So, I am looking for any ideas, help or head(s) smarter than mine.
This is my main problem for some time now, easiest way to replicate this is to run anything proton related. After that, some apps just keep spitting this via dmesg:
[ 1403.146954] eartag[5309]: segfault at 30 ip 000077f7d796491c sp 00007ffe20a35210 error 6 in libgtk-4.so.1.1400.4[56491c,77f7d74c7000+4c2000] likely on CPU 7 (core 3, socket 0)
[ 1403.146964] Code: c2 4c 8d 4d c0 48 8d 35 02 75 4a 00 45 31 c0 0f b6 d2 e8 c7 4c 01 00 f3 0f 10 03 48 8b 55 c0 4c 89 fe f3 41 0f 58 07 4c 89 f7 <f3> 0f 11 42 30 f3 0f 10 43 04 f3 41 0f 58 47 04 f3 0f 11 42 34 f3
[ 1408.104856] gnome-system-mo[5360]: segfault at 30 ip 00007c02762ec29e sp 00007ffc399829d0 error 6 in libgtk-4.so.1.1400.5[4ec29e,7c0275e81000+4b5000] likely on CPU 1 (core 1, socket 0)
[ 1408.104871] Code: 48 8d 35 c5 6f 62 00 45 31 c0 e8 fd 0f 01 00 48 8b 45 b0 48 8b 5d c0 4c 89 ff f3 41 0f 10 04 24 f3 0f 58 00 48 89 da 48 89 c6 <f3> 0f 11 43 30 f3 41 0f 10 44 24 04 f3 0f 58 40 04 f3 0f 11 43 34
[ 1434.535310] blackbox[5476]: segfault at 30 ip 00007d7ba0f6491c sp 00007ffcc9c2bfb0 error 6 in libgtk-4.so.1.1400.4[56491c,7d7ba0ac7000+4c2000] likely on CPU 4 (core 0, socket 0)
[ 1434.535318] Code: c2 4c 8d 4d c0 48 8d 35 02 75 4a 00 45 31 c0 0f b6 d2 e8 c7 4c 01 00 f3 0f 10 03 48 8b 55 c0 4c 89 fe f3 41 0f 58 07 4c 89 f7 <f3> 0f 11 42 30 f3 0f 10 43 04 f3 41 0f 58 47 04 f3 0f 11 42 34 f3
[ 1519.054496] tidal-hifi[5241]: segfault at 0 ip 0000776c9650bccc sp 00007ffce741f1a0 error 6 in libnvidia-glcore.so.560.35.03[b0bccc,776c95e00000+c00000] likely on CPU 1 (core 1, socket 0)
[ 1519.054505] Code: 41 0f 7e ce 89 fd 48 8b 80 80 00 00 00 c1 e2 12 66 41 0f 7e d5 48 b9 72 0e 05 a0 04 00 00 00 81 ca 00 0e 00 80 66 41 0f 7e dc <89> 10 48 83 c0 1c 48 89 48 e8 66 0f 7e 40 f0 66 0f 7e 48 f4 66 0f
After a restart, all apps run perfectly fine of course and I can do whatever I need to do, until they stop working again.
-
On one forum, someone was suggesting a problem with memory, so I memtested it hard — everything is fine
-
On another, the idea was to reinstall
intel-ucodepackage, this unfortunately didn't help neither -
Finally, I've tried
sudo pacman -Qnq | sudo pacman -S -because it was a suggestion somewhere too, no changes after that
My boat:
Arch Linux x86_64 / 6.10.9-arch1-2
Gnome 46.4 / Mutter (X11)
@
ASUS Z97-PRO (Wi-Fi ac)
INTEL i7-4790 + NVIDIA GTX 1070 Ti
Big, fat thanks for your time!
Update: Idea from jrgd did the trick! I am still testing it, but so far, I can't force anything to fail and everything seems to be working fine!
Well, I don't think that there's going to be a way to narrow it down just from that. Can maybe suggest some things to try.
Segfaults happen when an app is trying to write to memory that it shouldn't be. It tends to indicate either a bug (which shouldn't be the case at the application level, given multiple applications doing it), or corrupt memory.
Well, it's not the recent 13th+14th gen Intel hardware problems; I had two processors produce corrupt memory via that route recently. That's an ten-year-old processor, though.
Memory would be an obvious thing to blame, but memtest not hitting it is an argument against that.
Proton-related stuff -- Windows games running under Steam -- might use a fair bit of memory. Do you have swap space, maybe?
I can imagine a swap device corrupting memory. If you're using any swap, I'd probably do a fresh boot, use
sudo swapoff -ato disable swap space, and then try your repro case with proton.It's possible that there's some kind of kernel bug relating to memory allocation that you're tripping, I guess, but you're running an up-to-date Arch, so I assume that you've got a recent kernel, and you're saying that this has been going on for "some time". That would be consistent with problems continuing to happen once the problem occurs on a given boot and a reset making things go away, I suppose.
I guess it could hypothetically be a problem with the Nvidia drivers...if the Proton games you're playing are 3d games, that might be tripping it.
A game might also be stressing the CPU, triggering temperature or other issues. The
stressutility would let you generate sustained CPU load on cores (--cpu), maybe see if that reproduces your problem.Proton games might be loading the GPU...I don't know of a good way to artificially generate load there, unfortunately.
You could try checking the kernel log for any errors preceding the segfaults in the kernel log, maybe when you're running that Proton game; any issues there might give a clue.
Okay, tested this just now, disabling swap makes no difference but thanks for a new command for the future!
Which memtests have you tried? They all function a little differently, and passing one doesn't mean it will pass another. My rig passed OCCT and TM5 with flying colors, but it would fail every time on prime95 (until I eventually got it stable).
I just went for the memtest86 and let it do its thing.. Should I do some more testing? I really feel like this is not about memory, at least not in a “physically damaged” way.
It might not be a memory thing. If you run out of options and are down to trying memory again, take a look at the MemTestHelper test recommendations. You shouldn't have to run tests for more than ≈0.5–1.5 hours at a time (the 8hr+ testing regimen is pointless).
https://github.com/integralfx/MemTestHelper/blob/oc-guide/DDR4%20OC%20Guide.md#memory-testing-software