00:21RSpliet: martinus: depends. If you're running all upstream software then yes. If you're running older (or as distro's often misname it, "stable") software then you're better off in distro support channels.
00:27martinus: RSpliet: Cheers, I'll check in the distro channel then. :)
02:21tm512: I have an old PowerMac G4 with a Geforce 2 MX, which uses nouveau, and I decided to test out this OpenGL game I'm making. not totally sure whether it's a driver issue, but the window goes blank, seemingly after a certain number of calls to either glVertexPointer, glTexcoordPointer, or glDrawArrays
02:22tm512: it was happening rather quickly when drawing several things at once, but still eventually happened when drawing just one sprite
02:55imirkin__: tm512: support for nv17 is less-than-stellar, and support for ppc/be is less-than-steller. the combination may prove ... weak.
02:56imirkin__: tm512: that said, if you make a trace, i could take a look and/or make suggestoins
02:56imirkin__: [trace = apitrace]
02:57vpelletier: imirkin__: hi !
02:58imirkin__: vpelletier: hello. i see you've been knocking your head against the MSI thing...
02:58imirkin__: vpelletier: perhaps just let it go :)
02:58vpelletier: my head does not hurt too much yet :)
02:58tm512: I could do that. it's not a huge deal if this won't work properly on this machine, but I'm curious enough to try figuring it out
02:59imirkin__: vpelletier: well i think the deal with MSI is that you give some address to the device, and then when an interrupt happens, it writes the irq number to that address. the io-apic or whatever intercepts that write and triggers the proper interrupt on the cpu.
02:59imirkin__: tm512: well, first step would be to determine if things go wrong on a LE box as well
03:00imirkin__: tm512: note that various features in nouveau_vieux are outright unsupported, e.g. 3d textures
03:01tm512: this is a 2D game and afaik I'm not using any modern GL features
03:01tm512: it was rendering just fine, up to a point
03:01tm512: I was going to put a counter in the function to see if it failed around any particular number of calls
03:02imirkin__: apitrace makes this sort of stuff easier.
03:02imirkin__: things like VBO's with non-rgba8_unorm data could definitely be confusing.
03:03imirkin__: i have a nv34 in a ppc, i could test how it acts when emulating the nv25 interface. should be close enough to what you have.
03:04tm512: no VBOs in this code, I upload the vertices through VertexPointer
03:04imirkin__: hm ok
03:05tm512: also, I don't have an LE box with nvidia in it
03:05imirkin__: well if you post an apitrace, others (e.g. me) could test it
03:05imirkin__: not this second, but ... later.
03:06tm512: just booting up the G4
03:06imirkin__: you could try to work it out yourself too
03:06imirkin__: by tracing the driver, figuring out where things start going off
03:06imirkin__: you're more than welcome to do that =] happy to provide suggestions too.
03:07tm512: I'm assuming this could probably result in me needing to recompile nouveau on this machine?
03:07imirkin__: or on another machine ... yes.
03:07imirkin__: [the mesa portion of it]
03:09vpelletier: tm512: as I happen to be cross-compiling, if it can save you some time: make ARCH=powerpc CROSS_COMPILE=/usr/bin/powerpc-linux-gnu- menuconfig
03:09vpelletier: and similarly for other make targets
03:10tm512: I'm not too unfamiliar with cross-compilation, but I'd have to compile one
03:11tm512: and that is a bit of a chore
03:14ppc512: so, which package should I be looking in for apitrace?
03:14imirkin__: "apitrace" :)
03:15imirkin__: just records all the GL commands and all the data sent
03:15imirkin__: making it easy to replay on another machine
03:16ppc512: alright, I'll compile that
03:17ppc512: this machine sure is sluggish
03:18vpelletier: you may want to disable the build of qapitrace (the trace viewer GUI) to save on compilation time
03:19ppc512: I don't even think I have qt installed
03:19vpelletier: I believe there is a console-based replay, but I've never used it
03:20vpelletier: then it should be automatically disabled by configure
03:22vpelletier: mmh... gcc segfaulting during a kernel cross-build
03:22vpelletier: retrying and it passes
03:22imirkin__: glretrace works fine
03:22imirkin__: and has no QT requirements
03:23imirkin__: qapitrace is nice for debugging though
03:28vpelletier: imirkin__: about the MSI stuff, is there a way to be sure DMA writes work on the card ?
