01:48hakzsam: imirkin, request sent
02:42pmoreau: RSpliet: Should I test the v2 on my G96?
02:44RSpliet: pmoreau: nah, that's pointless
02:44RSpliet: I pushed them to the ML because they work on NVA0
02:44RSpliet: (N=3 statistics)
02:45pmoreau: That's not too bad
12:14kisak: hello, I was wondering a bit about Lauri Peltonen of nVidia. I saw a handful of patches by this dev in the git log and did not recognize the name.
12:16imirkin: what sort of thing were you wondering?
12:19kisak: mainly if she's the same dev group as Mr. Courbot focusing on the mobile initiative / gk20a
12:19kisak: s/she's/she's in/
12:20kisak: and if there's plans for future contributions
12:20pmoreau: s/she/he ;)
12:20imirkin: most likely 'he'
12:20kisak: ah, my mistake guessing a name ending in a vowel is a female name
12:21imirkin: dunno exactly what his status is re working with/next to/whatever gnurou, but i've seen his name pop up a few times
13:43flipmess: hi, do i have to use DRI3 for the nouveau card to switch off automatically?
13:44pmoreau: flipmess: You don't
13:44flipmess: it says 0:DIS: :DynPwr:0000:01:00.0
13:44pmoreau: What is your setup?
13:44imirkin: flipmess: that means the kernel thinks it's being used somehow
13:44imirkin: flipmess: e.g. if you have a secondary output that's plugged in
13:44flipmess: nv has vga and intel lvds
13:45flipmess: xrandr also says vga-1-1 disconnected
13:45imirkin: i think vga might just keep it always on (not by design, but by accident)
13:45imirkin: i've seen this happen to a few people in the past
13:45flipmess: shoudn't xrandr then say vga connected?
13:45imirkin: i never tracked it down =/
13:45imirkin: well, the poweroff isn't instant
13:45imirkin: it waits until the card is idle for a while
13:46flipmess: i tried to force it with echo OFF > /sys/kernel/debug/vgaswitcheroo/switch but nothing...
13:46imirkin: i think the VGA causes it to poll every short while to see if something's there
13:46imirkin: yeah that does nothing
13:46flipmess: it's idling for hours now :)
13:46pmoreau: Well, it should power off the card if X isn't running
13:47flipmess: hm... should i restart x or just kill it and wait?
13:47pmoreau: (Maybe also if X is running, don't remember)
13:48pmoreau: Do you have any message in dmesg saying Nouveau doesn't want to power off for some reason?
13:48tobijk: it also powers down if x is in use
13:49tobijk: happens after some seconds normally
13:49pmoreau: tobijk: Ok, thanks.. I guess I was mixing with DIS/IGD
13:50flipmess: pmoreau << its just suspending and resuming as i suspend and resume the pc
13:50joi: can't you disable polling by passing poll=0 to drm_kms_helper?
13:51imirkin: yeah that sounds familiar
13:51flipmess: can i check somehow if it's polling?
13:52flipmess: would it stop if i unload the module?
13:52flipmess: turn off i mean
13:52imirkin: flipmess: try booting with drm_kms_helper.poll=0
13:52imirkin: unloading the module will sadly *not* put the gpu to sleep
13:53pmoreau: flipmess: Ok, so either VGA or you're using the NVidia card to drive the screen :)
14:04flipmess: it's off now.. as soon as i activated "reverse prime" it turned on and then quickly off
14:04flipmess: it turned on 2 more times for a second or so
14:06flipmess: are there any negative consequences from "drm_kms_helper.poll=0"?
14:06imirkin: might miss some hotplug actions
14:07flipmess: hm.. haven't seen any so far :D
14:07imirkin: i mean connection/disconnection of displays
14:13flipmess: will i always have to do the xrandr reverse prime customisation manually (or by script) or should it work automatically since vga port is only connected to nvidia?
14:14imirkin: should work automatically assuming the hotplug detect works correctly
14:14imirkin: worst case you can poll it by running 'xrandr'
14:14flipmess: thx.. i'll test that
14:30flipmess: since i installed nouveau xbacklight stopped working. xrandr --verbose shows backlight property only on LVDS2 (intel). when i try xbacklight nvidia card powers on for short time.
