| FurretUber | Hi, what is the current situation on the fixes for Intel GPUs? Currently they are unusable with 5.4 from Focal. Are there plans to have them fixed for the release? | 00:14 |
|---|---|---|
| lotuspsychje | FurretUber: unusable how? | 00:16 |
| FurretUber | Always at full power, constant GPU hangs | 00:16 |
| lotuspsychje | FurretUber: wich intel graphics are you on please? | 00:16 |
| FurretUber | Intel HD Graphics 520 | 00:17 |
| lotuspsychje | i have a 620 here and got a few bugs about xorg, kernel oops and the dock | 00:18 |
| lotuspsychje | FurretUber: can you describe your symptons exactly please? | 00:18 |
| tomreyn | also, have you filed a bug or found an existing bug report on the specific issues you're experiencing? | 00:19 |
| FurretUber | Once a heavier task starts (the lightest one was the Audacious OpenGL viewer), GPU hangs happen endlessly until the task is closed. Any heavier task then will repeatedly caused GPU hangs | 00:19 |
| FurretUber | All kernels provided by Canonical have this problem. They are related to that GPU security issue | 00:20 |
| FurretUber | Here is the error log from the GPU with Kerbal Space Program on 5.4.0-9: https://paste.ubuntu.com/p/yBvT7gTdPC/ | 00:21 |
| lotuspsychje | you have a source for this FurretUber ? | 00:21 |
| FurretUber | About all kernels with the problem? I tested them all, after the fix for the security issue, I'm having to resort to older versions, patched versions with the fix removed or drm-tip to use my computer | 00:21 |
| lotuspsychje | a bug ID as tomreyn asked FurretUber | 00:22 |
| FurretUber | Hmmm... I have reported a related bug twice, but not this one...? | 00:24 |
| lotuspsychje | FurretUber: you say all intel gpu's have bugs, where did you pick that up? | 00:25 |
| FurretUber | The patches affected Intel GPUs starting from Ivy Bridge. It, at least, disabled all power management | 00:28 |
| lotuspsychje | FurretUber: your gpu hang results into what exactly? xorg crash? loginloop? system freeze to a reboot? black screen? | 00:30 |
| FurretUber | Constant frozen screen. Using Ctrl+Alt+F4 will access a TTY if I wait a few minutes. Killing the offending application makes the GPU hangs stop, but then any heavier task brings the endless GPU hangs again | 00:33 |
| lotuspsychje | FurretUber: can you see if this is you please? https://bugs.launchpad.net/ubuntu/+source/xorg-server/+bug/1853266 | 00:35 |
| ubottu | Launchpad bug 1853266 in xorg-server (Ubuntu) "Xorg/Xwayland segfaults in OsLookupColor() from funlockfile() from glamor_get_pixmap_texture() from glamor_create_gc()" [Undecided,Confirmed] | 00:35 |
| FurretUber | I don't think so, nothing crashes, not even the offending application (there was a significant work to remade the reset engines on Intel GPUs, which I found a critical problem on and delayed a kernel release :p ) | 00:37 |
| lotuspsychje | FurretUber: how about you file a bug for your system then, let us see whats all going there | 00:39 |
| FurretUber | I will report a bug on this GPU hang, as I'm not finding anything similar. That error file is the only thing I got, but I think it can be very useful | 00:39 |
| FurretUber | Hmm... should I trigger a GPU hang before reporting the bug to get a dmesg and card0error? | 00:40 |
| lotuspsychje | FurretUber: you didnt find anything similar, but you say all kernels have this issue? | 00:40 |
| lotuspsychje | or did you mean you tested several kernels on your system? | 00:41 |
| FurretUber | I meant on Launchpad. https://linuxreviews.org/Linux_Kernel_5.5_Will_Not_Fix_The_Frequent_Intel_GPU_Hangs_In_Recent_Kernels | 00:41 |