03:28vpelletier: ie, card to memory
03:28imirkin__: vpelletier: well ... dunno
03:43ppc512: well, I got the output of apitrace
03:45imirkin__: i assume the issue reproduces when replaying the trace?
03:46imirkin__: and when you replay the trace on a different box, it works fine?
03:46imirkin__: [could be your application messing things up...]
03:46ppc512: haven't replayed it elsewhere yet, I'll check
03:50tm512: is apitrace linux-only?
03:51imirkin__: in fact, in addition to GL it can also trace other api's, like DX
03:51tm512: hmm, it is failing to find this gstdint.h
03:51imirkin__: on windows you have to use scons, i suspect
03:52imirkin__: er no, probably not
03:52tm512: I'm on FreeBSD
03:52imirkin__: hm, dunno
03:52imirkin__: feel free to fix it up and/or send patches
03:53imirkin__: gtg... if you get anywhere useful, file a bug - i'll look later.
03:53tm512: alright, well thanks for the help
09:52debo: hi guys, back again trying to get my kde desktop working after upgrading to debian stretch
09:52debo: I get the following error in dmesg
09:52debo: plasmashell: segfault at 0 ip abae078c sp bfd99974 error 4 in nouveau_dri.so[ab67b000+a14000]
09:54debo: there is another error near the start:
09:54debo: ACPI BIOS Error (bug): \_SB.PCI0._OSC: Excess arguments - ASL declared 5, ACPI requires 4 (20160831/nsarguments-189)
09:55debo: on boot I have no GUI
09:55debo: i run: startx
09:56debo: there is no taskbar and I can not right-click the desktop
10:33pmoreau: debo: Could you please paste the whole dmesg somewhere?
10:35debo: hi pmoreau
10:35pmoreau: If you want the taskbar back, try restarting plasmashell
10:35debo: thanks for taking a look
10:35pmoreau: It might crash again though
10:35pmoreau: Thanks for the link
10:37pmoreau: Make sure to blacklist the NVIDIA driver, or uninstall it, if you want to use Nouveau
10:37pmoreau: It could conflict with Nouveau
10:42debo: i have purged with apt
10:43debo: i guess a while back I downloaded a tarball from the nvidia site but that's gone too I think
10:45pmoreau: As for, why you do not have a GUI on boot and need to run startx manually, I don’t think that is related to Nouveau as the first error Nouveau related happens ~230 seconds after boot.
10:46Tom^: "[ 17.584141] NVRM: No NVIDIA graphics adapter found!" something from nvidia still remains
10:47pmoreau: Are you using systemd? What does `systemctl get-default` return?
10:47pmoreau: And you have a display-manager configured?
10:48debo: how can I tell?
10:48pmoreau: Try `systemctl status display-manager`
10:49debo: active (running)
10:50debo: there are some dkms nvidia modules when I use locate nvidia
10:51pmoreau: Do you have anything on the kernel commandline reverting to text mode? `/proc/cmdline` should contain the cmdline of the current boot.
10:51pmoreau: Might want to remove those as well
10:52pmoreau: Tom^: Yes, the pastebin was from before he removed it.
10:52Tom^: i see
10:53debo: BOOT_IMAGE=/boot/vmlinuz-4.9.0-3-686-pae root=UUID=86dc975b-7529-422a-9624-f8155bba9f10 ro reboot=efi quiet
10:56debo: should I remove these files e.g.
11:00debo: or this one? /usr/lib/i386-linux-gnu/dri/nvidia_drv_video.so
11:02pmoreau: Hum… they are probably safe to remove manually if they aren't tracked by your package manager.
11:05debo: aptitude search nvidia | grep ^[ic] returns nothing
11:18karolherbst: mupuf: is there some kind of script to update the kernel on reator?
11:18karolherbst: mupuf: would be nice if we could just run something and it would take care of everything
13:32karolherbst: gnurou: is there by any chance another falcon we could use for ACR?
