12:01 <flowriser> Ok, so making some headway here; So it appears that even though I don't have the CONFIG_QCOM_LMH enabled, the cpufreq driver still tries to init the LMH and gets some IRQs missing

12:04 <flowriser> Trying to forcefully skip the LMH init (by commenting out the code) results in a black screen either way but some tuxes are present on the top of the screen very briefly

17:28 <flowriser> I got to this: https://ibb.co/vsMmTsp then the screen goes black :thinking:

18:20 <flowriser> And it now booted with ACPI

18:23 <flowriser> USB works, plugging an external keyboard works, testing ssh-ing from the outside

21:09 <flowriser> I'll need to wait until next week to test ssh-ing since my router is a bit too far from my workstation, but no reason why it shouldn't work :D

21:11 <flowriser> steev, What exactly are the next steps? So I managed to get the whole thing booting via ACPI, device tree is still so buggy for me and I gather that console does not work (sometimes)

21:12 <flowriser> Should I try to make a working dtb? Or focus somewhere else?

21:12 <steev> flowriser: yes, working dtb, if you want it to get into the kernel sources

21:15 <flowriser> Ok! Now that I'm already down to a linux terminal via ACPI, is there something I can do to make a working dtb easier? From what I can tell, the adreno gpu driver is somehow not working and that is why I get a blank screen when I boot via dtb

21:17 <steev> that i don't know, and can't say - you might want to look at bamse' next tree i pointed you to a few days/weeks back to see how he did the bringup of the flex5g

21:18 <flowriser> yes, using bamse' dtb for flex5g gets me furthest (alas to a blank screen)

21:18 <flowriser> Will keep hacking and see what works :D

21:19 <flowriser> Its been such a learning experience so far and the ARM efforts are amazing

21:21 <robclark> you won't get 3d accel with acpi.. and with dt display will die early (probably when arm-smmu tries to probe) without proper dtb (and *possibly* arm-smmu-qcom.c patch if the compat string has changed)

21:21 <robclark> basically efifb will work thru early boot (because configuration the fw leaves the hw in).. but once real smmu, etc, takes over it will die

21:31 <flowriser> robclark, exactly! It dies as arm-smmu tries to probe

21:31 <robclark> fwiw, see 1a7180ff and aded8c7c

21:32 <robclark> since fw takes smmu out of bypass, some special care is needed.. even then maybe that is not enough to keep display up (but it should come back once drm/msm probes and takes over)

21:33 <robclark> basically you need a lot of stuff working for display to come back successfully ;-)

21:36 <flowriser> robclark, thanks for the explanations!

21:40 <robclark> I guess in some sense the display "breaking" is progress.. you need quite a few pieces in place for display to work (but having just a subset of those pieces will break efifb)

22:09 <flowriser> The only notable difference I could find between the mmu on the acer-spin-7 vs the one on lenovo-flex-5g is that the acer spin 7 one has an extra perf control and an extra power state in the ACPI so I will focus on these differences

22:10 <flowriser> I just diffed the whole dsl files and these stood out

22:13 <robclark> bamse did more of the work on smmu handover.. I guess if smmu compat strings are the same as other sc8180 things that might not be the issue.. ie. maybe the issue is why display doesn't come back up?

22:14 <robclark> ie. it is expected to die.. the unexpected thing is not coming back

22:14 <robclark> likely thing to check is whether you have appropriate panel driver in dt?

22:15 <flowriser> makes sense, but how long does it usually take to see it come back? Hard to tell without any sort of serial what is going on except the screen going black :D

22:16 <flowriser> robclark, good idea about the panel driver! Thanks for your insight

22:17 <robclark> probably useful if you can get to the point of ssh'ing to the thing? But I guess as far as device specific things that might be missing panel driver and corresponding dt bits seems high on the list

22:20 <flowriser> I think I already can ssh into it when booting with ACPI, the problem with this is that it has 2 usb-c ports and I have to use a hub for the network cable; not entirely sure that when I boot it from dt the hub will continue to work

22:21 <robclark> usb stuff is pretty standard(ish).. if you have enough dt for usb to come up and have the right kernel driver for your usb-eth adapter, it should "just work"

22:29 <flowriser> robclark, it does indeed seem that the panels are different: B133HAN05.3 on the lenovo and B140HAN06.2_2A on the acer spin 7

22:31 <flowriser> and it seems bamse added the B14 one as well in panel-simple.yaml :D

22:33 <flowriser> oh wait, it is the other way around ... so if I make the change in the dtb I should be able to see something more I think

22:37 <flowriser> Same driver for both so what I'm missing most likely is the backlight thingy that could be different