System crash terminating wl_display_dispatch loop on Mali devices
Reported by
blu.d...@gmail.com,
Oct 5
|
||
Issue descriptionUserAgent: Mozilla/5.0 (X11; CrOS aarch64 11021.34.0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/70.0.3538.41 Safari/537.36 Platform: 11021.34.0 (Official Build) beta-channel bob Steps to reproduce the problem: 1. In a stand-alone app, set up a surfaceless EGL context using dma linux buffers for a double-buffered framebuffer, and pass those to wayland 2. Run a frame loop for an arbitrary number of frames 3. Terminate the loop and disconnect from wayland display What is the expected behavior? Clean exit of the app What went wrong? System crash, followed by a system reboot. Did this work before? N/A Chrome version: 70.0.3538.41 Channel: beta OS Version: 11021.34.0 Flash Version: See example code here: https://github.com/blu/hello-chromeos-gles2/blob/master/main_chromeos.cpp
,
Oct 16
Clarification: crash happens on RK3399 with Mali-T860.
,
Oct 19
dcastagna: can you investigate or triage to somebody else?
,
Oct 20
blu.dark@gmail.com: Can you please include logs? How do you run a binary that connects to exo? Is your chromebook in developer mode?
,
Oct 20
dcastagna: Yes, chromebook is in developer mode. Binary utilizes the DRM render-nodes support in ChromeOS and uses the following ChromeOS-supplied DSOs /usr/lib/libdrm.so /usr/lib/libEGL.so /usr/lib/libGLESv2.so to establish a surfaceless EGL context using dmabuf-based EGLImages which I then pass to wayland via the linux-dmabuf-clientprotocol for on-screen compositing. I use libwayland-client, ver 1.14.91, originally built from sources in the Chromebrew repo: https://github.com/skycocker/chromebrew. The full frameloop mechanism can be seen in the source code linked in the OP. It works as intended across all aarch64 and x86-64 chromebooks I've tried so far, save for RK3399-based Mali devices. I can provide a minimal repro code, if need be. Re logs -- what logs would you like me to include? Please, consider that the system crash leaves the filesystem unflushed, so in order to see a log on next reboot a deliberate sync of the fs will be needed at moments of interest, and also those logs would need not be reset at reboot. |
||
►
Sign in to add a comment |
||
Comment 1 by dtapu...@chromium.org
, Oct 16