04:18esperegu: hi. I booted into kde with nouveau but when I log in I just see my background color and when I click I get a full screen 'menu' only displaying some icons on the right. anybody a suggestion on how to solve this? thx!
04:19karolherbst: esperegu: maybe your kde session is somehow messed up?Or does it just works with nvidia?
04:20esperegu: karolherbst: yes. If I install latest nvidia it shows the desktop
04:21karolherbst: esperegu: okay, then dmesg and X log
04:21esperegu: karolherbst: I had a desktop before I rebooted though. cannot seem to get it back. Xorg took 100% cpu and a long time to start
04:22karolherbst: yeah maybe there is something wrong with your installation then and it tries to get nvidia again or something.
04:26esperegu: karolherbst: http://dpaste.com/1FKCV5A dmesg
04:27karolherbst: ohh you have two nvidia gpus
04:28karolherbst: k, dmesg looks fine though
04:31karolherbst: esperegu: so yeah, x log would be nice then ;)
04:31esperegu: karolherbst: yeah. to big. looking for other one ;-)
04:31karolherbst: ohh no, not the xsession one :D
04:31karolherbst: the x server log is in /var/log/xorg.0.log usually
04:31karolherbst: or .old for the last one
04:33esperegu: karolherbst: https://drive.google.com/file/d/0B8m6t6fjiR--X3d3M1hJODZCeEk/view?usp=sharing
04:35esperegu: karolherbst: I recall disabling one hdmi output (of the onboard hdmi)
04:35karolherbst: esperegu: do you have anyhting inside /usr/share/X11/xorg.conf?
04:37esperegu: karolherbst: dont have one
04:37esperegu: karolherbst: I disabled it in kde settings
04:37karolherbst: esperegu: do you still have the nvidia stuff isntalled?
04:37esperegu: karolherbst: no I purged it
04:37karolherbst: ohh wait
04:37karolherbst: this should be also fine
04:38karolherbst: ohhh wait
04:38karolherbst: did you set the tearing prevention to re-use screen content?
04:38esperegu: karolherbst: not that I am aware off
04:39karolherbst: I know that this might cause issues
04:39karolherbst: other then that, no idea. You could try to start a plain X with xterm from a tty and just start kwin and see where it gets you
04:42esperegu: karolherbst: k. I just killed X. then it shows me a login window properly. then logging in xorg goes: 2823 root 20 0 278748 32616 20400 R 98,0 0,1 0:27.96 Xorg
04:44esperegu: hmm. trying to select failsafe on the login seems to hang =)
04:48karolherbst: mhhh I had my fun with stuff like that on ubuntu :/ something was just misconfigured but no idea what it was
04:50esperegu: karolherbst: mja. I now logged in under a different user. get the desktop but with a plasma crash
04:50karolherbst: what does the crash say?
04:51esperegu: karolherbst: closed it.
04:51esperegu: I will relogin
04:51esperegu: hmm. laggy
04:54esperegu: hmm. now it does not crash
04:54esperegu: let me try and set it up proprely now with this user...
04:56esperegu: xorg going to 100% again
04:59esperegu: ok. now this user has the same.
04:59esperegu: what I did was turn one display clockwise, disabled one and put 2 others on the correct position
05:00karolherbst: ohh so you have like 3 displays?
05:01esperegu: karolherbst: yes
05:01esperegu: karolherbst: the one I disabled is the onboard one (which I only use for audio)
05:41esperegu: karolherbst: looks like it happens when I move the location of the monitors in in the screen (rotating it did not seem an issue)
05:42karolherbst: esperegu: do the resolutions differe in some way?
05:43esperegu: karolherbst: which?
05:43esperegu: karolherbst: after moving?
05:43karolherbst: in general
05:43karolherbst: or have all displays the same resolution?
05:44esperegu: karolherbst: no. I have one big one of 2560x1440 and 2 of 1680x1050
05:44esperegu: after I moved them the rotated big one does not use the whole screen anymore
05:45kugel: karolherbst: hello. I'm wondering about the state of your stable kepler reclocking branch. It's not mainline yet is it?
