add panels for sony-aries

[TheOrangeCat]

haven't yet checked if these work.

[z3ntu]

Replace {config,dtb} with config/dtb in the commit message, apart from that looks reasonable.

I'd like if you actually test the panel before we merge though, there's enough panels that need special care and don't just work in the auto-generated way.

[TheOrangeCat]

I'll try to test soon, probably with dumb-dcs as that seems like it could be needed.

[TheOrangeCat]

Sorry for not updating, it seems the panel driver somewhat works, but the backlight is weird, at least on phosh small adjustments to the level in the UI seem to completely randomly change the actual level. However, I do need to retest as this was back on kernel 6.9. I'm not sure when I'll have the time though.

[z3ntu]

No worries!

[TheOrangeCat]

Hmm, the kernel doesn't want to build with the generated panel.

drivers/gpu/drm/panel/msm8974-generated/panel-sony-novatek-jdi-720p-cmd.c: In function 'novatek_jdi_probe':
drivers/gpu/drm/panel/msm8974-generated/panel-sony-novatek-jdi-720p-cmd.c:125:15: error: implicit declaration of function 'devm_drm_panel_alloc'; did you mean 'devm_drm_panel_add_follower'? [-Wimplicit-function-declaration]
  125 |         ctx = devm_drm_panel_alloc(dev, struct novatek_jdi, panel,
      |               ^~~~~~~~~~~~~~~~~~~~
      |               devm_drm_panel_add_follower
drivers/gpu/drm/panel/msm8974-generated/panel-sony-novatek-jdi-720p-cmd.c:125:41: error: expected expression before 'struct'
  125 |         ctx = devm_drm_panel_alloc(dev, struct novatek_jdi, panel,
      |                                         ^~~~~~
drivers/gpu/drm/panel/msm8974-generated/panel-sony-novatek-jdi-720p-cmd.c: At top level:
drivers/gpu/drm/panel/msm8974-generated/panel-sony-novatek-jdi-720p-cmd.c:113:37: warning: 'novatek_jdi_panel_funcs' defined but not used [-Wunused-const-variable=]
  113 | static const struct drm_panel_funcs novatek_jdi_panel_funcs = {
      |                                     ^~~~~~~~~~~~~~~~~~~~~~~

This seems to be a consequence of msm8916-mainline/linux-mdss-dsi-panel-driver-generator commit c917fcc, but I'd guess that maybe it'd work on a newer kernel version maybe?

[z3ntu]

Go to an older generator commit then and regenerate. The msm8974-mainline/linux kernel hasn't been updated in a while unfortunately.

[TheOrangeCat]

Well, it compiled, but the panel doesn't seem to be working for some reason
dmesg.txt
From dmesg it seems it's a problem with the GPU itself not starting up? But I'm not sure
Edit: Actually, it seems I forgot to add the GPU firmware, which could explain the GPU not starting up :P

[z3ntu]

Please add cma=256m msm.vram=192m msm.allow_vram_carveout=1 to your kernel cmdline

[TheOrangeCat]

It works! Although only half of the backlight is controlled by the system for some reason, the other is constantly lit, and it still has the weird behavior when changing, but I think it is probably related to something in the DT

[z3ntu]

For backlight, double check your pm8941_wled config, iirc especially qcom,num-strings has an effect like that.

[TheOrangeCat]

I do remember playing around with that 2 years ago and I may have got it working, but theoretically the current config should have been fine as it matches the downstream dts. I'll maybe try removing num-strings completely. In terms of brightness adjustment I have to say it almost looks like the value gets truncated, as it flashes cyclically and full brightness seems to be full

[TheOrangeCat]

Well, completely removing num-strings produces a rather interesting result - now both halves change brightness, but the one that didn't adjust now changes smoothly and the one that flickered still flickers. I really don't know what the cause could be as downstream has num-strings=<2> and seemingly no other parameters that could change what strings are used

[z3ntu]

The values <1> and <3> are also valid fwiw, alternatively also qcom,enabled-strings is a valid property that may change some things there (but you can't use it at the same time as qcom,num-strings.