Dynamic linker is segfaulting

Engdyn · Yesterday 00:38:47

I updated my system a few hours ago and wanted to play something in Retroarch. I'm starting the core and it instantly segfaults.

Jul 29 00:42:44 EVA-01 kernel: retroarch[21510]: segfault at 0 ip 00007f07dab71689 sp 00007ffd0d026ec8 error 6 in libc.so.6[171689,7f07daa24000+175000] likely on CPU 8 (core 0, socket 0)
Jul 29 00:42:44 EVA-01 kernel: Code: 20 c5 fe 7f 07 c5 fe 7f 44 17 e0 c5 f8 77 c3 66 90 c5 fe 7f 47 c0 c5 fe 7f 47 e0 c5 f8 77 c3 66 90 48 3b 15 91 0b 0a 00 77 77 <c5> fe 7f 07 c5 fe 7f 47 20 48 01 d7 48 81 fa 80 00 00 00 76 d2 c5
Jul 29 00:42:44 EVA-01 systemd-coredump[21556]: Process 21510 (retroarch) of user 1000 terminated abnormally with signal 11/SEGV, processing...
Jul 29 00:42:44 EVA-01 systemd[1]: Started Process Core Dump (PID 21556/UID 0).
Jul 29 00:42:44 EVA-01 systemd[1]: Started Pass systemd-coredump journal entries to relevant user for potential DrKonqi handling.
Jul 29 00:42:45 EVA-01 systemd-coredump[21557]: [?] Process 21510 (retroarch) of user 1000 dumped core.

The first thing I always do is to compile the package in a chroot and test it. I'm going into my source directory and run "pkgctl build". Everything seems to work fine until KDE Plasma notifies me that the dynamic linker segfaulted too.

Jul 29 01:15:54 EVA-01 sudo[25729]:     engdyn : TTY=pts/1 ; PWD=/home/engdyn/.cache/paru/clone/libretro-wasm4-git ; USER=root ; COMMAND=/usr/bin/extra-x86_64-build -- -l engdyn-1 --
Jul 29 01:15:54 EVA-01 sudo[25729]: pam_unix(sudo:session): session opened for user root(uid=0) by engdyn(uid=1000)
Jul 29 01:15:56 EVA-01 systemd[1]: Started Container arch-nspawn-25775.
Jul 29 01:15:57 EVA-01 systemd[1]: arch-nspawn-25775.scope: Deactivated successfully.
Jul 29 01:15:57 EVA-01 systemd[1]: run-systemd-nspawn-arch\x2dnspawn\x2d25775-unix\x2dexport.mount: Deactivated successfully.
Jul 29 01:15:57 EVA-01 sudo[25884]:     root : TTY=pts/3 ; PWD=/home/engdyn/.cache/paru/clone/libretro-wasm4-git ; USER=engdyn ; COMMAND=/usr/bin/env SRCDEST=/home/engdyn/.cache/paru/clone/libretro-wasm4-git BUILDDIR=/tmp/makechrootpkg.kupmx39EL1 makepkg --config=/var/lib/archbuild/extra-x86_64/engdyn-1/etc/makepkg.conf --verifysource -o
Jul 29 01:15:57 EVA-01 sudo[25884]: pam_unix(sudo:session): session opened for user engdyn(uid=1000) by engdyn(uid=0)
Jul 29 01:15:59 EVA-01 sudo[25884]: pam_unix(sudo:session): session closed for user engdyn
Jul 29 01:15:59 EVA-01 sudo[27348]:     root : TTY=pts/3 ; PWD=/home/engdyn/.cache/paru/clone/libretro-wasm4-git ; USER=engdyn ; COMMAND=/usr/bin/bash -c '#012#011#011source PKGBUILD#012#011#011printf "%s\\n" "${pkgbase:-${pkgname}}" "${pkgname[@]}"#012'
Jul 29 01:15:59 EVA-01 sudo[27348]: pam_unix(sudo:session): session opened for user engdyn(uid=1000) by engdyn(uid=0)
Jul 29 01:15:59 EVA-01 sudo[27348]: pam_unix(sudo:session): session closed for user engdyn
Jul 29 01:16:00 EVA-01 systemd[1]: Started Container makechrootpkg-libretro-wasm4-git.build.25752.
Jul 29 01:16:08 EVA-01 kernel: ld-linux-x86-64[31150]: segfault at b8 ip 00007f404f8afb1b sp 00007ffecb349128 error 4 in libc.so.6[afb1b,7f404f824000+178000] likely on CPU 1 (core 1, socket 0)
Jul 29 01:16:08 EVA-01 kernel: Code: 85 e4 74 e7 4c 89 e2 4c 89 ef ff 15 07 64 16 00 43 c6 44 25 ff 00 eb d3 0f 1f 80 00 00 00 00 f3 0f 1e fa 48 8b 05 05 63 16 00 <8b> 90 b8 00 00 00 89 d1 f7 d1 81 e1 28 01 00 00 74 0b 48 8d 05 1c
Jul 29 01:16:08 EVA-01 systemd-coredump[31151]: Process 31150 (ld-linux-x86-64) of user 1000 terminated abnormally with signal 11/SEGV, processing...
Jul 29 01:16:08 EVA-01 systemd[1]: Started Process Core Dump (PID 31151/UID 0).
Jul 29 01:16:09 EVA-01 systemd[1]: Started Pass systemd-coredump journal entries to relevant user for potential DrKonqi handling.
Jul 29 01:16:09 EVA-01 systemd-coredump[31152]: Resource limits disable core dumping for process 31150 (ld-linux-x86-64).
Jul 29 01:16:09 EVA-01 systemd-coredump[31152]: [?] Process 31150 (ld-linux-x86-64) of user 1000 terminated abnormally without generating a coredump.
Jul 29 01:16:09 EVA-01 systemd[1]: systemd-coredump@3-20481-31151_45695-0.service: Deactivated successfully.
Jul 29 01:16:09 EVA-01 drkonqi-coredump-processor[31153]: Entry doesn't look like a dump. This may have been a vaccum run. Nothing to process.
Jul 29 01:16:09 EVA-01 drkonqi-coredump-processor[31153]: "/usr/lib/ld-linux-x86-64.so.2" 31150 ""
Jul 29 01:16:09 EVA-01 systemd[1]: drkonqi-coredump-processor@3-20481-31151_45695-0.service: Deactivated successfully.
Jul 29 01:16:09 EVA-01 systemd[1071]: Started Launch DrKonqi for a systemd-coredump crash (PID 31153/UID 0).
Jul 29 01:16:09 EVA-01 systemd[1]: makechrootpkg-libretro-wasm4-git.build.25752.scope: Deactivated successfully.
Jul 29 01:16:09 EVA-01 systemd[1]: makechrootpkg-libretro-wasm4-git.build.25752.scope: Consumed 10.735s CPU time over 8.260s wall clock time, 709.3M memory peak.
Jul 29 01:16:09 EVA-01 drkonqi-coredump-launcher[31160]: Unable to find file for pid 31150 expected at "kcrash-metadata/ld-linux-x86-64.so.2.9086bc88c4af46bdb583419a17b0a8d9.31150.ini"
Jul 29 01:16:09 EVA-01 drkonqi-coredump-launcher[31160]: KCrash metadata not marked complete. Please file a bug at bugs.kde.org giving as much detail about the crash as possible and maybe include the file ""
Jul 29 01:16:09 EVA-01 systemd[1]: run-systemd-nspawn-makechrootpkg\x2dlibretro\x2dwasm4\x2dgit.build.25752-unix\x2dexport.mount: Deactivated successfully.
Jul 29 01:16:09 EVA-01 sudo[25729]: pam_unix(sudo:session): session closed for user root

