#
#asahi-dev · 2025-10-09 19:09
sveni think the only way is an invalid dts
#asahi-dev · 2025-10-09 19:09
svenbut even then, nothing's gonna auto-probe the child devices
#asahi-dev · 2025-10-09 19:07
svenit's probably someone who's running a static analyzer or something and fixing "bugs"
#asahi-dev · 2025-10-09 19:06
sventhat too
#asahi-dev · 2025-10-09 19:01
jannaudon't the smc sub-drivers depend/select macsmc anyway? I guess you can still `rm macsmc.ko`
#asahi-dev · 2025-10-09 11:24
fl0_idt6030 is not even there
#asahi-dev · 2025-10-09 11:23
fl0_id(for t6031)
#asahi-dev · 2025-10-09 11:23
fl0_idjannau also found this comment in the cpufreq_init function: /* Unknown */
#asahi-dev · 2025-10-09 04:04
nnym./se
#asahi-dev · 2025-10-08 16:35
fl0_idok that I should be able to find out, just hope I do not have to reboot too often ^^
#asahi-dev · 2025-10-08 16:32
fl0_idyeah it does
#asahi-dev · 2025-10-08 16:31
jannaufl0_id: this looks like it's still in m1n1. I don't remember what the expected output of linux.py is but I wouldn't be surprised if this crash is in m1n1's cpufreq code
#asahi-dev · 2025-10-08 15:01
fl0_idbtw this is my error from kernel booting, likely doesn't tell anybody anything new but still, if it does, let me know :) https://www.pastery.net/jujenj/
#asahi-dev · 2025-10-08 02:08
adityaThere is a regression in the hid tree of asahi linux causing magic trackpad 1 to not work. I've sent a PR here: https://github.com/AsahiLinux/linux/pull/416
#asahi-dev · 2025-10-08 02:08
adityaHi @jannau
#asahi-dev · 2025-10-07 15:07
leioanyhow, so need to acquint myself more with the command stream and then see from there
#asahi-dev · 2025-10-07 15:05
leiovulkan is mainly frame based, VA is slice based
#asahi-dev · 2025-10-07 15:01
jannaumy assumption was/is that the command stream is per frame or slice
#asahi-dev · 2025-10-07 14:59
chaos_princessthe python driver is bypassing the firmware, but the hw itself is getting a command stream, and i am not sure you can drive it in a statless manner.
#asahi-dev · 2025-10-07 14:57
jannaudidn't eiln bypass the decoder firmware completely? That would make the decoder stateless. It also makes the "driver" more complex so that moving as much code as possible outside of the kernel a good idea. All from memory, it has been a while since I looked at the python PoC code
#asahi-dev · 2025-10-07 14:42
mripardfor a codec, you should absolutely use v4l2, no matter if it's stateful or stateless
#asahi-dev · 2025-10-07 14:42
leio"Vulkan does not have an interface to support stateful firmware based codecs, as an example Qualcomm CODECs can't as-is be supported through vulkan, you need a new firmware"
#asahi-dev · 2025-10-07 14:42
leioVulkan Video can't support stateful firmware
#asahi-dev · 2025-10-07 14:41
chaos_princessfrom a cursory look it will have to be stateful, but i am not 100% sure
#asahi-dev · 2025-10-07 14:40
leiochaos_princess: is it stateless or stateful for AVD then?
#asahi-dev · 2025-10-07 14:33
chaos_princessif it can be in userspace, it should probably go there. Especially with how complex video is
#asahi-dev · 2025-10-07 14:28
leioI like those arguments too, I wouldn't like working too much in kernel space, just playing devils advocate to get the points
#asahi-dev · 2025-10-07 14:27
jannaukeeping code out of the kernel which doesn't need to be in the kernel is another compelling argument against v4l2
#asahi-dev · 2025-10-07 14:26
jannauI wouldn't worry too much about vulkan video (or vaapi). Once there is a working driver supporting the other interface as well is a much smaller problem. It might be either through zink or by adding vaapi support to the user space part of the driver
#asahi-dev · 2025-10-07 14:22
leioit'll eventually get there, of course
#asahi-dev · 2025-10-07 14:22
leiomeanwhile firefox has v4l2 decode, but not vulkan video decode: https://bugzilla.mozilla.org/show_bug.cgi?id=1753129
#asahi-dev · 2025-10-07 14:21
leiobut ok, so counter-argument that they do it, and we want to be like desktop linux too
#asahi-dev · 2025-10-07 14:21
leiopart of the GPU?
#asahi-dev · 2025-10-07 14:20
chaos_princessgo ask amd, intel and nv as to why they do it that way
#asahi-dev · 2025-10-07 14:20
leiowe'll probably get some jabs at why such a driver when there's v4l2
#asahi-dev · 2025-10-07 14:20
chaos_princessalso, upstreaming a drm driver that just allocates buffers and executes them should be much easier than all the things you need for v4l2
#asahi-dev · 2025-10-07 14:18
leiohttps://bugzilla.mozilla.org/show_bug.cgi?id=1833354
#asahi-dev · 2025-10-07 14:18
chaos_princessthere is no v4l2 support in firefox
#asahi-dev · 2025-10-07 14:17
leiochaos_princess: v4l2 for AVD and AVE was mostly about looking not so great for stateful and too much work in kernel space to be convenient, and maybe less support for desktop linux stuff, or something more? I got various indications that v4l2 is a better fit politicially for upstreaming and such from some information face to face chats I had
#asahi-dev · 2025-10-07 14:16
leioanyways, to be more on topic,
#asahi-dev · 2025-10-07 14:14
leioI also got the wrong impression how far along 6.18 is, and thus how old 6.17 is
#asahi-dev · 2025-10-07 14:13
leiosomething about for each kernel release, but yes, I moved the overdue between quotation marks :)
#asahi-dev · 2025-10-07 14:11
LarstiQlast one was 2 months ago, while I do like reading them I would indeed not call that overdue ;P
#asahi-dev · 2025-10-07 12:59
leio"overdue"
#asahi-dev · 2025-10-07 12:59
leiochadmed: I think we are also overdue a progress report? ;)
#asahi-dev · 2025-10-07 11:53
chadmedyeah
#asahi-dev · 2025-10-07 11:52
svenit takes care of probing the device, but it doesn't do anything about the module
#asahi-dev · 2025-10-07 11:52
chadmedi mean it kinda looks like it does by registering the pdev and everything
#asahi-dev · 2025-10-07 11:52
chadmedi cant remember where i saw that mfd takes care of that for us
#asahi-dev · 2025-10-07 11:50
sveni can't find the documentation anymore but from what i remember that's some convention for userspace which gets some uevent or something and can then load the right module

plus ›