00:00 karolherbst: mupuf: ignore bits 0-2
00:00 mupuf: yep, but the WTF is for the inversion
00:01 mupuf: I will let the machine go through all the values and see what I get in the morning
00:01 karolherbst: mupuf: some silly signed value?
00:02 mupuf: we'll see if we get any sort of repeatition
00:02 karolherbst: mupuf: what is the first and the last value of the inversion part?
00:03 karolherbst: ohh
00:03 karolherbst: got it
00:05 karolherbst: okay...
00:06 karolherbst: it makes no sense
00:09 mupuf: so far, the inversion happened only once
00:09 mupuf: time to sleep, have fun guys!
00:11 airlied: imirkin: those fixes seem to fix the geom fails alright
00:11 imirkin: airlied: yay!
00:11 imirkin: airlied: do they break everything else in the process?
00:12 airlied: not sure, I've a bunch of regrsesions since my cts run a few week ago
00:13 airlied: imirkin: I still have a local patch for texture changes in my tree you gave me
00:13 airlied: was that fixed some other way?
00:13 imirkin: airlied: oh, which one?
00:13 imirkin: i have a recollection of giving you changes, but i don't remember the details
00:13 airlied: http://paste.fedoraproject.org/474428/84776131/
00:14 imirkin: oh yeah. so ... i think in practice, that may actually create problems =/
00:14 imirkin: for now it's abandoned, pending a cleverer change
00:14 airlied:nukes it
00:22 imirkin: airlied: although it did "fix" a bunch of tests that otherwise ended in RA failure
00:23 mwk: *sigh* the NV5 defeated me
00:23 saintdev: is the DP MST support published with the new atomic modesetting only for nv50?
00:24 mwk: somehow, the *value* of the chroma key is being taken into account when determining whether dithering should be enabled
00:24 imirkin: saintdev: was DP even a thing on pre-nv50?
00:24 imirkin: mwk: perhaps chroma key is multi-purpose?
00:24 imirkin: saintdev: even if it was, no way it was DP 1.2, which is what introduces MST
00:24 mwk: imirkin: I seriously doubt that
00:24 airlied: imirkin: a lot of the regressions had WARNING: value %4 not uniquely defined
00:25 saintdev: imirkin: i mean i have a maxwell card, will it work on that too
00:25 imirkin: saintdev: yes
00:25 imirkin: saintdev: nv50+ :)
00:25 airlied: imirkin: let me know when you have a real fix without RA fails :)
00:25 imirkin: although actually MST itself is probably for kepler+
00:25 imirkin: airlied: will do. the fix will be to implement the same idea, just after SSA when it can be done reliably
00:26 imirkin: airlied: i doubt i'll be fixing the underlying RA failures
00:26 saintdev: I may have to give it a test then
02:32 mwk: that was a bitch.
02:48 saintdev: are there any config options i need to specify for testing the 4.10 tree (like, debug options i should have enabled?)
02:48 imirkin: nope
02:48 imirkin: unless it's not working :)
02:48 saintdev: kk
03:09 imirkin: airlied: updated version of that patch... almost entirely untested: https://github.com/imirkin/mesa/commit/0d84d9c183c8bb94df7cc9f61fc43bf1da65c13e.patch
03:13 imirkin: oh, just noticed it has a minor problem for fermi (but not kepler+) - fixed in github, but you'll need a diff commit id...
03:14 imirkin: and here it is - https://github.com/imirkin/mesa/commit/edf832d936c4679aa1a8fc7d89f8c541819ca3ab.patch
04:38 airlied: imirkin: that patch does a lot better
04:39 N0p3: hello? Is it right place to ask about current state of nouveau?
07:51 mupuf: imirkin, karol: http://fs.mupuf.org/mupuf/nvidia/6b13_sawtooth-2.png
08:00 mupuf: karolherbst: http://fs.mupuf.org/mupuf/nvidia/6b13_sawtooth-2.png
08:35 karolherbst: :D
08:36 karolherbst: what the heck
10:05 karolherbst: mupuf: this seems like a lot of fun somewhat
10:05 karolherbst: mupuf: you know what? you should try that with another GPU, just in case
10:36 mwk: well, fuck
10:36 mwk:managed to hang the test machine
11:06 mupuf: karolherbst: well, I could, but I don't see why it would help
11:06 mupuf: this is a sw scaling
11:06 mupuf:is not monitoring the fan speed here
11:06 mupuf: but yeah, I would love to know what the fuck this value is
11:07 mupuf: must be quite an interesting bitfield to behave like this :D
11:11 karolherbst: maybe there are other factors affecting that curve
11:11 mupuf: yep, possibly
11:12 mupuf: but, can you come up with a good function that would approximate this curve? :D
11:21 karolherbst: uhhm
11:21 karolherbst: without crazy stuff?
11:22 karolherbst: mupuf: maybe scan what values are used within the vbios, maybe there is a pattern matching with the graph?
