01:56 oblio: imirkin_: eh im fine w highres
01:56 oblio: imirkin_: (days later lol)
02:08 imirkin_: oblio: nice
02:08 imirkin_: i suspect you could get up to 4k if need be
02:11 oblio: imirkin_: i feel like i'll wait for a new video card
02:11 oblio: to run full res
02:11 oblio: rather than @30hz but *shrug*
02:13 oblio: imirkin_: how do you think i could get 4k tho?
02:13 imirkin_: bumping up hdmimax :)
02:14 oblio: oh lol
02:14 imirkin_: i based it on the mode that was reported over DP
02:14 oblio: do some trial and errorr and see what happens
02:14 imirkin_: but i suspect the one reported via HDMI has a higher bandwidth
02:14 oblio: maybe i'll try
02:14 imirkin_: nah, just like look at the xorg log
02:14 imirkin_: should be obvious
02:14 oblio: the old one from when it was on dp?
02:14 oblio: (dunno if it's still around but i'll check)
02:15 imirkin_: no
02:16 imirkin_: the new one from hdmi :)
02:17 oblio: ahh OK
02:17 oblio: [ 18.861] (II) NOUVEAU(0): Supported detailed timing:
02:17 oblio: [ 18.861] (II) NOUVEAU(0): clock: 533.2 MHz Image Size: 597 x 336 mm
02:17 oblio: 533.2?
02:17 imirkin_: no.
02:17 imirkin_: pastebin the whole thing.
02:17 oblio: haha ok
02:17 imirkin_: that's for 4k@60
02:17 imirkin_: which ain't happening
02:18 oblio: oh it's the next entry then
02:18 oblio: 297
02:19 imirkin_: maybe. just pastebin it :)
02:19 oblio: imirkin_: https://hastebin.com/keqokapuja
02:20 imirkin_: this is via DP
02:20 imirkin_: oh, i see. later with hdmi.
02:20 imirkin_: yeah, looks like 297 should work
02:21 imirkin_: but it might not _actually_ work
02:21 imirkin_: like i said before, it's a wee bit out of spec
02:21 oblio: i'm down to try
02:21 oblio: lol
03:31 oblio: imirkin_: lol
03:32 oblio: now the monitor isnt coming on during post
03:32 imirkin_: during _POST_?
03:32 oblio: haha yeah
03:33 oblio: it's been flakey so hoping it's just that
03:33 oblio: but might even try pulling one of the nvs out and removing the line
03:33 oblio: switching it to the other one
03:38 oblio: imirkin_: it's also possible it's unrelated b/c things have been flakey
03:38 oblio: but it had been fine
03:38 oblio: tbh i can't recall if i saw a POST screen on reboot after upping the mhz
03:42 oblio: imirkin_: got it up on the other card
03:42 oblio: i only need one anyway
03:42 oblio: it says 297 is an invalid value
03:44 oblio: apparently i put 297.0
03:45 oblio: imirkin_: it's up-- i think that card may have been flaking, the one i pulled
03:45 oblio: 3840x2160 @30hz
03:46 oblio: (hope you enjoy the play by play, ha)
03:48 imirkin_: cool
03:50 oblio: i wonder if it wasnt seating properly
03:50 oblio: no need to run the extra heat/power anyway
03:51 oblio: need to find a blank plate tho
04:21 oblio: imirkin_: im curious what your thoughts on wayland are
04:24 imirkin_: it solves some nasty problems, but i don't really have those problems
04:24 imirkin_: i don't expect to use wayland until ... not sure when
04:26 oblio: lol
04:27 oblio: was just curious, i've never used it
04:27 oblio: my newest problem i was dealing w was zoom screensharing had stopped working
04:27 oblio: which actually started working just now when i switched off of nouveau
04:27 oblio: anyway everything talks about enabling wayland support to fix your zoom screensharing problems but of course im not running wayland
04:28 imirkin_: zoom screenshare worksforme
04:28 imirkin_: with chrome
04:28 imirkin_: with firefox it hangs, but there's documented hangs in firefox + webrtc
04:28 oblio: i'm using the native app
04:28 imirkin_: i only ever share windows, not the whole screen though
04:28 imirkin_: ah yeah - i never do that
04:29 oblio: i dont think i realized i could use it in browser
04:29 imirkin_: yeah. they really want you to use their app
04:29 imirkin_: go to the invite link
04:29 imirkin_: then click on the "start the app" link
04:29 imirkin_: then DONT START THE APP
04:29 oblio: ha
04:29 imirkin_: that link becomes "open in web browser"
04:29 imirkin_: great system, right?