14:32imirkin: xbacklight must be deciding to use the backlight on the nvidia card, probably because it has LVDS-1
14:33imirkin: you could ensure that nvidia gets loaded after intel
14:33imirkin: it should probably check for connectedness, but... oops
14:34flipmess: hm.. that would make intel use LVDS1 again and nvidia lvds2?
14:34flipmess: it looks like every program "thinks" nvidia is the primary screen
14:37flipmess: i also get at the same time when nvidia turnes on and off some ACPI warnings
14:37flipmess: kernel: ACPI Warning: \_SB_.PCI0.PEG_.VID_._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20150410/nsarguments-95)
14:38flipmess: does this have to do with nouveau?
14:38imirkin: sssort of
14:38imirkin: it has to do with the acpi method nouveau calls
14:38imirkin: tbh i'm not sure if it's nouveau's fault or not
14:39flipmess: k. i couldn't find anything on google. thx
14:40imirkin: in the past i've blamed it on dumb acpi code
14:40imirkin: but could be that nouveau is invoking things wrong
14:42pmoreau: ACPI specifies the 4th argument of _DSM to be a package, however old BIOSes used a Buffer
14:43imirkin: and by "old" you mean "lots of brand-new hardware being produced"
14:43pmoreau: So for compatibility, ACPI specification allow a buffer as 4th arg.
14:43pmoreau: Probably :D
14:44pmoreau: But iirc, that's the term used by the specification of ACPI v4.0? Maybe even v3.x
14:45pmoreau: Nouveau could use another ACPI method to probe _DSM and avoid the warning :)
14:46imirkin: that'd be nice... it's super-common to see it in logs
14:47pmoreau: I did that when I added detection of Apple Gmux, but could change the existing Optimus dectection as well
14:48pmoreau:should re-submit those patches one day...
14:49RSpliet: tomorrow is one day
14:49pmoreau: Yeah, I've been saying that since Friday
14:51pmoreau: I'll try to push that and the EVO doc during this week
15:46meeeee: someone want to see a my dmesg dump? before i bother with bugzilla
15:48imirkin: what's the problem/
15:49meeeee: black screen in doomsday engine in ubuntu 14.04. unity with compiz on. G94/nv50 fam
15:50imirkin: what version of mesa?
15:53meeeee: 10.1.3. coreprof 3.3
15:53meeeee: [16173.047336] nouveau E[ PFB][0000:03:00.0] trapped read at 0x00207dd800 on channel 0x0001fbe9 [compiz] PGRAPH/TEXTURE/00 reason: PAGE_NOT_PRESENT [16173.064008] nouveau E[ PFB][0000:03:00.0] trapped read at 0x00207dd600 on channel 0x0001f764 [doomsday] PGRAPH/TEXTURE/00 reason: PAGE_NOT_PRESENT [16173.070298] nouveau E[ PFB][0000:03:00.0] trapped read at 0x00207dd800 on channel 0x0001fbe9 [compiz] PGRA
15:53imirkin: hm, well, a boatload of stuff has been fixed since then
15:54imirkin: mesa 10.5.6 was just released
15:54tobijk: mh looks somehow like the oom i had on nv50 a few weeks ago :)
15:55imirkin: the PAGE_NOT_PRESENT thing is weird indeed... sounds like some sort of resource tracking is messed up =/
15:57meeeee: the log is about 1000 lines of those
15:57meeeee: so recompile mesa?
15:59imirkin: that'd be the first thing i try
15:59imirkin: (do grab 10.5.6 -- it fixes a small regression in the 10.5 series for pre-G200 nv50 gpu's)
16:01meeeee: do i need to recompile the drm /ddi?
16:01meeeee: so just mesa?
16:03meeeee: well actuallz..mz mesa is old (default ver on my distro) but the drm is from about week ago ....
16:05imirkin: oh, make sure you don't have libdrm 2.4.60 -- that one's bad for nouveau
16:05imirkin: 2.4.59 or 2.4.61 is fine
16:06meeeee: gonna check
16:07imirkin: skeggsb: did swtnl ever actually work for nv30?