11:23 mwk: yay, nv4 runs methods
11:23 mwk: well, at least one method, at any rate
11:37 mupuf: karolherbst: I need to scan again, I scanned already the entire table, but for PWM = 100%
11:38 mupuf: now that I know that 0% is different, I will scan again
13:17 imirkin: airlied: do you have a newer cts piglit thing you could upload? (i'm mostly just curious)
14:56 imirkin_: karolherbst: iirc you sent comments about having a emitTARG() in gm107 emitter? or was it hakzsam_? either way: https://github.com/imirkin/mesa/commit/8321766aaca7f9ad3f271bbea43970992e48b067
14:56 hakzsam_: imirkin_: I did
14:56 hakzsam_: looks much better, thanks
14:57 imirkin_: hakzsam_: arguably the surrounding if/else should go as well...
14:57 imirkin_: not sure if we should do that.
14:57 hakzsam_: yeah, not sure
18:50 airlied: imirkin: will try and upload later
19:02 karolherbst: buying tickets is hard sometimes...
19:09 imirkin_: airlied: cool thanks
19:11 karolherbst: mupuf: currently wondering if I will be able to buy my ticket before midnight or after :D
19:17 Ancient: karolherbst, I was just about to report an issue with X + atomic modesetting blacking the screen permanently at screen-sleep. But I see you've already reported that on the mailing list.
19:17 Ancient: Is there a good place to follow this issue or otherwise be useful?
19:18 karolherbst: Ancient: I guess you could try out updated trees
19:21 karolherbst: well, this worked out great, mupuf, all tickets are unavailable :D
19:27 mark4: i cannot get nvidia-driver to work with my hybrid graphics
19:27 mark4: does nouveau work with a nvidia GPU that has NO ACCESS to a display other than thru the intel GPU garbage proxy?
19:28 karolherbst: yes, through the intel garbage proxy
19:28 mark4: nvidia does not
19:28 mark4: it used to
19:28 mark4: if i revert to kernel 3.x.x and the appropriate nvidia-drivers it works
19:28 mark4: nothing since
19:28 karolherbst: 3.x that is like ancient
19:28 mark4: how do i configure nouveau fpor hubrid
19:28 imirkin_: mark4: should work fine.
19:28 mark4: no shit
19:29 imirkin_: mark4: https://nouveau.freedesktop.org/wiki/Optimus/
19:29 mark4: its going to turn my $2500 gaming laptop into a $400 gaming laptop tho :/
19:29 imirkin_: i'd recommend the DRI3 route if possible
19:29 mark4: what is DRI3?
19:29 imirkin_: read the guide
19:29 imirkin_: it's in reference to the instructions on that wiki page
19:30 mark4: no
19:30 mark4: thats still talking about kernel 3.x.x
19:30 imirkin_: no?
19:30 mark4: oh maybe i miss read that
19:31 mark4: guess i have to rebuild my kernel with nouveau here too. its srsly fscking annoying that i cannot get nvidia-drivers to function properly :(
19:31 imirkin_: you do need a 3.15 or later kernel for the DRI3 stuff to work, i think...
19:32 mark4: and nobody in #n vidia can help
19:32 mark4: yea thats what i miss read
19:32 mark4: i want to use a kernel 4.8.x
19:32 mark4: but i cant with nvidia-drivers and my card
19:32 mark4: 01:00.0 VGA compatible controller: NVIDIA Corporation GM204M [GeForce GTX 965M] (rev a1)
19:33 imirkin_: ah, well you should expect nouveau to support accel on that (kernel 4.6+) but it'll be pretty slow
19:33 imirkin_: as we can't reclock that gpu
19:33 imirkin_: although arguably it's a laptop, probably has separate fans, we could rig something up there
19:33 imirkin_: [the #1 reason we can't reclock is we can't control the fans]
19:33 mark4: like i said. will turn my $2500 gaming laptop into a $400 gaming laptop
19:33 mark4: why cant you control the fans?
19:34 karolherbst: because nvidia doesn't allow us
19:34 mark4: wait
19:34 karolherbst: only a signed PMU firmware image can
19:34 imirkin_: because nvidia will only allow fan control from securely-loaded firmware
19:34 mark4: doesnt ALLOW you?
19:34 karolherbst: can you crach hardware crypto?
19:34 karolherbst: *crack
19:34 imirkin_: and in order to securely-load firmware you need to have a signed payload
19:34 mark4: right. a srs fuck you to all linux users
19:34 mark4: guess im never going to buy another nvidia product ever again
19:34 karolherbst: well
19:34 karolherbst: but you are lucky
19:35 karolherbst: your fan is controlled by the EC
19:35 mark4: how do you know?
19:35 karolherbst: and I managed to full reclock those GPUs
19:35 imirkin_: karolherbst: you should try to rig something up for people with laptops
19:35 karolherbst: because it is a laptop
19:35 karolherbst: mark4: clevo?
19:35 mark4: yes
19:35 karolherbst: press fn + 1
19:35 mark4: i dont have an FN button
19:35 karolherbst: huh?
19:35 mark4: oh lol
19:35 mark4: /facepalm
19:35 mark4: yes i do
19:36 karolherbst: fn+1 should turn your fans to max (auto if you press it again)
19:36 karolherbst: both fans actually
19:36 mark4: ive been using a diff laptop and the FM button is on the left. on the laptop in question its on the right lol
19:36 karolherbst: cpu and gpu
19:36 karolherbst: this way you know for sure
19:36 mark4: amg!