04:29 oblio: yep
04:31 imirkin_: running their app is pretty much a non-starter for me
04:32 imirkin_: (i don't like blobs running on my cpu...)
04:32 oblio: yeah
04:32 oblio: chromium a/v can kill tho
04:32 imirkin_: dunno. the webrtc stuff works fairly well
08:47 VanackSabbadium: Good morning karolherbst, i've got 2 logs for you
08:48 VanackSabbadium: this one should be referring to the latest crash: https://cloud.disroot.org/s/cMpp54daqSrpTtt
08:49 VanackSabbadium: this one should be after reboot: https://cloud.disroot.org/s/8epGCt3eZWCTXZx
08:51 VanackSabbadium: or maybe it's the opposite, i don't understand
08:52 VanackSabbadium: however these are the 2 logs inside .local/share/xorg, after the latest freeze
08:52 VanackSabbadium: i also updated to the latest git nouveau, but it crashed anyway with gimp+chrome+firefox+mpv opened together
08:57 karolherbst: those "failure to allocate surface 1x8@32 (pitch 64): -2" messages are odd
08:58 VanackSabbadium: what does that mean?
08:58 karolherbst: -2 is ENOENT and that I think is the error code we get if the gpu context is gone..
08:59 karolherbst: but why does it take minutes until X is gone as well..
08:59 karolherbst: strange
08:59 karolherbst: ohh wait
08:59 karolherbst: that's not true
08:59 VanackSabbadium: this time X did not crash, it rebooted automatically, it's the first time that happens, but i waited a looooooooong time
08:59 karolherbst: channel gone is ENODEV
08:59 karolherbst: mhhh
09:00 karolherbst: yeah.... there is something super funky going on
09:00 VanackSabbadium: oh yeah, funky XD
09:00 karolherbst: VanackSabbadium: do you have a kernel log as well>?
09:01 VanackSabbadium: is it in /var/log?
09:02 karolherbst: VanackSabbadium: journalctl --dmesg --boot -1 or so...
09:02 karolherbst: you'd need to find the correct boot
09:02 VanackSabbadium: i think i have it
09:02 VanackSabbadium: i have a very big log since june 5
09:02 VanackSabbadium: but i see the latest crashes also
09:02 VanackSabbadium: wait
09:03 VanackSabbadium: alright this should be enough stuff for you to work on: https://cloud.disroot.org/s/pykZD7DYMBezrka
09:04 karolherbst: you can replace -1 by a different number to find the one correct one.. -1 just means the last one where 0 ist the current, etc...
09:05 karolherbst: mhhhh
09:06 karolherbst: anyway.. I think scanning through the olds could help with something.. let's see
09:06 VanackSabbadium: this is the result of journalctl etc..... https://cloud.disroot.org/s/poaBLHQwyzmaz5J
09:07 karolherbst: ahh yeah
09:08 karolherbst: mhhh
09:08 karolherbst: okay. seems like we fail to recover the GPU
09:09 karolherbst: what's weird is that sometimes it seems like with compton with upload a broken shader
09:09 karolherbst: not sure why that is
09:10 VanackSabbadium: i could try using my desktop without Compton and see what happens
09:10 VanackSabbadium: i gotta disable it at start
09:10 karolherbst: yeah.. I'd even assume that works better
09:11 VanackSabbadium: alright, i do it
09:11 karolherbst: xfce4-wm might be a good alternative for now
09:11 karolherbst: uhm
09:11 karolherbst: xfwm4 is the name :D
09:11 VanackSabbadium: i disabled compton
09:11 VanackSabbadium: killed the process XD
09:11 karolherbst: yeah well..
