[PATCH v4 2/4] spi: Split spi message into max_dma_len size chunks

[kernel at martin.sperl.org]

> On 12.04.2019, at 13:16, Lukas Wunner <lukas at wunner.de> wrote:
> 
> On Fri, Apr 12, 2019 at 12:09:34PM +0100, Mark Brown wrote:
>> On Fri, Apr 12, 2019 at 12:54:48PM +0200, Lukas Wunner wrote:
>>> On Fri, Apr 12, 2019 at 10:47:21AM +0100, Mark Brown wrote:
>>>> I *think* we managed to fix all the architectures to at least stub out
>>>> the DMA interfaces, it's such a pointless thing to have conditional -
>>>> it really only makes a difference for build coverage.
>> 
>>> My point was that the call to spi_split_transfers_maxsize() shouldn't
>>> be called on non-DMA-capable platforms in the first place (because it's
>>> DMA-specific).
>> 
>> It's not a DMA specific problem - there can be SPI controller
>> limitations on transfer sizes too.
> 
> The call does pass in ctlr->max_dma_len though, so is clearly motivated
> by a DMA limitation.
The limitation is in this register: BCM2835_SPI_DLEN

Where the bcm2835-SDK (Brcm_Android_ICS_Graphics_Stack.tar.gz)defines:
bit 0-15 for DLEN

See transformed data at:
https://github.com/msperl/rpi-registers/blob/master/md/Region_SPI.md

Or the Arm Periperial Document on page 156 with regards to DLEN register.

This is where the DMA length limit comes from - so it is NOT
really DMA block related but SPI block related.

More specifically to DMA requests are only triggered when this counter
is > 0.

That is unless the Documentation has another errata and more bits are
actually allowed.

Martin