00:53 orbea: imirkin_: did you ever have any luck with the sampler changes for gallium nine?
00:59 imirkin: orbea: yeah
00:59 imirkin: er
00:59 imirkin: well, it's still on my list
00:59 imirkin: and that list is on my list of lists
00:59 imirkin: ;)
00:59 orbea: heh :)
01:01 imirkin: too bad i lost the list of lists...
01:01 imirkin: i should use valgrind to find it :)
01:01 orbea: good one :P
02:09 deevz: Does nouveau support proper 3d acceleration? I'm getting an opengl error when I try to launch any opengl app: https://pastebin.com/Hetam6Se
02:27 gnarface: that doesn't look like an opengl error to me
02:27 gnarface: that looks like an X error
02:27 gnarface: or maybe glx
04:19 imirkin: skeggsb: any idea why i might not be getting vblank events?
04:19 imirkin: and i have [2161247.218] (WW) NOUVEAU(0): Wait for VBlank failed: Cannot allocate memory
04:19 imirkin: in my xorg logs
04:21 imirkin: which would be drmWaitVBlank failing with ENOMEM
04:23 gnarface: hey i've seen that one before
04:23 gnarface: i think a moment or two before the whole system froze up because it had apparently actually run out of memory
04:26 imirkin: skeggsb: does any of this strike you as odd? https://hastebin.com/egumiqidip.pas
04:26 imirkin: one observation is that current > 2^32
04:27 imirkin: looks like it's truncating to the 32-bit value...
04:27 imirkin: i unfortunately don't know how any of this is supposed to work =/
04:34 imirkin: lol ok. drm_vblank_count returns a u32, but drm_vblank_count_and_time returns a u64
04:34 imirkin: airlied: does this strike you as inconsistent?
04:35 gnarface: heh, not if the second one is just the 32-bit drm_vblank_count and drm_vblank_time values concatenated
04:35 gnarface: that would be stupid, but not inconsistent
04:35 imirkin: it's not. the time is filled into a pointer.
04:39 imirkin: anyways, the widen_32_to_64 appears to do nothing, since seq is a 32-bit value. i think it's meant to be the 64-bit value. but there seems to be some confusion around whether vblank counter is 32- or 64-bit...
04:39 imirkin: i think that's why vblanks are fucked for me right now
04:40 airlied: imirkin: there were some changes from keithp and some fixes for those changes
04:40 airlied: maybe there are more changes needed
04:41 imirkin: airlied: my "tree" is 4.15 + nouveau patches
04:41 imirkin: airlied: do you happen to remember when such fixes would have gone in?
04:42 imirkin: (more specifically, my tree is ben's linux-4.16 branch)
04:42 imirkin: + a handful of patches that don't affect this either way
04:42 airlied: 3b765c0b765d2cc03ef02276f1af2658a03b3ced
04:43 airlied: drm/vblank: Data type fixes for 64-bit vblank sequences.
04:43 imirkin: i guess i don't have it: fatal: bad object 3b765c0b765d2cc03ef02276f1af2658a03b3ced
04:43 airlied: that might be drm-next only
04:43 imirkin: yeah, let me fetch that
04:43 airlied: not sure if it went to stable or not
04:44 imirkin: would have been under a diff hash :)
04:44 imirkin: ok, looks like that would fix at least the thing that i obviously see right now (since it makes drm_vblank_count return u64)
04:45 imirkin: but could well be that more is needed
04:45 imirkin: i'd suggest a way of initializing the vblank count to something high
04:45 imirkin: so that it can be tested without waiting for weeks of uptime
04:46 imirkin: [hrm ... not sure how i hit it so fast. i've only been up for 30 days, which according to my calculations means just over 150 million vblanks
04:46 imirkin: (at 60fps, full time, which wouldn't count dpms)
04:47 imirkin: skeggsb: --^
04:47 imirkin: my current number is 9.5 billion
04:50 gnarface: dropped frames?
