[PR #55] [CLOSED] Warn when nvidia_drm modeset is off (fixes #53) #68

New issue

Closed

opened 2026-10-15 12:55:20 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-15 12:55:20 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/games-on-whales/unraid-plugin/pull/55
Author: @JBailes
Created: 6/14/2026
Status: ❌ Closed

Base: main ← Head: fix/issue-53-nvidia-egl-modeset

📝 Commits (1)

30f42ee Warn at deploy time when nvidia_drm modeset is off (#53)

📊 Changes

3 files changed (+47 additions, -0 deletions)

View changed files

📝 docs/FAQ.md (+22 -0)
📝 gow.plg (+4 -0)
📝 scripts/deploy.sh (+21 -0)

📄 Description

Issue

#53 — No Video Received from Host on Moonlight (RTX A4500, driver 595.80).

The Wolf log in the report is the classic NVIDIA-Wayland-without-modeset failure:

ERROR smithay::backend::egl::ffi: [EGL] 0x3003 (BAD_ALLOC) eglQueryDevicesEXT: EGL_BAD_ALLOC ...
libEGL warning: egl: failed to create dri2 screen
MESA: error: ZINK: failed to choose pdev
panicked ... Failed to create EGLDisplay: InitFailed(Unknown(0))

EGL can't enumerate the NVIDIA device, falls back to Mesa/ZINK, and the
compositor thread panics, so the video pipeline never starts → no video.

Root cause

nvidia_drm loaded without modeset=1. The settings page already detects this
and warns on the setup/dashboard, but deploy.sh — whose log the user watches
during Install — was silent, and the exact error string wasn't documented
anywhere, so anyone googling eglQueryDevicesEXT EGL_BAD_ALLOC found nothing.

Changes

scripts/deploy.sh: non-fatal warning when an NVIDIA GPU is selected and
/sys/module/nvidia_drm/parameters/modeset is not Y. The driver volume still
builds so the user can apply the modeset fix and reboot without re-running.
docs/FAQ.md: add the exact EGL error signature under the existing modeset
section, plus a note covering the multi-GPU case where the wrong render node
produces the same EGL failure even with modeset on.
gow.plg: changelog.

Notes

Scoped to surfacing/documenting the dominant cause (per the maintainer's call:
deploy-log warning + FAQ, not auto-render-node detection). I couldn't reproduce
on this dev box (no NVIDIA hardware); the warning logic was unit-tested for the
Y / N / missing-file cases.

🤖 Generated with Claude Code

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/games-on-whales/unraid-plugin/pull/55 **Author:** [@JBailes](https://github.com/JBailes) **Created:** 6/14/2026 **Status:** ❌ Closed **Base:** `main` ← **Head:** `fix/issue-53-nvidia-egl-modeset` --- ### 📝 Commits (1) - [`30f42ee`](https://github.com/games-on-whales/unraid-plugin/commit/30f42ee47fda9814f568d8580f2e1ab62eecba50) Warn at deploy time when nvidia_drm modeset is off (#53) ### 📊 Changes **3 files changed** (+47 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `docs/FAQ.md` (+22 -0) 📝 `gow.plg` (+4 -0) 📝 `scripts/deploy.sh` (+21 -0) </details> ### 📄 Description ## Issue [#53](https://github.com/games-on-whales/unraid-plugin/issues/53) — *No Video Received from Host on Moonlight* (RTX A4500, driver 595.80). The Wolf log in the report is the classic NVIDIA-Wayland-without-modeset failure: ```text ERROR smithay::backend::egl::ffi: [EGL] 0x3003 (BAD_ALLOC) eglQueryDevicesEXT: EGL_BAD_ALLOC ... libEGL warning: egl: failed to create dri2 screen MESA: error: ZINK: failed to choose pdev panicked ... Failed to create EGLDisplay: InitFailed(Unknown(0)) ``` EGL can't enumerate the NVIDIA device, falls back to Mesa/ZINK, and the compositor thread panics, so the video pipeline never starts → no video. ## Root cause `nvidia_drm` loaded without `modeset=1`. The settings page already detects this and warns on the setup/dashboard, but **`deploy.sh` — whose log the user watches during Install — was silent**, and the exact error string wasn't documented anywhere, so anyone googling `eglQueryDevicesEXT EGL_BAD_ALLOC` found nothing. ## Changes - `scripts/deploy.sh`: non-fatal warning when an NVIDIA GPU is selected and `/sys/module/nvidia_drm/parameters/modeset` is not `Y`. The driver volume still builds so the user can apply the modeset fix and reboot without re-running. - `docs/FAQ.md`: add the exact EGL error signature under the existing modeset section, plus a note covering the multi-GPU case where the wrong render node produces the same EGL failure even with modeset on. - `gow.plg`: changelog. ## Notes Scoped to surfacing/documenting the dominant cause (per the maintainer's call: deploy-log warning + FAQ, not auto-render-node detection). I couldn't reproduce on this dev box (no NVIDIA hardware); the warning logic was unit-tested for the `Y` / `N` / missing-file cases. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-15 12:55:20 +00:00