00:29 <Viciouss> I'm currently trying to get lima running on my note 10.1 (exynos4) with android 10 but drm_hwc throws the following error: failed to import prime fd 13 ret=-1, I'm using a 5.4 kernel with libdrm master, mesa 20.0 and drm_hwcomposer master and I have set gralloc and hwc to use card0. Could it be an issue that I get the error 'Failed to set DPMS property for connector 56' some time before the other one?

00:31 <anarsoul|2> no idea, you'll have to debug it

00:33 <Viciouss> with buildroot and kmscube everything looked ok but getting it to run in android is a real pain

00:34 <anarsoul|2> so debug what's going on with android?

00:35 <MoeIcenowy> anarsoul|2: no

00:35 <MoeIcenowy> Viciouss: what's card0?

00:35 <MoeIcenowy> is card0 lima or exynos drm?

00:35 <Viciouss> the exynos drm

00:38 <MoeIcenowy> I think it's not related to DPMS

00:39 <anarsoul|2> MoeIcenowy: I'll grab it then

00:39 <Viciouss> I have rather limited experience with C and never debugged before so this would be interesting :-)

00:39 <MoeIcenowy> BTW please recheck what device card0 is

00:39 <MoeIcenowy> by checking sysfs

00:39 <MoeIcenowy> on my PineTab card0 is lima and card1 is sun4i-drm

00:40 <Viciouss> ok let me see

00:43 <MoeIcenowy> Viciouss: BTW you are using gralloc.gbm, right?

00:44 <Viciouss> card0 is exynos, card1 is lima for me

00:44 <Viciouss> indeed gralloc.gbm that is

00:45 <MoeIcenowy> oh maybe you need some kernel permission hack

00:45 <MoeIcenowy> https://github.com/glodroid/glodroid_forks/commit/2663eac9fe9c021f028fb29e36afe3954ebad44b

00:46 <Viciouss> could it be related to my display driver? I only have a gpio backlight node which is sufficient to get the display powered up, the rest is handled by the samsung fimd driver it seems

00:47 <MoeIcenowy> check the hack I pointed out first

00:47 <Viciouss> ok I'll do so, thanks

00:48 <MoeIcenowy> Try to apply it on top of your kernel

01:05 <Viciouss> the prime fd error message is gone but now I get the following: Commit test failed for display 0, FIXME

01:19 <Viciouss> the return value of the CommitFrame call in drmdisplaycompositor.cpp is -22, there is none of the other log messages though so no idea where this is coming from

01:23 <Viciouss> it's the same error message though as if I set the configs to card1

08:28 <anarsoul|2> rellla: can you review https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4051 please?

17:42 <anarsoul|2> we really need to print constant buffer for gpir :\

17:43 <anarsoul|2> otherwise it's pretty hard to read disassembly

18:31 <anarsoul|2> hm

18:32 <anarsoul|2> I have strong suspicion that we have only 12 regs in GP

18:32 <anarsoul|2> whenever shader uses $12 it fails

18:41 <anarsoul|2> hm, probably just a coincidence

21:10 <Viciouss> I did some more digging and found that the issue for the Commit test failed error message is in the kernel in drm_atomic_uapi.c in the call to drm_mode_atomic_ioctl, didn't have the time to go deeper there. the other thing I found was this trace in the kernel log: https://vomitb.in/nBNebi1yBV

22:31 <anarsoul|2> Viciouss: regarding last backtrace -- you're likely using 5.4 kernel and it has broken recovery from gpu errors

22:31 <anarsoul|2> switch to 5.5

22:31 <anarsoul|2> that's not lima bug, it's bug in drm_scheduler

22:43 <Viciouss> it doesn't look like this is related to my other issue, is that correct?

22:44 <anarsoul|2> yeah

22:47 <anarsoul|2> rellla: can you grep your new dump with disassembly for $12, $13, $14, $15?

22:52 <Viciouss> ok thanks, my goal currently is to get anything displayed on the lcd before updating the kernel any further, I'm stuck on an empty boot screen for quite some time now and I need a visual success for the motivational part

22:55 <anarsoul|2> keep in mind that lima is render-only gpu, it's up to your display driver to display stuff

22:56 <Viciouss> well I had it running on buildroot, so this part works