02:09 <alyssa> looks like you've found yourself a graphics driver bug

03:03 <steev> :(

03:41 <steev> alyssa: fwiw, seems to only happen on the radxa zero, which i believe is a mali g31 (amlogic s905y or something), doesn't happen on e.g. the pinebook pro

05:05 <flowriser> steev, how would I write a dts?

05:08 <flowriser> Because as far as I can tell, all the little bits of drivers are covered so I just need something to unify them like the lenovo flex 5 dts

05:14 <flowriser> Oh, it is documented in the kernel: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/usage-model.txt?h=v5.2-rc5

05:15 <flowriser> Well, I've made it this far; no turning back now

06:17 <steev> keep in mind that not everything needed for sc8180x is in the kernel yet either

06:18 <flowriser> But I suppose somebody managed to boot the kernel at least xD

06:31 <steev> there is bamse's tree that i pointed to the other day, that has the most complete support, but some timings and such may be off for your device, but you could take that tree

08:16 <flowriser> gonna give a try to 5.16 too

12:35 <steev> 5.16 won't have it

12:36 <steev> https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/arch/arm64/boot/dts/qcom?h=linux-5.16.y notice how there is no sc8180x in there

12:37 <steev> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/arch/arm64/boot/dts/qcom nor is it in next :(

12:52 <HdkR> I should really just take my tweet down with me booting Linux on the ProX. It's just misleading.

13:12 <steev> it was true at the time

13:26 <alyssa> steev: mali-g31 and mali-t860 (pinebook pro) are different architectures

13:26 <alyssa> could easily be e.g. a compiler bug

13:26 <steev> alyssa: ah, good point

13:26 <alyssa> Can you install the latest Mesa (git clone https://gitlab.freedesktop.org/mesa/mesa) and reproduce there?

13:27 <steev> i can try latest mesa, i'm on 21.3.3 iirc

13:27 <alyssa> Thanks... 21.3 "should" be new enough but just in case

14:33 <steev> alyssa: so, somewhat better, but also, somewhat worse?

14:34 <steev> https://usercontent.irccloud-cdn.com/file/hYa1WFqd/2022-01-10-083251_screenshot.png

14:35 <alyssa> steev: ....I'll take what I can get? 😅

14:35 <steev> "meson build/ -Dgallium-drivers=lima,panfrost,kmsro,swrast -Dplatforms=wayland,x11 -Dvulkan-drivers=panfrost -Ddri-drivers= -Dllvm=true" are my settings

14:35 <alyssa> that's fine

14:35 <steev> the weird "fuzzy" lettering is gone

14:36 <steev> i'm rebuilding alacritty though because it's not working properly

14:36 <alyssa> dmesg, I imagine, is complaining a lot about panfrost

14:36 <alyssa> Page faults or GPU timeouts, I imagine

14:37 <steev> oh

14:37 <steev> yes, yes it is

14:37 <alyssa> so good news, this sounds like a mesa bug

14:37 <alyssa> ("how is that good news again?")

14:37 <steev> [ 50.748262] panfrost ffe40000.gpu: js fault, js=0, status=INSTR_INVALID_ENC, head=0x7209900, tail=0x7209900

14:37 <steev> [ 50.853067] panfrost ffe40000.gpu: js fault, js=0, status=IMPRECISE_FAULT, head=0x6f96900, tail=0x6f96900

14:37 <alyssa> bad news, -ENOTIME to f---

14:37 <alyssa> INSTR_INVALID_ENC you say?

14:37 <alyssa> now you have my attention :p

14:38 <alyssa> a compiler bug in a fragment shader, spicy

14:39 <steev> it's not a huge deal, that system *mostly* sits idle (though i have 2 more on the way.... eventually... whenever ameridroid actually gets their stock in and ships them....)

14:41 <steev> i'd paste the dmesg but the stupid audio is broken on there too, so the dmesg prior to the panfrost stuff is very spammy about dai links being missing

14:41 <steev> https://bpa.st/Z52Q is all the panfrost bits

14:41 <steev> i suppose i could check if firefox is now broken heh

14:41 <steev> this is using debian's llvm 13 btw

14:42 <alyssa> this is definitely a Mesa bug

14:42 <alyssa> (or possible panfrost kernel)

14:42 <alyssa> definitely not firefox or llvm

14:43 <alyssa> likely a Bifrost compiler bug

14:43 <flowriser> HdkR, whaa you booted Linux on the ProX? How did you manage that?

14:43 <alyssa> we should probably move this to #panfrost and stop spamming this channel

14:47 <HdkR> flowriser: Badly

15:01 <flowriser> HdkR, more details please :D I am trying to boot on an acer spin 7 which has the same cpu mostly so I am very curious

21:40 <HdkR> flowriser: I used ACPI boot as a hack to bring it up

21:41 <flowriser> HdkR, ooh you mean the acpi=off flag?

21:41 <HdkR> nah, Just didn't use device tree

21:43 <flowriser> Did you use grub and any other parameters? Also, what patches did you wind up using?

21:45 <HdkR> I used Fedora's grub. https://github.com/Sonicadvance1/linux/tree/wip/sq1_surface_prox This hacky AF branch.

21:48 <flowriser> Amazing, will try it myself :D How did you create the bootable usb? For me, using Ubuntu's grub on arm is missing some modules

21:51 <HdkR> ProX uses UEFI so just a USB flash drive with some boot flags set was enough

21:53 <flowriser> ooh, I've been struggling for days with the grub thing; I've been trying with 1 Mib partition, ESP partition, DATA partition and then using grub-install --target arm64

21:54 <flowriser> HdkR, am I on the right track? I've even dropped to the Shell efi and tried to boot the kernel but didn't work

21:58 <flowriser> Looking over the Fedora grub docs is very nice, much more readable and understandable for me

22:02 <HdkR> I didn't do any grub-install things. Just yoinked the partition from their image

23:40 <flowriser> ok, so I've figured out why the kernel was simply restarting and not doing anything

23:40 <flowriser> I think, at least; it appears that when EFI loads a certain driver it reboots

23:42 <flowriser> and that driver appears to be GraphicsConsole