| lotuspsychje | right, now we getting somewhere | 00:42 |
| lotuspsychje | FurretUber: so all the bugs presented on that link, aint happening to your system? | 00:43 |
| FurretUber | The high temperature issue is present due to the disabled power management, but none of the linked one there is the exact problem I'm facing. This is why I'll report on Launchpad now | 00:45 |
| lotuspsychje | will you use ubuntu-bug FurretUber ? | 00:46 |
| FurretUber | Yes. I'll just reproduce it to get a dmesg log | 00:47 |
| lotuspsychje | great, tnx for taking your time to file it | 00:47 |
| lotuspsychje | FurretUber: is your bug creating? | 01:00 |
| FurretUber | No, just got the dmesg and card0 error | 01:05 |
| lotuspsychje | whats your launchpad account FurretUber ? | 01:06 |
| lotuspsychje | ... | 01:13 |
| FurretUber | Sorry, here is the bug report: https://bugs.launchpad.net/ubuntu/+source/linux-signed-5.4/+bug/1861590 | 01:15 |
| ubottu | Launchpad bug 1861590 in linux-signed-5.4 (Ubuntu) "Intel GPU hangs constantly under load. After closing offending application, all GPU intensive programs run degraded" [Undecided,New] | 01:15 |
| FurretUber | No dmesg was uploaded? I used ubuntu-bug linux | 01:15 |
| lotuspsychje | FurretUber: you can still add it now if you like | 01:16 |
| FurretUber | I added them. I thought when reporting against linux it would upload the files :p | 01:17 |
| lotuspsychje | FurretUber: you upgraded to 20.04 from 19.10? | 01:18 |
| FurretUber | From 18.04 | 01:18 |
| lotuspsychje | FurretUber: did you test if you can reproduce your bug on a daily 20.04? | 01:19 |
| lotuspsychje | FurretUber: usually in this stage of development, we try to test bugs on daily iso, this to avoid alot of work for the devs | 01:20 |
| FurretUber | Well, I can do this test, but I will have to download the daily ISO | 01:20 |
| lotuspsychje | please do FurretUber so we can rule out a too early lts upgrade problem | 01:21 |
| FurretUber | I will test with the Xubuntu ISO, as it's Xubuntu I'm using now. It will take a while to download | 01:22 |
| lotuspsychje | FurretUber: your dmesg looks like it has errors about it allright | 01:25 |
| FurretUber | So, after 2 hours, I got the GPU hang on Focal Daily. Same symptoms | 03:43 |
| FurretUber | I have added the dmesg and card0/error to the bug report | 03:59 |
| tomreyn | FurretUber: is it still running, and if so, what's in journalctl -b -p3 | 04:02 |
| tomreyn | maybe -p4 also if that's not more than you want to share | 04:03 |
| FurretUber | No, I stopped it. I have to go sleep, but I can do this tomorrow, good night and thank you for the support | 04:13 |
| tomreyn | so related bug reports are https://bugs.freedesktop.org/show_bug.cgi?id=111812 https://gitlab.freedesktop.org/drm/intel/issues/451 and maybe https://gitlab.freedesktop.org/drm/intel/issues/673 | 04:33 |
| ubottu | Freedesktop bug 111812 in DRM/Intel "i915 0000:00:02.0: GPU HANG: ecode 9:0:0x00000000, hang on rcs0" [Major,Resolved: moved] | 04:33 |
| tomreyn | combining i915.enable_dc=0 and intel_idle.max_cstate=1 seems to give folks acceptable uptime. kernel 5.5 also seems to help (not sure whether it fully fixes it). | 04:36 |
| tomreyn | loads of reports there indeed, this looks bad. | 04:38 |
| tomreyn | and using those kernel parameters means lesser power savings. | 04:39 |
| lotuspsychje | FurretUber: tnx for testing a xubuntu daily on your bug | 14:25 |
| FurretUber | The journalctl -b -p4 log: https://termbin.com/yxfn it seems there is nothing new related to this problem | 17:56 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!