19:36 mark4: fn+1 put my fans in high speed
19:36 karolherbst: it is totally bypassing the kernel by the way
19:37 karolherbst: mark4: yeah, check if the gpu fan too
19:37 imirkin_: xexaxo1: https://bugs.freedesktop.org/show_bug.cgi?id=98629
19:37 mark4: both
19:37 karolherbst: k, nice
19:37 mark4: so i CAN overclock the gpu?
19:37 imirkin_: well - reclock, not overclock...
19:38 karolherbst: imirkin_: the thing, you need like two nouveau images: 1. boot nouveau without secboot and clock to max 2. unload and load nouveau with secboot enabled#
19:38 mark4: ok reclock i meant that
19:38 mark4: i never OC a gpu
19:38 imirkin_: karolherbst: lol
19:38 mark4: there is a tool to do it?
19:38 karolherbst: imirkin_: exactly
19:38 mark4: or will it reclock automatically?
19:38 karolherbst: imirkin_: the PMU is used for secbooting the other falcons
19:38 imirkin_: mark4: "echo" :)
19:38 imirkin_: karolherbst: right makes sense =/
19:38 imirkin_: karolherbst: which also means no reclocking, right?
19:39 mark4: echo?
19:39 karolherbst: imirkin_: basically
19:39 imirkin_: karolherbst: could we get the nouveau logic to load a second pmu after the first one is "done"?
19:39 karolherbst: imirkin_: gnurou told me though that there is a way to properly reset the PMU and load our image on it
19:39 karolherbst: I somewhat wait until his rework is done
19:39 karolherbst: mark4: yeah, echo 0f into a pstate file within debugfs
19:39 karolherbst: mark4: I have a prototype for automatic reclocking though
19:40 karolherbst: mark4: it's comming somehow, but there is a lot of stuff to be done before
19:40 mark4: i can wait
19:40 karolherbst: things like overheating protections, over budget protection
19:40 mark4: if switching to nouveau makes this system even WORK at all... im switching
19:40 imirkin_: karolherbst: he's sent his stuff out...
19:40 karolherbst: I hope that in the end of 2017 most of the things are in place for kepler/maxwell
19:40 karolherbst: imirkin_: right, I know
19:40 mark4: and im never EVER buying another nvidia based product ever again
19:40 mark4: ever
19:40 imirkin_: good idea.
19:41 karolherbst: mark4: we still make fun about it though.. being able to reclock, but not to control the fans, that is. Makes totally sense
19:42 mark4: karolherbst, i honestly think both intel and nvidia are taking kickbacks from M% to fsck over linux users
19:42 mark4: m$
19:42 karolherbst: intel does?
19:43 mark4: intel is part of this whole hybrid gfx scheme
19:43 karolherbst: well sure
19:43 imirkin_: intel's drivers right now are by far the best...
19:43 mark4: no
19:43 karolherbst: but we have support for that on intel for a long time
19:43 imirkin_: they also have by far the largest team
19:43 karolherbst: yeah, on linux
19:43 imirkin_: which is what i'm talking about.
19:43 mark4: intel hardware is about 30 years behind nvidia
19:43 karolherbst: nope
19:43 mark4: their gpu's are GARBAGE
19:43 karolherbst: not even close
19:44 imirkin_: they do pretty well given their power budget...
19:44 karolherbst: they beat pretty much everything below x50
19:44 karolherbst: 940 or recent intel? intel is faster
19:44 karolherbst: hw wise
19:44 mark4: Model name: Intel(R) Core(TM) i7-4710MQ CPU @ 2.50GHz
19:44 karolherbst: their only problem is that of shared memory
19:45 mark4: run pure intel accelerated and i get 15fps in world of warcrack
19:45 karolherbst: mark4: you won't believe it, but that intel gpu will run faster than your nvidia gpu with nouveau
19:45 karolherbst: on lowest clocks that is
19:45 karolherbst: different story with higher ones
19:46 mark4: exactly and why would i want to keep my expensive toy in slow mode
19:46 karolherbst: on paper the 965m is 2.5x as fast as your intel one
19:46 karolherbst: right
19:46 mark4: you compare intels fastest fast mode with the nvidia fastest fast mode
19:46 karolherbst: but nvidia screwed us over, that's why you have to stick with those
19:46 mark4: not the fastest with the slowest
19:46 mark4: i restate: intell gpu's are GARBAGE
19:46 mark4: yup nvidia are a bunch of #$(%^#*(Q
19:47 karolherbst: allthough the hsw intel gpus are indeed kind of slow
19:47 karolherbst: skylake iris is real fast
19:47 karolherbst: especially the 580
19:48 mark4: the week after i bouth this laptop m-tech updated its CPUs to skylakes :/
19:48 karolherbst: :(
19:48 mark4: yup lol
19:48 karolherbst: I guess you would have gotten one with the HD 530?
