On Wed, Oct 16, 2019 at 07:44:51PM +0200, Daniel Vetter wrote:
On Wed, Oct 16, 2019 at 6:13 PM Andy Shevchenko
On Tue, Oct 15, 2019 at 05:57:20PM +0200, Daniel Vetter wrote:
Something like this, as a test patch.
max_transfer_size should be a function. In that case it works.
Why do you want to make it a function? At least from my reading of the code, the dma vs pio decision seems to be done once. So no need to change this at runtime. Changing at runtime would also be a pretty big surprise I think for users of spi.
Yeah, I'd expect it to be a fixed property of the hardware that doesn't vary at runtime though I'm sure there must be some innovation out there which challenges that assumption.
However I'm not sure it's the best approach, thus, Cc to SPI PXA people.
Hm didn't spot the pxa people, added them. Mark, should I just go ahead and bake this into a proper patch for discussion? Or fundamentally wrong approach?
That seems sensible enough, it should certainly fix the immediate issue.