04:51 imirkin: shouldn't cause vblank counter to go up afaik
05:01 gnarface: what if something weird happened like DPMS shut the screen off but the card somehow didn't react right and kept trying to play?
05:02 imirkin: well, clearly "something weird" happened
05:02 imirkin: question is what could it be
05:03 imirkin: and probably best answered by someone with intimate knowledge of the machinery
08:08 hrw: morng
08:33 hrw: is there a way to get rid of graphical artifacts? gtx 1050 ti, fedora 27, 4.15.10 kernel
10:44 pendingchaos: imirkin_: the third revision of the patches have been sent
10:45 pendingchaos: I think I'll be sending a new revision of the tests once I figure out the -fbo thing
11:31 karolherbst: imirkin: we have quite a lot of new fails in the CTS I think
11:31 karolherbst: KHR-GL45.packed_pixels.pbo*
11:31 karolherbst: or KHR-GL45.packed_pixels.rectangle.rg32ui
12:29 deevz: #join looking-glass
13:07 pendingchaos: imirkin_: a new revision of the tests have been sent, though I messed up the In-Reply-To header.
13:16 pmoreau: deevz: What’s the output of `glxinfo | grep OpenGL`? If you have Tesla cards, you should get OpenGL 3.3, if you have Kepler+ then you would 4.3 (not entirely what you get for Fermi, but something between 3.3 and 4.3).
13:20 deevz: pmoreau, BadValue (integer parameter out of range for operation)
13:20 deevz: Major opcode of failed request: 154 (GLX)
13:20 deevz: Minor opcode of failed request: 24 (X_GLXCreateNewContext)
13:21 deevz: Every opengl executables returns the same
13:21 pmoreau: `LIBGL_DEBUG=verbose glxinfo`?
13:22 pmoreau: Also, could you please attach the output of dmesg and your Xorg.0.log?
13:24 deevz: Gleaned this from my Xorg log: AIGLX: Screen 0 is not DRI2 capable, (EE) AIGLX: reverting to software rendering
13:25 deevz: And my dmesg output: https://pastebin.com/tqZKuVzT
13:27 pmoreau: Thanks. Something seems to be going wrong when configuring the DP output.
13:28 pmoreau: Lyude: You have been working on DP IIRC, have you seen something like this before? https://pastebin.com/tqZKuVzT
13:29 deevz: Graphics driver have been giving me hell since I switched to a custom 4.15.12 kernel
13:30 deevz: Gave up on nvidia because compiling the module is hell
13:31 pmoreau: Could you share the whole output from the Xorg log , please?
13:32 deevz: https://pastebin.com/dw7x1Ysp
13:35 pmoreau: Try installing the xf86-video-nouveau package (or however it is called on your distribution), to get the Nouveau DDX instead of the modesetting one; it should give you acceleration.
13:35 pmoreau: imirkin: On Maxwell+, do you only get acceleration with the Nouveau DDX?
13:37 deevz: Ok, I'll try to reboot with that change
13:39 deevz: Didn't solve the glxinfo errors
13:40 deevz: That said, my Xorg log is looking better
13:40 pmoreau: :-/
13:40 imirkin: pmoreau: nah, should work ok
13:40 pmoreau: Okay
13:40 deevz: [ 4.294]
13:40 deevz: X.Org X Server 1.19.6
13:40 deevz: Release Date: 2017-12-20
13:40 deevz: [ 4.294] X Protocol Version 11, Revision 0
13:40 deevz: [ 4.294] Build Operating System: Linux 4.14.12-1-ARCH x86_64
13:41 deevz: [ 4.294] Current Operating System: Linux david-pc 4.15.10-1-vfio #1 SMP PREEMPT Sat Mar 24 18:37:27 EDT 2018 x86_64
13:41 deevz: [ 4.294] Kernel command line: BOOT_IMAGE=/boot/vmlinuz-linux-vfio root=UUID=5a42250e-62e3-45ca-8b3c-a2540c555e65 rw quiet intel_iommu=on pcie_acs_override=downstream
13:41 deevz: [ 4.294] Build Date: 26 January 2018 10:25:18AM
13:41 deevz: [ 4.294]
13:41 deevz: [ 4.294] Current version of pixman: 0.34.0
13:41 deevz: [ 4.294] Before reporting problems, check http://wiki.x.org
13:41 deevz: to make sure that you have the latest version.
13:41 pmoreau: Hum
13:41 deevz: [ 4.294] Markers: (--) probed, (**) from config file, (==) default setting,
13:41 deevz: (++) from command line, (!!) notice, (II) informational,
13:41 deevz: (WW) warning, (EE) error, (NI) not implemented, (??) unknown.
13:41 deevz: [ 4.294] (==) Log file: "/var/log/Xorg.0.log", Time: Wed Mar 28 09:37:44 2018
13:41 deevz: [ 4.295] (==) Using config directory: "/etc/X11/xorg.conf.d"
13:41 deevz: [ 4.295] (==) Using system config directory "/usr/share/X11/xorg.conf.d"
13:41 deevz: [ 4.296] (==) No Layout section. Using the first Screen section.
13:41 deevz: [ 4.296] (==) No screen section available. Using defaults.
13:41 deevz: [ 4.296] (**) |-->Screen "Default Screen Section" (0)
13:41 deevz: [ 4.296] (**) | |-->Monitor "<default monitor>"
13:41 pmoreau: deevz: Please use pastebin/similar for more than 3 lines
13:42 pmoreau: imirkin: Thanks :-)
13:42 deevz: [ 4.302] (WW) Open ACPI failed (/var/run/acpid.socket) (No such file or directory)
13:42 deevz: [ 4.302] (II) LoadModule: "glx"
13:42 deevz: [ 4.302] (II) Loading /usr/lib/xorg/modules/extensions/libglx.so
13:42 imirkin: seriously?
13:43 deevz: [ 4.686] (II) Module glx: vendor="X.Org Foundation"
13:43 deevz: [ 4.686] compiled for 1.19.6, module version = 1.0.0
13:43 deevz: [ 4.686] ABI class: X.Org Server Extension, version 10.0
13:43 deevz: [ 4.702] (II) Unloading nv
13:45 pmoreau: deevz: Please don’t paste more than 3 lines in chat.
13:45 imirkin: deevz: your client is flooding. pm me when your client stops, and i will unquiet you in this chan.
13:46 deevz: test
13:46 pmoreau: pong
13:46 deevz: Thanks
13:46 deevz: hehe
13:46 imirkin: don't flood channels.
13:46 deevz: thats common sense
13:46 deevz: I thought I had the pastebin link in my clipboard, but it was the text I just put in said bin
13:46 deevz: Sorry about that XD
13:47 deevz: I'm surprised I was even able to post that much stuff at once
13:47 imirkin: well - it comes in one line per second
13:47 imirkin: which was why i kicked you
13:47 deevz: I see
13:47 imirkin: but ... your client appears to be extra-clever
13:47 deevz: Client displayed it all at once
13:47 imirkin: so when you rejoined, it just kept on flooding it.
13:47 deevz: Not sure if bug or feature :P
13:48 deevz: Using hexchat
13:48 pmoreau: Yeah, it almost happened to me a few times, but thankfully my chat client asks me whether I really want to paste X lines before doing it, so I can always cancel it.
13:48 deevz: your client seems more sensible than mine
13:48 pmoreau:uses WeeChat
13:49 deevz: haven't set my mind on learning the weechat shortcuts to properly use it yet
13:49 imirkin: so you got that pastebin link?
13:49 deevz: Even though I use my keyboard for most things including browsing, transition is still ongoing
13:49 deevz: https://pastebin.com/GRgGhHMB
13:49 deevz: There it is
13:50 pmoreau: I haven’t learned that many shortcuts yet. Mainly using the ones for switching between buffers, and scrolling up and down; that’s about it. :-D
13:50 imirkin: deevz: well that seems much happier
13:50 imirkin: deevz: pastebin glxinfo ?
