10:44 jjsa: I have an Acer Predator helios 300 with a geForce GTX 1050 Ti Mobile and Intel UHD Graphics 630 (Mobile). My distribution is fedora 30. Booting is only possible if I give nomodeset or nouveau.modeset=0. I have tried the nvidia components but this worse, The login screen ist placed within the not attached external display. I have not found a way for
10:44 jjsa: correcting this.I reverted finally to the nouveau driver which is open source. With the kernel command line nouveau.modeset=0 i915.modeset=1 The system work on the notebook display but of course the hdmi output don't work. and i get often error messages: "WARNING: CPU: 9 PID: 472 at drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmmgf100.c:221
10:44 jjsa: gf100_vmm_invalidate+0x1a6/0x1c0 [nouveau]". I want to use the hdmi output which is attached to the nvidia card occasionnely. Is there a solution.
10:52 karolherbst: jjsa: mhhh.. I could create a copr build which should contain some experimental fixes to solve all the issues
10:54 jjsa: karolherbst COPR, what do you mean?
10:54 karolherbst: yum repository
10:55 jjsa: Okay. I can perform some tests if i get the build.
10:56 karolherbst: yeah.. it will just take a while unless I figure out how to build locally
11:00 jjsa: Building locally shall not be complicated if only the kernel driver is to be modified. If some other dynamic libraries are also to be compiled there will be more to do.
11:01 karolherbst: yeah.. it's just a speed optimization. The copr infra needs like 10 hours to do a kernel build
11:07 jjsa: I see! the problem must no be resolved within a day, firthermore I will not have time on the next days. Booting a manjaro Linux from an iso image worked with the i915 and nouveau modeset=1 parameters mostly. The image contain a 4.x kernel, Fedora has actually 5.2.15-200.fc30.x86_64
11:10 karolherbst: jjsa: ohh, it won't be a good long term solution as a kernel upgrade will just give you a newer kernel without the fixes. We are in the process of getting those upstream, but it will still take some time and I doubt it will happen before a 5.4 kernel
11:10 karolherbst: maybe we get those backported
11:12 jjsa: karolherbst This is not really a problem. I may compile module locally and test them if I get the sources.
11:19 karolherbst: jjsa: I trigered a build here: https://copr.fedorainfracloud.org/coprs/karolherbst/Nouveau_Testing/build/1037764/
11:19 karolherbst: it's usually easier to just install a new kernel to test GPU drivers
11:20 karolherbst: otherwise you mess around with dracut and everything else
11:23 jjsa: Karolherbst Okay I have see this. As soon the build is finished, I will download and installt it. I have put your copr repo withn yum/dnf I assume that i will be able to see the update notification or am I wrong?
11:23 karolherbst: I really don't have anything else on the copr, so if you " dnf copr enable karolherbst/Nouveau_Testing" the repo you will get it as normal system updates
11:24 jjsa: Okay.
11:24 karolherbst: but.. the fedora updates repo may install a newer update later on
11:24 karolherbst: I am just mainly interested if those patches fixes all of your issues
11:25 jjsa: I can understood this. If this happen there will be nice and I should not maintain my old notebook.
11:25 karolherbst: jjsa: what you could do in the meantime is, if "nouveau.runpm=0" is enough to fix your issue or not
11:26 karolherbst: (which will kill battery lifetime, but that might be good enough for now untill we get those patches merged)
11:26 jjsa: I can test this, but I assume thatthis can only fix the error I mentioned in my first text.
11:28 karolherbst: it will probably also fix our external display issue.. but you will see if that will do the trick or not
11:28 jjsa: Okay. I am away fot the next hour, i have a lot to do.
11:33 alkisg: Here's a screenshot of the artifacts that nouveau shows with marco, that I was talking about last week: https://imgur.com/a/Vbh7nBw
11:33 alkisg: This was with: Subsystem: ASUSTeK Computer Inc. GT218 [GeForce 210] [1043:852d]
11:33 alkisg: Sometimes they're a lot more severe; and as I said they're fixed either by toggling the resolution, or by disabling xpresent
11:34 alkisg: Is there any program that uses xpresent a lot, that I could test with, to reproduce the problem more easily?
11:36 alkisg: Or, where could I file a bug report for this, to be able to track the issue better?
11:38 karolherbst: alkisg: weird
11:41 diogenes_: alkisg, the newest xfwm4 uses xpresent as a backend too.
11:41 alkisg: diogenes_: i tested it and this too has the issue
11:41 alkisg: Actually the marco code was copied from xfwm4, so... it's like testing 1 program
11:42 diogenes_: hehe i see, what about compiz?
11:42 alkisg: I haven't tested compiz, is it using xpresent, or just opengl/glx/whatever?
11:42 diogenes_: opengl (afaik)
11:43 alkisg: I guess it's not affected then, mutter and metacity that are not using xpresent aren't affected
11:44 diogenes_: i wish i could test it on my system but i have no clue how since i have optimus, is there a way to run xfwm4 with DRI_PRIME=1 or something?
11:45 alkisg:has no clue about optimus
11:46 diogenes_: i'll try to run my entire session with DRI_PRIME=1 and see what i get.
14:33 dashs: Debian 10 - where can I get GV100 firmware? How to install it?
14:39 karolherbst: ask debian devs
14:40 phillipsjk: dash it may be in the non-free repo if it exists,
14:48 phillipsjk: dash: https://packages.debian.org/buster/firmware-misc-nonfree
14:50 phillipsjk: For installion help, I suggest #debian
15:33 nvz: I got an old HP Pavilion dv9410us here with GeForce Go 6150 (10de:0244) and it starts the lightdm fine using noveau on buster, but the screen is all wrung out looking like diagonal lines. It looks normal for a brief second when switching VT to tty1, I'm thinking maybe there is a tweak of some kind that can fix the display issue?
15:56 phillipsjk: is this a CRT dsiplay? Does it look like a frequency mis-match?
15:57 phillipsjk: IIRC [Crt] [Alt] [+/-] lets you cycke through resolutions.
16:14 diogenes_: phillipsjk, he fixed it by turning off pageflip.
16:17 phillipsjk:plans to do a "reverse prime" thing: Vega FB -> NV FB -> VGA
16:18 phillipsjk: new cards don't have VGA outputs :P
16:19 phillipsjk: Oh, you were tlaking about the other thing.
16:27 phillipsjk: In troubleshooting freeBSD's radeon KMS, I ran into a similar issue: dragged out a CRT to cofirm that yeas, that is really what the video output looked like (looked like a frequency mis-match but was more a buffer resolution mis-match).