You are not logged in.
I was trying to install gnome for a friend on his new machine. Install looks fine but both GDM and Gnome fail immediately with the "Oh no! Something has gone wrong" screen. Nothing in the Xorg log, but this appears in the dmesg output:
[ 421.162474] traps: gnome-session-c[934] general protection ip:7f4662953080 sp:7ffdbb0a1328 error:0 in libpthread-2.21.so[7f4662941000+18000]
[ 421.191727] traps: gnome-session-c[946] general protection ip:7f724768a080 sp:7fffef81f718 error:0 in libpthread-2.21.so[7f7247678000+18000]
I searched around but didn't find much. Any ideas?
Thanks in advance.
#binarii @ irc.binarii.net
Matrix Server: https://matrix.binarii.net
-------------
Allan -> ArchBang is not supported because it is stupid.
Offline
Hey,
I seem to habe the exact same issue, but only with the nvidia-libgl package, using mesa-libgl I can start and work with GNOME.
Can you post, you post your configuration?
Offline
Hej,
Same issue on my machine with latest nvidia drivers. Only workaround I found so far is to use the legacy drivers (nvidia-340xx-*).
Regards, Tom
Offline
If you guys are on Intel processors (especially Haswell), make sure you are updating your CPU microcode.
Online
I am getting this error on intel skylake i5-6600k. I have ensured that I have the latest cpu microcode and added the cpu microcode updater to my boot config. I tested this with linux 4.1.6 from the core repo and linux 4.2-3 from the testing repo. I also tested the nvidia 355 beta drivers from the testing repo. All of these give the OP error.
Right now the best way around the error seems to be tokudd's solution, install the nvidia legacy drivers nvidia-340xx
Sole Construct
Offline
Has anyone figured out a solution for this issue that doesn't involve rolling back to nvidia-340xx? I have a GTX 970 so that's not an option. Is the lack of an update to Intel's microcode the culprit here?
Offline
Same here with a i7-6700K. As the microcode firmware was last updated in January 2015 it did not contain any Skylake microcode. On my ASUS mainboard, I installed the latest BIOS but that did not help either.
Still getting many "traps: .... general protection ip: .... sp: .... error:0 in libpthread-2.22.so"
Everything working though so far, except gvim.
Hopefully Intel releases new Microcode for Skylake soon, I also tried using a 4.2.x kernel but that did not change anything either.
See also the related kernel source: http://lxr.free-electrons.com/source/ar … aps.c#L486
And: https://en.wikipedia.org/wiki/General_protection_fault
Update:
Hm, interesting. Rolling back to nvidia-340xx fixed the issue for me, so is it not related to the Intel microcode at all and only a Nvidia driver bug?
Last edited by clemens (2015-09-24 12:22:26)
Offline
Well, I'm working around the issue for now by installing nvidia-346xx-dkms from the AUR https://aur.archlinux.org/packages/nvidia-346xx-dkms/ since that has support for GTX 970 and doesn't have the issue that 352xx and 355xx have.
Offline
Well, I'm working around the issue for now by installing nvidia-346xx-dkms from the AUR https://aur.archlinux.org/packages/nvidia-346xx-dkms/ since that has support for GTX 970 and doesn't have the issue that 352xx and 355xx have.
I can confirm, this works for me with i7-6700k and a GTX 970.
A better fix, apparantly this is TSX related, compiling glibc without `--enable-lock-elision` seems to fix this issue (just grab it from ABS and edit the PKGBUILD). After installing the self-compiled glibc you can reinstall the latest nvidia drivers.
Thanks to BlaXpirit from IRC for this.
Last edited by D4v1d (2015-10-16 16:31:00)
Offline
A better fix, apparantly this is TSX related, compiling glibc without `--enable-lock-elision` seems to fix this issue (just grab it from ABS and edit the PKGBUILD). After installing the self-compiled glibc you can reinstall the latest nvidia drivers.
Thanks to BlaXpirit from IRC for this.
This is exactly what I was looking for. I fought with very similar symptoms over the weekend (segfaults all over the place due to libpthread and nvidia-utils) setting up Arch with my new Core i5-6500 and GTX 970. I dropped back to mesa-libgl to get Gnome running, but I confirmed last night that recompiling glibc without --enable-lock-elision allows me to run Gnome with the actual Nvidia drivers. \o/ Thanks so much for posting this!
