Re: Issue 384654 in chromium: Chrome quits repeatedly (Linux, AMD GPU, Ubuntu 14.04)

403 views
Skip to first unread message

chro...@googlecode.com

unread,
Jun 18, 2014, 9:18:05 PM6/18/14
to chromi...@chromium.org
Updates:
Summary: Chrome quits repeatedly (Linux, AMD GPU, Ubuntu 14.04)

Comment #17 on issue 384654 by k...@chromium.org: Chrome quits repeatedly
(Linux, AMD GPU, Ubuntu 14.04)
http://code.google.com/p/chromium/issues/detail?id=384654

(No comment was entered for this change.)

--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

chro...@googlecode.com

unread,
Jun 19, 2014, 10:11:47 PM6/19/14
to chromi...@chromium.org

Comment #19 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
Just to be clear, tab crashing with this error occurs even when only
software rendering is enabled. In this case, no hardware acceleration
should be taking place. How could the watchdog timer cause a tab crash
while software rendering if it's an AMD driver issue?

Log Messages
[5244:5244:0615/195724:ERROR:sandbox_linux.cc(308)] : InitializeSandbox()
called with multiple threads in process gpu-process
[5244:5254:0615/200642:ERROR:gpu_watchdog_thread.cc(253)] : The GPU process
hung. Terminating after 10000 ms.
GpuProcessHostUIShim: The GPU process crashed!
[6148:6148:0615/200654:ERROR:sandbox_linux.cc(308)] : InitializeSandbox()
called with multiple threads in process gpu-process
[6148:6154:0615/200706:ERROR:gpu_watchdog_thread.cc(253)] : The GPU process
hung. Terminating after 10000 ms.
GpuProcessHostUIShim: The GPU process crashed!
[6167:6167:0615/200707:ERROR:sandbox_linux.cc(308)] : InitializeSandbox()
called with multiple threads in process gpu-process
[6167:6170:0615/200719:ERROR:gpu_watchdog_thread.cc(253)] : The GPU process
hung. Terminating after 10000 ms.
GpuProcessHostUIShim: The GPU process crashed!

Attachments:
chromegpu_software.txt 10.1 KB

chro...@googlecode.com

unread,
Jun 20, 2014, 8:25:42 PM6/20/14
to chromi...@chromium.org

Comment #20 on issue 384654 by k...@chromium.org: Chrome quits repeatedly
That looks like a bug in Chrome. How exactly did you disable GPU access?

chro...@googlecode.com

unread,
Jun 20, 2014, 8:28:42 PM6/20/14
to chromi...@chromium.org

Comment #21 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
I didn't ignore-gpu-blacklist. After repeated crashes, it appears to
blacklist the driver automatically and the chrome://gpu always shows
software rendering.

chro...@googlecode.com

unread,
Jun 20, 2014, 9:04:50 PM6/20/14
to chromi...@chromium.org

Comment #22 on issue 384654 by k...@chromium.org: Chrome quits repeatedly
If you are still seeing tab crashes ("Aw, snap") after Chrome switches to
software rendering mode, could you please provide some of the associated
crash IDs from about:crashes?

chro...@googlecode.com

unread,
Jun 20, 2014, 9:07:51 PM6/20/14
to chromi...@chromium.org

Comment #23 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
Here are 50. Take your pick.

Attachments:
chrome_crashes.txt 4.4 KB

chro...@googlecode.com

unread,
Jun 20, 2014, 9:09:51 PM6/20/14
to chromi...@chromium.org

Comment #24 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
I'll just add that some of the earliest may have some hardware rendering
enabled. The most recent dozen, at minimum, should all be with software
rendering (at least according to the chrome://gpu tab).

chro...@googlecode.com

unread,
Jun 20, 2014, 9:47:52 PM6/20/14
to chromi...@chromium.org
Updates:
Cc: pi...@chromium.org jba...@chromium.org

Comment #25 on issue 384654 by k...@chromium.org: Chrome quits repeatedly
I went through the first 20 or so of those crash IDs and they're all from
the GPU process. An "Aw, snap" would be caused by a renderer process
crashing. If you're seeing Aw, Snaps then for some reason they're not
getting uploaded to our crash server.