19:48 cyndis: my work machine has a skylake, i get a constant spew of kernel warnings from the intel drm driver >_>
19:48 mark4: if intel ever pulls their head out of their ass and makes a FAST gpu i have no problems switching
19:48 karolherbst: :D
19:48 mark4: i did not get an HD machine
19:49 karolherbst: nah, I meant the recent skylake ones
19:49 mark4: oh
19:49 mark4: hd would have been real nice but i already spent over $2000 ON THIS laptop
19:50 mark4: building 4.8.2 and swithing to nouvau
19:50 karolherbst: mark4: use this X config: https://gist.github.com/karolherbst/1f1bdd1a3822df74097f
19:50 karolherbst: easieast way to reclock with later on
19:51 mark4: kk
19:51 mark4: one step at a time :)
19:51 mark4: kernel build then reboot to config everything else
19:51 karolherbst: yeah, first one is kernel patch
19:52 karolherbst: basically what you need is one nouveau module to bootsrap reclocking
19:52 mark4: gotta check the gentoo nouvau page for kernel options
19:52 karolherbst: and then a second one you can load in after reclocking
19:52 karolherbst: mark4: drivers/gpu/drm/nouveau/nvkm/engine/device/base.c
19:52 mark4: first one in the initramfs?
19:52 karolherbst: yeah
19:52 karolherbst: you have to disable those ".secboot = gm200_secboot_new" lines
19:52 karolherbst: yours is a GM204
19:52 karolherbst: so it might be enough to just do it for the GM204 struct
19:53 karolherbst: nv124_chipset = { ...
19:53 karolherbst: uhh wait
19:53 karolherbst: 4.8 doesn't haave the reclocking fixes
19:53 karolherbst: ohh well
19:53 karolherbst: doesn't matter for engines
19:53 karolherbst: fast memory real nice
19:54 karolherbst: mark4: and boot with nouveau.config=NvClkMode=15
19:54 karolherbst: ohh wait
19:54 karolherbst: crap
19:54 mark4: heh
19:54 karolherbst: the 4.8 module won't do it
19:54 mark4: dont use kernel 4.8?
19:54 karolherbst: just leave it unpatched then
19:54 karolherbst: nah, it is fine
19:54 karolherbst: I have branches
19:54 karolherbst: just blacklist nouveau
19:54 karolherbst: and that X config is important
19:55 karolherbst: and x11-drivers/xf86-video-dummy
19:56 mark4: erm im lost lol
19:56 mark4: ok let me get the kernel built first
19:56 karolherbst: yes
19:56 mark4: that doesnt take long, just keep getting side tracked lol
19:56 mark4: oh and emerge -C nvidia-drivers
19:57 mark4: should i set my VIDEO_CARDS="dummy" ?
19:57 mark4: erm "intel dummy"
19:59 mark4: or do i set it to nouveau?
19:59 mark4: and do i need modesetting?
20:01 karolherbst: nope
20:01 karolherbst: no nouveau, no modesetting
20:01 mark4: ok just dummy?
20:01 karolherbst: and intel
20:01 karolherbst: ohh wait
20:01 karolherbst: for mesa you need nouveu
20:01 karolherbst: ...
20:01 karolherbst: what a crap
20:01 karolherbst: well, if you use the X config file on mine, you can also add nouveau
20:02 mark4: ok
20:02 mark4: the config you posted?
20:02 karolherbst: or "media-libs/mesa video_cards_nouveau" in package.use
20:02 karolherbst: yes
20:02 karolherbst: and "x11-libs/libdrm video_cards_nouveau"
20:03 mark4: so VIDEO_CARDS="dummy" and add those use flags to the specified package.use ?
20:04 karolherbst: and intel
20:04 imirkin_: bleh. $17 for a PCI nv4x =/
20:04 imirkin_: not worth it.
20:04 karolherbst: :D
20:06 karolherbst: mupuf: it isn't like I didn't expect this, but it seems to be sold out after one hour...
20:06 mark4: 01:00.0 VGA compatible controller: NVIDIA Corporation GT218M [NVS 3100M] (rev a2)
20:06 mark4: can nouveau upclock that one?
20:06 karolherbst: well sure
20:06 karolherbst: I think
20:06 imirkin_: hm. or an AGP one actually - it'd mean i lose OF, but should still work...
20:06 mark4: thats THIS laptop, its a dell latitude e6410
20:07 karolherbst: mark4: but that gpu is pretty slow
20:07 mark4: FN+1 does not do anything on this one
20:07 imirkin_: mark4: it should work, although some cards have issues.
20:07 mark4: yea lol i know
20:07 mark4: which is why i did not mind switching to nouveau one iota here
20:07 karolherbst: mark4: that fn+1 is clevo only I think, not quite sure
20:07 mark4: nouvau LOOKS better on everytghing ive run it on
20:08 mark4: what is the DCP button?
20:08 mark4: does not seem to do anything lol
20:11 mark4: just doing an emerge -uDN world to make sure
20:11 mark4: yea #x11-modules-rebuild did not rebuild libdrm or mesa
20:11 mark4: any other clever clevo tricks i should know?
20:12 karolherbst: none I am aware off
20:12 karolherbst: *of
20:12 mark4: that one was kinda neat lol
20:13 mark4: mesa takes a few mins to build
20:14 mark4: wish all this was coded in forth instead of c. my compiler would do an emerge -e world in 2 minutes :P
20:15 amluto: should I expect everything on my nouveau GM206 desktop to be laggy?
20:15 Calinou: no power management there yet, so yes
20:15 amluto: it seems like most of the desktop runs at maybe 10fps. the mouse pointer is still fast
20:15 amluto: :(
20:15 amluto: can i force everything to full power?
