> 
> Yes, that's what I would prefer and would like to see reccomended.
> Would you like to see that order listed (like you have above)
> in the pci.txt file?
> 

I think that list would be very useful. But as you said, there are
other steps remaining than ones I came up with at once. I can't
deal with the steps of all of them...

BTW, Section 3 says "Before you do anything with the device you've
found, you need to enable it by calling pci_enable_device()...". I
think it would be one of the causes of misunderstanding the order
between pci_request_regions() and pci_enable_device().

Thanks,
Kenji Kaneshige


> A less precise list is in the first section of Documentation/pci.txt.
> 
> [ TODO: Can someone define which kernel versions implement "new style"? ]
> 
> There's more to this list unfortunately:
> 	DMA mask settings, MSI support, power state
> 
> And probably a few more that I'm not thinking of right now.
> 
> Restructing the document to list the steps, indicate which
> are optional, and describe each step in order is more than
> I can deal with right now.  Section 3 and 5 cover most of
> the material but aren't as clear as Kenji's list.
> 
> thanks,
> grant
>