Hello,

On Thu, Feb 01, 2024 at 06:25:26PM +0200, Viorel Suman wrote:
> On 2/1/24 15:38, Uwe Kleine-König wrote:
> > > +	/* Resets all internal logic and registers */
> > > +	writel(PWM_IMX_TPM_GLOBAL_RST, tpm->base + PWM_IMX_TPM_GLOBAL);
> > > +	writel(0, tpm->base + PWM_IMX_TPM_GLOBAL);
> > > +
> > This opposes the use case that the bootloader setup the pwm-backlight to
> > show a splash screen that is simply taken over by Linux without
> > flickering, right?
> 
> Yes, I was not aware of such use case. Is it acceptable if I'll update
> 
> the patch in a such way so that the software reset happens as function
> 
> of a property in the related DTS node ?

That sounds wrong. Why do you reset at all? If at all, only reset if all
channels are disabled (if at all).

> > Otherwise the commit log should motivate why "the module takes the
> > default state" is better than the status quo and what this default state
> > is.
> 
> The default state above means IP internal logic being reset to the initial
> state and
> 
> registers values being set to their reset values. We're facing a situation
> on iMX95 when
> 
> A core may be reset independently from the rest of the SoC, this triggers a
> new
> 
> SPL->U-Boot->Linux boot, and in Linux probe phase PWM will inherit its state
> from
> 
> previous Linux runtime - this leads to some issues in suspend/resume
> functionality.

What kind of issues?

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |