11:27xvilka_: hi! I want to help with this bug https://bugs.freedesktop.org/show_bug.cgi?id=100228 So what should I do? Is there any super-short crash course in nouveau hacking? I have big C, assembly and RE experience, so hopefully I can help a bit with this one
11:27xvilka_: and I am supermotivated, because I have a laptop prone to this error
11:33karolherbst: xvilka_: do you have only the error message or does something not work as well?
11:45xvilka_: karolherbst: when I do startx Xorg starts very slowly, also graph mode is unusable - everything is very-very-very slow
11:45xvilka_: and this error appears all the time, not just once
11:45karolherbst: xvilka_: well if the graph firmware isn't loaded, eveything is expected to be slow
11:45karolherbst: I see
11:49karolherbst: is the number 409800 for you as well? or is it a different one?
11:51xvilka_: yes, same number
11:52karolherbst: xvilka_: gp108 as well or a different GPU?
11:54xvilka_: this one is originally mine https://bugs.freedesktop.org/show_bug.cgi?id=101553
12:04karolherbst: mupuf: ........ I think I fixed the memory reclocking bug on your nve6
12:04karolherbst: I set the fN pll parameter to 0xf000
12:05karolherbst: ohh nice, another hang, but it was more stable so far
12:07karolherbst: I think memory is stable now and I hit the clock instability
12:07karolherbst: at least it was able to recover, what isn't possible if the memory messed up afaik
12:12karolherbst: xvilka_: did you try that acpi_rev_override=1 thing?
12:15xvilka_: karolherbst: yes, without it even just stuck without booting
12:26NikitaK3: Hello, I am a complete beginner with open source and wish to contribute to Nouveau. I went through tasks on trello.com/nouveau. I am not understanding though how to start contributing to the task and some cards have been added long back so is work still going on there. Can someone please guide me in details how to go about a task?
12:33anEpiov: damn, as soon as i knock/knock something in Metal Gear I get instant crash.
12:33anEpiov: wasn't happening before.
12:33anEpiov: something to do with GLES.h
12:34anEpiov: and dmesg segfault in lib.c
12:34anEpiov: wait, segfault in lib.so
12:34anEpiov: what am I doing wrong?
13:00karolherbst: NikitaK3: find something which you find interesting and ask about this task
13:01karolherbst: there is no "general" way of doing this, it always depends on the task
14:02BarryButsers: Hi, I have heavy flickering in an application. Is there anything I can do to work around this?
14:42orbea: BarryButsers: Can you make an apitrace?
14:42orbea: might help the devs here debug it
14:45BarryButsers: it`s on another (old) laptop. How do i mak
14:45BarryButsers: eHow do i make an apitrace?
14:50orbea: BarryButsers: try reading their github wiki, its pretty helpful. IN short: apitrace trace application
14:50orbea: then: xz -9 application.trace
14:50orbea: and upload it somewhere to share
14:50orbea: you can make sure the trace replays with: apitrace replay application.trace
14:51orbea: note that the traces can get big pretty fast
14:52BarryButsers: ok, i`ll give it a try, could take a while though
14:55BarryButsers: there`s one problem: i am not 100% sure it`s due to nouveau. The application is an app i run in wine. I tried pass nouveau.noaccel=1 to boot commandline and then the flickering is gone
14:55BarryButsers: How can i be sure that it`s not due to wine?
14:56orbea: does it occur if you use the llvmpipe instead of nouveau?
14:58orbea: you will need a 32-bit apitrace to trace wine btw
15:02BarryButsers: When i run with LIBGL_ALWAYS_SOFTWARE=1 the flickering is gone as well
15:02orbea: sounds like it could be related to nouveau then
15:03orbea: just to be sure, this command reports llvmpipe, right?
15:03orbea: LIBGL_ALWAYS_SOFTWARE=1 glxinfo | grep 'OpenGL renderer'
15:03BarryButsers: It`s a very old GPU, Geforce Mobile 440
15:03anEpiov: use to work by the way
15:04imirkin: xvilka: talk to rhyskidd -- he's been looking into issues adjacent to that one
15:04imirkin: xvilka: my impression is that some kind of power-up procedure is missing, and so parts of the chip are remaining asleep
15:05imirkin: xvilka: there's a chance that it's the fault of the firmware that nvidia released, in which case it'll be much trickier to fix -- that firmware has to be signed
15:05imirkin: xvilka: however one might hope that it's a simpler issue of just not disabling block-level gating or something along those lines
15:06imirkin: xvilka: basically look at an mmiotrace of the blob, and try to do things that the blob is doing. it's all pretty difficult though, esp if you don't know your way around; hard to tell what's important and what's not.
15:07BarryButsers: LIBGL_ALWAYS_SOFTWARE=1 glxinfo | grep 'OpenGL renderer' gives Gallium 0.4 (LLVM 4.0 128 bits)
15:10orbea: sounds right
15:10orbea: so try to get an apitrace, share it here and hopefully a real dev will have time to take a look :)
15:11Fedev: is HDMI audio supposed to work right away or does it need to be enabled somehow? I'm getting video but the HDMI audio device does not show up. This is on a GTX 1060
15:11BarryButsers: orbea: ok thanks for the help!
15:11Fedev: that is NV130
15:12imirkin: Fedev: should work, but likely less-tested on GP10x
15:12imirkin: Fedev: do you see eld?
15:13Fedev: imirkin: where can I see that?
15:14imirkin: er wait, are you saying that you don't have an HDA pci device?
15:14Fedev: I dpm
15:14imirkin: pastebin `lspci -nn -d 10de:`
15:14Fedev: I don't see eld imirkin
15:14Fedev: ok, one sec...
15:16Fedev: that didn't amount to much. imirkin here is the paste https://pastebin.com/exYPsS5X
15:16imirkin: you have a weird system
15:17imirkin: you're *also* supposed to see a device like
15:17imirkin: 02:00.1 Audio device : NVIDIA Corporation GK208 HDMI/DP Audio Controller [10de:0e0f] (rev a1)
15:17imirkin: there must be some logic in the smm or whatever which disconnects that device from the pci bus when there's no hdmi cable
15:17imirkin: and when an hdmi cable is plugged in, linux doesn't rescan the pci bus
15:18imirkin: so ... with the hdmi cable plugged in, try triggering a pci bus rescan
15:18Fedev: it doesn't. I don't see other audio devices except the intel card...
15:18imirkin: e.g. echo 1 > /sys/bus/pci/rescan
15:18Fedev: let me try that
15:18imirkin: er wait, is the HDMI port on the nvidia
15:18imirkin: or is it on intel?
15:18imirkin: if it's on intel, then this is all bogus obviously
15:18Fedev: on the nvidia
15:19karolherbst: Fedev: did you check or do you assume?
15:19imirkin: well presumably he got output working...
15:20karolherbst: just want to make sure
15:20Fedev: I have the video output working, that goes on nouveau
15:20imirkin: reverse prime and all that jazz, right?
15:20Fedev: there is no intel video card, just audio
15:21karolherbst: Fedev: is it a laptop where you can disable it in the vbios?
15:21BarryButsers: I made an apitrace : it`s at https://we.tl/qfh3vn8Ara
15:21Fedev: it is already disabled. If not mistaken I read it had something to do with the gsync support planned for it
15:21BarryButsers: (back in an hour)
15:22Fedev: imirkin: for audio I only get "00:1f.3 Audio device : Intel Corporation Device [8086:a171] (rev 31)"
15:22karolherbst: Fedev: okay, but in this case we have a totally different situation going on here
15:22karolherbst: because the nvidia GPU never suspends
15:23Fedev: so there should be an audio device for nvidia no matter what? karolherbst
15:23karolherbst: well, I dunno? Maybe there is some crazy way to use the intel one with the nvidia GPU? Since pascal I am not sure anymore what nvidia wouldn't do
15:23karolherbst: but if there is a nvidia audio device, it shouldn't disappear, right
15:24Fedev: would it make any difference to try the nvidia driver to see if there is anything showing there?
15:24karolherbst: maybe, it's worth a try at least
15:25karolherbst: except imirkin knows something here
15:28Fedev: ok, let me try that, test and get back
15:42rhyskidd: xvilka: I've been looking at the mobile GTX 1050 (GP107) issues. No solution so far
15:43rhyskidd: replicating the behaviour of the blob via mmiotraces *should* help us work out what we are missing
15:43rhyskidd: it is looking increasingly like changes between Maxwell2 -> Pascal were larger than we previously expected
15:45Fedev: ok, back again. The nvidia drivers don't make any difference for me. I still don't get audio and there is no audio device listed as nvidia in lspci
15:46Fedev: would there be any useful information the nvidia driver could provide before I switch back to nouveau? karolherbst / imirkin ?
15:47wovcf: rhyskidd: drop this crappy nonfree hardware that requires closed source software signed with some private key that only people you dont know own
15:51karolherbst: Fedev: try to use intel as main
15:51karolherbst: Fedev: if you want to use nouveau anyway, then it makes sense to let intel handle your desktop, due to stability
15:53Fedev: karolherbst: are you referring to intel for graphics? Then there is none. Any info I could provide which might help understand this setup better?
15:57karolherbst: Fedev: I am quite sure there is one, even if it is disables by some means
15:57Fedev: karolherbst: if lspci doesn't show it, then disabled in BIOS?
15:59NikitaK3: Okay I will select a task first. Thanks karolherbst.
16:02karolherbst: Fedev: most likely
16:02karolherbst: Fedev: sometimes they are even hw disabled, but afaik all mobile intel CPUs come with an intel GPU as well
16:03karolherbst: Fedev: the only thing where using intel main and nouveau second matters would be if you want to play something at 4K res or drive an external 4k display
16:03Fedev: I don't have windows installed but I'll try to figure out how HDMI works with it
16:04Fedev: thanks for your help karolherbst
16:08imirkin: on some laptops, the intel gpu is entirely inaccessible
16:10imirkin: Fedev: i assume rescanning the pci bus didn't help?
16:10Fedev: imirkin: so the problem is not really that the intel gpu is no accessible but that the nvidia audio car is not showing
16:10imirkin: i.e. still no HDA device?
16:10Fedev: imirkin: no HDA for nvidia after the refresh
16:11imirkin: what if you boot with the hdmi cable plugged in?
16:12imirkin: but like karolherbst says -- if you can convince your bios to go into hybrid mode, and the hdmi port ends up on intel, you'll be much much happier
16:12Fedev: tried that and tried to unplug and plug again. Now on the nvidia driver wanted to see if that made a difference. It didn't
16:12Fedev: let me get rid of the nvidia driver and check the bios settings again. I'll be back soon
16:20karolherbst: imirkin: is the audio device shown in the vbios?
16:26imirkin: Fedev: any luck?
16:26Fedev: imirkin: I was reviewing the BIOS options and found mentioned the Azalia driver. Turning it off killed all sound, looking at it, it seems it is a driver from intel which routes audio over HDMI
16:26Fedev: could that be it?
16:27Fedev: No option to turn off/on the intel gfx
16:27imirkin: wtf? no clue.
16:27imirkin: could be that hdmi audio is handled somehow differently for you...
16:27imirkin: anyone with a GP10x around? if so could you check if you have an audio pci device function attached to it?
16:28karolherbst: Fedev: did you check if HDMI audio works with the nvidia driver?
16:29Fedev: it didn't karolherbst . All I found was the mention of the Azalia driver
16:30imirkin: ah - i think that's in reference to the "main" audio function
16:34karolherbst: well, sounds like the HDMI thing though, not main
16:35Fedev: was wondering if this setup has some weird configuration in which the audio over hdmi goes over the intel card mixed with the gfx from nvidia. Crazy but I'm running out of explanations
16:36imirkin: i've never heard of such a thing.
16:36imirkin: since GT21x, there's a dedicated HDA function that hangs off the main pci device
16:36imirkin: and that is how audio is passed to the HDMI encoder
16:37imirkin: prior to that, you had to have a SPDIF cable from the audio card to the GPU, and the GPU would do its own sampling/resampling.
16:37karolherbst: Fedev: do you see the intel HDMI audio device?
16:37karolherbst: I mean in pulse or whereever
16:38Fedev: let me get more info about that
16:38karolherbst: Fedev: pavucontrol -> configuration tab
16:38imirkin: Fedev: just to make sure you're not misinterpreting things... mind pastebinning 'lspci -nn' and 'grep . /sys/class/drm/card*-*/status'
16:38karolherbst: there might be an aduio device with profile set to off
16:39karolherbst: and where the available profils are HDMI related
16:40Fedev: imirkin: https://pastebin.com/Xqsvbaks
16:41imirkin: yep, no misinterpretation on your end.
16:42imirkin: this is a high-end laptop i take it?
16:42imirkin: i think lenovo has disabled the onboard gpu in the past on some of those
16:42Fedev: karolherbst: pavucontrol doesn't have anything but built-in audio
16:43Fedev: imirkin: yes it is, high end but with lots of problems for the simplest of things haha
16:43karolherbst: Fedev: do you have that hdmi audio disabled in the bios?
16:43karolherbst: ohh wait
16:43Fedev: karolherbst: no, I did disable it just a while ago for testing but ended up with no sound device at all
16:44karolherbst: well I have two intel audio devices, one for the main thing, the other for HDMI
16:44Fedev: had hopes of the nvidia audio showing by then. Since it didn't I had reverted that back
16:44Fedev: let me show you what aplay gives...
16:44karolherbst: 00:02.0: intel GPU, 00:03.0: HDMI audio
16:45karolherbst: Fedev: can you check what profiles your audio device has?
16:45karolherbst: maybe it also has HDMI profiles
16:45Fedev: karolherbst: https://pastebin.com/GFCG8Ky4
16:46karolherbst: super weird
16:46karolherbst: there is no hdmi audio
16:46imirkin: i think it's the issue where the HDA device needs to be turned on somehow
16:46imirkin: we've seen it before
16:46karolherbst: most likely
16:46imirkin: Fedev: oh, are you booting with weird acpi params
16:46karolherbst: imirkin: and let me guess, nvidia didn't care to implement it as well?
16:46Fedev: profiles are: Analog Stereo Duplex, Analog Stereo Output and Analog Stereo Input
16:47imirkin: e.g. this one: https://bugs.freedesktop.org/show_bug.cgi?id=75985
16:48imirkin: Fedev: lspci -H1
16:48imirkin: does that show an HDA device hanging off the nvidia?
16:48Fedev: no, I've been trying some weird acpi parameters but not running them since a while ago
16:49Fedev: imirkin: https://pastebin.com/B40gZcYa
16:51Fedev: it is so weird. Why can't they keep a simple design.
16:52karolherbst: because they want to be "smart"
16:58Fedev: I'll try a couple of things and let you know if there is any improvements. Thanks to both of you for you help on this
17:06imirkin: Fedev: you may want to grab the acpi_call module
17:07imirkin: and try calling some acpi methods to see if you can get it to appear
17:07Fedev: how do I do that ?
17:07imirkin: by grabbing the acpi_call module :)
17:07imirkin: i don't remember exactly how to operate it though
17:07Fedev: any documentation about it you might remember?
17:07imirkin: i hear there are these things called 'search engines' though, and they can help you find things
17:08imirkin: i can always send a lmgtfy.com link if needed :p
17:15karolherbst: well, it ain't so easy though
17:15karolherbst: usually you have to know what you wanna call and all that
17:17xvilka_: sorry, I have bad connection, so if you said something to me - I could have missed it
17:18xvilka_: will setup bouncer tomorrow
17:25karolherbst: xvilka_: link to logs in the topic
17:31xvilka_: karolherbst: ah, thx
21:16Tom^: karolherbst: oh yeah saw your phoronix forum post, i run/ran 144hz on my 780ti just fine over dvi. so it is supported :p
21:16Tom^: if its of any worth knowing.
21:44karolherbst: Tom^: well depending on the EDID, and how we parse the bios, we might not expose @144 res
21:44Tom^: mmh oh well