09:11 VanackSabbadium: i'm using compton kawase, for kawase blur
09:11 karolherbst: you need some compositor though
09:12 VanackSabbadium: no i don't need it
09:12 VanackSabbadium: i have a wm without compositor, that's not a problem
09:12 VanackSabbadium: it's just less beautiful
09:12 karolherbst: okay
09:13 karolherbst: VanackSabbadium: mind creating a bug with that nouveau crash though?
09:13 VanackSabbadium: no problem, what should i attach?
09:13 karolherbst: maybe skeggsb knows what to do
09:14 karolherbst: VanackSabbadium: just the full kernel log showing the crash and say something like after that you won't be able to start X anymore and need to reboot
09:14 karolherbst: that issue is serious enough on its won
09:14 karolherbst: *own
09:14 karolherbst: with that fixed you should be able to just restart everything
09:14 VanackSabbadium: fine
09:15 karolherbst: so we have essentially two bugs here: compton crashing the GPU, which is probably a mesa bug and now idea what is going on there
09:15 karolherbst: the second is that nouveau fails to recover the GPU
09:15 VanackSabbadium: here? https://gitlab.freedesktop.org/xorg/driver/xf86-video-nouveau/-/issues
09:23 VanackSabbadium: sorry, it disconnected, i'm opening a bug
09:25 karolherbst: cool, thanks
09:26 VanackSabbadium: https://gitlab.freedesktop.org/xorg/driver/xf86-video-nouveau/-/issues/538
09:26 VanackSabbadium: see if that's enough
09:27 karolherbst: VanackSabbadium: I meant the log with the crash though
09:27 karolherbst: that one: https://cloud.disroot.org/s/poaBLHQwyzmaz5J
09:28 VanackSabbadium: oh, ok
09:28 VanackSabbadium: gonna change it
09:28 karolherbst: thanks
09:30 VanackSabbadium: done
09:30 VanackSabbadium: fuck all these captchas XD
09:30 karolherbst: there are captchas? :D
09:31 VanackSabbadium: yeah, if you modify your post after a few minutes it asks for captchas to check if you're a bot
09:31 karolherbst: VanackSabbadium: mhh.. mind uploading the log as a plain text file? md kind of messes up all underscores
09:31 karolherbst: ehhh
09:31 karolherbst: annoying
09:31 VanackSabbadium: damn XD
09:31 VanackSabbadium: alright
09:32 VanackSabbadium: done
09:32 VanackSabbadium: check it out
09:33 karolherbst: cool, thanks
09:33 VanackSabbadium: no prob
09:33 VanackSabbadium: gonna do a test right now
09:34 VanackSabbadium: stressing nouveau without compositor to see what happens
09:45 VanackSabbadium: karolherbst it crashed again, even without compton
09:45 VanackSabbadium: waited 2 minutes, same errors as before, than rebooted automatically
09:46 VanackSabbadium: you mind give me again that journalctl command so i can give you the latest kern log?
09:52 karolherbst: VanackSabbadium: journalctl --dmesg --boot -1
09:55 VanackSabbadium: nope, nothing interesting
09:56 VanackSabbadium: i also used 0
09:56 karolherbst: VanackSabbadium: then the reboot is strange
09:58 karolherbst: VanackSabbadium: normally there should be a reason for the reboot to happen
09:59 VanackSabbadium: oh wait
09:59 VanackSabbadium: here is the log, don't now why the first time it did not show
10:00 VanackSabbadium: https://cloud.disroot.org/s/Psb2qGkFBCZg3zP
10:01 VanackSabbadium: also modified the issue opened on gitlab
10:18 karolherbst: VanackSabbadium: I have an idea.. let me check something
10:20 VanackSabbadium: i'm going to work now, but feel free to write me at vanacksabbadium@disroot.org
10:21 VanackSabbadium: or, if you use Telegram, t.me/vanacksabbadium
10:22 VanackSabbadium: i will connect here in the late afternoon from my workplace
16:17 imirkin_: skeggsb: https://github.com/skeggsb/nouveau/commit/2b9345f98be41e2145f561f6cfecc99686f4ef71 -- probably want to update the comment about it waiting for 1ms
16:35 Tom^: excuse my poor memory which is fading, what tool did you use to "record" a game and be able to send that dump/recording and let others view/play it? glrecordsomethingsomething... or similiar. :P
16:35 Tom^: gltrace maybe? hm
16:36 RSpliet: apitrace ?
