Re: [chromium-dev] chrome tainted:G O, level1 translation fault in arm64

65 views
Skip to first unread message

Robert Kroeger

unread,
May 23, 2016, 1:42:34 PM5/23/16
to binoy...@gmail.com, ozon...@chromium.org
[-chromium-dev, +ozone-dev]



On Monday, May 23, 2016, binoy mv (melatt vythakkatt) <binoy...@gmail.com> wrote:
Hello all

I am trying to run chromium in an arm64 based processor, with Wayland graphics back end using ozone Wayland and chromium 45.

so.. ozone/wayland is not a highly supported platform yet. And it's not really done. So you're in mostly uncharted waters here. 

That being said: I'd expect that you're seeing a sandbox issue here. See if it runs better with the sandbox turned off.

That your kernel crashes is however a Linux bug external to Chrome. Chrome should perhaps crash. Or revert to software rendering if my surmise is correct. I'd try running with the sandbox deactivated. And make sure that the drm drivers are fresh and known workable on your device.

Rob.
 

Chrome Content shell I am able to build and run hardware accelerated output (tested  webgl) but when trying to build and run chromium observing a kernel crash with a level 1 translation fault followed by chrome tainted (G O), this happen when gpu process trying to load gpu drivers and initialize..and later when an egl call Is made its failing at that point..

Using same code and setting I built both chrome and content shell..only chrome fails here..any suggestion is greatly appreciated




--
--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
    http://groups.google.com/a/chromium.org/group/chromium-dev
---
You received this message because you are subscribed to the Google Groups "Chromium-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-dev...@chromium.org.

binoy mv

unread,
May 24, 2016, 4:00:58 AM5/24/16
to Robert Kroeger, ozon...@chromium.org
Thanks Rob,

Agree ozone Wayland is getting improved, but thanks to the team worked on this, i could run hardware accelerated hTML5 with this (in content shell), 

regarding the issue, indeed i run both shell and chrome without sand box only (--no-sandbox) otherwise getting error as below (even thought the SUID sand boxing procedure followed)
[1:1:0101/005048:3048054588:FATAL:setuid_sandbox_client.cc(126)] Check failed: IsFileSystemAccessDenied().

i assume the drivers are intact because i could run other EGL wayland clients and as i mentioned the content shell is running fine here. 

Indeed your assumption is correct, chrome gpu process crash, and exits as ozone wayland does not support software rendering mode,
root@salvator-x:/usr/bin# google-chrome --no-sandbox
[2798:2798:0101/011212:ERROR:browser_main_loop.cc(189)] Running without the SUID sandbox! See https://code.google.com/p/chromium/wiki/LinuxSUIDSandboxDevelopment for more information on developing with the sandbox on.
[2831:2831:0101/011212:ERROR:renderer_main.cc(200)] Running without renderer sandbox
[2798:2798:0101/011213:FATAL:display.cc(296)] The browser process has attempted to start the GPU process in software rendering mode. Software rendering is not supported in Ozone-Wayland, so this is fatal. Usually this error occurs because the GPU process crashed in hardware rendering mode, often due to failure to initialize EGL. To debug the GPU process, start Chrome with --gpu-startup-dialog so that the GPU process pauses on startup, then attach to it with 'gdb -p' and run the command 'signal SIGUSR1' in order to unpause it. If you have xterm then it is easier to run 'chrome --no-sandbox --gpu-launcher='xterm -title renderer -e gdb --eval-command=run --args''
/usr/bin/google-chrome: line 12:  2798 Aborted                 /usr/bin/chromium/chrome ${CHROME_EXTRA_ARGS} $@

and there is kernel failure happen exactly when the GPU process trying to load the GPU drivers (EGL/ GLES etc.), 
root@salvator-x:/usr/bin/chromium# [ 4623.065823] chrome[2908]: unhandled level 1 translation fault (11) at 0xffffff9a58c160, esr 0x92000005
[ 4623.075193] pgd = ffffffc4f510b000
[ 4623.078590] [ffffff9a58c160] *pgd=000000053d243003, *pud=000000053d243003, *pmd=0000000000000000
[ 4623.087563] 
[ 4623.089086] CPU: 3 PID: 2908 Comm: chrome Tainted: G           O    4.4.0-yocto-standard #1
[ 4623.097511] Hardware name: Renesas Salvator-X board based on r8a7795 (DT)
[ 4623.104327] task: ffffffc5bb866c00 ti: ffffffc5b82cc000 task.ti: ffffffc5b82cc000
[ 4623.111834] PC is at 0x7fa37ed78c
[ 4623.115146] LR is at 0x7fa37ed7f0
[ 4623.118481] pc : [<0000007fa37ed78c>] lr : [<0000007fa37ed7f0>] pstate: 60000000
[ 4623.125892] sp : 0000007ffc1bdc30
[ 4623.129224] x29: 0000007ffc1bdc30 x28: 00000055949f8000 
[ 4623.134563] x27: 00000001138e2c3c x26: 00000055948ec000 
[ 4623.139911] x25: 0000007ffc1bdcf0 x24: 00000055949f8000 

later when a EGLgetDisplay call made by chrome (gl_surface_egl.cc), EGL library fails to get the right reference for the window handle and crash, surprisingly it does not fail for content shell. 


best regards,
Binoy
Reply all
Reply to author
Forward
0 new messages