Re: [PATCH 1/20] libata: kill ata_host_stop()

[Tejun Heo <htejun@gmail.com>]

Jeff Garzik wrote:
> Tejun Heo wrote:
>> As all the current LLDs are PCI, ata_pci_host_stop() should be used,
>> which BTW is effectively identical to ata_host_stop().  Convert all
>> LLDs to use ata_pci_host_stop() and kill ata_host_stop().
> 
> NAK, they are not equivalent replacements:
> 
> * ata_host_stop checks mmio_base before calling iounmap.  Most of the 
> drivers you change do not use ioremap (thus it's really a no-op call for 
> them)
> 
> * ata_pci_host_stop not only unmaps unconditionally, it calls 
> pci_iounmap().  Only drivers which have previously used pci_iomap() 
> should use this helper.
> 
> This patch may make sense later in the patch series, when 
> ata_pci_host_stop presumably works for all iomap-converted drivers, but 
> not really as the first patch?

It's the first patch because I wanted to free the function name 
ata_host_stop(), which will be used for another purpose in new init 
model.  As the new use isn't a port op, there shouldn't be too much 
confusion.

Hmmm.. I'll push this backward and use different name for it.

Thanks.

-- 
tejun