16:08skeggsb: i *believe* so, but it's entirely possible that i'm misremembering that
16:08imirkin: i'm thinking of doing a mega-bisect
16:11tobijk: 6.0 - today? :D
16:11imirkin: well the current nv30 driver only appeared in 9.0 or 9.1
16:11imirkin: so there's a limit to the quantity of bisect
16:14imirkin: i think there's like a *core* step missing, but i'm not seeing it
16:14imirkin: with llvm disabled, it never even hits the nv30 draw functions
16:14tobijk: and with llvm it does get hit? :O
16:15imirkin: skeggsb: btw, i had a weird hang with my nv44... the notify object never increased its fence count. i unbound the pci device from nouveau and rebound it, and it was all fine
16:15imirkin: skeggsb: is doing something like that in any way reasonable automatically?
16:15imirkin: (not the literal unbind/bind, but the equivalent, when we detect channel hangs)
16:16skeggsb: imirkin: the "plan" would be to kill the channel, and report it to userspace. that seems to be the approach taken, reporting a "lost device" to the app
16:17meeeee: my libdrm seems to be 2.4.0
16:17skeggsb: imirkin: we need to track/deal with semaphores between channels too
16:17imirkin: meeeee: that's not possible
16:17skeggsb: or we end up hanging more than we should
16:17imirkin: skeggsb: nnnnot sure what you mean. but as long as you do, it's ok ;)
16:17meeeee: well libdrm..also have libdrm_nouveau
16:18imirkin: meeeee: 2.4.0 is an *ancient* version, well before the mesa 10.x releases
16:18meeeee: so which one exactly do u mean
16:18imirkin: they should be the same version i should hope
16:18meeeee: i compiled some of them from source a week ago..the rest is from the old ubuntu mesa
16:18skeggsb: imirkin: btw, hans (just noticed he commented in a nv4x bug) is going to take a look at trying to resurrect the nv30/40 gallium driver
16:19imirkin: skeggsb: oh awesome. i've been looking at it a bunch too
16:19imirkin: i have some patches i haven't posted yet
16:19imirkin: but i was hoping to defeat swtnl first
16:20skeggsb: cool :)
16:21imirkin: swtnl is only hit in pretty rare cases, but it's pretty annoying that it's totally broken
16:21skeggsb: were you going to tackle the colour/zeta restrictions by any chance? :P
16:21imirkin: haha no
16:21imirkin: i'll leave that to someone with actual time
16:21imirkin: is hans on irc?
16:22skeggsb: yes, but i'm not sure if he hangs around here yet.. i can't see him on internal channels atm either, i still need to figure out exactly when he's around
16:22skeggsb: i'll point him towards you when i see him
16:23imirkin: sounds good
16:26meeeee: libdrm.so.2 -> libdrm.so.2.4.0 libdrm_intel.so.1 -> libdrm_intel.so.1.0.0 libdrm_nouveau.so.2 -> libdrm_nouveau.so.2.0.0 libdrm_radeon.so.1 -> libdrm_radeon.so.1.0.1 /sbin/ldconfig.real: /lib/x86_64-linux-gnu/ld-2.19.so is the dynamic linker, ignoring libdrm.so.2 -> libdrm.so.2.4.0 libdrm_intel.so.1 -> libdrm_intel.so.1.0.0 libdrm_nouveau.so.2 -> libdrm_nouveau.so.2.0.0 libdrm_radeon.so.1 -> libdrm_radeon.so.1.0.1 /sbin/ldcon
16:27meeeee: thats what i get from ldconfig
16:27imirkin: meeeee: that's not the version number
16:27imirkin: meeeee: check the installed package version
16:31meeeee: Desired=Unknown/Install/Remove/Purge/Hold | Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad) ||/ Name Version Architecture Description +++-====================-===============-===============-============================================= ii libdrm2:amd64 2.4.60-2~ubuntu amd64 Userspace interface to kernel
16:32imirkin: meeeee: right. get one that's *not* that version ;)
16:35imirkin: ... and of course i can't build mesa 9.1 because my llvm is too new!
16:35imirkin: now i remember why i hate llvm
16:36tobijk: now get an older llvm, little package, no compile time at all :)
16:36imirkin: yeah that won't happen. the solution is to not build llvm
16:37glennk: someone should point out to the llvm devs that other compiler which doesn't have a problem being installed with multiple versions...
16:37imirkin: oh, i wouldn't be surprised if gentoo let me
16:37imirkin: but... i'm not about to mess with that
16:37imirkin: esp when it's not central to the issue
16:38meeeee: is a downgrade to 2.4.56 sensible?