20:15 Calinou: nope
20:15 karolherbst: Calinou: well if the EC can control the GPU fan, then there isn't a big reason not to reclock on those :p
20:18 amluto: i thought the whole point of the signed firmware was that a driver *can't* overheat the card
20:20 amluto: also, how big is the performance ratio between minimum clock and full speed? i feel like a card that can nominally play fancy games should be able to get 60fps text rendering at even a small fraction of its speed
20:21 karolherbst: lol
20:21 karolherbst: nope
20:21 karolherbst: not gonna happen
20:21 mark4: ok i have exactly the same issue with nouveau so far
20:22 karolherbst: for most cases you get like 30% of max perf at most
20:22 amluto: shouldn't 30% be plenty?
20:22 mark4: screen is on for .5 of a second. off for 1.5 seconds. reopeat over and over
20:22 amluto: oh
20:22 karolherbst: amluto: depends on the workload
20:22 karolherbst: mark4: odd
20:22 karolherbst: mark4: I doubt it has anything to do with your nvidia gpu
20:23 mark4: thats in console not x
20:23 karolherbst: mark4: or did you select nvidia only in the bios?
20:23 amluto: typing letters into hexchat for IRC?
20:23 mark4: cant
20:23 mark4: cannot turn off the intel gpu
20:23 karolherbst: mark4: k, then it should be an intel issue actually
20:24 mark4: i can startx now but the display still does the on off on off thing
20:24 mark4: but now its on for .5 of a second, off for 2.5 seconds
20:24 mark4: so whats wrong with my intel driver and how do i fix :(
20:25 airlied: imirkin_: https://people.freedesktop.org/~airlied/piglit/cts_gl45/
20:25 mark4: if i dont use an accelerated desktop it does not flicker
20:25 mark4: this is a opengl issue
20:25 airlied: column on the right is with that patch and newest CTS
20:25 imirkin_: airlied: thanks =]
20:25 airlied: the colum on the left is older CTS and older driver
20:26 imirkin_: airlied: those crashes are weird. looking at the regressions page.
20:26 Calinou: amluto: Linux 2D acceleration is a wreck
20:26 Calinou: unlike Windows or macOS
20:26 Calinou: it's usually nonexistent (Chromium) or crap (the rest)
20:27 amluto: Calinou: is even glamor a wreck?
20:27 Calinou: probably
20:27 Calinou: if you compare it to the proprietary OS APIs
20:27 amluto: i suppose i could try Wayland. no 2D at all
20:27 amluto: or no ddx baggage at least
20:27 airlied: imirkin_: oh khr debug ignore those
20:28 airlied: those tests are broken
20:28 airlied: and my "fix" didn't port
20:28 amluto: somehow my laptop manages excellent performance on glamor, and I'm pretty sure my GM206 is *way* faster at gl than my laptop
20:29 imirkin_: airlied: ok, figured it was something like that.
20:29 airlied: imirkin_: the arrays of arrays crashes are wierdes tones I think
20:29 airlied: since I think those tests pass on radeonsi
20:30 airlied: ah spill stuff
20:31 hakzsam_: airlied: does radeonsi pass all CTS tests now?
20:31 imirkin_: on the bright side, lots of stuff got fixed
20:32 airlied: hakzsam_: don't think so
20:33 imirkin_: that's a lot of enhanced layouts failures =/
20:33 imirkin_: does intel pass all those?
20:33 airlied:has no intel results
20:33 airlied:will try and get radeonsi results sometime this week
20:34 hakzsam_: airlied: ah okay, because nha enabled GL 4.5 the past week
20:34 airlied: I know there was bugs in it
20:34 airlied: hakzsam_: they have a plan to fix all theirs by next release
20:34 hakzsam_: makes sense
20:35 airlied: imirkin_: but I think some of the enhancde tests might be buggy
20:35 airlied: I'm also running CTS master, not the exact tests for GL 4.5
20:35 imirkin_: fantastic...
20:36 imirkin_: that's what i like in life - buggy conformance tests.
20:36 hakzsam_: that's sad :)
20:36 airlied: well they aren't offical CTS tests
20:36 airlied: as I said I'm running master
20:37 mark4: im also getting my boot messages spammed with kernel log messages still
20:37 airlied: for official conformance there is a branch per GL release, master gets all the new tests
20:37 airlied: and they have to be reviewed and validated
20:37 imirkin_: airlied: ah ok. any reason you're not running the branch?
20:38 airlied: imirkin_: because it has other fun properties I don't want to deal with
20:39 airlied:isn't sure why anyone is surpriseed that conformance tests have bugs
20:39 airlied: it's code like any other
20:40 imirkin_: but it's more annoying
20:40 airlied: and the spec is vague in so many ways
20:41 hakzsam_: "unable to find root device" after building a fresh kernel, does this sound familiar to someone? first time I get that message...
20:42 mark4: did you use genkernel?
20:42 hakzsam_: no
20:43 hakzsam_: I used the "traditional" way
20:43 imirkin_: hakzsam_: did you accidentally forget to compile in support for your disk controller?
20:43 hakzsam_: I doubt, because I use the same config as my current kernel (ie. zcat /proc/config.gz)
20:44 towo`: is it 4.9-rc?