I rebooted into a snapshot and restored it. I tried to recompile it outside the chroot and had the same result. I tried running my wasm4 core again (the one before I recompiled it) and it works fine. I checked paclog and I compiled it on the 10th of July with gcc 16.1.1+r346+g4e03491b401d-1. So this seems to be the last working version.
I wrote the PKGBUILD myself and it's in the AUR. I also tried running "pkgctl build" with other PKGBUILDs from completely different projects and every time ld-linux-x86-64 segfaults. I'm using the default kernel.

I honestly have no idea what I should try or do. The moment gcc or (dynamic) linker breaks my brain turns off. Any help to isolate the problem and/or fix it would be much appreciated.

seth · Yesterday 06:44:14

glibc updated to 2.44 on July 25th and gcc was rebuild for that - assuming you're not holding back any versions, does moving back in time to before that fix things?
https://wiki.archlinux.org/title/Arch_L … cific_date
libretro/retroarch might require a rebuild, no idea whether building wasm4 using gcc15 would (still) work

Lone_Wolf · Yesterday 09:31:48

On july 25 there was a toolchain rebuild for glibc 2.44 , on july 28 there was another toolchain rebuild for binutils 2.47 .

You need to check atleast 2 dates : one before july 25 and one between july 25 and july 28 .

Engdyn · Today 09:36:10

I tried some different dates and it didn't help. I investigated a bit more and it's actually 2 different problems that I coincidentally encountered at the same time.

The first problem was an update to a submodule of the wasm4 core that caused the segfault. I submitted an upstream PR that got merged and it's working fine again.

The other problem is isolated to "pkgctl". I setup a VM with a minimal arch install to make sure it's not my install and it's happening there too. The dynamic linker segaults when building a package with "pkgctl build", but only the debug part.
If we take libretro-wasm4-git as an example, if you use "makepkg" you'll get, without any issues, libretro-wasm4-git-version.tar.zst and libretro-wasm4-git-debug-version.tar.zst. If you use "pkgctl build" instead, you'll get libretro-wasm4-git-version.tar.zst but not libretro-wasm4-git-debug-version.tar.zst because the dynamic linker segfaults while trying to build it.

I'm unsure where I should report it. The dynamic linker is part of gcc, but it only segfaults when using "pkgctl". Should I report this to the gcc or devtools package on the arch gitlab?

Arch Linux

#1 Yesterday 00:38:47

Dynamic linker is segfaulting

#2 Yesterday 06:44:14

Re: Dynamic linker is segfaulting

#3 Yesterday 09:31:48

Re: Dynamic linker is segfaulting

#4 Today 09:36:10

Re: Dynamic linker is segfaulting

Board footer