Re: dma_sync_*_for_cpu and direction=TO_DEVICE (was Re: [PATCH 02/20] dma-mapping: provide a generic dma-noncoherent implementation)

[Vineet Gupta <Vineet.Gupta1@synopsys.com>]

On 05/18/2018 10:50 AM, Russell King - ARM Linux wrote:
> On Fri, May 18, 2018 at 10:20:02AM -0700, Vineet Gupta wrote:
>> I never understood the need for this direction. And if memory serves me
>> right, at that time I was seeing twice the amount of cache flushing !
> It's necessary.  Take a moment to think carefully about this:
>
> 	dma_map_single(, dir)
>
> 	dma_sync_single_for_cpu(, dir)
>
> 	dma_sync_single_for_device(, dir)
>
> 	dma_unmap_single(, dir)

As an aside, do these imply a state machine of sorts - does a driver needs to 
always call map_single first ?

My original point of contention/confusion is the specific combinations of API and 
direction, specifically for_cpu(TO_DEV) and for_device(TO_CPU)

Semantically what does dma_sync_single_for_cpu(TO_DEV) even imply for a non dma 
coherent arch.
Your tables below have "none" for both, implying it is unlikely to be a real 
combination (for ARM and ARC atleast).

The other case, actually @dir TO_CPU, independent of for_{cpu, device}  implies 
driver intends to touch it after the call, so it would invalidate any stray lines, 
unconditionally (and not just for speculative prefetch case).


> In the case of a DMA-incoherent architecture, the operations done at each
> stage depend on the direction argument:
>
> 	map		for_cpu		for_device	unmap
> TO_DEV	writeback	none		writeback	none
> TO_CPU	invalidate	invalidate*	invalidate	invalidate*
> BIDIR	writeback	invalidate	writeback	invalidate
>
> * - only necessary if the CPU speculatively prefetches.
>
> The multiple invalidations for the TO_CPU case handles different
> conditions that can result in data corruption, and for some CPUs, all
> four are necessary.

Can you please explain in some more detail, TO_CPU row, why invalidate is 
conditional sometimes.