20:44 hakzsam_: and I removed the UUID thing in fstab, I use /dev/sdaXX now
20:44 hakzsam_: yeah, 4.9-rc2
20:44 towo`: then you have to deactivate CONFIG_MODVERSIONS
20:44 towo`: or you have to apply 2 little patches
20:45 towo`: btw, why building rc2 if rc4 is out?
20:45 imirkin_: hakzsam_: just grab drm-next - that has -rc4 backmerged in
20:46 hakzsam_: ah, I cloned airlied's repo just before rc4
20:46 hakzsam_: will update
20:49 hakzsam_: no, it was not that old, I cloned it like 3 days ago, well I will re-build rc4 that should be fixed hopefully
20:52 towo`: no, is not ficed in rc4
20:52 towo`: fixed
20:53 hakzsam_: but disabling CONFIG_MODVERSION fixes it?
20:53 towo`: yes
20:53 towo`: or 2 patches
20:53 hakzsam_: cool, thanks
20:53 hakzsam_: which patches, btw?
20:53 towo`: http://paste.debian.net/893346/ this and http://paste.debian.net/893347/ this
20:54 hakzsam_: k
20:56 karolherbst: mark4: anything in dmesg?
20:56 imirkin_: skeggsb: is there some quick easy way of disabling GART on nv4x?
20:56 mark4: no im trying to boot to a fedora live dvd atm
20:56 mark4: but it seems to NOT work
20:57 imirkin_: skeggsb: i'm looking for the equivalent of agpmode=0 on agp boards
20:57 mark4: so i just reburned the dvd
21:01 mark4: ok it took a while but the fedora live dvd worked
21:01 mark4: its kernel 4.5.5 tho
21:02 mark4: is there a way to see if i have nouveau accelerated (ish) gfx enabled?
21:05 mark4: how do i make glxgears not sync to vblank?
21:05 hakzsam_: vblank_mode=0 glxgears
21:06 mark4: ok so i get 4k fps when not syncing. 60 when syncing
21:06 karolherbst: well
21:07 mark4: thats with nouveau
21:07 karolherbst: *intel
21:07 mark4: thats with intel?
21:07 karolherbst: yes
21:07 mark4: how do i make nouvau active?
21:07 karolherbst: no way on 4.5
21:07 mark4: crap
21:07 karolherbst: you should figure out what is wrong with 4.8
21:07 mark4: utterly fucking usless live dvd then
21:07 mark4: utterly FUCKING useless
21:07 karolherbst: try to build nouveau as a module
21:08 karolherbst: and blacklist it
21:08 mark4: it is built as a module on the gentoo boot
21:08 mark4: why blacklist it?
21:09 imirkin_: have faith
21:09 mark4: heh
21:09 imirkin_: eventually all will be clear
21:09 mark4: kk
21:10 mark4: rebooting to 4.8.2
21:10 mark4: i forget how to blacklist a module lol
21:11 imirkin_: add "blacklist nouveau" in modprobe.conf
21:13 mark4: there is no modprobe.conf
21:16 mark4: i cannot seem to get nouveau to be blacklisted
21:16 mark4: using any method i find on the net and reboot and... nouveau is still loaded
21:19 mark4: modprobe.conf does not exist
21:20 imirkin_: do you have a /etc/modprobe.d ?
21:20 imirkin_: if so, stick a file in there with "blacklist nouveau" in it
21:22 mark4: yes
21:22 mark4: i did
21:23 mark4: i called the file "blacklist"
21:23 mark4: just told need to call it whatever.conf
21:23 mark4: not just whatever
21:23 imirkin_: yea
21:23 imirkin_: thought that was implied :p
21:25 mark4: ok, no nouveau is loaded now
21:25 mark4: whats the next stage :)
21:26 mark4: startx still gives me on/off/on/off
21:28 karolherbst: uhm
21:29 karolherbst: then I assume it is an intel issue
21:29 karolherbst: anything in dmesg?
21:29 karolherbst: or something in the x log?
21:29 mark4: https://paste.pound-python.org/show/qvaKCrkSMaId13d2POIl/
21:30 mark4: https://paste.pound-python.org/show/vxfQlYE4ff6pBMBdYalE/
21:30 imirkin_: no nouveau in there..
21:31 mark4: should there be?
21:31 imirkin_: dunno
21:31 imirkin_: not 100% sure what instructions karol is getting you to follow
21:32 hakzsam_: karolherbst: could you check if that series doesn't break some games? I tested all UE4 demos, heaven and valley, they look fine, but I don't have a ton of games like you... https://cgit.freedesktop.org/~hakzsam/mesa/log/?h=refactoring&showmsg=1
21:57 mark4: just set VIDEO_CARDS="intel" and rebult x11 modules
21:57 mark4: removed all references to nvidia and nouveau and am rebuilding world
21:57 mark4: see if i get a stable X with pure intel only guaranteed
21:57 mark4: i do on the fedora live dvd
21:58 barteks2x: um... I got a hard nouveau crash
21:58 barteks2x: in java
21:58 barteks2x: using lwjgl
22:01 barteks2x: should I report it somewhere?
22:01 barteks2x: I have no idea if it's something using opengl incorrectly or if it's nopuveau bug
22:03 imirkin_: incorrect opengl usage shouldn't end in hard hangs, so either way, it's a nouveau bug
22:03 barteks2x: all I have is hs_err file from java
22:03 imirkin_: that said ... are you using qt5 or kde plasma?