16:36 Tom^: that might be it
16:36 Tom^: ah yes seems so
16:36 Tom^: thanks
16:39 imirkin_: Tom^: there's also renderdoc now
16:40 imirkin_: which also works with vulkan
16:40 Tom^: imirkin_: oh nice
17:49 VanackSabbadium: hi karolherbst, you were telling me about an idea...
18:25 karolherbst: VanackSabbadium: ahh yeah.. you could try if it's better with the nvidia firmware, meaning booting with nouveau.config=NvGrUseFW=1 and extracting the firmware
18:27 VanackSabbadium: yeah i've seen this around, someone says it works. I gotta try. What do you mean by extracting the firmware? And how do i do that?
18:31 karolherbst: VanackSabbadium: some distributions have packages for that
18:32 karolherbst: VanackSabbadium: or you can use this script: https://github.com/envytools/firmware/blob/master/extract_firmware.py
18:33 karolherbst: although..
18:33 karolherbst: I think it only extracts video firmware? mhh let me check
18:33 karolherbst: ahh no, it also extracts the graph stuff
18:34 karolherbst: imirkin_: do you need to use the 325.15 driver for extracting the graph firmware?
18:34 imirkin_: es
18:34 imirkin_: yes*
18:34 karolherbst: okay
18:35 VanackSabbadium: this is where i read about it, very old topic -> https://131.252.210.165/show_bug.cgi?id=93629
18:35 VanackSabbadium: someone says the trick do not work anymore
18:36 imirkin_: it works.
18:36 VanackSabbadium: sure?
18:36 imirkin_: there was a period of time when it stopped.
18:36 imirkin_: but it should still work now.
18:36 imirkin_: and no, i'm never sure of anything
18:37 imirkin_: the pgraph firmware helps under fairly specific conditions IME
18:37 imirkin_: i.e. the GTX 660 for which our thing doesn't work
18:37 karolherbst: imirkin_: yeah.. but it seems like VanackSabbadium has an issue where the graph engine literally stops working
18:37 karolherbst: not even reseting works...
18:38 karolherbst: but yeah..
18:38 karolherbst: I also see the chances of the blob firmware to improve the situation quite low.. but at least that's something to try out
18:38 karolherbst: imirkin_: log is here: https://gitlab.freedesktop.org/xorg/driver/xf86-video-nouveau/uploads/d2a6260fe68bc3e6ecc8c3b5d62f9cf0/kernel_latestest.txt
18:39 karolherbst: but I guess our recovery code is simply busted
18:39 imirkin_: so it breaks like an hour after boot?
18:39 karolherbst: or faster
18:39 VanackSabbadium: it's totally random
18:39 karolherbst: or slower
18:40 VanackSabbadium: i can trigger it playing a game which resize my desktop window 3 times in a row
18:40 VanackSabbadium: (which is Desperados, BTW)
18:40 RSpliet: It's a Fermi, that gen had the least love of all from nouveau devs.
18:40 imirkin_: VanackSabbadium: can you pastebin /proc/interrupts?
18:40 VanackSabbadium: not now, i'm @ work
18:40 imirkin_: ah ok
18:40 karolherbst: imirkin_: after it happens or before?
