06:25 <krzk> belgin: hi, yeah, looks like that. You actually diagnosed possible issues - haptic registers are left in reset state or power is not supplied by regulator. PWM I think no, it's well tested part of kernel - I am assuming you use mainline kernel.

06:25 <krzk> belgin: I wasn't touching much of max77693 haptic driver, though.

06:28 <krzk> belgin: I would propose to dump all haptic registers before and after suspend to be sure they did not reset (0x00 - 0x10).

06:32 <krzk> belgin: the regulator is ldo26 from PMIC. You can check in kernel whether it is kept enabled after resume. Then check it's registers. In PMIC I2C it will be 0x59 and 0x79

06:36 <krzk> belgin: have in mind that regulator is being turned on when input device is opened. If you close it (for whatever reason in suspend), the regulator will be off.

06:37 <krzk> belgin: if all these look fine (all haptic registers, regulator LDO26 status in kernel, LDO26 registers), then maybe something with PWM... Anyway I don't have a device with max77693 anymore, so I cannot debug here anything.

09:24 <Grimler> belgin, krzk: looks like there might be some differences between pmic regulator config in trats2 and midas. always-on and {on,off}-in-suspend seem to differ for a bunch of ldo:s, for ldo26 it is set as regulator-off-in-suspend in vendor kernel (if I read source correctly), but not in mainline. Maybe that affects haptic driver in this case

09:25 <Grimler> belgin: also, let me know if you want me to test it on i9300 (unless you already have multiple i9300 to test on)

09:40 <krzk> Grimler: Trats2 is almost the same as S3, at least in mainline. What do you refer to when saying trats2 is different than midas?

09:49 <krzk> Grimler: vendor kernel 3.0 did not support DTS and regulator-off-in-suspend. Do you refer to parameters passed to REGULATOR_INIT in midas-power.c?

09:51 <Grimler> krzk: yes, I mean that pmic config in vendor kernel for midas seem to differ a bit from what we have in exynos4412-midas.dtsi (and the pmic config there was added when trats2 support was back in the day). Yeah, downstream passes a bool for always-on and disabled/enabled in state-mem as part of REGULATOR_INIT it seems

09:53 <adjtm> krzk, I think that belgin is using replicant's linux 5.10

09:55 <adjtm> or maybe 5.9

10:03 <krzk> Grimler: I understand. Could be, vendor kernel was often keeping many things always-on because drivers that time did not take reference to regulator. To test if this is the issue here you can change ldo26 to always-on... or always-on plus off-in-suspend or on-in-suspend (although this does not make particular sense... just to try out).

10:04 <krzk> One can also test whether this is really ldo26 by removing reference (usage) in DTS. It should get disabled by core because no one else is using it and then motor should stop working after boot (before suspend even).