09:32alterjsive: Is there anyone who got hybrid graphics to work?(amd/nvidia)
09:39alterjsive: or at least get it to powerdown so my laptop fans don't make so much noise that it bothers my team mates
09:39karolherbst: alterjsive: the issue is, that you are running into a nouveau bug, it has nothing to do with runpm not working
09:39alterjsive: does anyone know how to find out the ascpi code?
09:40karolherbst: alterjsive: what you could do is to blacklist nouveau and turn on runpm on all devices neccassary
09:40karolherbst: (writing auto to control)
09:40alterjsive: what is runpm?
09:40karolherbst: runtime power management
09:41karolherbst: you could eg use powertop to enable it for the GPU, all sub devices and the root bus
09:41karolherbst: powertop -> Tunables -> search for "Runtime PM"
09:41karolherbst: could just enable it for all devices
09:41alterjsive: karolherbst: https://01.org/linuxgraphics/gfx-docs/drm/ch05s02.html this page uses it as a kernel param
09:42karolherbst: nouveau as well
09:42karolherbst: but that's not the issue
09:44alterjsive: what does it mean if it says bad
09:44karolherbst: that it's disabled
09:45alterjsive: karolherbst: I don't see my nvidia gtx card in there
09:46karolherbst: is nouveau loaded?
09:46alterjsive: oh wait I found it sorry
09:46alterjsive: I just tuned it off
09:46karolherbst: well, for it to work you need to blacklist nouveau first, reboot and then turn it on
09:47alterjsive: karolherbst: my goal is to use less power so it generates less heat
09:47alterjsive: karolherbst: and my fans don't spin constantly
09:48alterjsive: but power usage is still the same 40W
09:48karolherbst: did you do what I suggested to do?
09:48alterjsive: I will sure, just checking, so you want me to blacklist neauvea so I can turn it off
09:49alterjsive: I understand, thx
09:51alterjsive: cat /etc/modprobe.d/blacklist-nvidia-nouveau.conf
09:51alterjsive: blacklist nouveau
09:51alterjsive: options nouveau modeset=0
09:52alterjsive: brb reboot
09:55alterjsive: nouveau is still loaded
09:55dcomp: So I've just woken up and started trying this card again
09:55alterjsive: should I try a grub kernel parameter?
09:56karolherbst: alterjsive: you might need to regenerate your initramfs
09:56alterjsive: karolherbst: ok, will do thanks
09:59dcomp: I think there is a regression. Because I couldn't get NvClkMode and runpm to stop the faults.
10:00dcomp: I've found the original bug]
10:02karolherbst: dcomp: mind sharing dmesg with your latest try?
10:03dcomp: That was run with: COMMAND=/usr/sbin/modprobe nouveau config=NvClkMode=0x7 runpm=0
10:05dcomp: I was waiting for your patches to hit mainline, so I could just edit my module.conf and unblacklist nouveau. I guess I waited too long
10:08alterjsive: karolherbst: it's off now, when I check it to bad/good it has no effect on the power usage
10:08karolherbst: but is the GPU actually suspended?
10:08alterjsive: how can I check?
10:08karolherbst: grep . /sys/bus/pci/devices/*/power/runtime_status
10:09alterjsive: karolherbst: I think it's still active
10:10karolherbst: does "grep . /sys/bus/pci/devices/*/power/control" show auto for the GPU?
10:11alterjsive: 01:00.0 VGA compatible controller : NVIDIA Corporation TU106M [GeForce RTX 2060 Max-Q] [10de:1f12] (rev a1) (prog-if 00 [VGA controller])
10:12alterjsive: karolherbst: so it's off then?
10:13karolherbst: should be
10:13karolherbst: mind worth checking the root port as well
10:13alterjsive: so the fan noise is not the root cause of the load fans
10:13alterjsive: the nvidia card*
10:13karolherbst: lspci -t should show (even though the view is quite bad) where the GPU is connected to
10:13karolherbst: probably 00:01.0, but I saw others on other laptops
10:13karolherbst: this device needs to be suspended as well
10:14karolherbst: (and all parent device up to 00:00.0)
10:14karolherbst: except 00:00.0 of course
10:14alterjsive: how can I check?
