log and screenshot:
https://pantheon.corp.google.com/storage/browser/chromiumos-test-logs/bugfiles/cr/608024/?pli=1

there is no problem if suspend and resuming by running "powerd_dbuss_suspend" 

Facing same issue on samus device with build 8172.18.0

+marcheu, dbehr

Any ideas on where this could be broken?

Is this repro on the last approved dev 8172.16.0?

I'm building ToT, I'll test and see if it happens there.

no repro on ToT on auron.

helenzhang@ and sontis@ tested .16, .17 and the issue appears with M51- 8172.17.0, where we have https://chromium-review.googlesource.com/#/c/339705/ fixing  issue 589122 .

This bug doesn't repro, so it's a P2 at best.

it's still reproducible in 8172.21.0, 51.0.2704.32

AS per c#12, this seems to be a regression between https://crosland.corp.google.com/log/8172.16.0..8172.17.0

Helen, Are there any special steps to repro? Is this confirmed to be present on both auron and samus? 

CL got also merged to M50, can you check if problem is there too?

Tested lulu 7978.68.0, 50.0.2661.94, it's reproducible. 

It's 100% reproducible : 
- sign in 
- close lid 
- open lid 
- to see it more obviously, just move mouse cursor or ctr-t to launch new tab 
if go to vt2 (alt-ctr-f2) at this point, there is no way to go back to vt1 (alt-ctr-f1) anymore. 

helenzhang@ Did this CL land before 7978.66.0?

ketakid@, please see https://bugs.chromium.org/p/chromium/issues/detail?id=589122#c92

It landed on .64 -https://crosland.corp.google.com/log/7978.63.0..7978.64.0

Interesting. We had test verify this and hadn't seen the issue with the fix before we approved and merged it to 50.

Just chatted with dbehr@ about this issue. Per him frecon doesnt do anything in background on suspend and resume and is only active in background in developer mode. He doesn't think his frecon changes are causing this. Should we have marcheu@ take a look?

marcheu@ any idea what could be causing this if not the frecon changes?

Is this reproductible on ToT?
Does it happen both on R50 and R51?
Does it only happen in test image or dev mode?

I see these messages in UI log

[30535:30542:0429/185335:ERROR:gbm_buffer.cc(84)] Failed to export buffer to dma_buf: Permission denied
[30244:30527:0429/115335:ERROR:browser_gpu_memory_buffer_manager.cc(587)] Failed to create GpuMemoryBuffer.