16:38meeeee: the newest ubuntu package is 2.4.60
16:38imirkin: meeeee: yea that's fine
16:39meeeee: however there is still one thing that bugs me
16:40meeeee: i i overwrote some of the pckage with sustom bins compiled from the nouveau sources...
16:41meeeee: so reverting to a precompiled package may screw my video drv completely
16:41imirkin: my recommendation would be to keep manually compiled binaries separate from distro ones
16:52imirkin: well, mesa 9.1.7 doesn't work with swtnl forced on
16:53meeeee: imi..if i do that downgrde...thats a GL apokalypse
16:53meeeee: The following packages will be REMOVED: libdrm-dev libdrm-intel1:i386 libdrm-nouveau2:i386 libdrm-radeon1:i386 libdrm2:i386 libgl1-mesa-dri:i386 libgl1-mesa-glx:i386 libqt4-opengl:i386 libqtwebkit4:i386 mesa-common-dev xorg-dev xserver-xorg-dev The following packages will be DOWNGRADED: libdrm2
16:53imirkin: meeeee: enjoy your distro binary package fun :)
16:54meeeee: basicly i need to reinstal..i.e. recompile the whole driver...
16:56meeeee: nouveau is disabled in the precompiled packages...they are just dummies in ubuntu (the nouveau portion) so one needs to compile and overwrite the bins oneself
16:56imirkin: ah ok
16:58Karlton: meeeee: ubuntu dropped support for nouveau?
16:58imirkin: ../../../src/mesa/main/bufferobj.c:2340:4: error: implicit declaration of function 'SET_BindBufferARB' [-Werror=implicit-function-declaration]
16:59imirkin: was building in the 9.0 days not a supported operation?
17:00meeeee: they never had it..just the 2d part is on. just enough to run unity and compiz... without extra 3d. mesa stays at softGL
17:01meeeee: had to recompile the drm and 3d part myself... GL works great in quake (darkplaces engine with relief mapping and so on) and in nexuiz.... just that doom now screwed up
17:02Karlton: I thought they added it years ago but if that is true, it's another reason to hate ubuntu :P
17:02meeeee: the 3d part of the driver of ubuntu is just dummies with NO 3D support
17:03meeeee: they plan to add it...some day
17:05meeeee: so imi..before i go to recompile hell..are u 100% positive..that libdrm is the one screwing thing up?
17:06tobijk: it screws up many thnigs but no, not 100% sure
17:06imirkin: ok. 9.0 just doesn't build.
17:06tobijk: maybe 9.1?
17:06imirkin: meeeee: i'm positive that 2.4.60 causes issues. no idea if it causes *your* issues.
17:06imirkin: tobijk: yeah, i got 9.1.7. no go.
17:07tobijk: seems unlilkely...
17:08tobijk: maybe stable was a used different back then ;-)
17:09imirkin: i mean, 9.1.7 built... eventually
17:09imirkin: but it didn't work with swtnl
17:41meeeee: hows gonna praise the fourth link? > https://www.google.de/search?q=install+nouveau+guide&ie=UTF-8&sa=Search&channel=fe&client=browser-ubuntu&hl=en&gws_rd=cr,ssl&ei=tldiVcSfEK6N7Ab-3YHYBw
17:41meeeee: i mean *who*
17:42meeeee: and also the second now the fourth link that appears
17:42meeeee: totally not concentrated
18:20meeeee: rebuilding mesa... :P imi its ur fault
19:03meeeee: imi, still there?
19:05meeeee: how the fuck are u using ipv6?
19:22meeeee: ....and rebooting
20:23meeeee: cannot install `es1api/libGLESv1_CM.la' to a directory not ending in /usr/lib
20:23meeeee: any idea?
20:23meeeee: ..building mesa
20:25Karlton: does nouveau need gles?
20:29meeeee: according to that ..yes
20:35Karlton: you did --prefix=/usr ?
20:36meeeee: yes but as i read make clean should solve it
20:36meeeee: rebuilding now...again
22:02meeee: just installed the newest mesa...looks TERIBLE!
22:06meeee: take a look at my new look> http://snag.gy/xJArM.jpg
22:07meeee: the black squares are supposed to be shadows...
22:13meeee: someone there?