13:51 deevz: I still get the same error as before
13:51 imirkin: the X opcode thing?
13:52 deevz: https://pastebin.com/3MWVbSA6
13:52 deevz: Yes
13:52 imirkin: mind doing 'LIBGL_DEBUG=verbose strace -f -e open glxinfo |& somefile'
13:52 imirkin: er
13:52 imirkin: >& somefile
13:52 imirkin: and pastebinning that?
13:53 deevz: is 'strace' a real command?
13:54 imirkin: yes.
13:54 imirkin: syscall trace
13:54 deevz: Cool, its packaged on my distro
13:54 imirkin: oh, and in case hrw comes back - tell him to use xf86-video-nouveau if he wants to avoid artifacts on maxwell+.
13:55 deevz: imirkin, it's no more verbose than glxinfo. Exactly the same input, in fact
13:55 deevz: output*
13:56 imirkin: can you tell me EXACTLY what command you ran?
13:56 imirkin: i.e. copy & paste from the shell
13:56 deevz: sudo LIBGL_DEBUG=verbose strace -f -e open glxinfo >& glxinfo.log
13:56 imirkin: remove 'sudo', continue.
13:57 deevz: Same thing
13:58 imirkin: strace should have caused a ton of 'open(foo)' lines to get printed
13:58 imirkin: you don't see those?
13:58 deevz: Nope
13:58 imirkin: instead of "open", can you try "openat"?
13:58 imirkin: (perhaps glibc is doing something clever)
13:59 deevz: Getting something more verbose now: https://pastebin.com/eHqm3YsY
13:59 imirkin: openat(AT_FDCWD, "/usr/lib/nvidia/libGL.so.1", O_RDONLY|O_CLOEXEC) = 3
13:59 imirkin: figured out what's wrong yet? :)
13:59 deevz: Honestly have no clue what that entails
13:59 imirkin: you're using blob libGL
14:00 deevz: Equals 3 what exactly? three schmeckles? XD
14:00 imirkin: that's why glamor didn't load either
14:00 imirkin: more like the path
14:00 imirkin: 3 is just the file descriptor number, it's meaningless here
14:00 imirkin: openat(AT_FDCWD, "/usr/lib/libnvidia-glcore.so.340.106", O_RDONLY|O_CLOEXEC) = 3
14:00 imirkin: etc
14:00 imirkin: you're using blob userspace
14:00 deevz: So it refers to an enum of sorts?
14:00 imirkin: won't work with nouveau
14:01 imirkin: either install kernel blob driver, or switch to mesa-provided userspace libraries.
14:01 deevz: Wait wait... you wait I'm using binary libGl?
14:01 deevz: you mean*
14:01 imirkin: i do wait indeed :p
14:01 deevz: XD
14:02 imirkin: (or just end your troubles and get an AMD gpu...)
14:02 deevz: Not going to happen considering I have two perfectly good nvidia cards in my case atm XD
14:02 imirkin: apparently not *perfectly* good ...
14:02 deevz: hehe
14:03 deevz: Well, it'd all work fine if I just stuck to my distro stock kernel
14:03 deevz: But I'm being a crafty bastard and trying to do gpu passthrough with a bunch of patches
14:04 deevz: Besides, I heard AMD was actually worse in terms of open source support. Maybe the situation has changed
14:05 imirkin: worse than nouveau? that may have been true at some point, but hasn't been in years.
14:05 deevz: lib32-mesa sounds like a promising package to try
14:05 imirkin: most importantly, they have a team of developers, with access to documentation, hardware, and internal resources.
14:06 imirkin: i think on debian-based things, there's a "update-alternatives" which lets you switch between things?
14:06 deevz: dedicated to foss?
14:06 imirkin: eys.
14:06 imirkin: yes.
