Pi 5: V3D hang with 'MMU error from client PSE ... pte invalid' in Chromium (cage kiosk) on 6.18.50

未关闭
  • 哪里不对？
#7,721 2 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
35/100
Issue 类型
缺陷
描述清晰度
基本清楚
活跃度
活跃
技术栈
c, linux
领域
computer-graphics, operating-systems

调研方向

Start by reading the V3D hang and MMU error logs in the report, then compare the earlier fixes in #7400 and #7503 with the prior reports #7397 and #7381. The issue does not identify a source file or provide an on-demand reproduction, so investigation will likely require tracing V3D kernel driver state and may need the suggested GPU tests or state dump. Done means identifying and fixing the cause of the intermittent hang, then confirming the GPU no longer resets under the reported workload.

由索引模型根据 Issue 内容生成。

描述

Describe the bug

On a Pi 5 running Chromium in a cage kiosk session, the V3D GPU hangs while zooming out on a map page. The GPU is reset 6 times in about 6 s, with one MMU error from client PSE ... pte invalid. Chromium's GPU process crashes three times, falls back to --use-gl=disabled, and the screen stays black until the kiosk session is restarted.

The page is Freeboard-SK, an OpenLayers chart plotter. A radar overlay that renders with WebGL2 was off at the time, so the GPU load was Chromium's normal compositing and rasterization of a map while zooming out.

The kernel already includes #7400 and #7503, which fixed the earlier pte invalid hangs in #7397 and #7381.

Steps to reproduce the behaviour
  1. Pi 5, Raspberry Pi OS Lite (trixie) arm64, with cage -s -d -- chromium --kiosk --ozone-platform=wayland <url> on HDMI-A-2 (1280x800).
  2. Open a map page (Freeboard-SK / OpenLayers) and zoom out repeatedly with the touchscreen.
  3. After several hours of uptime it hung once, on a zoom-out. I can't reproduce it on demand yet.
Device (s)

Raspberry Pi 5

System
  • Raspberry Pi 5 Model B Rev 1.0, 8 GB (revision d04170)
  • Kernel 6.18.50+rpt-rpi-2712 #1 SMP PREEMPT Debian 1:6.18.50-1+rpt1 (2026-09-11)
  • Firmware 1accd665, bootloader EEPROM 2026-09-25 (up to date), raspi-firmware 1:1.20260915-1
  • Mesa 26.2.2-1~bpo13+0~rpt1, chromium 1:155.0.8059.79-1~deb13u1+rpt1, rpi-chromium-mods 20260211
  • config.txt: stock (dtoverlay=vc4-kms-v3d, max_framebuffers=2, disable_fw_kms_setup=1, arm_boost=1)
  • vcgencmd get_throttled = 0xe0000 (no under-voltage); SoC at about 65 °C at the time of the hang
Logs
[Tue Oct 27 14:08:45 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* Resetting GPU for hang.
[Tue Oct 27 14:08:45 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* V3D_ERR_STAT: 0x00000000
[Tue Oct 27 14:08:45 2026] v3d 1002000000.v3d: MMU error from client PSE (57) at 0x1b100320, pte invalid
[Tue Oct 27 14:08:46 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* Resetting GPU for hang.
[Tue Oct 27 14:08:46 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* V3D_ERR_STAT: 0x00001000
[Tue Oct 27 14:08:46 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* Resetting GPU for hang.
[Tue Oct 27 14:08:46 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* V3D_ERR_STAT: 0x00001000
[Tue Oct 27 14:08:47 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* Resetting GPU for hang.
[Tue Oct 27 14:08:47 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* V3D_ERR_STAT: 0x00001000
[Tue Oct 27 14:08:50 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* Resetting GPU for hang.
[Tue Oct 27 14:08:50 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* V3D_ERR_STAT: 0x00001000
[Tue Oct 27 14:08:51 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* Resetting GPU for hang.
[Tue Oct 27 14:08:51 2026] v3d 1002000000.v3d: [drm:v3d_reset [v3d]] *ERROR* V3D_ERR_STAT: 0x00001000

Chromium GPU process afterwards: --gpu-recent-crash-count=3 ... --render-node-override=/dev/dri/card1 --use-angle=gles ... --use-gl=disabled

Additional context
  • Earlier in the same boot, starting Chromium with --use-angle=gles-egl under cage produced 16 similar resets within a minute (09:14–09:15). That flag was then removed, and the GPU ran normally for about five hours before the hang above, with stock flags.
  • The two previous boots had no GPU hangs. After a reboot there have been none so far.
  • The same software stack on another Pi doesn't show this, so a board-specific cause (as discussed in #7657) is possible. I'm happy to run a GPU test (e.g. piglit / deqp) or collect a V3D state dump if that helps.
主要语言
C
星标
13.3k
派生
5.5k
平均合并
1 天 7 小时
30 天内合并 PR
16

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

raspberrypi/linux 的其他 Issue

查看 raspberrypi/linux 的全部 Issue

相似的 Issue

更多 C Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。