Re: [PATCH] pci_set_dac helper

[Jeff Garzik <jgarzik@pobox.com>]

Matthew Wilcox wrote:
> On Wed, Dec 31, 2003 at 11:12:42AM -0500, Jeff Garzik wrote:
> 
>>It seems to me like a lot of drivers wind up getting their 
>>pci_set_dma_mask stuff wrong, occasionally in subtle ways.  So, I 
>>created a "give me 64-bit PCI DMA" helper function.
> 
> 
> I like it, but I think it could be better.  A lot of drivers want
> 64-bit streaming DMA but 32-bit consistent DMA.  So how about this:

I like the direction of your patch:  the driver informs us ahead of time 
what it wants (even though this isn't necessarily true with PCI_DAC_EN, 
which falls back instead of fails).

However, the one big failing of your version is that the driver _must_ 
know if PCI DAC succeeded or not.  Therefore, two pieces of information 
must be returned (error value, DAC flag(s)), which lends itself more to 
leaving my version as-is ;-)


> I note ithat both this and your patch will lead to two errors being
> printed on 64-bit consistent failure; one by tg3 and one by the PCI
> layer; this seems suboptimal.  I suspect you want to do away with the
> error printk in the tg3 driver.

That was intentional in my patch, as it's a warning not an error in my 
pci_set_dac.  In your version I would agree.

	Jeff