Re: [PATCH 3/8] parisc: implement dma_mmap_coherent()

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Fri, 2009-07-10 at 19:16 +0100, Russell King wrote:
> On Fri, Jul 10, 2009 at 03:11:29PM +0000, James Bottomley wrote:
> > The design of coherent memory was for memory based device mailboxes
> > managed by the kernel ... trying to give userspace coherent access to
> > the same mailbox is problematic because it gives a direct way for the
> > process to interfere with a device function ... shouldn't whatever
> > you're trying to do be better accomplished by using an API to control
> > the device and keeping the coherent mailbox fully in the kernel address
> > space?
> 
> As far as sound DMA goes, it's not about mailboxes.  It's about a circular
> buffer which you want the device to DMA from direct to/from the DAC/ADC
> and have the application write/read data directly to/from that same
> buffer.

But that makes it sound like ordinary streaming DMA from a device to
user space ... that's what the dma_map_xx APIs are already designed to
handle:  I don't understand why you need coherent memory for this (which
can be a scarce resource on some platforms).

> Without this, you end up having to copy the sound data - at something
> around 200KB/s from applications into a driver managed buffer, which is
> quite an unnecessary overhead for the CPU.

Why?  The dma_map_xx API is designed to be zero copy.

James