With these kernel log messages to go with it

016-04-29T12:00:14.387956-07:00 WARNING kernel: [   34.564989] WARNING: CPU: 2 PID: 6940 at /mnt/host/source/src/third_party/kernel/v3.14/drivers/gpu/drm/i915/i915_gem.c:5125 i915_gem_obj_to_ggtt+0x4b/0x4f()
2016-04-29T12:00:14.387972-07:00 NOTICE kernel: [   34.565002] Modules linked in: rfcomm evdi i2c_dev uinput x86_pkg_temp_thermal iwlmvm iwl7000_mac80211 snd_hda_codec_realtek snd_hda_codec_generic uvcvideo snd_hda_codec_hdmi memc_x86 videobuf2_vmalloc aesni_intel iwlwifi zram aes_x86_64 glue_helper lrw gf128mul ablk_helper cryptd iio_trig_sysfs
 snd_hda_intel snd_hda_controller snd_hda_codec snd_hwdep fuse snd_soc_sst_acpi cfg80211 nf_conntrack_ipv6 nf_defrag_ipv6 ip6table_filter ip6_tables btusb btbcm btintel bluetooth cros_ec_accel kfifo_buf industrialio joydev snd_seq_midi snd_seq_midi_event snd_rawmidi snd_seq snd_seq_device ppp_async ppp_generic slhc tun
2016-04-29T12:00:14.387980-07:00 NOTICE kernel: [   34.565106] CPU: 2 PID: 6940 Comm: CompositorTileW Tainted: G        W    3.14.0 #1
2016-04-29T12:00:14.387985-07:00 NOTICE kernel: [   34.565115] Hardware name: GOOGLE Auron_Yuna, BIOS Google_Auron_yuna.6301.59.8 04/02/2015
2016-04-29T12:00:14.387990-07:00 NOTICE kernel: [   34.565123]  0000000000000000 00000000dfe23e84 ffff880164f83de8 ffffffffa5f974d3
2016-04-29T12:00:14.387995-07:00 NOTICE kernel: [   34.565135]  0000000000000000 ffff880164f83e20 ffffffffa5a3df6b ffffffffa5ce1b87
2016-04-29T12:00:14.387998-07:00 NOTICE kernel: [   34.565146]  ffff880077a47680 0000000000000000 ffff8801795d07c8 0000000040086200
2016-04-29T12:00:14.388001-07:00 NOTICE kernel: [   34.565157] Call Trace:
2016-04-29T12:00:14.388005-07:00 NOTICE kernel: [   34.565166]  [<ffffffffa5f974d3>] dump_stack+0x4d/0x6f
2016-04-29T12:00:14.388008-07:00 NOTICE kernel: [   34.565176]  [<ffffffffa5a3df6b>] warn_slowpath_common+0x7f/0x98
2016-04-29T12:00:14.388010-07:00 NOTICE kernel: [   34.565184]  [<ffffffffa5ce1b87>] ? i915_gem_obj_to_ggtt+0x4b/0x4f
2016-04-29T12:00:14.388013-07:00 NOTICE kernel: [   34.565193]  [<ffffffffa5a3e07d>] warn_slowpath_null+0x1a/0x1c
2016-04-29T12:00:14.388016-07:00 NOTICE kernel: [   34.565201]  [<ffffffffa5ce1b87>] i915_gem_obj_to_ggtt+0x4b/0x4f
2016-04-29T12:00:14.388019-07:00 NOTICE kernel: [   34.565209]  [<ffffffffa5ce1c6b>] i915_gem_object_set_to_gtt_domain+0xe0/0x113
2016-04-29T12:00:14.388022-07:00 NOTICE kernel: [   34.565219]  [<ffffffffa5d318eb>] i915_gem_end_cpu_access+0x2e/0x42
2016-04-29T12:00:14.388025-07:00 NOTICE kernel: [   34.565228]  [<ffffffffa5d5fd97>] dma_buf_end_cpu_access+0x3f/0x44
2016-04-29T12:00:14.388028-07:00 NOTICE kernel: [   34.565236]  [<ffffffffa5d60090>] dma_buf_ioctl+0x8d/0xc5
2016-04-29T12:00:14.388030-07:00 NOTICE kernel: [   34.565244]  [<ffffffffa5b1c708>] do_vfs_ioctl+0x355/0x416
2016-04-29T12:00:14.388033-07:00 NOTICE kernel: [   34.565252]  [<ffffffffa5b24ec7>] ? __fget+0x6f/0x79
2016-04-29T12:00:14.388035-07:00 NOTICE kernel: [   34.565259]  [<ffffffffa5b1c820>] SyS_ioctl+0x57/0x79
2016-04-29T12:00:14.388038-07:00 NOTICE kernel: [   34.565268]  [<ffffffffa5f9c89d>] tracesys+0xda/0xdf

So I think exporting dma buf fails and then Chrome cannot lock object and upload texture.

Frecon reprobes all the monitor on hotplug, which I think also happens on resume.
So maybe it somehow steals master from Chrome and Chrome is not able to create and export buffers?

I have series of changes in ToT that fixes frecon behavior when frecon is in background and at resume time, so I can merge them to R50 -> see here
https://chromium-review.googlesource.com/#/c/342100/1

and if this happens in R51 as well, I will need to port the same changes to R51.

And here we have merge for R51 https://chromium-review.googlesource.com/#/c/342141/

Is this reproductible on ToT? - NO

Does it happen both on R50 and R51? - YES

Does it only happen in test image or dev mode? - YES

Before merging to 51 (and on to 50) is there anything we can do to validate these on ToT?

It sounds like this is only happening on the branches which is a bit unfortunate :(, especially for R50 which is just now going to stable. 

If there is anything we can do on ToT to validate these for say the next 24 hours we can land in R51 tomorrow to get into the second beta later this week.