05:45karolherbst: kugel: no, volting has to be fixed first
05:54whompy: Quick question: do the nouveau.config=NvClkMode kernel command line strings still parse in 4.5? Looks like no change is required there.
05:55karolherbst: whompy: it should afaik
05:55karolherbst: whompy: do you ask because of the pstate thing?
05:57whompy: Yeah. Cleaning up post change (thanks for working on that BTW!)
05:59karolherbst: mupuf: I guess you won't find time to look into the nvafakebios stuff the next days? Because then I could try to do that so that we can take care of that volting stuff faster :) But I have no idea how that vbios faking works at all, so is there somewhere a nice thing to read up on?
06:01esperegu: [ 89.036284] show_signal_msg: 30 callbacks suppressed
06:01esperegu: [ 89.036289] kactivitymanage: segfault at 7f5122f76cd0 ip 00007f51091ce1b1 sp 00007ffc1da61f88 error 4 in libQt5Sql.so.5.4.2[7f51091ba000+3f000]
06:02karolherbst: uhhh right
06:03karolherbst: qt5 5.4 is somehow broken with multiple displays
06:03karolherbst: might want to try out either 5.3 or 5.5
06:03karolherbst: but you won't get either with your ubuntu version :/
06:03karolherbst: like switching displays crashes every qt5 application
06:04karolherbst: you might think they would backport patches or update to 5.5 ;)
06:04karolherbst: but guess what, 5.5 is not tested and backporting patches isn't that easy :D
06:05esperegu: sucks. grrr
06:06karolherbst: esperegu: maybe this helps: https://wiki.ubuntu.com/Touch/QtTesting
06:08esperegu: karolherbst: will try that than. thx
06:35esperegu: karolherbst: still not functioning properly. I suppose I have to wait a bit more till it is still a bit more common =)
06:36karolherbst: could be also a nouveau issue in the end or both
06:57mupuf: karolherbst: the code is pretty straightforward
06:57mupuf: I can plug the GM206 GPU also
06:59karolherbst: I can test with my gpu as well
06:59karolherbst: I still have two laptops here so I will be fine
07:00mupuf: hmm hmm
07:00mupuf: well, your problem seemed different than mine
07:00karolherbst: ohh I doubt it though
07:01karolherbst: I am pretty sure it will be the same in the end
07:05karolherbst: k, now the crash test
07:06mupuf: nope, it for sure is not
07:06mupuf: In my case, it is just the blob not picking up my custom one
07:06mupuf: you get a crash
07:07mupuf: in my case, there are additional checks done at the end
07:08not_karolherbst: soo, this will make suff easier I guess
07:08not_karolherbst: ... or not?
07:09karolherbst: mupuf: k, so I thought I could generate a stack or something, but even pstore is empty
07:09karolherbst: after faking the vbios, kernel completly freezes
07:10mupuf: karolherbst: stop writing to random addresses :D
07:10karolherbst: or maybe not, at least it doesn't give me any keyboard input anymore, sysrq keys doesn't work, ... ohh wait, I can check if my uefi still works
07:10karolherbst: at least the acpi parts, cause I can put my fans into full blast mode with a key combination
07:10karolherbst: without involving the kernel
07:11karolherbst: ... which sounds scary if I put it that way
07:11esperegu: karolherbst: did not change anything but I have a desktop now. lets see if it keeps working ;-)
07:11karolherbst: mupuf: k, so why does it work with other gpus then? Because the random address works for most gpus?
07:12karolherbst: or most systems?
07:12esperegu: hmmm... mplayer crashed now =(
07:12mupuf: nope, because the address is not random on other cards
07:12karolherbst: ahhh okay
07:12mupuf: it is likely initialized by the BIOS
07:12karolherbst: but I thought nvafakebios always uploads to the same memory region?
07:13karolherbst: ohh right, which makes it non random
07:13mupuf: just look at the code :)
07:14karolherbst: I see nva_wr8 calls
07:15karolherbst: i reads the address out from nva_cards[card]->bar0
07:17esperegu: Failed to open VDPAU backend libvdpau_nouveau.so: cannot open shared object file: No such file or directory
07:17esperegu: [vdpau] Error when calling vdp_device_create_x11: 1
07:17esperegu: Error opening/initializing the selected video_out (-vo) device.