13:45karolherbst: maxwell2 reclocking without hacks!
13:49karolherbst: kind of hacky, but less hacky than having to load/unload/load nouveau: https://github.com/karolherbst/nouveau/commit/b16f4a89940c395302a3bc3dd2f425a2cff737ab
13:49karolherbst: should probably call nvkm_subdev_fini() first
13:52karolherbst: imirkin: if somebody asks about external controlled fans on maxwell2 and reclocking, two top commits: https://github.com/karolherbst/nouveau/commits/maxwell2_playground
14:52pmoreau: karolherbst: \o/ Good job! Too bad I don’t have one lying around… :-/
14:57karolherbst: pmoreau: well, I need to add a flag to disable it by default for now though
14:57karolherbst: and messing around witht he subdev is also kind of messy
14:57karolherbst: maybe skeggsb_ has a better idea on how to handle this
14:58karolherbst: but secboot also puts the pmu into a state the pmu subdev can't use the pmu anymore
15:51KOLANICH: Hi everybody. Xorg shows black screen. Xorg.0.log seems to be OK.
15:51KOLANICH: What should I do to make it work?
15:52karolherbst: how do you start Xorg?
15:54KOLANICH: 1 sddm, 2 manually from root. In both cases it doesn't work.
15:55karolherbst: could you upload your Xorg.0.log file somewhere and dmesg?
16:14KOLANICH: karolherbst: https://pastebin.com/XW4PcyDh https://pastebin.com/26XAt3L7
16:14KOLANICH: they are sanitized a bit
16:21karolherbst: KOLANICH: mhh seems fine, maybe your sddm is kind of borked? you should check the logs there as well
16:22KOLANICH: karolherbst: it seems it's not sddm. I tried to start xorg manually and then try to start any gui app, it says it cannot connect to xorg
16:22Tom^: if im reading that Xorg.log right its a nv11 card. isnt that uh quite old :p ?
16:23KOLANICH: p.s. I have set the variable $DISPLAY to :0
16:24KOLANICH: Tom^: it is, I have found a really old PC and want make it work
16:34karolherbst: KOLANICH: try "DISPLAY=:0 glxinfo"
16:36KOLANICH: karolherbst: >unable to open disppay
16:37karolherbst: maybe your Xorg server opens another one? not quite sure, but if x is running it should work
16:37karolherbst: try to start x like this; "Xorg :0"
16:39KOLANICH: karolherbst: the same
16:42karolherbst: KOLANICH: could it be that X just exists?
16:42KOLANICH: karolherbst: what do you mean?
16:43karolherbst: I mean it shouldn't quit, but maybe it does and that's why starting something fails
16:43KOLANICH: karolherbst: no, I have to kill it manually, so it's working
16:44karolherbst: maybe something really odd is going on, also I have no idea about the status of nv11 gpus with nouveau
16:44karolherbst: it shouldn't prevent the X display to be accessable
16:56KOLANICH: karolherbst: have installed xinit, typed startx and got an empty screen with mouse cursor. Still unable to connect to it using glxinfo
16:56KOLANICH: karolherbst: thank you all anyway
18:00karolherbst: nvidia doesn't read the TEMP_CAL_OK fuse anymore
18:00karolherbst: on pascal
18:02karolherbst: okay and now why can't I read out the fan speed on maxwell2
18:06karolherbst: doesn't matter anyway
18:08Tom^: you mean i can replace this 780ti for a maxwell soon? =D
18:09karolherbst: Tom^: only if the fan isn't controlled by the GPU itself
18:09Tom^: put the fan pin on the motherboard?
18:09Tom^: or simply just watercool it heh
18:09karolherbst: well there are multiple ways
18:10Tom^: interesting, gonna see if there are any second hand or refurbished cheaply here, instead of waiting for that amd restock.
18:10karolherbst: laptops usually don't have GPU controlled fans
21:28karolherbst: odd, calling the pmu fini function causes some serious messup, oh well
21:31karolherbst: anyhow top three patches: https://github.com/karolherbst/nouveau/commits/maxwell2_playground