[PATCH] spi: stm32: use runtime PM to enable/disable controller

[Ben Wolsieffer]

On Fri, Dec 15, 2023 at 07:27:39PM +0100, Alain Volmat wrote:
> Hi,
> 
> sorry for the delay.
> 
> On Thu, Dec 14, 2023 at 10:58:54AM +0000, Mark Brown wrote:
> > On Mon, Dec 04, 2023 at 03:20:55PM -0500, Ben Wolsieffer wrote:
> > > Instead of disabling the SPI controller between each message, do it
> > > as part of runtime PM.
> > 
> > This doesn't apply against current code, please check and resend.
> 
> I rapidly gave a try on this patch on top of the spi/for-next branch
> (manually fixing the conflict due to the MASTER->HOST renaming).
> It turns out that with that applied, transfers on the MP13
> (compatible: st,stm32h7-spi) are not working anymore while simply
> removing it back it works again.
> (test is simply doing loopback spidev_test)

That's unfortunate; I was worried about something like this because I
only have an STM32F7 to test. If you can't easily determine what's going
wrong, it would be interesting to know if the original version of this
patch has the same problem:
https://lore.kernel.org/lkml/ZWpoKEcM0ZeYAsBa@dell-precision-5540/T/

> 
> spi mode: 0x0
> bits per word: 8
> max speed: 500000 Hz (500 kHz)
> TX | 8D D6 73 8B 9D 8B 1C 7D 8D 80 EC 32 F9 0D BA AD 9F 88 A5 9B 3F AA 48 8C 21 35 0D C1 C8 E5 6A 81  |..s....}...2........?.H.!5....j.|
> RX | 8D 00 00 00 D6 00 73 00 8B 00 00 00 9D 00 00 8B 1C 00 00 00 7D 00 00 8D F9 00 00 00 BA 00 00 00  |......s.............}...........|
> 
> The RX data contains lots of 00 between each byte.  Moreover it seems
> that with this patch applied non-dma transfer (when there is no dmas
> properties within the node) are now failing.
> 
> I'll check that and give more details but could you avoid applying this
> patch for the time being ?
> 
> Thanks.
> Alain