Re: [PATCHv3 2/3] drivers: spi: Add qspi flash controller

[Nishanth Menon <nm@ti.com>]

On 07/09/2013 05:05 AM, Mark Brown wrote:
> On Tue, Jul 09, 2013 at 09:51:43AM +0300, Felipe Balbi wrote:
>
>>> Looks like a case to use regmap?
>>> Dumb q: why cant we use regmap_spi? worst case, you should be able to
>
>> read regmap-spi and you'll see why it can't be used in this case.
>
>> regmap-spi is for SPI clients who want to read their register map
>> through SPI commands. This is a driver for the SPI master which has its
>> registers memory mapped.
>
> Indeed, regmap-spi would be a client of this driver.  Though there is
> regmap-mmio which may be helpful, on the other hand it's relatively
> heavyweight and SPI can be a bit performance sensitive so perhaps it's
> not awesome here.  Russell did mention the idea of some helpers along a
> similar style though, they may be a good idea but nobody wrote them yet.
>
Fair enough. I am starting to see at least a second driver (internally 
for DRA7 IRQ/DMA crossbar) doing the same handling for different bit 
sized registers. maybe now is the time for us to create a lighter 
version of mmio. mmio at least the last I looked uses spinlocks and is 
fairly fast. But, I agree, when the client of regmap uses locks to 
protect operations of their own, regmap might be a bit of an overhead. 
that said, replicating regmap logic in the driver looks a bit weird and 
might allow the basis for other drivers to do the same as well as some 
users of mmio would consider themselves performance sensitive :(.

-- 
Regards,
Nishanth Menon