Offline
Also a i5-6600k owner here, suffering from the same problem. I can't seem to apply any of the suggested workarounds: glibc compilation fails with
a - elf/dl-libc.os
a - elf/dl-sym.os
a - elf/dl-tsd.os
a - elf/dl-vdso.os
gcc -Wl,-O1,--sort-common,--as-needed,-z,relro -nostdlib -nostartfiles -r -o /tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/librtld.map.o '-Wl,-(' /tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/dl-allobjs.os /tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/libc_pic.a -lgcc '-Wl,-)' -Wl,-Map,/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/librtld.mapT
/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/libc_pic.a(init-first.os):(.data+0x0): definiciones múltiples de `__libc_multiple_libcs'
/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/dl-allobjs.os:(.bss+0xf8): primero se definió aquí
/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/libc_pic.a(libc_fatal.os): En la función `__GI___libc_fatal':
libc_fatal.c:(.text+0x310): definiciones múltiples de `__GI___libc_fatal'
/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/dl-allobjs.os:(.text+0x16c50): primero se definió aquí
/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/libc_pic.a(libc_fatal.os): En la función `__GI___libc_fatal':
libc_fatal.c:(.text+0x310): definiciones múltiples de `__libc_fatal'
/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/dl-allobjs.os:(.text+0x16c50): primero se definió aquí
/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/libc_pic.a(dl-addr.os): En la función `_dl_addr_inside_object':
dl-addr.c:(.text+0x2e0): definiciones múltiples de `_dl_addr_inside_object'
/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/dl-allobjs.os:(.text+0x13490): primero se definió aquí
collect2: error: ld devolvió el estado de salida 1
Makefile:323: fallo en las instrucciones para el objetivo '/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/librtld.map'
make[2]: *** [/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-build/elf/librtld.map] Error 1
make[2]: se sale del directorio '/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-2.22/elf'
Makefile:213: fallo en las instrucciones para el objetivo 'elf/subdir_lib'
make[1]: *** [elf/subdir_lib] Error 2
make[1]: se sale del directorio '/tmp/yaourt-tmp-nobill/abs-glibc/src/glibc-2.22'
Makefile:9: fallo en las instrucciones para el objetivo 'all'
make: *** [all] Error 2
Going to older nvidia drivers doesn't help either, GDM will show the: "Oh no! something has gone wrong" message. I've also tried using nouveau (only option that seems to work) but I get segfaults pretty often when trying to browse or play videos.
Recompiling glibc seems the best option, but I can't get to that point, any ideas?
Offline
This is not an issue with Intel TSX. It appears to be a programming error in /usr/lib/libEGL_nvidia.so.0. The crash occurs when a program erroneously tries to unlock a free lock. For max performance, glibc does not check for this case.
I created a patch to add the check to glibc:
View: http://gbpaste.org/zYn3q
Download: http://gbpaste.org/download/zYn3q/l
Replacement glibc-2.22 PKGBUILD with patch: http://gbpaste.org/gcv01
With this patch you can use the latest nvidia driver without disabling lock-ellision.
Last edited by LordReg (2015-10-24 01:10:27)
Offline
@LordReg I attempted to compile glibc with your patch and I get the exact same results. glibc fails to compile with the same error I posted previously.
Any ideas on that compile error?
Offline
Nevermind, its compiling now . Had been a while since I used makepkg and my /etc/makepkg.conf was pretty outdated.
Offline
@LordReg I attempted to compile glibc with your patch and I get the exact same results. glibc fails to compile with the same error I posted previously.
Any ideas on that compile error?
Are you trying to use yaourt?
Try following the steps here: https://wiki.archlinux.org/index.php/Ar … to_use_ABS
On step 3/4, copy /var/abs/core/glibc and replace the PKGBUILD with the one I created.
Offline
same problem. Nvidia drivers. Gnome 3
Offline
@LordReg: The patch is nice. You should probably bump the package version number though
Offline
@LordReg: The patch is nice. You should probably bump the package version number though
Thanks, glad it works for you too.
I feel it is better to keep the original version/rel when modifying official packages. A cleaner way to preserve the package: https://wiki.archlinux.org/index.php/Ar … d_packages
I updated the PKGBUILD on my previous post with a 'modified' group. You can repack the package without recompiling by running
$ makepkg -Rf
Offline
Thanks for the patch, its working perfectly here too. Do you have any upstream bugreport so we can follow this issue?
Offline
I just added a 960 card and I'm still having this problem. I can't access the patch either.
Does anyone have any resolution to this issue ?
Offline