On Fri, Feb 12, 2021 at 06:35:30PM +0200, Andy Shevchenko wrote:

Hi, Mika!

There is an interesting discussion [1] going on about necessity of the
_irqsave/_irqrestore variants of spin lock in the IRQ handler.

It appears that in our driver(s) we have also such variants of spin locks in
use. Do you have any idea why is it so?

Okay, this is me, who did it in a copy'n'paste manner in the commit
e64fbfa51e8f ("pinctrl: intel: Protect IO in few call backs by lock").

A bonus question, why do we use _NO_THREAD flag explicitly as per Thomas [2]
this won't work well with RT kernels?

Okay, this part is explained in the commit 1a7d1cb81eb2 ("pinctrl: intel:
Prevent force threading of the interrupt handler").

[1]: https://lore.kernel.org/linux-gpio/1612774577-55943-1-git-send-email-luojiaxing@huawei.com/T/#u (local)
[2]: https://lore.kernel.org/linux-i2c/87zh2s8buh.fsf@nanos.tec.linutronix.de/ (local)

-- 
With Best Regards,
Andy Shevchenko