07:17karolherbst: esperegu: you need nvidia firmware for that and the file names and stuff have changed
07:18mupuf: karolherbst: oh dear, you are missing it :D Give me a sec
07:19mupuf: 0x619f04 <-- this is the reg that should contain the base address for the VBIOS shadow copy
07:19esperegu: karolherbst: I don't have XXX FUCK.. I even don't have a /lib/firmware/nouveau folder
07:19karolherbst: mupuf: ohh it is 0x1 for me
07:20mupuf: over-writing the first 32K of memory sounds ... bad :D
07:20karolherbst: why does the kernel allow this :D
07:20esperegu: karolherbst: apt-getted tehfirmware
07:20mupuf: wait, sorry, I am telling shit
07:20karolherbst: esperegu: well you need to reboot an all then
07:20karolherbst: mupuf: well bar0 is also pretty random for me, if that matters
07:21mupuf: not exactly sure what would be mapped at address 1 from the GPU PoV, but pretty sure it is bad :D
07:22karolherbst: 0x1700 is also like 0x0, but that figures I guess
07:23esperegu: karolherbst: oeps. then my desktop was gone again. lol
07:23karolherbst: I think I slowly understand how that works now
07:23karolherbst: okay, so there is stuff written into maped vram and we will the gpu where to look and then nvafakebios uploads the stuff there
07:25mupuf: well, it does not matter really matter for the gpu
07:26mupuf: it is for the driver and the BIOS
07:26karolherbst: ohh okay
07:26mupuf: which is probably why it is not set up for you
07:26mupuf: because the BIOS does not need it
07:26mupuf: it uses the intel gpu for displaying stuff
07:26karolherbst: mhhh I have a switch in my vbios though for some stuff I don't understand
07:27karolherbst: dgpu related
07:27mupuf: hmm, maybe it would work if you forced it to use the dgpu
07:27mupuf: but that would mena you have a mux
07:28karolherbst: I can toggle PX 5.0 between BAC0 and something else
07:28karolherbst: but that somehow sounds fan related when I look for this
07:29karolherbst: no I can't enable only the dgpu though
07:29karolherbst: mupuf: is there some pattern to the value inside 0x619f04?
07:30karolherbst: maybe I could just use a value from another kepler card and it would just work
07:31karolherbst: mupuf: funny, on my fermi laptop it is also 0x1
07:32mupuf: well, I don't think it is a good idea
07:32mupuf: what I think really is supposed to happen is that the bios would allocate memory for shadowing the vbios
07:32mupuf: the question is, how does it work when we try to fake the bios of a secondary gpu
07:34karolherbst: maybe we can't?
07:34karolherbst: or maybe another question: how did somebody found how how to shadow the vbios in the first place?
07:46karolherbst: imirkin, mupuf: by the way: I got in touch with the tracer maintainer of the kernel to look into the mmiotraceing issue, will send a official bug report to the linux email soon enough, I will probably CC the nouveau mailing list as well
08:01esperegu: karolherbst: don't think this is normal? http://dpaste.com/14FWK64
08:14karolherbst: esperegu: usually not
08:26esperegu: hmm still cant find libvdpau_nouveau.so
08:27RSpliet: esperegu: did you ask your package manager for the corresponding package?
08:27esperegu: RSpliet: who is? ;-)
08:28RSpliet: idk, is that ubuntu? something like apt-file update, apt-file find libvdpau_nouveau.so ?
08:29RSpliet: double-check you have the right package installed for the same arch as your video player
08:29RSpliet: esperegu: ^
08:30esperegu: RSpliet: all there is: http://dpaste.com/3JXK62H
08:31RSpliet: esperegu: I gave you those particular commands for a reason
08:31RSpliet: the package might not have nouveau in the name
08:33esperegu: RSpliet: u right... by the way I did not knew that apt-file command. thx a lot. thats pretty handy
08:35RSpliet: geen probleem
08:45esperegu: desalnieteminbedankt =)
08:50esperegu: hmm. now I get: [vdpau] Failed creating VDPAU decoder: An invalid/unsupported VdpDecoderProfile value was supplied.