18:40 imirkin_: well i have this vague recollection that there were some weird MSI issues with some Fermi boards
18:40 karolherbst: ohh, interesting
18:40 imirkin_: you could try booting with nouveau.config=NvMSI=0 to disable it
18:41 VanackSabbadium: i'll try that too
18:41 imirkin_: i thought we accounted for those issues in our logic, but who knows
18:41 imirkin_: i think it was with nvc3 though, not nvcf
18:41 imirkin_: or nvc4 maybe?
18:42 karolherbst: VanackSabbadium: anyway, thanks for being so patient about this.. bugs like this can just be super painful to debug, especially if we don't own the hardware to reproduce it :/
18:42 VanackSabbadium: what? are you kidding me?
18:43 VanackSabbadium: i didn't expect to find so much helpful guys and also the same nouveau developers
18:43 karolherbst: :)
18:43 VanackSabbadium: YOU are patient, not me!
18:44 VanackSabbadium: i began using nouveau drivers last year, when i installed Trisquel for the first time
18:44 karolherbst: VanackSabbadium: ohh btw, what might help is to state what exact model you have.. I have no fermi GPUs so I might just go ahead and buy some in the future for work anyway
18:45 VanackSabbadium: GeForce GTX 550 Ti
18:45 karolherbst: yeah.. but I also meant the vendor + fancy marketing name + model number and stuff
18:45 VanackSabbadium: Oh, yeah, i can't remember, i gotta check that
18:46 karolherbst: the shity thing is, that 100 550 ti could work flawlessly and you just happen to own the one being broken
18:46 VanackSabbadium: This GPU has 8 fucking years XD
18:46 VanackSabbadium: i don't know... seems a very common issue
18:47 karolherbst: well
18:47 VanackSabbadium: i could simply install proprietary drivers, but that's not the point
18:47 karolherbst: what does "common" even mean here? how many users are there having no problems
18:47 karolherbst: 1000? 10.000?
18:47 karolherbst: more?
18:47 VanackSabbadium: maybe, who knows? i'm not so lucky in my life XD
18:48 karolherbst: well, skeggsb has a few fermi cards but he is so "lucky" always getting the ones working just fine
18:48 karolherbst: :/
18:48 RSpliet: one does not simply install proprietary drivers. Fermi is "legacy", those drivers receive very little love from NVIDIA to stay compatible with new kernels/X.org/...
18:48 VanackSabbadium: oh well, the last time i used proprietary drivers, i had to install a very old version of them
18:48 karolherbst: yeah.. fermi
18:49 karolherbst: 340.x I think?
18:49 karolherbst: or was fermi 404?
18:49 VanackSabbadium: hmm... yeah, something similar
18:49 VanackSabbadium: no no i think 340.x
18:49 karolherbst: 390 actually
18:49 VanackSabbadium: if i remember correctly
18:49 VanackSabbadium: i don't know, the suggested ones by nvidia on their site
18:49 karolherbst: 390 seems like the latest one supporting fermi
18:49 VanackSabbadium: exactly
18:50 karolherbst: 340 was tesla
18:51 VanackSabbadium: you know, i was a little bothered when i saw that power management for Fermi cards was "mostly working but still WIP", and then discovering that it doesn't work XD
18:51 karolherbst: there are some patches around
18:51 karolherbst: but uff...
18:51 karolherbst: "untested" and "unfinished
18:51 karolherbst: "
18:51 VanackSabbadium: but i said... well, i mostly play old games, i don't care about 60 fps or motion blur... i don't care about performances, if that game works and it's playable, i'm good
18:51 karolherbst: RSpliet: btw, what would you need to finish it :p
18:52 VanackSabbadium: come on! i will be with this card for years! I can't afford a new one!
18:52 VanackSabbadium: XD
18:52 RSpliet: karolherbst: time and a good salary.
18:53 RSpliet: Oh, and access to those GPUs, which are currently boxed up in the Netherlands
18:53 karolherbst: ohh, I have a stupid idea :D
18:53 VanackSabbadium: RSpliet are you the one working on it?
18:54 RSpliet: VanackSabbadium: many many moons ago. Another dev did a lot of work on it too
18:54 karolherbst: RSpliet: when will your time problem be fixed?
23:57 grgalex: q