22:04 mark4: yea they suck donkey
22:04 barteks2x: I use cinnamon desktop environment
22:04 mark4: plasma destroys its own config files every couple of days
22:04 imirkin_: ok. i haven't heard of that one, which is good
22:04 imirkin_: (if i've heard of it, that means people have complained about it)
22:04 mark4: cinnamon is a fork of gnome 3 to look like gnome 2
22:04 barteks2x: but I do have nvidia optimus
22:05 barteks2x: which I know could cause weird issues
22:05 imirkin_: are you running off the intel gpu or the nvidia gpu?
22:05 barteks2x: that was when running on nvidia GPU
22:05 mark4: its part of linux mint
22:05 imirkin_: only that one java app, or everything?
22:05 barteks2x: only the java app
22:05 imirkin_: k. and which java app is this? minecraft or something?
22:06 barteks2x: minecraft (in forge dev environment in debug mode)
22:06 imirkin_: k. that does sound familiar
22:06 imirkin_: which is bad :)
22:06 imirkin_: iirc it does some multithreading stuff
22:06 imirkin_: which ends in a pile of fail on nouveau
22:06 barteks2x: I think it may be related to the forge loading screen which does some mlultithreading
22:06 mark4: lol im ssh'ed from laptop a to laptop b. laptop b somehow got unplugged and shut down during an emerge
22:07 mark4: i plugged it back in and powered it back up
22:07 mark4: and the ssh didnt break and the build continues lol
22:07 imirkin_: yay for TCP
22:07 mark4: inorite!
22:07 mark4: its almost like they knew what they were doing
22:07 mark4: unlike nvidia
22:07 mark4: and me :/
22:08 barteks2x: according to them, this is a valid way to do it (as I remember), but it caused some weird issues even with intel drivers
22:08 barteks2x: (something with some xkn_something error)
22:08 barteks2x: *xkb
22:09 barteks2x: should I still report it or is it going to be classified as "not a bug" anyway?
22:11 karolherbst: barteks2x: report it or you won't know
22:19 mark4: how can i tell what gen version my integrated gpu is?
22:24 RSpliet: karolherbst: likely a threading-related problem?
22:27 karolherbst: RSpliet: didn't look into it
22:27 karolherbst: mark4: haswell
22:27 karolherbst: mark4: I think it is gen 7.5? or 7?
22:28 karolherbst: not _quite_ sure
22:29 mark4: yea i figured it
22:29 mark4: i added i965 to my vide cards line
22:29 mark4: but i still have the same problem
22:29 mark4: i took out nvidia and nouveau entirely
22:29 karolherbst: yeah, as I expected
22:29 karolherbst: ask in #intel-gfx
22:29 mark4: i have a problem with my intel for sure
22:29 mark4: that channel exists?
22:29 mark4: !
22:29 karolherbst: yes
22:32 barteks2x: some of that crash even appeared in dmesg: gr: DATA_ERROR 0000000c [INVALID_BITFIELD] ch 4 [007fae1000 java[7326]] subc 0 class a197 mthd 1c04 data 200681c3
22:32 imirkin_: barteks2x: yeah, concurrency + nouveau = fail
22:32 imirkin_: not worth trying to debug it - it just plain doesn't work. at all.
22:33 barteks2x: but should I report it or not?
22:33 imirkin_: if you're interested, i have a branch where i attempted to fix it and failed. it covers a lot of situations, but it'll deadlock occasionally, and the overall approach is buggered
22:33 imirkin_: this is the branch: https://github.com/imirkin/mesa/commits/locking
22:33 barteks2x: if there is no chance it will be fixed - I probably won't even bother reporting it
22:34 imirkin_: there's a high chance it'll be fixed. the issue is well-known.
22:34 imirkin_: it's just not particularly easy to fix
22:34 barteks2x: is there anything in issue tracker for it?
22:34 naptastic_: Hi, so, I upgraded to a 1080, and on Linux 4.8, KMS works fine, *but* my mouse cursor doesn't move.
22:34 imirkin_: barteks2x: there are a ton of issues filed about concurrency + nouveau failing
22:35 imirkin_: search around in the bug tracker for them
22:35 naptastic_: 4.8.5, to be precise. Is there a fix, or do I need to just wait?
22:35 imirkin_: naptastic_: are you sure you're using nouveau? pastebin dmesg
22:35 naptastic_: In Linux 4.7, KMS doesn't work and I'm stuck at 1024x768.
22:36 imirkin_: i'm surprised the mouse doesn't work... seems like something skeggsb would have tested in his bringup :)
22:36 naptastic_: imirkin_, right now it will be wrong because I'm on 4.7. I may not be able to get dmesg on 4.8.
22:36 naptastic_: Should I reboot and attempt it?
22:36 imirkin_: naptastic_: well, wtvr. if you're sure you're using nouveau, no need to do that
22:36 imirkin_: naptastic_: it's just very surprising that the mouse would be broken
22:37 imirkin_: but it's not like i've personally tested
22:37 naptastic_: imirkin_, I agree. I'm wondering if there's something specific to my setup, but it's pretty ordinary Debian Sid + custom newer kernel
22:37 naptastic_: I'm looking at dmesg on this boot anyway.