It's interesting that the crashes come in groups of 3 spaced about 3 hours
apart. That indicates to me that the aging of the GPU process crash count
(Chrome allows 1 per hour) is retrying the GPU accelerated code path every
3 hours; it rapidly crashes 3 times and disables it for another 3 hours.

Please send in some more crash IDs that occur over the next couple of days.
We really need to get a stack trace from the renderer process.

chro...@googlecode.com

unread,
Jun 20, 2014, 9:53:51 PM6/20/14
to chromi...@chromium.org

Comment #26 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
That is odd. I'm fairly certain that the about:gpu tab hasn't indicated any
hardware acceleration for at least a few days...

chro...@googlecode.com

unread,
Jun 20, 2014, 11:09:19 PM6/20/14
to chromi...@chromium.org

Comment #27 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
Ah, I take that back. When "Use hardware acceleration when available" is
checked under Settings > Advanced > System, some hardware acceleration is
being used initially on first load. After a tab crash or two, only
software rendering is enabled here.

chrome://gpu snippet:

Graphics Feature Status
Canvas: Software only, hardware acceleration unavailable
Flash 3D: Hardware accelerated
Flash Stage3D: Unavailable. Hardware acceleration unavailable
Flash Stage3D Baseline profile: Unavailable. Hardware acceleration
unavailable
Compositing: Hardware accelerated and threaded.
Rasterization: Software only, hardware acceleration unavailable
Video Decode: Software only, hardware acceleration unavailable
Video Encode: Hardware accelerated
WebGL: Hardware accelerated

I'll uncheck the hardware acceleration option and see if tab crashes
persist. Otherwise, it may indeed be a driver issue.

chro...@googlecode.com

unread,
Jun 20, 2014, 11:47:32 PM6/20/14
to chromi...@chromium.org

Comment #28 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
The quick answer is yes, I am very much still getting Aw-snap tab crashes
with exclusively software rendering. In fact, once the first one occurs,
numerous others tabs crash in succession (i.e. no isolation). Most tabs
are thereafter incredibly unstable until I exit and restart the entire
browser. No single action consistently causes the initial crash. I see no
more recent Crash ID on chrome://crashes after a bout of Aw-snaps, nor do I
see any dumps in '~/.config/google-chrome/Crash Reports.'

Aside: why does google-chrome-unstable use ~/.config/google-chrome for
crash dumps?

chro...@googlecode.com

unread,
Jun 24, 2014, 9:02:54 PM6/24/14
to chromi...@chromium.org

Comment #29 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
@kbr, I just want to reiterate that these crashes continue, but they are
not producing any crash IDs viewable in chrome.

Incidentally, the very top ranked crash on errors.ubuntu.com arises from
the watchdog thread crashing in chromium.

See https://errors.ubuntu.com/

I don't have access to the bug reports, but it states the crashed function
as:
chromium-browser (11) content → → GpuWatchdogThread → →
DeliberatelyTerminateToRecoverFromHang → Run → base → → MessageLoop → →
RunTask → base → → MessageLoop → → DeferOrRunPendingTask → base → →
MessageLoop → → DoDelayedWork

It appears to have spiked in Trusty 14.04.

There are a couple of unfixed bug reports at Ubuntu on this:

https://bugs.launchpad.net/ubuntu/+source/chromium-browser/+bug/1250174
https://bugs.launchpad.net/ubuntu/+source/chromium-browser/+bug/1229021

https://errors.ubuntu.com/problem/8b463e589278636b308c3a617c73a5477ec458bf

chro...@googlecode.com

unread,
Jun 24, 2014, 9:47:58 PM6/24/14
to chromi...@chromium.org

Comment #30 on issue 384654 by k...@chromium.org: Chrome quits repeatedly
@lelyveld: sorry for the continued trouble and thanks for the pointers. Our
(small) group has still not had the time to install a machine here with the
associated drivers and try to catch the hang or crash in the debugger. I
hope that Canonical will be able to.

