Hi Uwe,
Le mardi 17 janvier 2023 à 22:35 +0100, Uwe Kleine-König a écrit :
Hello Paul,
On Fri, Nov 18, 2022 at 09:55:40AM +0000, Paul Cercueil wrote:
Le jeu. 17 nov. 2022 à 14:29:27 +0100, Uwe Kleine-König u.kleine-koenig@pengutronix.de a écrit :
Hello Paul,
On Tue, Oct 25, 2022 at 11:02:00AM +0100, Paul Cercueil wrote:
Le mar. 25 oct. 2022 à 08:21:29 +0200, Uwe Kleine-König u.kleine-koenig@pengutronix.de a écrit : > Hello, > > On Mon, Oct 24, 2022 at 09:52:09PM +0100, Paul Cercueil wrote: > > The "duty > cycle" trick to force the pin level of a disabled TCU2 > > channel would only work when the channel had been enabled > > previously. > > > > Address this issue by enabling the PWM mode in jz4740_pwm_disable > > (I know, right) so that the "duty > cycle" trick works before > > disabling > > the PWM channel right after. > > > > This issue went unnoticed, as the PWM pins on the majority of the > > boards > > tested would default to the inactive level once the corresponding > > TCU > > clock was enabled, so the first call to jz4740_pwm_disable() would > > not > > actually change the pin levels. > > > > On the GCW Zero however, the PWM pin for the backlight (PWM1, which > > is > > a TCU2 channel) goes active as soon as the timer1 clock is enabled. > > Since the jz4740_pwm_disable() function did not work on channels not > > previously enabled, the backlight would shine at full brightness > > from > > the moment the backlight driver would probe, until the backlight > > driver > > tried to *enable* the PWM output. > > > > With this fix, the PWM pins will be forced inactive as soon as > > jz4740_pwm_apply() is called (and might be reconfigured to active if > > dictated by the pwm_state). This means that there is still a tiny > > time > > frame between the .request() and .apply() callbacks where the PWM > > pin > > might be active. Sadly, there is no way to fix this issue: it is > > impossible to write a PWM channel's registers if the corresponding > > clock > > is not enabled, and enabling the clock is what causes the PWM pin > > to go > > active. > > > > There is a workaround, though, which complements this fix: simply > > starting the backlight driver (or any PWM client driver) with a > > "init" > > pinctrl state that sets the pin as an inactive GPIO. Once the > > driver is > > probed and the pinctrl state switches to "default", the regular PWM > > pin > > configuration can be used as it will be properly driven. > > > > Fixes: c2693514a0a1 ("pwm: jz4740: Obtain regmap from parent node") > > Signed-off-by: Paul Cercueil paul@crapouillou.net > > Cc: stable@vger.kernel.org > > OK, understood the issue. I think there is another similar issue: The > clk is get and enabled only in the .request() callback. The result is (I > think---depends on a few further conditions) that if you have the > backlight driver as a module and the bootloader enables the backlight to > show a splash screen, the backlight goes off because of the > clk_disable_unused initcall.
I will have to verify, but I'm pretty sure disabling the clock doesn't change the pin level back to inactive.
Given that you set the clk's rate depending on the period to apply, I'd claim that you need to keep the clk on. Maybe it doesn't hurt, because another component of the system keeps the clk running, but it's wrong anyhow. Assumptions like these tend to break on new chip revisions.
If the backlight driver is a module then it will probe before the clk_disable_unused initcall, unless something is really wrong.
I'd claim the clk_disable_unused initcall is called before userspace starts and so before the module can be loaded. Who is wrong here?
Probably me.
So the backlight would stay ON if it was enabled by the bootloader, unless the DTB decides it doesn't have to be.
Don't understand that. How could hte DTB decide the backlight can be disabled?
I don't remember what I meant by that :)
Anyway, I can try your suggestion, and move the trick to force- disable PWM pins in the probe(). After that, the clocks can be safely disabled, so I can disable them (for the disabled PWMs) at the end of the probe and re-enable them again in their respective .request() callback.
I really lost track of the problem here and would appreciate a new submission of the remaining (and improved?) patches.
Sure. I still have the patchset on the backburner and plan to (eventually) send an updated version.
If you are fishing for patches I think you can take patches 3/5 and 4/5 of this patchset. Then I won't have to send them again in v2.
Cheers, -Paul