03:02 <SolidHal> @alyssa Hey, dev of PrawnOS here, trying to get kernel 5.2 running on the asus c201 but it just boots to a black screen and doesn't seem to throw anything out the serial port. I don't have panfrost enabled, so it isn't at fault but I was wondering if anyone that worked on it experienced this issue?

04:10 <wens> SolidHal: might want to ask over in #linux-rockchip

04:10 <wens> SolidHal: first you might want to check if all drivers needed are enabled. One hard-to-find one is CONFIG_ROCKCHIP_IODOMAIN

04:53 <SolidHal> Thanks wens, I'll checkout that channel. IODOMAIN is enabled, doesn't look new to 5.2 so probably left over from my 4.19 config

07:19 <tomeu> anarsoul|c: the x2 is probing the gpu just fine: https://lava.collabora.co.uk/scheduler/job/1781708#L643

07:19 <tomeu> the display can be a problem, as weston with the headless backend doesn't support the gl renderer, afaik

07:19 <tomeu> and using the pixman renderer will cause mesa to use swrast

07:20 <tomeu> I think Xvfb should work, though

07:21 <tomeu> though it would be nice for both CIs to not diverge

07:21 <tomeu> daniels: any ideas?

07:46 <daniels> tomeu: i think we should add GL support to the headless backend

07:47 <daniels> shouldn't be too much work

07:52 <tomeu> great, virgl and tracie could drop Xvfb usage then

11:42 <shadeslayer> oh nice, I have a new crasher https://paste.ubuntu.com/p/Sn5Xvj8QwW/

12:56 <daniels> shadeslayer: so, taking a step back - I think before going to debug more asserts, you need to first form a model of how the job lifetimes should be working, and then analyse how they _are_ in reality

13:38 <robher> daniels, tomeu: Couldn't vkms be used for headless? We'd have to add it to kmsro.

13:39 <daniels> robher: hmm yeah, I suppose you could use vkms + Panfrost with linear surfaces

14:24 <tomeu> robher: good idea, and could be useful for testing gnome-shell, etc

14:33 <alyssa> SolidHal: Not sure :/ Definitely do try to get things going with Panfrost/GL disabled first

14:34 <alyssa> I've never had a magic touch with kernels

14:59 <shadeslayer> daniels: sounds like a plan

20:40 <alyssa> Planning to cleanup scheduling-adjacent stuff

20:52 <mmind00> is there some howto available on what setup the cool collabora supertuxkart-demo at ELC ran? ... somehow both x11 and wayland select llvmpipe for me right now, while glmark2-es2-drm does the right thing

20:53 <alyssa> robertfoss: ^^

21:15 <alyssa> This scheduler code is just

21:15 <alyssa> entirely

21:15 <alyssa> suspecy.

21:18 <alyssa> Gah

21:33 <shadeslayer> alyssa: http://paste.ubuntu.com/p/ZrDFNRWZHW/

21:39 <alyssa> shadeslayer: Seems sane :+1:

21:39 <shadeslayer> <3

21:40 <alyssa> shadeslayer: In general, Panfrost shouldn't know what "scanout" means

21:40 <alyssa> Everything should be an FBO

21:40 <alyssa> That's not what we actually do -- we infer scanout and then mess with stuff

21:40 <alyssa> but conceptually we.. should not.. :)

21:40 <Lyude> yikes

21:41 <alyssa> Lyude: yikes?

21:41 <Lyude> alyssa: @ inferring what the scanout is :p

21:41 <alyssa> Ah :p

21:41 <shadeslayer> alyssa: gotcha

21:41 <alyssa> Lyude: I mean it's just checking if the bound framebuffer is PIPE_BIND_SCANOUT but still

21:41 <shadeslayer> I'm just going through pan_job stuff to understand job lifecycles, it's a bit dense :P

21:42 <alyssa> Sorry :x

21:42 <alyssa> Lyude: If you want something to yikes at it should be trying to write a Midgard scheduler

21:42 <Lyude> alyssa: ahhh, that's referring to something else other than what I thought it was referring to :p

21:42 <Lyude> alyssa: hehe, i'm sure

21:42 <shadeslayer> alyssa: nah, it's just growing pains, at some point it'll all make sense ;)

21:42 <alyssa> shadeslayer: You sure? Doesn't make sense to me yet ;)

21:46 <shadeslayer> :D

21:46 <shadeslayer> I would hope so :P

21:48 <shadeslayer> alyssa: worst case, let me enjoy my naive innocent enthusiasm :P

21:50 <alyssa> Hehehe

21:54 <shadeslayer> alyssa: is any of this used anywhere? https://gitlab.freedesktop.org/mesa/mesa/blob/master/src/gallium/drivers/panfrost/pan_job.c#L375-392

21:54 <shadeslayer> I think we can kill it with fire?

21:55 <alyssa> shadeslayer: Please keep that function.

21:55 <alyssa> It is not used at the moment, but it will be.

21:55 <alyssa> Its purpose will be to allow job pipelining.

21:56 <shadeslayer> I see

21:56 <shadeslayer> alyssa: should I just flesh it out instead then?

21:57 <alyssa> Sure :)

21:58 <shadeslayer> alyssa: http://paste.ubuntu.com/p/Z87v86MJmx/

21:59 <shadeslayer> something to that extent, but you see where I'm going with it :)