I have tested it on my Samus on ToT by building frecon R50 and R51 branches and reproducing the bug on ToT and then applying these patches and they fixed it.

You can also verify it yourself on ToT by cros_workon start frecon and switching between master, R50 and R51 branches and building and deploying frecon.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/platform/frecon/+/602b601cc0a605b255f984a5355b44dc8485b9e4

commit 602b601cc0a605b255f984a5355b44dc8485b9e4
Author: Dominik Behr <dbehr@chromium.org>
Date: Thu Apr 14 21:57:21 2016

frecon: keep master till first modeset

Create DRM objects after daemonizing frecon.
Force clearing terminal when splash screen clears screen.
Add PID to frecon tag in log utility to differentiate between potential frecon
instances.

BUG= chromium:589122 , 608024 
TEST=boot Chrome OS in normal mode. Edit chromeos_startup and\
chromeos-boot-alert to force block_devmode screen at startup and boot again.

Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/339060
Reviewed-by: Ilja Friedel <ihf@chromium.org>
(cherry picked from commit b1abcba880c8e62bc4b2ddf2409fc919b744db14)
Change-Id: I4a93c33c1201018ae1d404dc261a5cc6dc2f0280
Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/342141
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>
Reviewed-by: Bernie Thompson <bhthompson@chromium.org>
Commit-Queue: Bernie Thompson <bhthompson@chromium.org>
Tested-by: Bernie Thompson <bhthompson@chromium.org>

[modify] https://crrev.com/602b601cc0a605b255f984a5355b44dc8485b9e4/util.c
[modify] https://crrev.com/602b601cc0a605b255f984a5355b44dc8485b9e4/drm.c
[modify] https://crrev.com/602b601cc0a605b255f984a5355b44dc8485b9e4/drm.h
[modify] https://crrev.com/602b601cc0a605b255f984a5355b44dc8485b9e4/term.c
[modify] https://crrev.com/602b601cc0a605b255f984a5355b44dc8485b9e4/splash.h
[modify] https://crrev.com/602b601cc0a605b255f984a5355b44dc8485b9e4/splash.c
[modify] https://crrev.com/602b601cc0a605b255f984a5355b44dc8485b9e4/term.h
[modify] https://crrev.com/602b601cc0a605b255f984a5355b44dc8485b9e4/main.c

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/platform/frecon/+/ad2d7048367554923da3c85d0eae730b219e3b50

commit ad2d7048367554923da3c85d0eae730b219e3b50
Author: Dominik Behr <dbehr@chromium.org>
Date: Thu Apr 21 19:35:08 2016

frecon: do not process hotplug events when frecon is in background

Also, refactor frecon activation/deactivation so it uses a set of two
functions.

BUG= chromium:608024 
TEST=plug in external monitor when frecon is in background, see dmesg

Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/340156
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>
(cherry picked from commit 83864df591406cf058264328dbecd114a2357697)
Change-Id: I6a1b8e9b1ae04a5e977ef52608adb9346b5f6274
Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/342142
Reviewed-by: Bernie Thompson <bhthompson@chromium.org>
Commit-Queue: Bernie Thompson <bhthompson@chromium.org>
Tested-by: Bernie Thompson <bhthompson@chromium.org>

[modify] https://crrev.com/ad2d7048367554923da3c85d0eae730b219e3b50/dbus.h
[modify] https://crrev.com/ad2d7048367554923da3c85d0eae730b219e3b50/dbus.c
[modify] https://crrev.com/ad2d7048367554923da3c85d0eae730b219e3b50/input.c
[modify] https://crrev.com/ad2d7048367554923da3c85d0eae730b219e3b50/term.c
[modify] https://crrev.com/ad2d7048367554923da3c85d0eae730b219e3b50/main.h
[modify] https://crrev.com/ad2d7048367554923da3c85d0eae730b219e3b50/term.h
[modify] https://crrev.com/ad2d7048367554923da3c85d0eae730b219e3b50/main.c

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/platform/frecon/+/1f6b36f3eaeaf6faa98e6b904ee8d3a5c7405675

commit 1f6b36f3eaeaf6faa98e6b904ee8d3a5c7405675
Author: Dominik Behr <dbehr@chromium.org>
Date: Wed Apr 27 19:31:02 2016

frecon: add a hook for SuspendDone DBUS signal

And use it to trigger monitor hotplug so frecon could detect monitors that
were connected or disconnected during suspend.

