Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Sat, Aug 28, 2010 at 06:30:05PM +0200, Marek Vasut wrote:
> Dne So 28. srpna 2010 18:18:58 Sascha Hauer napsal(a):
> > Nope, this has nothing to do with static mappings vs. ioremap. The
> > difference is that read[b,w,l] do little endian accesses suitable for
> > PCI whereas the __raw_* functions do accesses in CPU endianess.
> > Peripherals integrated into a SoC like the SDMA engine here are
> > normally accessible in native endianess and thus need the __raw_*
> > functions. An external network controller (for example a LAN9117) will
> > probably need the non raw functions. Note that 99% of the arm users use
> > little endian only and thus cpu_to_le* is a noop, so both types will work
> > for most people.
> 
> Sorry, you got me here. Why do we have io{read,write} then btw. ? That's for 
> x86's io space, right ?

io{read,write} are for use with ioremap/ioport_map, and allow drivers to
be written which can access registers via MMIO or the PC IO space.

If your driver doesn't support the PC IO space (iow, doesn't use
ioport_map) there's no point using the io{read,write} APIs.

Also note that there's one very big difference between read[bwl] and
__raw_read[bwl].  The former have a barrier to ensure correct ordering
for drivers doing DMA, the latter do not.  So if you use the latter and
you care about data being visible to a DMA agent, you have to ensure
you have proper barriers in place.