protonscr

steamwebhelper crash loop: NOTREACHED in libcef.so (Chromium 126) on NVIDIA RTX 5060 Ti (Blackwell) with nvidia-open driver

steamopen NVIDIA driversWeb ComponentFlathub-provided steam package
ValveSoftware/steam-for-linux#13240 · opened 2026-05-22 by sergarsilla · updated 2026-07-10 · 1 comments · github
Ssergarsilla 2026-05-22 github

Your system information

  • Steam client version (build number or date): 1778281814 (May 8 2026)
  • Distribution (e.g. Ubuntu): CachyOS (Arch-based), kernel 7.0.9-1-cachyos (also tested 6.18.32-1-cachyos-lts)
  • Opted into Steam client beta?: No
  • Have you checked for system updates?: Yes — system is fully updated
  • Steam Logs: See key excerpts below (Flatpak install — logs at ~/.var/app/com.valvesoftware.Steam/.local/share/Steam/logs/)
  • GPU: Nvidia — GeForce RTX 5060 Ti (Blackwell, GB206), driver 595.71.05 (nvidia-open / open-source kernel module)

Additional info:

  • Desktop: Hyprland 0.55.1 (Wayland)
  • Steam installed via: Flatpak (com.valvesoftware.Steam, stable, v1.0.0.85)
  • Also tested with native pacman Steam — identical crash

Please describe your issue in as much detail as possible:

Expected: Steam launches and shows the login window.

Actual: steamwebhelper (the CEF-based UI renderer) enters an infinite crash loop immediately after launch. The Steam UI never appears. Each instance crashes within seconds and Steam restarts it indefinitely (startcount=0, startcount=1, startcount=2... repeating).

The crash is a SIGTRAP (signal 5) caused by a NOTREACHED() assertion failure inside libcef.so (Chromium 126.0.6478.183, bundled with Steam, dated January 2026). The crash happens inside cef_execute_process — before any GPU initialization, windowing, or rendering takes place.

Key log from webhelper-linux.txt:

[ERROR:check.cc(376)] Check failed: false. NOTREACHED log messages are omitted in official builds. Sorry!

This repeats 3–4 times per launch cycle, then Steam restarts the webhelper.

Key log from cef_log.txt:

GPU process launch failed: error_code=1002
GPU process isn't usable. Goodbye.

Stack trace (from coredumpctl info):

Signal: 5 (TRAP)
Executable: steamwebhelper (ubuntu12_64/steamwebhelper)
libcef.so version: Chromium 126.0.6478.183

Stack trace of thread (zygote process):
#0  libcef.so + 0x61f589c
#1  libcef.so + 0x61efd3e
#2  libcef.so + 0x61f0979
#3  libcef.so + 0x61f1715
#4  libcef.so + 0x61ef8d0
#5  libcef.so + 0x293039c
#6  libcef.so + 0x2919107
#7  cef_execute_process (libcef.so + 0x2887b11)

What I have tried (none of these change the crash):

  • Fresh Flatpak reinstall with --delete-data
  • Both LTS (6.18.x) and bleeding-edge (7.0.x) kernels
  • Gamescope wrap (gamescope -W 1920 -H 1080 -- flatpak run com.valvesoftware.Steam)
  • Clearing GPU crash cache (~/.var/app/com.valvesoftware.Steam/.local/share/Steam/config/htmlcache/)
  • --no-sandbox already passed by Steam by default
  • LIBGL_ALWAYS_SOFTWARE=1 env var
  • Removing WAYLAND_DISPLAY to force X11-only mode (made it worse)
  • Native Steam (pacman) — same identical crash in same location

Steps for reproducing this issue:

  1. Install Flatpak Steam on CachyOS (Arch-based) with NVIDIA RTX 5060 Ti (Blackwell) and nvidia-open driver 595.71.05, in a Hyprland/Wayland session
  2. Run flatpak run com.valvesoftware.Steam
  3. Observe: Steam bootstrap completes, then steamwebhelper starts crashing in a loop — Check failed: false. NOTREACHED in libcef.so — UI never appears

The crash appears specific to the NVIDIA Blackwell architecture (RTX 5060 Ti, GB206) with the open-source nvidia-open kernel module. The libcef.so bundled with Steam (Chromium 126, January 2026) triggers a NOTREACHED() assertion in cef_execute_process on this hardware/driver combination.

DDanaGoyette 2026-07-06 github

I'm getting a similar issue with an RTX 4080 and a non-flatpak Steam, and on another machine wth a Radeon Pro W5700 and a non-flatpak Steam.

It might be good for Valve to build their libcef with those NOTREACHED messages enabled, so we can see what expectation wasn't met, rather than having it just shrug and make us guess.

I'll file a new bug for my issues with my machine, because I have NVIDIA and/or PRIME offloading in the mix, unlike in this bug report.

Nothing extracted yet.