14:06 deevz: sounds great
14:06 imirkin: whereas nvidia has locked down their power management unit and isn't sharing firmware.
14:07 deevz: Alright, I really have to fly to work
14:08 deevz: It's past 10am already
14:08 imirkin: who doesn't :)
14:08 deevz: Any normal person on the east coast I suppose
14:08 deevz: Because they'd already be at work XD
14:09 deevz: thanks for helping, imirkin & pmoreau
14:09 deevz: have a good day
14:09 pmoreau: Same to you!
14:09 imirkin: normal's in the eye of the beholder...
14:10 imirkin: i like to think most people are paid to do more than keep seats warm
14:10 imirkin: and once you have that, strict start/end times are less important
14:54 hrw: hello
14:55 imirkin_: hrw: you were asking about artifacts on maxwell+?
15:01 annadane: the 5 day work week (or even 6) is honestly insane and a new model would significantly help with people's mental health
15:01 annadane: doesn't help that there's economic inequality
15:13 hrw: imirkin_: moving from 4.15.10 -> 4.16-rc4 solved problem
15:52 imirkin_: hrw: ah cool
15:55 hrw: imirkin_: decided to make f27 -> f28-beta update and it solved problem
15:56 imirkin_: glad it worked out for you
15:56 imirkin_: i think f27+ also defaults to wayland?
15:56 hrw: maybe distro is. my destkop is not
15:56 imirkin_: ah ok
15:56 hrw: tried wayland once, was not satisfied
15:57 imirkin_: yeah, i'm avoiding making changes to my setup until i truly have to
15:57 imirkin_: (which is hopefully never)
15:57 hrw: kde is wayland ready they say but last time fontsizes were different and some apps misbehaved
15:57 hrw: I am 42, have some habits, finger memory etc
15:58 imirkin_: woohoo, people older than me exist!
15:58 imirkin_: but yeah, i'm pretty happy with my current setup and see no need to change
15:58 imirkin_: haven't since the late 90's
15:59 hrw: ;D
16:23 karolherbst: the heck.. you can stop the glxgears animation by pressing "a"???
16:23 skeggsb: you can rotate the gears too! ;)
16:23 mupuf: karolherbst: feature creep :p
16:23 hrw: glxgears...
16:23 karolherbst: skeggsb: :D
16:23 karolherbst: skeggsb: didn't know that
16:23 skeggsb: use the arrow keys
16:23 imirkin_: i had no idea either
16:26 imirkin_: skeggsb: any thoughts on the vblank counter thing?
16:26 imirkin_: like ... why did it get so high?
16:26 skeggsb:doesn't even know how that insane code works
16:26 skeggsb: like... at all
16:34 imirkin_: didn't you have a hand in writing it?
16:34 imirkin_: i think my direct issue was because of u32 vs u64 confusion
16:34 imirkin_: but ... how'd the counter get so high?
16:34 imirkin_: shouldn't it map 1:1 with vblank refreshes?
16:35 imirkin_: after only 30 days of uptime and a 60hz screen
16:56 Lyude: pmoreau: I think I have
16:57 Lyude: yes I ddefinitely has, seeing as I was the one who reported the issue that added that line :p
16:58 Lyude: that's something I see with the chamelium, where the default settings on the i2c aux chip will make it so that it reads out bytes on the i2c aux line as it gets them, but does so without actually properly indicating how many bytes it read back
16:58 Lyude: iirc it does something like 1 byte return for a 16 byte read. Which is -technically- valid, albeit annoying and confusing enough to scare some MST hubs
16:59 Lyude: to be fair though, having WARN_ON() seems like a bit much
16:59 Lyude: pmoreau: does it actually manage to train and bring up the display how you'd expect it to?
18:16 pmoreau: Lyude: I guess it did, as deev did not complain about a monitor not coming up, just about not getting any OpenGL application to work.