chro...@googlecode.com

unread,
Jun 25, 2014, 7:26:26 PM6/25/14
to chromi...@chromium.org

Comment #33 on issue 384654 by pi...@chromium.org: Chrome quits repeatedly
jbauman spent some time looking at FD usage, and in theory we should now
try very hard to stay under the limit for graphics resources - unless you
have a very low ulimit (is this stock Ubuntu? which version?)

one thing that could help is to check the list of fds open, e.g. with 'ls
-la /proc/$pid/fd' where $pid is the pid of the browser process (seen in
Chrome's task manager, or from the logs, 32081 in this case).

chro...@googlecode.com

unread,
Jun 26, 2014, 4:04:57 PM6/26/14
to chromi...@chromium.org

Comment #36 on issue 384654 by pi...@chromium.org: Chrome quits repeatedly
@#34: I was hoping for the actual list of open files :) Trying to figure
out if it's font files, shm files, or something else (we've had issues with
all of these in the past).

chro...@googlecode.com

unread,
Jun 26, 2014, 5:54:16 PM6/26/14
to chromi...@chromium.org
Updates:
Owner: jba...@chromium.org

Comment #38 on issue 384654 by pi...@chromium.org: Chrome quits repeatedly
502 shm
274 sockets
150 others

sockets is high, maybe due to some network pool of some sort, but still shm
is the dominant one.
That it happens while scrolling suggests it's definitely due to software
tiles. While we aggressively release resources for hidden tabs, should we
have a per-renderer limit?

If it prevents crash reports from being sent, it's kinda bad :(

chro...@googlecode.com

unread,
Jun 26, 2014, 6:04:20 PM6/26/14
to chromi...@chromium.org

Comment #39 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
I can confirm that it's shm that spikes when rapid scrolling with:

while true ; do ls -la /proc/$pid/fd | grep shm | wc -l ; done

Just simply scrolling rapidly up and down (twiddling my mouse wheel) on
this page can spike the count by several hundred. That number goes up and
stays up.

chro...@googlecode.com

unread,
Jun 30, 2014, 5:27:04 PM6/30/14
to chromi...@chromium.org

Comment #45 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
Would you kindly check Crash IDs fab4d1acd5cefa63 or c66418932a35301e for
relevance?

chro...@googlecode.com

unread,
Jun 30, 2014, 5:53:14 PM6/30/14
to chromi...@chromium.org

Comment #46 on issue 384654 by jam...@chromium.org: Chrome quits repeatedly
@lelyveld - those are the same thing, the process is out of PIDs.

chro...@googlecode.com

unread,
Jun 30, 2014, 5:55:13 PM6/30/14
to chromi...@chromium.org

Comment #47 on issue 384654 by jam...@chromium.org: Chrome quits repeatedly
Sorry not PIDs, it's out of FDs.

chro...@googlecode.com

unread,
Jun 30, 2014, 6:04:15 PM6/30/14
to chromi...@chromium.org

Comment #48 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
OK, thanks. It somehow it managed to make dumps for those anyway.

chro...@googlecode.com

unread,
Jul 16, 2014, 2:46:52 AM7/16/14
to chromi...@chromium.org

Comment #50 on issue 384654 by bugdro...@chromium.org: Chrome quits
repeatedly (Linux, AMD GPU, Ubuntu 14.04)
http://code.google.com/p/chromium/issues/detail?id=384654#c50

The following revision refers to this bug:

https://chromium.googlesource.com/chromium/src.git/+/721532e2b9f182f026ecd31055c133429464adaa

commit 721532e2b9f182f026ecd31055c133429464adaa
Author: jba...@chromium.org
<jba...@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98>
Date: Wed Jul 16 05:57:45 2014

Reduce resource limit with software compositor.

Every resource takes an FD, so we want to limit this to avoid running out.