08:52RSpliet: that could imply that the specific video (codec, encoding settings, resolution, whatnot) is not supported by VDPAU on your device
08:53RSpliet: with nouveau
08:53esperegu: RSpliet: ok. thx. back to nvidia then :-(
08:53RSpliet: ubuntu packages are notorious for not being up to date
08:55RSpliet: https://nouveau.freedesktop.org/wiki/VideoAcceleration/ gives an overview of what's supported roughly
08:56RSpliet: also: double-check your firmwares, otherwise none of the profiles will be supported
09:03esperegu: GF108 [GeForce GT 520] (rev a1) & GF119 [GeForce GT 520] (rev a1). # dmesg | grep -i chipset
09:03esperegu: [ 1.528786] nouveau [ DEVICE][0000:01:00.0] Chipset: GF108 (NVC1)
09:03esperegu: [ 3.113440] nouveau [ DEVICE][0000:02:00.0] Chipset: GF119 (NVD9)
09:04karolherbst: esperegu: there should be something in dmesg after trying to get video acceleration
09:04esperegu: karolherbst: u right:
09:04esperegu: [ 214.236500] nouveau 0000:01:00.0: Direct firmware load for nouveau/nvc1_fuc084 failed with error -2
09:04esperegu: [ 214.236516] nouveau 0000:01:00.0: Direct firmware load for nouveau/nvc1_fuc084d failed with error -2
09:04esperegu: [ 214.236520] nouveau E[ PMSVLD][0000:01:00.0] unable to load firmware data
09:04esperegu: [ 214.236523] nouveau E[ PMSVLD][0000:01:00.0] init failed, -2
09:05karolherbst: RSpliet: http://packages.ubuntu.com/trusty/all/nouveau-firmware/filelist :D
09:05karolherbst: ofc there is no newer version....
09:06karolherbst: but this is for the context stuff only
09:06karolherbst: the hell
09:07karolherbst: esperegu: https://nouveau.freedesktop.org/wiki/VideoAcceleration/
09:07karolherbst: "Firmware" section
09:13karolherbst: pmoreau: 18th feb 5pm UTC, important :p
09:13mupuf: karolherbst: great news for mmiotrace!
09:13mupuf: Was it pq? Pekka Paalanen?
09:15mupuf: the maintainer
09:15karolherbst: no clue
09:15karolherbst: wasn't really touched this decade I think
09:16karolherbst: mupuf: https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/log/kernel/trace/trace_mmiotrace.c?id=refs/tags/next-20160208
09:16karolherbst: yeah seems to be pekka
09:17karolherbst: I wrote to steven and ingo though, because there may be more active. I am in touch with steven now anyway, so maybe he has any good idea
09:18karolherbst: stevens seems happy though to see it is still used :D
09:26karolherbst: okay now back to something we have to fix ourselves :D
09:26karolherbst: any idea what we could try out regarding faking vbios?
09:34mupuf: well well well, we would need to alocate memory to hold the bios
09:34mupuf: but ... that may require a kernel driver
09:35mupuf: a very simple one, but still
09:35karolherbst: maybe we could use hugetables for that?
09:35karolherbst: no idea how that can be pushed into kernel space though
09:35karolherbst: there ise hugetblfs or something
09:36mupuf: maybe nouveau could be nice and allocate memory for us
09:36mupuf: that is actually a good idea, checking if the value returned is invalid
09:37mupuf: and then allocate memory for it on loading the module
09:37mupuf: then, nvafakebios could work
09:37mupuf: but ... how to guarantee that the blob will not try to overwrite the memory later :s
09:38karolherbst: mupuf: https://lwn.net/Articles/375096/ "2 HugeTLBFS" wouldn't that be somehow possible to use too?
09:38mupuf: no idea how this actually works ... we should write to system memory instead
09:38mupuf: you will not know the physical address
09:38mupuf: and that's what the gpu sees
09:38mupuf: and we may need ot configure the IOMMU too :s
09:38karolherbst: ohh right
09:39karolherbst: it is getting ugly now
09:39Guest66899: I am confused about the various components in the graphics stack.
