Re: [PATCH 2/3] spi: pxa2xx: Fix controller unregister order

[Tsuchiya Yuto <kitakar@xxxxxxxxx>]

On 5/28/20 6:31 PM, Lukas Wunner wrote:
> On Thu, May 28, 2020 at 11:41:21AM +0300, Andy Shevchenko wrote:
>> Thank you very much for testing, I will figure out what can be done
>> more there, but it's minor now.
>> For input and touchscreen I guess you may ask Dmitry (input subsystem
>> maintainer) and Benjamin (HID, but he might have an idea as well).
>
> This might not be an input issue, perhaps the spi-pxa2xx.c driver
> cannot cope with s2idle on this particular platform.
>
> E.g., pxa2xx_spi_suspend() zeroes the SSCR0 register.  It seems this
> disables or resets the controller.  But pxa2xx_spi_resume() isn't
> touching the register at all.  Maybe the register contains crap when
> coming out of s2idle, so needs to be set to a sane value on resume?

Thanks everyone. I later found that touch input (surface3_spi) crashing
is reproducible by just putting display off/on on recent kernels. So,
this is rather not related to s2idle. Also it seems that runtime_pm is
not related, too.

> Tsuchiya Yuto says that reloading the SPI controller driver makes
> the touch driver work again, so I'd check what's done on ->remove()
> and ->probe() both in the touch driver as well as in the SPI controller
> driver that fixes the problem.  The SSCR0 register is zeroed on
> ->remove() and re-initialized on ->probe(), so that register may
> indeed play a role.

I looked into what is done on probe()/remove() in the SPI controller
and it seems that release/setup DMA helps to get surface3_spi driver
working again with DMA mode.

I added details to that bugzilla [1], since this crashing itself is not
relevant to this series.

[1] https://bugzilla.kernel.org/show_bug.cgi?id=206403#c4