protonscr

steamwebhelper crashes (SIGSEGV in GL Composer thread) on RADV/Navi 23, workaround is -cef-disable-gpu

steamclosed
ValveSoftware/steam-for-linux#13360 · opened 2026-06-29 by MartyTremblay · updated 2026-06-29 · 2 comments · github
MMartyTremblay 2026-06-29 github

Your system information

  • Steam client version (build number or date): 1782533657
  • Distribution: CachyOS (Arch-based) 7.1.2-2, COSMIC DE, Wayland
  • Opted into Steam client beta?: No
  • Have you checked for system updates?: Yes
  • GPU: AMD Radeon RX 6600 (eGPU via Thunderbolt/USB4, Navi 23, RADV) + Intel Iris Xe (Raptor Lake-P, integrated)
  • Kernel drivers: amdgpu (eGPU Thunderbolt 3), i915 (iGPU)

Please describe your issue in as much detail as possible:

After launching steam, the client starts normally and authenticates, but steamwebhelper immediately crashes and respawns in a loop (-startcount=0, 1, 2, ...), with no Steam UI window ever appearing. The Steam client process itself stays alive; only steamwebhelper repeatedly dies and restarts.

coredumpctl confirms a real crash, not a clean exit:

Sun 2026-06-28 20:10:12 PDT  8069 1000 1000 SIGSEGV present  /home/marty/.local/share/Steam/ubuntu12_64/steamwebhelper
TID: 8186 (GL Composer Thr)
Signal: 11 (SEGV) si_code: SEGV_MAPERR
Stack trace of thread 8186:
#0  0x0000000000000000 n/a (n/a + 0x0)

The crash is in the GL Composer Thr thread (CEF's GPU compositor thread for Steam's own UI), faulting at address 0x0, consistent with a null GL function pointer call. The coredump has no symbols available, so I could not get a fully symbolized backtrace identifying the specific call.

cef_log.txt shows no fatal error logged for any of the restarts. Each attempt cleanly starts Chromium, sets up crash reporting, and the log simply stops after the line ERROR:atom_cache.cc(229)] Add STEAM_GAME to kAtomsToCache, with no further output before the process is gone and the next restart begins.

Confirmed workarounds (both stop the crash loop):

  1. LIBGL_ALWAYS_SOFTWARE=1 steam — forces software GL rendering, Steam opens normally.
  2. steam -cef-disable-gpu — disables GPU compositing specifically for Steam's CEF-based UI, Steam opens normally and games are unaffected (games render through a separate process, not through steamwebhelper).

Both workarounds point at GPU-accelerated compositing in Steam's own CEF UI shell as the trigger, specifically when rendering on this AMD RADV/Navi 23 device. The Intel iGPU was not tested as the rendering target for steamwebhelper specifically, since the system is configured to prefer the discrete GPU for Steam via .desktop switcheroo-control hints (PrefersNonDefaultGPU=true); glxinfo confirms DRI_PRIME GPU selection itself works correctly and is not the cause (see "Other things ruled out" below).

Steps for reproducing this issue:

  1. On a system with an AMD RADV/Navi 23 GPU (RX 6600) as the discrete/preferred GPU, launch steam normally (no special flags).
  2. Observe steamwebhelper crash and restart repeatedly (-startcount incrementing), no UI window ever appears.
  3. Confirm via coredumpctl info <pid>: SIGSEGV, SEGV_MAPERR, in thread GL Composer Thr, at address 0x0.
  4. Confirm workaround: steam -cef-disable-gpu launches successfully.

Other things ruled out during troubleshooting:

  • A global DRI_PRIME=1 set via ~/.config/environment.d/ was suspected initially, but removing it entirely (confirmed via systemctl --user show-environment) did not stop the crash loop, the identical startcount sequence and identical coredump signature occurred with DRI_PRIME completely unset.
  • A separate, unrelated GTK2 theme parsing issue (Breeze-Dark theme causing gtk-theme-name parse errors) was found and fixed independently; it caused an earlier, different crash loop with explicit theme-parser errors in stdout. That fix did not resolve this issue.
  • glxinfo confirms GPU selection itself (via DRI_PRIME or PCI bus ID) correctly targets the RX 6600 and returns a working OpenGL context outside of Steam.
Kkisak-valve maintainer 2026-06-29 github

Hello @MartyTremblay, this issue is already being tracked at #9383. Closing as a duplicate.

MMartyTremblay 2026-06-29 github

Thanks for taking a look. I think this may not actually be a duplicate of #9383, and wanted to share the specific evidence before this gets lost in a closed issue.

#9383 describes Steam crashing when DRI_PRIME is set to any non-zero value, with DRI_PRIME=0 (or unset) reported as a working fix.

In my case, I tested that directly: I removed DRI_PRIME from my environment entirely (confirmed via systemctl --user show-environment | grep -i prime returning empty), restarted my session, and launched Steam clean. The crash loop was identical: same -startcount increment pattern (0, 1, 2...), same coredumpctl signature (SIGSEGV, SEGV_MAPERR, thread GL Composer Thr, fault address 0x0) as when DRI_PRIME=1 was set. So the #9383 fix (don't set DRI_PRIME to non-zero) did not resolve this, and the crash isn't dependent on DRI_PRIME having any particular value.

What did fix it was steam -cef-disable-gpu, which disables GPU-accelerated compositing specifically for Steam's own CEF UI shell. That points to the crash being in GPU-accelerated UI compositing on this driver stack (AMD RADV, Navi 23 / RX 6600), not in GPU selection logic the way #9383 describes.

Happy to provide the coredump, cef_log.txt, and full journalctl output if useful for re-opening or filing separately.

Nothing extracted yet.