22:38 naptastic_: [ 1.962469] nouveau 0000:01:00.0: unknown chipset (134000a1)
22:38 naptastic_: That's kernel 4.7, which is what I expected.
22:38 imirkin_: right
22:38 naptastic_: I'mma reboot into 4.8 and see if I can get that information. BRB (I hope).
22:40 naptastic_: ok, I'm on 4.8 now, and relying on keyboard + blind mouse navigation. This will be entertaining.
22:41 naptastic_: https://dpaste.de/TQsC
22:44 imirkin_: yeah, that all seems reasonable
22:44 naptastic_: Now I notice the cursor is *changing* as appropriate, just not moving from the top-left of the screen.
22:44 imirkin_: i wonder if the "failed to create kernel channel" bit (which means it failed to init accel - fully expected) somehow causes the mouse pointer to die
22:44 imirkin_: huh, ok
22:45 imirkin_: probably the x/y regs moved and we didn't update the code
22:45 imirkin_: bad skeggsb :p
22:45 naptastic_: lawl
22:45 naptastic_: I'm trying to get to the 4.8.6 changelog in the hopes that it will contain a fix. But web browsing without a mouse...
22:45 imirkin_: eh... the mouse is there
22:45 imirkin_: just not displaying where it should :)
22:46 naptastic_: I have to infer the cursor's location. It's an interesting challenge.
22:47 naptastic_: and there's nothing in that changelog for nouveau, so... hopes dashed, I'll reboot back into Windows and just play Overwatch instead of coding like a good little boy.
22:48 naptastic_: Guess I could try 4.9-rc4?
22:50 karolherbst: naptastic_: ohh actually, there was a fix for concurrency stuff
22:50 karolherbst: I don't know if it helps though
22:50 naptastic_: karolherbst, in 4.9?
22:50 karolherbst: nope
22:50 karolherbst: out of tree
22:50 karolherbst: https://github.com/skeggsb/nouveau/commits/master
22:50 karolherbst: newest commit
22:51 karolherbst: uhhh
22:51 karolherbst: wrong
22:51 karolherbst: second newest commit
22:51 naptastic_: lol
22:51 karolherbst: https://github.com/skeggsb/nouveau/commit/f2a5685ed0dba31e25a9f9783440f7fafb0d49ef
22:51 naptastic_: lemme see if I can operate github in this state. ;)
22:52 barteks2x: I'm trying to report it, and I have no idea which product should I choose...
22:53 mark4: well #intel-gfx is a total loss
22:53 barteks2x: why there is no "I'm not sure" option?
22:53 mark4: they want me to use git disect and bullshit like that to "Track down the bug"
22:53 naptastic_: karolherbst, if I'm not seeing that specific error message, can I conclude that that patch won't help?
22:53 karolherbst: naptastic_: no clue
22:53 karolherbst: maybe not, maybe it does
22:53 naptastic_: :) It's all kind of a crapshoot at this point.
22:53 karolherbst: sure
22:54 naptastic_:compiles a 4.9 kernel
22:54 imirkin_: naptastic_: no nouveau changes in 4.9
22:54 naptastic_: oh damn
22:54 karolherbst: sometimes life is tough
22:54 naptastic_: Yeah.
22:54 naptastic_: Sometimes the answer is to play Overwatch.
22:54 naptastic_: KTHXBYE <3
22:54 karolherbst: don't know about that
23:13 airlied: imirkin_: yeah skylake fails those enhanced layout tests as wel
23:15 imirkin_: airlied: excellent :)
23:20 hakzsam_: gnurou: skeggsb: any ideas why I can't launch NVIDIA gr firmwares on gm107? I extracted those from a mmiotrace but failed when reloading nouveau http://hastebin.com/zebewoxulu.sql
23:25 skeggsb: imirkin_: config=NvPCIE=0 will give you the same as agp=0, you'll still have "GART", but it'll be a paged ctxdma instead of the proper page tables
23:27 imirkin_: skeggsb: ok cool
23:33 hakzsam_: wtf, fecs_inst.bin is empty...
23:36 hakzsam_: makes sense, no occurences of 0x409184 in the mmiotrace
23:38 imirkin_: script isn't perfect :)
23:38 imirkin_: skeggsb: did you see the thing above - cursor's not moving on pascal
23:38 imirkin_: skeggsb: did you test the mouse on yours?
23:39 hakzsam_: imirkin_: yup
23:42 skeggsb: umm, no, i didn't notice that behaviour
23:42 skeggsb: i'm adding gp102 support now, so, i'll see what happens there
23:44 imirkin_: skeggsb: well - were you testing with a mouse?
23:44 skeggsb: yes
23:44 imirkin_: ok. so presumably that's something you would have picked up on :)
23:44 skeggsb: one would hope
23:44 skeggsb: but, perhaps i didn't re-test pascal after some other changes
23:45 imirkin_: was it the primary gpu?
23:45 imirkin_: otherwise you would have had a software mouse
23:45 skeggsb: yes, it was
23:45 imirkin_: (just thinking of simple obvious reasons why it would have failed)
23:45 imirkin_: er, not failed.
23:45 skeggsb: that seems like a fail to me :P