09:40Guest66899: Right now what does it work with up to date nouveau? DRI3, glamor, gallium...
09:41esperegu: karolherbst: thx! vdpau works. still that strange desktop behaviour though :-( probably not yet fixed in 5.5 I suppose
09:41karolherbst: mupuf: I guess on the desktop gpus the gpu already claims such a memory area which just works?
09:41mupuf: karolherbst: I am sure it is the vbios doing so
09:41Guest66899: AIGLX etc.
09:41mupuf: and the kernel do not touch some parts of the menory?
09:41imirkin_: Guest66899: you shouldn't have to worry about any of that
09:42karolherbst: mupuf: mhhh
09:42imirkin_: Guest66899: just make sure you have the latest xf86-video-nouveau, latest mesa, and latest libdrm.
09:42imirkin_: Guest66899: in rare situations, the version of X also matters, but usually not.
09:42karolherbst: *cough kernel cough*
09:42Guest66899: yeah, I got all the latest.
09:43imirkin_: Guest66899: if you want to understand how it all plugs in, i think danvet had a thing somewhere.....
09:44danvet: that was someone else
09:44danvet: and fairly old
09:44danvet: might need an update
09:44imirkin_: i tried searching, but couldn't find it
09:45imirkin_: there's this one http://dri.sourceforge.net/doc/images/control_flow_poster.jpg
09:45imirkin_: but not the one i was thinking of
09:46imirkin_: oh, and it's describing DRI1
09:51esperegu: nizzzz. now I get 100% cpu on systemsettings5
10:15pmoreau: karolherbst: Yep!! Saw that, and secured the evening! :-)
10:18karolherbst: pmoreau: :) but it seems like we might be able to ask questions as well
10:19pmoreau: Well, I'm waiting to learn more about if before thinking about possible questions. :-)
10:23pmoreau: How is the mmiotrace thing going? I saw you sent an email and mentioned talking with one of the devs, in the logs?
10:42karolherbst: pmoreau: yeah well, still bad, but I am in touch with the trace maintainer now
10:43karolherbst: pmoreau: funny that they weren't aware of the fact that we are still using it :)
11:01pq: mupuf, wuh, what?
11:03pq: no, I deny all maintainership on mmiotrace, I havent' touched it in a decade :-P
11:05orbea: is it right that this dmesg spam with mpv is another sign of concurrent gl+vdpau calls not working with nouveau? http://dpaste.com/2ZR5RZ8
11:06imirkin_: orbea: could be. or could be mpv doing something funky.
11:07imirkin_: orbea: just use mplayer :)
11:07orbea: i found myself using both now
11:07imirkin_: mplayer decodes all videos, that's why i use it.
11:07orbea: mpv has all that ytdl integeration that is hard to replace, mplayer like you said does hdwdec better
11:07imirkin_: every other player fails at decoding some group of videos
11:08orbea: i very much like being able to play bandcamp links with mpv for istance, well dont need hwdec for that anyways
11:11linkmauve1: Oh, mplayer does hwdec properly now?
11:11imirkin_: it always has
11:12linkmauve1: Maybe I should try it again, I switched to mpv partly due to its support of vaapi on wayland.
11:12imirkin_: vaapi is broken
11:12imirkin_: use vdpau
11:12linkmauve1: Works pretty fine on all of my Intel machines. :p
11:12imirkin_: and only intel.
11:12linkmauve1: Sorry, I don’t use Nouveau.
11:13orbea: yea, vaapi is all that works on my intel too
11:13linkmauve1: But AFAIK mpv’s vdpau also works quite perfectly.
11:13imirkin_: apparently va-api has some shortcoming that prevents it from doing mpeg4 properly
11:13orbea: not in nouveau
11:13imirkin_: linkmauve1: maybe. but it fails at decoding or maybe seeking or maybe aspect ratio correction. i forget which one i ran into trouble with it.
11:14imirkin_: every player other than mplayer had issues with at least one of those
11:14imirkin_: on occasion mplayer also has issues, but no other player handles those videos correctly either
11:15orbea: too bad since mpv has come a long way in a good direction with other regards
11:15linkmauve1: I have yet to encounter any issue of those kinds with mpv.