18:18 Lyude: huh, I'd make sure it actually came up
18:19 pmoreau: Well, (s)he is gone for now, dunno if (s)he will come back. So we might never know more about it :-/
18:19 pmoreau: Oh, maybe we can detect that from the Xorg logs.
18:43 pendingchaos: imirkin_: should I put spaces around the * in "(uint32_t)(cso->conservative_raster_dilate*4) << 8"?
18:53 imirkin_: yes
18:57 pendingchaos: oh and what about the < in "(post_snap || class_3d<GP100_3D_CLASS)"?
18:57 booti386: Hello
18:58 booti386: I often get: nouveau 0000:01:00.0: swiotlb: coherent allocation failed, size=2097152
18:58 imirkin_: pendingchaos: yes
18:58 imirkin_: booti386: you need a patch
18:59 booti386: Ah! Which one? :)
18:59 imirkin_: the one in my tree at home
18:59 imirkin_: ;)
18:59 imirkin_: let's see if i can find it...
19:00 booti386: I'm sure i can grab it from here :)
19:00 imirkin_: right. question is which patch.
19:01 imirkin_: 40b4667babaf73218e62cd73a17184a7c2d50434
19:02 imirkin_: looks like it's in v4.16-rc1+
19:04 booti386: I'm on rc5
19:05 pendingchaos: imirkin_: does the "Reviewed-by: ..." in https://lists.freedesktop.org/archives/mesa-dev/2018-March/190383.html apply for patches 1 to 3, 2 and 3 or just 3?
19:05 pendingchaos: and should I resend patches that have been replied with "Reviewed-by: ..."?
19:06 imirkin_: it applies to patches 2 and 3
19:06 imirkin_: (since the response is on patch 3 and it mentions that 2 and 3 are good)
19:06 imirkin_: you should resend them if you resend the whole series (and with the R-b tag on them)
19:06 imirkin_: but don't resend them just for the sake of resending
19:12 booti386: imirkin_: Also the display freezes for 1 second when it appears in the logs
19:14 pendingchaos: sorry, could you rephrase that?
19:14 pendingchaos: I think I should resend patch 1, 3 and 4 since they changed though I don't know about patch 2.
19:17 imirkin_: booti386: i think it's unrelated
19:18 imirkin_: pendingchaos: if you made a bunch of changes to 1 and 4, go ahead and resend 2+3
19:18 imirkin_: but include the R-b
19:18 imirkin_: but if you weren't going to resend anything, theoretically, no point in resending *with* the R-b just to add the r-b
19:18 pendingchaos: ah ok
19:19 imirkin_: but when you resend, definitely include the r-b
19:19 imirkin_: otherwise people get mad (and do extra work since they don't realize it's been reviewed)
19:33 imirkin_: pendingchaos: aaaaah much better. thanks for adding the spaces =]
19:33 imirkin_: i can breathe now
22:53 Lyude: https://paste.fedoraproject.org/paste/~q1CwSoXaAAUbqGGLDGrMw anyone seen this before? GPU is a GP104
22:53 Lyude: ( skeggsb ?)
22:53 imirkin_: files missing?
22:54 Lyude: ah, might be that
22:54 Lyude: i'll double check I've got the necessary firmware in a sec
22:54 imirkin_: oh hm no
22:55 imirkin_: if (!nvbios_pmuRm(bios, type, &pmu)) {
22:55 imirkin_: nvkm_error(subdev, "VBIOS PMU fuc %02x not found\n", type);
22:55 imirkin_: so it hates your vbios.
22:55 Lyude: ah
22:56 Lyude: knowing which company made this laptop makes that fact far less surprising
22:56 imirkin_: HP?
22:56 Lyude: good guess
22:56 imirkin_: =]
22:57 imirkin_: entirely possible that the vbios is fine, and the code just didn't anticipate some situation
22:57 imirkin_: you'll have to dump it out and have a look though
22:57 Lyude: yeah
22:57 Lyude: ...oh yeahhhh, we have the vbios docs now as well :)
22:58 imirkin_: ... some limited ones