22:15Karlton: no :-)
22:16Karlton: only thing I can suggest is to do a git bisect and find what is causing the bug
22:16Karlton: if it is an issue with mesa
22:19meeee: worked fine before i updated
22:20meeee: quake still looks grate
22:20meeee: havent still trzed the game i wanted to fix prior to the update
22:23Karlton: yeah, it is possible someone broke something in mesa
22:27Karlton: meeee: you updated from 10.6-dev to latest git and it broke after you restarted X?
22:28meeee: updated from precompiled 10.1 or something lake that
22:32Karlton: might be your compositor that is drawing the shadows and doesn't like the different mesa version
22:33Karlton: I am not sure what (ubuntu?) uses
22:34meeee: unity desktop with compiz
22:35meeee: btw the performance also seem to be a bit worse in general
22:35meeee: especially the GUI
22:36Karlton: can you turn compiz off and see what happens?
22:39meeee: offcourse the "shadows" will be gone the ugly black squares...but unfortunately ubuntu 14.04 does not provide 2D alternative to compiz anymore
22:40meeee: and if u look carefully u will see that even the window border with the resize and close buttons is NOT drawn!
22:40meeee: the new mesa definitelz broke the GUI
22:40meeee: quake is fine http://snag.gy/kmi3t.jpg
22:41Karlton: if unity has special transparent widow boarders, they would be using compiz also
22:42Karlton: need to find out is compiz is complaining about missing shared libaries or something
22:42meeee: transparancz is fine in the starter i.e. the HUD to the left
22:43meeee: the shadows and the window borders and resize buttons are gone or fucked
22:43Karlton: s/is compiz/if compiz/
22:44meeee: s/is? what do u mean?
22:44Karlton: correcting my typo
22:45meeee: its not
22:45meeee: i guess
22:46Karlton: if compiz is spamming debugging messages in the background, that could slow you down
22:46meeee: i just have the basic UI on..no extra effects like cuybe3d or wobbling windowws
22:46meeee: will check dmesg
22:47meeee: had to anyway to see if the problem i wanted to fix in the first place is gone
22:51Karlton: if gui is acting slow, check your cpu usage and see if it is abnormal
22:53meeee: dmesg is clean..no obvios errors...but the game i wanted to fix still shows only black screen..and now the GUI is also ...slow and "drawn"
22:57meeee: cpu is freeee> http://snag.gy/I7VpI.jpg
22:58meeee: so the question is who broke the shadows in NV50... :D and how
23:04Karlton: what happens if you just run 'compiz' in the termianl emulator?
23:07Karlton: or compiz --replace
23:09meeee_: just crashed the whole UI
23:10Karlton: using compiz --replace? (I am reading an old debian wiki on compiz) xD
23:11Karlton: I don't know how to get debugging errors from it
23:18meeee: [ 4055.688891] traps: compiz general protection ip:7fc187dce5d0 sp:7ffdd8ec8940 error:0 in libunitymtgrabhandles.so[7fc187db1000+27000] [ 4105.873000] nouveau E[ PGRAPH][0000:03:00.0] TRAP DISPATCH_QUERY [ 4105.873014] nouveau E[ PGRAPH][0000:03:00.0] no stuck command? [ 4105.873034] nouveau E[ PFB][0000:03:00.0] trapped write at 0x0020276000 on channel 0x0001fa52 [Xorg] PGRAPH/DISPATCH/QUERY reason: PAGE_NOT_PRESENT
23:19meeee: gfault at 7 ip 00007fd4897e950d sp 00007ffe44ba5718 error 4 in libcompiz_core.so.0.9.11.3[7fd489785000+ae000]
23:26meeee: and for those who havent yet seen it..my new fashionable UI... perfectlz drawn > http://snag.gy/I7VpI.jpg
23:29meeee: how do we trace the missing window borders buttons and shadows?
23:47Karlton: did you compile mesa from source you did you get that version from ubuntu?
23:50meeee: off course i just compiled it
23:51meeee: its the newest..ubuntu came with 10.1...or so
23:52Karlton: 10.1 seems kind of old though, the latest stabled is 10.5.5
23:52meeee: ..but it was working..to a point..where i tryed doom and got a kernel dump and i was adviced here to get the newest and i compiled all night long..and got what i got
23:54meeee: anz idea how to trace the gl commands from compiz?