11:15imirkin_: being able to properly play videos is kind of a killer feature for me
11:16linkmauve1: Same here. ^^
11:16imirkin_: if you can't play the video, any other features you might have don't matter
11:16orbea: linkmauve1: when using nouveau and mpv and hwdec, -vo opengl or vdpau, do you get a lot of dmesg spam?
11:16linkmauve1: orbea, I don’t have any machine with a Nouveau-capable GPU. :/
11:16linkmauve1: You should use --vo=opengl --hwdec=vdpau, which is the default anyway.
11:17linkmauve1: (Well, --hwdec=auto is the default, which will try vaapi and vdpau, I don’t remember in which order.)
11:17imirkin_: linkmauve1: nouveau fails at concurrent access from multiple threads, which is the issue orbea ran into.
11:19linkmauve1: imirkin_, oh, and mpv does that?
11:20imirkin_: well it's perfectly legal behavior
11:20linkmauve1: Here it seems to only do GL calls from the vo thread.
11:20imirkin_: it's actually an issue that's starting to get hit by real software out there, e.g. warsow 2.0, and i noticed even grid autosport does it.
11:20imirkin_: linkmauve1: and it does vdpau calls from the vdpau thread...
11:21linkmauve1: Ah, I see.
11:24Guest66899: wow, that chart is madness.
11:25Guest66899: os one thing, is ok for nouveau to use AIXGL?
11:25imirkin_: Guest66899: complex things are complex.
11:25imirkin_: Guest66899: AIGLX is indirect glx i guess? that _ought_ to work, yes
11:25imirkin_: but it's not a nouveau thing
11:25imirkin_: it's an X thing
11:26Guest66899: it's what's using as default.
11:29imirkin_: but perhaps you're seeing something in the logs which makes you think that.
11:29imirkin_: even though it's not the case.
11:37Guest66899: Xorg.0.log several lines
11:40Guest66899: AIGLX, yes.
11:40Guest66899: by the way, is it wrong if there's no KMS lines in Xorg.0.log?
11:42imirkin_: if you like, i can look at your log and see if there's anything odd
11:42Guest66899: so let me get this straight. GLAMOR is NOT needed for nouveau? Because the driver I am using has the option to compile it with or not.
11:42imirkin_: if you pastebin it somewhere, i'll have a look
11:42imirkin_: glamor is only required for maxwell gpu's, and only with nouveau 1.0.11
11:43imirkin_: 1.0.12 just dropped all that junk in favor of the much more appropriate xf86-video-modesetting + glamor integration
11:43imirkin_: (i mean xf86-video-nouveau 1.0.12 of coruse)
11:47Guest66899: I have 1.0.12
11:47Guest66899: why is it odd the AIGLX lines? What should it be instead?
11:47imirkin_: those lines are fine
11:48imirkin_: but it doesn't mean that you're actually using indirect glx :)
11:48Guest66899: oh I see.
11:50Guest66899: imirkin_: do you know how to force nouveau to use DRI3? I get a message about 'module no used'
11:51Guest66899: some places the option is "DRI3" and others "DRI" "3"
11:51imirkin_: Guest66899: Option "DRI" "3" iirc
11:51imirkin_: should be in the man page
11:52Guest66899:rushes to change it
11:53karolherbst: pq: ahh so you _are_ there sometimes :p
11:54karolherbst: pq: we have a serious problem though, really serious
12:01pmoreau: l1k: Ping
12:02l1k: pmoreau: pong
12:02pmoreau: l1k: No lockup using nofbaccel, which makes sense since the lockup function is called by `nouveau_fbcon_imageblit()`
12:03pmoreau: And to answer your other question: a bit of freeze and corruption when the lockup happens, but nothing major.
12:05l1k: pmoreau: interesting. this looks like a bug that's been there anyway and is just exposed by the switching patches, but not actually caused by them.
12:06pmoreau: *I think* I hit those a long time ago (by long time ago, probably ~3 years ago while I was beginning on Linux and having a few troubles with Nouveau)
12:07pmoreau: But I don't remember how it went + the laptop started locking up which was even more inconvenient.
