06:22 <aka_[m]> Marijn: not sure but is amount of vblanks fired main issue there?

06:22 <aka_[m]> I don't see them in that amount on 7125

10:19 <Marijn[m]> aka_: Where do you see an "amount" of vblanks being fired?

10:20 <Marijn[m]> Keep in mind that there is typically a vsync line going from the panel to the board, letting the DPU (the pingpong block, AFAIR) know when it is ready for the next frame / finished presenting the current frame

10:20 <Marijn[m]> Make sure you're turning that on with a DCS command 😬

13:31 <aka_[m]> <Marijn[m]> "aka_: Where do you see an "..." <- it keeps saying "increase count"

13:31 <aka_[m]> up to 186 or something

13:34 <aka_[m]> i have changed dpu caps because somehow i used qcm2290 linewidth which says 2160 and panel i use is 2340 and limits for vig(?) says 2560

13:34 <aka_[m]> tho not sure about mixer because downstream says 2048 for both qcm2290/bengal but qcm2290 has it defined as 2560

13:39 <aka_[m]> https://pastebin.com/rUsT2ddm

13:40 <aka_[m]> [drm:drm_update_vblank_count] updating vblank count on crtc 0: current=95, diff=0, hw=94 hw_last=94

15:04 <aka_[m]> Marijn: have you tried bringing a610 on trinket yet?

15:04 <Marijn[m]> We have DPU but I don't remember if we also started on the GPU

15:05 <aka_[m]> i tried on bengal and it crash with -19, some dts mention it uses zap shader however no region to be found

15:05 <aka_[m]> entry point says 0x5000

17:49 <aka_[m]> Any idea how to hit on a6xx gpu without GMU?

17:49 <aka_[m]> I added node but it appears to reboot

17:49 <aka_[m]> pstore is blank

17:53 <Mis012[m]> imho if gpu is needed for display out, that's a bug

18:31 <aka_[m]> so i was missing power-domains for GPU but even with them it reboot

18:33 <Mis012[m]> sa

18:33 <Mis012[m]> d

18:44 <aka_[m]> at first -19 so i think it was because i was lacking adreno smmu

18:44 <aka_[m]> so it was crying ENODEV

18:46 <aka_[m]> but cannot really find where it request iommu

18:54 <aka_[m]> Mis012: so i commented iommus and it fails but not reboot

18:59 <Mis012[m]> does display care?

18:59 <aka_[m]> robclark: do a6xx_gpu.c by any chance assume every a6xx has LLCC?

19:00 <aka_[m]> https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/msm/adreno/a6xx_gpu.c#L1896

19:00 <aka_[m]> There is bengal without LLCC

19:03 <robclark> it at least hasn't been tested on any device w/out LLCC that I am aware of

19:41 <aka_[m]> quite sad

19:41 <aka_[m]> bengal/trinket/khaje are ones without llcc

19:42 <aka_[m]> 662/665/680

19:49 <robclark> patches welcome ;-)

20:19 <robclark> aka_[m]: fwiw, we had a6xx running for a quite a while before llcc was supported upstream.. so if it doesn't work on devices without llcc then it should only be trivial fixes

20:19 <aka_[m]> im not exactly up to the task xD

20:19 <aka_[m]> i know somewhere after it probably goes for iommu it die

20:20 <aka_[m]> a610 has no gmu so maybe it needs some special care

20:21 <aka_[m]> im still quite stuck with blank output from dpu...

20:22 <robclark> not having gmu is probably a much more significant challenge

20:23 <robclark> gpu doesn't even try to load until you have rootfs (and it can load it's firmware files).. so you are probably not even getting as far as gpu stuff yet

20:23 <aka_[m]> it just do something then goes straight to cleanup after printing revision

20:24 <robclark> with bootloader leaving the display enabled, there are some workarounds needed on arm-smmu-qcom.c .. basically you'll need to add your device to a list of compatible strings in there to tell arm-smmu to not fiddle with things and break display

20:24 <robclark> if it gets to point of trying to load gpu, probably bails if no gmu device? The no-gmu config is completely unsupported

20:25 <aka_[m]> just cleanup and nothing more in log

20:25 <aka_[m]> after i added iommu it reboots

20:26 <aka_[m]> otherwise it was doing ENODEV

20:27 <aka_[m]> sm6115 dpu is very similar to qcm2290 but im not sure if it works just like that without additonal patches

20:27 <aka_[m]> on mdss front i added compatible "qcom,sm6115-mdss"

20:27 <aka_[m]> logs says it does some stuff but as long as it not show anything its not really anything worth mentioning

20:28 <aka_[m]> sm6115 dpu is very similar to qcm2290 but im not sure if it works just like that without additonal patches

20:28 <aka_[m]> * on mdss front i added compatible "qcom,sm6115-mdss" to arm-smmu-qcom

20:31 <robclark> I think you probably also need to update qcom_smmu_impl_of_match[] ... maybe that isn't needed yet

20:31 <robclark> defn need display compat in qcom_smmu_client_of_match

20:35 <aka_[m]> https://github.com/Kiciuk/linux/blob/rc6/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c#L415

20:35 <aka_[m]> already added

20:36 <aka_[m]> im bit too dumb to read drm.debug logs but its something like that:

20:36 <aka_[m]> https://pastebin.com/rUsT2ddm