soc: imx: gpcv2: regulator issue with system suspend on imx8mq

[Martin Kepplinger]

hi Lucas and all interested in suspend to ram on imx8mq,

This is slighly repeating my previous observations that still apply,
but summarizing the situation:

s2idle should work on mainline when looking at the implementations of
platform drivers. With the missing bits
https://source.codeaurora.org/external/imx/linux-imx/commit/?h=imx_5.10.35_2.0.0_imx8ulp_er&id=ab850d655c22df562c27c9d6775a26b6df6865b5
and
https://lore.kernel.org/linux-arm-kernel/1631554694-9599-7-git-send-email-abel.vesa@xxxxxxx/
suspend to ram should work too,
and it does for me, except when a power domain is using a board-
regulator as power-supply that is not always-on, but controlled by a
driver. (when I describe these as "always-on", things are fine (except
for unrelated edgecases)) here's the example I'm running where I don't
describe "buck3" as always-on, but etnaviv runtime pm is controlling
it:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/freescale/imx8mq-librem5.dtsi#n1161
When starting to resume, the following happens:

[  139.985440] Enabling non-boot CPUs ...
[  139.990363] Detected VIPT I-cache on CPU1
[  139.990413] GICv3: CPU1: found redistributor 1 region
0:0x00000000388a0000
[  139.990487] CPU1: Booted secondary processor 0x0000000001
[0x410fd034]
[  139.991445] CPU1 is up
[  140.011836] Detected VIPT I-cache on CPU2
[  140.011852] GICv3: CPU2: found redistributor 2 region
0:0x00000000388c0000
[  140.011876] CPU2: Booted secondary processor 0x0000000002
[0x410fd034]
[  140.012284] CPU2 is up
[  140.032739] Detected VIPT I-cache on CPU3
[  140.032756] GICv3: CPU3: found redistributor 3 region
0:0x00000000388e0000
[  140.032780] CPU3: Booted secondary processor 0x0000000003
[0x410fd034]
[  140.033310] CPU3 is up
[  140.158791] imx-pgc imx-pgc-domain.5: failed to enable regulator: -
110

trying runtime-resume in system-suspend for i2c busses didn't help me
here for example. What's your idea for solving this? (regulator always-
on is not an acceptable workaround :) I'm always happy to test concrete
ideas and would like to know from anyone who uses system suspend on
imx8mq.

history
-------
last time this came to my attention via the mainling lists was the VPU
addition:
https://lore.kernel.org/linux-arm-kernel/8ed3a28d59b442b531e68e95d83b187bb3392940.camel@xxxxxxx/
but for the above logs and all current tests, I ignore the VPU (set the
power-supply always-on) simply because the the driver is in staging and
seems to create a different problem when suspending, and the GPU power-
supply example is very well suited to highlight the problem.

but before that, "gpcv2: support systemd suspend/resume"
( https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=da4112230f86
) didn't work for me, see:
https://lore.kernel.org/all/a20ecd639f8e8b7fa4a9bed7a8e9590225262784.camel@xxxxxxx/

thanks a lot,

                               martin