12:07l1k: pmoreau: without the patches, the driver allocates a framebuffer with 1024x768 but probably doesn't configure a crtc, which would explain why it doesn't occur in that case.
12:07pmoreau: Dunno if I had opened a bug report or talked about it at the time, I'll try to dig it up
12:09l1k: pmoreau: so basically we're missing something on the G96 when waking it from sleep that breaks fbcon acceleration. something that's not missing on newer cards like the GK107.
12:11l1k: pmoreau: we're probably missing some kind of init step. I'll see which classes are used by G96 versus GK107 and will try to find any differences.
12:12l1k: pmoreau: thanks a lot for testing this.
12:13pmoreau: Hum, init thingy… I don't remember when I stopped testing without ForcePost=1… Need to try that again
12:13pmoreau: And maybe PCI reset as well
12:14l1k: pmoreau: so with nofbaccel switching works just fine? did you try a couple of times?
12:15pmoreau: Yep, just put it in a loop and did at least 10-15 switches
12:15pmoreau: And tried starting X, running glxgears, killing X, reswitching a few times, restarting X, and so on
12:16l1k: pmoreau: cool! great to hear. well not *that* great since the choice is now between switching and fbcon accel, but it feels like a step forward. :)
12:17pmoreau: Indeed! We have a *solution* to *fixing* the lockup while switching :-D
12:19Guest66899: imirkin_: thanks! so far has done the INSTANT crash.
12:20karolherbst: why do I still have internet connection
12:20imirkin_: Guest66899: sorry, not sure how to interpret that... "has done"?
12:20pmoreau: I do remember that bug https://bugs.freedesktop.org/show_bug.cgi?id=58556, was some time ago. :-)
12:20karolherbst: mupuf: first time I did the mmiotrace with getting the issue without my wifi adapter dying
12:20imirkin_: Guest66899: are you saying the situation has improved, or deteriorated?
12:21pmoreau: karolherbst: Oo Too bad… :-p
12:30not_karolherbst_: does anybody know how I can safe /proc/kcore ?
12:35l1k: pmoreau: so on boot, both cards are initially turned on, right? if you switch to the G96 for the first time, fbcon accel works just fine, is that correct? (if nofbaccel is not specified of course)
13:04Guest66899: improved, now I have to try with the dreaded ppsspp :(
13:10pmoreau: l1k: Sorry, had been fighting to connect to the Internet. Worked fine over ethernet on Mac OS, didn't got any reply from the router on Linux…
13:11pmoreau: l1k: Correct, both are on upon boot. And yes, switching the first time to the G96 works fine.
13:43l1k: pmoreau: hm, this can't be that hard, we're just missing some init step for fbcon accell when waking the card...
14:05Guest66899: dead on the same spot!
14:05Guest66899: is not even 3D
14:17imirkin_: what mesa version?
14:36AlexAltea: does anyone know what commands *are* allowed between a pair of NV4097_SET_BEGIN_END's?
14:36AlexAltea: I assume things like changes in the PGRAPH state is not allowed
14:36imirkin_: AlexAltea: a very restricted set... but i def don't have an exhaustive list
14:37imirkin_: AlexAltea: in general just sending down vertices/vbo/idxbuf-related items
14:37AlexAltea: imirkin_, I see.. thanks! :)
14:38drathir: there is any known issue with xfce4 and freezing after few sec of login with Asus R556L with 00:02.0 VGA compatible controller: Intel Corporation Broadwell-U Integrated Graphics (rev 09) and 04:00.0 3D controller: NVIDIA Corporation GM108M [GeForce 940M] (rev a2) onboard... the crashing occur with sna and glamor and only one possible to fix is run with uxa, but also freeze with 3d app or glx runin, but normal os operations possible...
14:39imirkin_: drathir: nouveau should be refusing to init the GM108 gpu anyways, so it's unlikely to be nouveau-related
14:39imirkin_: drathir: unless you're running with a patch to enable init of GM108?