10:14alterjsive: 01:00.0 VGA compatible controller : NVIDIA Corporation TU106M [GeForce RTX 2060 Max-Q] [10de:1f12] (rev a1) (prog-if 00 [VGA controller])
10:14alterjsive: 04:00.0 VGA compatible controller : Advanced Micro Devices, Inc. [AMD/ATI] Renoir [1002:1636] (rev c5) (prog-if 00 [VGA controller])
10:15karolherbst: lspci -t
10:15alterjsive: I don't know how to read this one so i'll post it on pastebin
10:16alterjsive: karolherbst: https://pastebin.ubuntu.com/p/jvT867KcVg/
10:16karolherbst: ahh okay
10:16karolherbst: 00:01.1 is the root port
10:16karolherbst: and you have all three sub devices on it: GPU, HDA, USB, UCSI
10:17karolherbst: for all of those 5 control has to be auto and runtime_status suspended
10:18karolherbst: you can also do lspci -tv to show the names, but it won't show the names of parent devices.. it's a bit annoying to read that
10:19alterjsive: hmmz https://pastebin.ubuntu.com/p/fNK8w8Dy6K/
10:19karolherbst: yeah.. anyway :)
10:19karolherbst: we already know what devices need to get this enabled
10:19karolherbst: I have no idea if that's done automatically for the USB/UCSI ones
10:20karolherbst: it's quite rare for those to exists on laptops
10:20alterjsive: lucky me =)))
10:21alterjsive: how can I know if it's auto and runtime_status suspended?
10:21karolherbst: the grep commands :)
10:27alterjsive: grep . /sys/bus/pci/devices/0000:00:01.1/0000:01:00.0/power/runtime_status
10:27alterjsive: ➜ ~ grep . /sys/bus/pci/devices/0000:00:01.1/0000:01:00.0/power/control
10:29alterjsive: karolherbst: that's really bad news for me, that means the problem is elsewhere
10:29dcomp: Is there a race between NvClkMode and bios/devinit I just got FIFO sched error 20 when trying to run with debug=bios=trace,devinit=trace
10:29karolherbst: alterjsive: you need to check .1 .2 and .3
10:30karolherbst: dcomp: mh.. could be actually
10:31karolherbst: alterjsive: soo.. 0000:01:00.0, 0000:01:00.1, 0000:01:00.2 and 0000:01:00.3 need to be auto/suspended
10:34alterjsive: karolherbst: check https://pastebin.ubuntu.com/p/g3BnFkCH3T/
10:34karolherbst: okay, and same for 0000:00:01.1?
10:36alterjsive: lucky me =)))
10:36alterjsive: grep . /sys/bus/pci/devices/0000:00:01.1/power/runtime_status
10:36alterjsive: ➜ ~ grep . /sys/bus/pci/devices/0000:00:01.1/power/control
10:36karolherbst: yeah.. you can just echo "auto" into control
10:36karolherbst: maybe that makes it work
10:37karolherbst: if not, there might be some bits missing in the pci or acpi core code somewhere
10:38alterjsive: karolherbst: I turned it to auto but still active
10:38karolherbst: might take a while
10:38karolherbst: but usually the root ports have quite a low delay
10:39karolherbst: smells like another bug :/
10:40karolherbst: anyway. I have located another laptop with a similiar or the same nouveau issue, so maybe I will be able to figure something out there
10:40karolherbst: but for the root port not suspending I have no idea..
10:40karolherbst: probably worth filing a kernel bug against pci
10:40alterjsive: karolherbst: ok thanks
10:40alterjsive: is there an acpi chatroom?
10:40karolherbst: but I bet the ACPI stuff is just not wired up
10:40karolherbst: I don't think so
10:41karolherbst: or maybe it doesn't exist on your laptop at all, which would be odd..
10:41karolherbst: but having the GPU suspended might give you a lower power consumption, even though I suspect it's not that much lower
10:49alterjsive: karolherbst: do you think that if the root device is enabled it will consume more power?
10:54alterjsive: karolherbst: I guess you just awnsed that question, thanks anyways, think I should go back to windows then maybe :(