Re: [PATCH 4/4] Make Emulex lpfc driver legacy I/O port free

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

Christoph Hellwig wrote:
> On Wed, Jun 07, 2006 at 12:15:34PM +0900, Kenji Kaneshige wrote:
> 
>>This patch makes Emulex lpfc driver legacy I/O port free.
> 
> 
> Your interface for this is really horrible ;-)
> 
> 
>>+	int bars = pci_select_bars(pdev, IORESOURCE_MEM);
>> 
>>-	if (pci_enable_device(pdev))
>>+	if (pci_enable_device_bars(pdev, bars))
>> 		goto out;
>>-	if (pci_request_regions(pdev, LPFC_DRIVER_NAME))
>>+	if (pci_request_selected_regions(pdev, bars, LPFC_DRIVER_NAME))
>> 		goto out_disable_device;
> 
> 
> Please make this something like:
> 
> 	if (pci_enable_device_noioport(pdev))
> 		goto out;
> 	if (pci_request_regions(pdev, LPFC_DRIVER_NAME))
> 		goto out_disable_device;
> 
> as in:
> 
>  - get rid of this awkward pci_select_bars function, the pci_enable* function
>    should do all the work and add a flag to struct pci_dev so that all other
>    functions do the right thing.
> 
> 

No. Your idea is very similar to the idea of previous version of my patche
which had a bug. The problem is that it doesn't work if pci_request_regions()
is called before pci_enable_device*() (This is the correct order, though so
many drivers breaks this rule).

Thanks,
Kenji Kaneshige