14:41drathir: imirkin_: even tried blacklist nouveau but its the same, if not blacklisted kernel boot report error nouveau 0000:04:00.0: unknown chipset (118070a2)...
14:42imirkin_: drathir: yeah, that's expected
14:42imirkin_: drathir: nouveau basically ignores the GPU
14:42drathir: there is no patches performed by me only clean install of arch official repo...
14:42drathir: that mean its intel related?
14:44drathir: it detect prime/optimus platform, but i guess if boot error is reported not load sucessfully drivers..
14:44imirkin_: i'm fairly sure nouveau has nothing to do with your issues
14:46drathir: imirkin_: really thanks a lot i will search at intel side than, btw there is a way to nouveau handle that gpu or only nvidia drivers to use?
14:52imirkin_: drathir: it should be pretty trivial to enable nouveau to load for it, but it'll do you little good - without reclocking, it won't be faster than the intel gpu
14:55Guest66899: imirkin_: 1.11.1 and it's just ONE game, I've tested others on the ppsspp and it's just one game.
14:55Guest66899: and this single game kills nouveau real hard
14:55orbea: what game?
14:55Guest66899: metal gear
14:56orbea: i haven't tested that one here
14:56Guest66899: passed the second mission with bad guys, and it crashes hard not on 3D game but the interactive cartoon!!
14:57drathir: imirkin_: that intel in uxa mode is so slow that even in ff watching videoclip makin a problems to switch into fullscreen...
14:59imirkin_: drathir: probably no accel... you need to update something
15:05drathir: imirkin_: that is funny it have uxa accel with noaccel it even not refresh the screen but looks like no crashing...
15:06drathir: probably drivers messed in some way...
15:06drathir: no refresh screen mean only screen update when mouse moved...
15:12imirkin_: Guest66899: you might try getting mesa from git, but i doubt it'll help your issues
15:12imirkin_: Guest66899: chances are the most expedient way to work it out would be for me to be able to repro at will.
15:16Guest66899: imirkin_: get ppsspp, play Metal Gear, I can send you the save state up to the point leading to hard crash.
15:16imirkin_: save state is probably good
15:16Guest66899: Metal Gear Solid: Peace Walker
15:17Guest66899: ppsspp 1.1.1
16:38orbea: i get this with setting the high pstate only, I'm using linux 4.4.1 http://dpaste.com/3PC9GHN is it expected?
16:43imirkin_: orbea: not _really_ -- it means we don't do something properly. i think karol has a slew of patches to fix it up
16:44orbea: cool, so perhaps in the next kernel?
16:44vita_cell: I don't know if it will work, I think that not,
16:45vita_cell: kernel 4.3 is not ready for reclock, karol gave me some fixes, that I needes for compile and throw that in the 4.3
16:46vita_cell: but, if you install 4.4, you need to do nothing, all ready for reclock, only add nouveau.pstate=1 to you grub and run update grub, then, reboot and reclock
16:47orbea: im on 4.4.1
16:48vita_cell: you can not reclock, you need fixes, and then compile them
16:48vita_cell: the easyest way is download 4.4, install it, and you no need to do nothing more
16:49vita_cell: 4.4 ready for reclock
17:03imirkin: orbea: you need additional fixes. like i said, karol has a bunch that could address your problem, esp in a hacky way.
17:23orbea: do you know where they were posted?
17:24imirkin: they were not posted
17:25imirkin: they in one of a multitude of his github branches
17:25imirkin: you'll have to wait until he appears to figure out which
17:25orbea: yea, I can wait :)
17:27imirkin: you have a gtx780ti right?
17:27imirkin: i think those in particular have some sort of voltage misunderstanding situation
17:38vita_cell: gtx770 and gtx650, work fine with karol's fixes for 4.3
17:39vita_cell: but fixes no needes more, if use 4.4
17:39imirkin: perhaps not for you :)
17:39vita_cell: he need to install 4.4, the easyest way
17:41imirkin: i'll say it for the last time -- lots of situations are unhandled in 4.4 still. 4.4 helps a lot of situations, but hardly all.
19:43drathir: imirkin: k fixed... like alwaus user fault... for install used intel install script which one have little customized mkinitcpio which one have included intel agp module...