BUG=384654

Review URL: https://codereview.chromium.org/357423002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@283370
0039d316-1c4b-4281-b951-d872f2087c98

chro...@googlecode.com

unread,
Jul 16, 2014, 2:49:54 AM7/16/14
to chromi...@chromium.org

Comment #51 on issue 384654 by bugdro...@chromium.org: Chrome quits
repeatedly (Linux, AMD GPU, Ubuntu 14.04)
http://code.google.com/p/chromium/issues/detail?id=384654#c51

------------------------------------------------------------------
r283370 | jba...@chromium.org | 2014-07-16T05:57:45.099633Z

Changed paths:
M
http://src.chromium.org/viewvc/chrome/trunk/src/content/renderer/gpu/compositor_output_surface.cc?r1=283370&r2=283369&pathrev=283370

Reduce resource limit with software compositor.

Every resource takes an FD, so we want to limit this to avoid running out.

BUG=384654

Review URL: https://codereview.chromium.org/357423002
-----------------------------------------------------------------

chro...@googlecode.com

unread,
Aug 29, 2014, 1:55:06 AM8/29/14
to chromi...@chromium.org

Comment #53 on issue 384654 by jorg...@chromium.org: Chrome quits
repeatedly (Linux, AMD GPU, Ubuntu 14.04)
https://code.google.com/p/chromium/issues/detail?id=384654

FWIW I can repro this on my home machine with a Radeon 6850:
cfe9abee02a574ed is me, and all the other crashes from that client id too.
This was also happening on stable on that same machine but I don't have
crash ids handy.

chro...@googlecode.com

unread,
Sep 1, 2014, 5:21:14 AM9/1/14
to chromi...@chromium.org

Comment #57 on issue 384654 by maheswar...@gmail.com: Chrome quits
repeatedly (Linux, AMD GPU, Ubuntu 14.04)
https://code.google.com/p/chromium/issues/detail?id=384654

can some one mention clear steps to reproduce the issue

i have tried with the below config but no crash , every thing is running as
expected
OS : Ubuntu 12.04.4 64 bit
chrome version 37.0.2062.94
AMD 14.6 beta driver

chro...@googlecode.com

unread,
Sep 1, 2014, 11:18:43 AM9/1/14
to chromi...@chromium.org

Comment #58 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
You need to be in Ubuntu 14.04 to repro.

chro...@googlecode.com

unread,
Sep 2, 2014, 4:32:11 PM9/2/14
to chromi...@chromium.org

Comment #61 on issue 384654 by eric.bla...@digitalmatrices.com: Chrome
quits repeatedly (Linux, AMD GPU, Ubuntu 14.04)
https://code.google.com/p/chromium/issues/detail?id=384654

I tried disabling all plugins and running it in Incognito mode. Same thing.

[4735:4764:0902/160114:ERROR:value_store_frontend.cc(62)] Error while
writing pioclpoplcdbaefihamjohnefbikjilc.browser_action to
/home/eblanchette/.config/google-chrome/Default/Extension State
[4735:4765:0902/160115:ERROR:host_shared_bitmap_manager.cc(144)] Cannot
create shared memory buffer
[4735:4774:0902/160115:ERROR:crash_handler_host_linux.cc(371)] Failed to
write crash dump for pid 5565
Cannot upload crash dump: failed to open
Failed to create temporary file in /tmp: cannot upload crash dump

chro...@googlecode.com

unread,
Sep 22, 2014, 4:18:22 PM9/22/14
to chromi...@chromium.org

Comment #65 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
This may be simply the unfixed file descriptor bug. Try a higher ulimit -n

chro...@googlecode.com

unread,
Sep 23, 2014, 6:34:37 AM9/23/14
to chromi...@chromium.org

Comment #66 on issue 384654 by a...@chromium.org: Chrome quits repeatedly
Issue 416166 has been merged into this issue.

chro...@googlecode.com

unread,
Nov 3, 2014, 3:42:39 PM11/3/14
to chromi...@chromium.org

Comment #68 on issue 384654 by dam...@addepar.com: Chrome quits repeatedly
I'm getting this error on a Nvidia GPU as well.

uname -a:
Linux damien-W530 3.13.0-36-lowlatency #63-Ubuntu SMP PREEMPT Wed Sep 3
21:56:12 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux

Chrome: Version 38.0.2125.111 (64-bit)

chro...@googlecode.com

unread,
Nov 12, 2014, 12:13:18 AM11/12/14
to chromi...@chromium.org

Comment #70 on issue 384654 by daniel.d...@gmail.com: Chrome quits
repeatedly (Linux, AMD GPU, Ubuntu 14.04)
https://code.google.com/p/chromium/issues/detail?id=384654

Same issue here, Intel GPU, Dell XPS 15

Linux gorgel-XPS-15-9530 3.16.0-24-generic #32-Ubuntu SMP Tue Oct 28
13:07:32 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux

Version: Chromium 38.0.2125.111 Ubuntu 14.10

chro...@googlecode.com

unread,
Nov 12, 2014, 7:16:37 AM11/12/14
to chromi...@chromium.org

Comment #71 on issue 384654 by george.m...@gmail.com: Chrome quits
repeatedly (Linux, AMD GPU, Ubuntu 14.04)
https://code.google.com/p/chromium/issues/detail?id=384654

Chrome crashes all the time on Ubuntu.

chro...@googlecode.com

unread,
Nov 12, 2014, 10:16:47 AM11/12/14
to chromi...@chromium.org

Comment #72 on issue 384654 by m...@rahidelvi.ca: Chrome quits repeatedly
I'm seeing the same issue. The terminal shows the 'Too many open files'
message.

As others have mentioned, it's easy to reproduce. To trigger the crash I
just need to open several tabs in two or more user profiles.

Version 38.0.2125.111 (64-bit) on Ubuntu 14.04.

Wasn't able to tinker with a higher ulimit.

chro...@googlecode.com

unread,
Jan 13, 2015, 8:22:57 AM1/13/15
to chromi...@chromium.org

Comment #73 on issue 384654 by yoann.ma...@gmail.com: Chrome quits
repeatedly (Linux, AMD GPU, Ubuntu 14.04)
https://code.google.com/p/chromium/issues/detail?id=384654

I have exactly the same issue. Always crashing since Google Chrome 39

[11146:11178:0113/141423:ERROR:host_shared_bitmap_manager.cc(144)] Cannot
create shared memory buffer
[11146:11187:0113/141423:ERROR:crash_handler_host_linux.cc(371)] Failed to
write crash dump for pid 12029
Cannot upload crash dump: failed to open
Failed to create temporary file in /tmp: cannot upload crash dump
[11146:11178:0113/141425:ERROR:host_shared_bitmap_manager.cc(144)] Cannot
create shared memory buffer
[11146:11175:0113/141425:ERROR:crash_handler_host_linux.cc(371)] Failed to
write crash dump for pid 12543
Cannot upload crash dump: failed to open
Failed to create temporary file in /tmp: cannot upload crash dump


chro...@googlecode.com

unread,
May 4, 2015, 9:56:30 PM5/4/15
to chromi...@chromium.org
Updates:
Status: WontFix

Comment #74 on issue 384654 by jba...@chromium.org: Chrome quits
repeatedly (Linux, AMD GPU, Ubuntu 14.04)
https://code.google.com/p/chromium/issues/detail?id=384654

The open files issue should be fixed. There's not much we can do about the
original GPU driver crash that caused the problem, but the software
fallback should work fine.

chro...@googlecode.com

unread,
May 5, 2015, 1:35:45 AM5/5/15
to chromi...@chromium.org

Comment #75 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
You were getting two basic bug reports here:

1) A file descriptor leak.

2) AMD graphics driver crash.

I confirm that you have indeed fixed the first bug (at least in the canary
channel). The second bug should remain open. It is not responsible to
just punt it upstream, since upstream claims it's downstream, and software
rendering does not give equivalent functionality (e.g. it will not support
WebGL, etc.). Furthermore, you have reports of it in this thread on both
Intel and AMD graphics platforms.

chro...@googlecode.com

unread,
May 5, 2015, 1:37:46 AM5/5/15
to chromi...@chromium.org

Comment #76 on issue 384654 by lelyv...@gmail.com: Chrome quits repeatedly
Just to add very briefly, in case it isn't obvious, I am the original
reporter.
Reply all
Reply to author
Forward
0 new messages