BUG= chromium:608024 
TEST=boot panther, suspend, switch monitor to other port, resume, switch to\
console using CTRL-ALT-F2

Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/340991
(cherry picked from commit 1883c040bfef8304bb6963396de50405564387e1)
Change-Id: Ie5bd3c86f2f0f03f4ba6ae08fa68c3c776a58948
Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/342143
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>
Reviewed-by: Bernie Thompson <bhthompson@chromium.org>
Commit-Queue: Bernie Thompson <bhthompson@chromium.org>
Tested-by: Bernie Thompson <bhthompson@chromium.org>

[modify] https://crrev.com/1f6b36f3eaeaf6faa98e6b904ee8d3a5c7405675/dbus.h
[modify] https://crrev.com/1f6b36f3eaeaf6faa98e6b904ee8d3a5c7405675/dbus_interface.h
[modify] https://crrev.com/1f6b36f3eaeaf6faa98e6b904ee8d3a5c7405675/dbus.c
[modify] https://crrev.com/1f6b36f3eaeaf6faa98e6b904ee8d3a5c7405675/term.c
[modify] https://crrev.com/1f6b36f3eaeaf6faa98e6b904ee8d3a5c7405675/term.h
[modify] https://crrev.com/1f6b36f3eaeaf6faa98e6b904ee8d3a5c7405675/main.c

it's not reproducible in ChromeOS 8172.25.0 / 51.0.2704.37 on paine 

Fix is verified on ChromeOS 8172.25.0 / 51.0.2704.37 with Samus and Auron Paine.

Still able to reproduce this issue on M50 build 7978.74.0

Screenshot and logs are present at https://pantheon.corp.google.com/storage/browser/chromiumos-test-logs/bugfiles/cr/608024/new/?debugUI=CLOUD

Yeah, the changes are ready to go here
https://chromium-review.googlesource.com/#/c/342100/
https://chromium-review.googlesource.com/#/c/342102/
https://chromium-review.googlesource.com/#/c/342101/

ketakid@ was going to decide whether she wants them in M50 after they are verified in M51. So far I haven't heard from here.

Does this behavior happen every single time the user opens /closes the lid? This issue is in the current stable that is out there since this was not RBS before and we had not yet baked the fix in 51 at that time. 

Re#39
Yes happens every time the user opens /closes the lid.

and not repro'd on last stable release(7978.66.0-lulu) but repro'd on 7978.74.0-lulu.

Yes, the issue reproduces most of the time, and has pretty bad UX impact. 
If it had same impact in M50 prod build we should have heard about it.
We tested the M-50 7978.66.0 build on lulu and issue does not reproduce. 
c#17 points to M50-7978.68.0 build where it is reproduced on lulu.

Correction
...We tested the M-50 7978.66.3 build on lulu and issue does not reproduce...

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/platform/frecon/+/eaa8325e4ce7468553e331699666bdab043af4c3

commit eaa8325e4ce7468553e331699666bdab043af4c3
Author: Dominik Behr <dbehr@chromium.org>
Date: Thu Apr 14 21:57:21 2016

frecon: keep master till first modeset

Create DRM objects after daemonizing frecon.
Force clearing terminal when splash screen clears screen.
Add PID to frecon tag in log utility to differentiate between potential frecon
instances.

BUG= chromium:589122 , 608024 
TEST=boot Chrome OS in normal mode. Edit chromeos_startup and\
chromeos-boot-alert to force block_devmode screen at startup and boot again.

Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/339060
Reviewed-by: Ilja Friedel <ihf@chromium.org>
(cherry picked from commit b1abcba880c8e62bc4b2ddf2409fc919b744db14)
Change-Id: I45455dffbd408e7dfafe51f91b408a92f2f62b75
Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/342100
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>

[modify] https://crrev.com/eaa8325e4ce7468553e331699666bdab043af4c3/util.c
[modify] https://crrev.com/eaa8325e4ce7468553e331699666bdab043af4c3/drm.c
[modify] https://crrev.com/eaa8325e4ce7468553e331699666bdab043af4c3/drm.h
[modify] https://crrev.com/eaa8325e4ce7468553e331699666bdab043af4c3/term.c
[modify] https://crrev.com/eaa8325e4ce7468553e331699666bdab043af4c3/splash.h
[modify] https://crrev.com/eaa8325e4ce7468553e331699666bdab043af4c3/splash.c
[modify] https://crrev.com/eaa8325e4ce7468553e331699666bdab043af4c3/term.h
[modify] https://crrev.com/eaa8325e4ce7468553e331699666bdab043af4c3/main.c

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/platform/frecon/+/023c0f88aa6bc15b783eeae8c57b132f00a3b446

commit 023c0f88aa6bc15b783eeae8c57b132f00a3b446
Author: Dominik Behr <dbehr@chromium.org>
Date: Thu Apr 21 19:35:08 2016

frecon: do not process hotplug events when frecon is in background

Also, refactor frecon activation/deactivation so it uses a set of two
functions.

BUG= chromium:608024 
TEST=plug in external monitor when frecon is in background, see dmesg

Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/340156
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>
(cherry picked from commit 83864df591406cf058264328dbecd114a2357697)
Change-Id: Id5e4689f8d4f21afa5ad73c81717632b3d7e85ce
Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/342101

[modify] https://crrev.com/023c0f88aa6bc15b783eeae8c57b132f00a3b446/dbus.h
[modify] https://crrev.com/023c0f88aa6bc15b783eeae8c57b132f00a3b446/dbus.c
[modify] https://crrev.com/023c0f88aa6bc15b783eeae8c57b132f00a3b446/input.c
[modify] https://crrev.com/023c0f88aa6bc15b783eeae8c57b132f00a3b446/term.c
[modify] https://crrev.com/023c0f88aa6bc15b783eeae8c57b132f00a3b446/main.h
[modify] https://crrev.com/023c0f88aa6bc15b783eeae8c57b132f00a3b446/term.h
[modify] https://crrev.com/023c0f88aa6bc15b783eeae8c57b132f00a3b446/main.c

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/platform/frecon/+/e746c881d75092ec5db2ef44f7dc7ce7b7cd551e

commit e746c881d75092ec5db2ef44f7dc7ce7b7cd551e
Author: Dominik Behr <dbehr@chromium.org>
Date: Wed Apr 27 19:31:02 2016

frecon: add a hook for SuspendDone DBUS signal

And use it to trigger monitor hotplug so frecon could detect monitors that
were connected or disconnected during suspend.

BUG= chromium:608024 
TEST=boot panther, suspend, switch monitor to other port, resume, switch to\
console using CTRL-ALT-F2

Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/340991
(cherry picked from commit 1883c040bfef8304bb6963396de50405564387e1)
Change-Id: Ifbd0e8d5e89554c94eceb9997fa5dc830a1277a7
Signed-off-by: Dominik Behr <dbehr@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/342102
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>

[modify] https://crrev.com/e746c881d75092ec5db2ef44f7dc7ce7b7cd551e/dbus.h
[modify] https://crrev.com/e746c881d75092ec5db2ef44f7dc7ce7b7cd551e/dbus_interface.h
[modify] https://crrev.com/e746c881d75092ec5db2ef44f7dc7ce7b7cd551e/dbus.c
[modify] https://crrev.com/e746c881d75092ec5db2ef44f7dc7ce7b7cd551e/term.c
[modify] https://crrev.com/e746c881d75092ec5db2ef44f7dc7ce7b7cd551e/term.h
[modify] https://crrev.com/e746c881d75092ec5db2ef44f7dc7ce7b7cd551e/main.c

 Issue 608808  has been merged into this issue.

The issue is verified as Fixed in M50-STABLE-18 (7978.76.0, 50.0.2661.104)

The fixed is also verified with webgl, video and flash. Looks good.

Confirmed latest fix (#45) was in M-51 8172.25.0. Fix was verified per #34.