On Mon, Oct 21, 2019 at 01:18:47PM +0200, Uwe Kleine-König wrote:
> On Mon, Oct 21, 2019 at 12:58:30PM +0200, Thierry Reding wrote:
> > It turns out that commit 01ccf903edd6 ("pwm: Let pwm_get_state() return
> > the last implemented state") causes backlight failures on a number of
> > boards. The reason is that some of the drivers do not write the full
> > state through to the hardware registers, which means that ->get_state()
> > subsequently does not return the correct state. Consumers which rely on
> > pwm_get_state() returning the current state will therefore get confused
> > and subsequently try to program a bad state.
> > 
> > Before this change can be made, existing drivers need to be more
> > carefully audited and fixed to behave as the framework expects. Until
> > then, keep the original behaviour of returning the software state that
> > was applied rather than reading the state back from hardware.
> 
> I would really prefer to fix that in the framework instead. This is

There's nothing to fix in the framework. The framework isn't broken, the
drivers are.

> something that affects several drivers (cros-ec, imx27, atmel, imx-tpm,
> lpss, meson, sifive, sprd and stm32-lp). This is im my eyes really
> sufficient to justify a framework wide solution instead of adapting
> several drivers in a way that doesn't affect the values programmed to
> hardware.

Can you come up with a proposal for how to want to implement this in the
framework?

Thierry

> > Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
> 
> Reviewed-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
> 
> Thanks
> Uwe
> 
> -- 
> Pengutronix e.K.                           | Uwe Kleine-König            |
> Industrial Linux Solutions                 | http://www.pengutronix.de/  |