gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI buffers.

gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel

Committed: https://crrev.com/1ff18bae4738ee3d62c2310d9e03211054be0582
Cr-Commit-Position: refs/heads/master@{#406973}

[reveman, 2016-07-21 14:57:00]

Description was changed from

==========
gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI
buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
==========

to

==========
gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI
buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[hshi1, 2016-07-21 20:49:08]

The CQ bit was checked by hshi@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-07-21 20:49:38]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[hshi1, 2016-07-21 20:58:37]

Description was changed from

==========
gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI
buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI
buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[hshi1, 2016-07-21 20:59:30]

lgtm but this appears to be the right thing to do and not really a workaround?

the behavior of kernel seems to be working as intended

[reveman, 2016-07-21 21:08:26]

reveman@chromium.org changed reviewers:
+ piman@chromium.org

[reveman, 2016-07-21 21:08:26]

+piman

[reveman, 2016-07-21 21:11:48]

On 2016/07/21 at 20:59:30, hshi wrote:
> lgtm but this appears to be the right thing to do and not really a workaround?
> 
> the behavior of kernel seems to be working as intended

I think it's appropriate for dma-buf shared buffers to survive when shared with
another process even if the original process died. This works as expected with
other drivers.

[piman, 2016-07-21 21:15:25]

lgtm

https://codereview.chromium.org/2171713004/diff/1/gpu/ipc/service/gpu_memory_...
File gpu/ipc/service/gpu_memory_buffer_factory_ozone_native_pixmap.h (right):

https://codereview.chromium.org/2171713004/diff/1/gpu/ipc/service/gpu_memory_...
gpu/ipc/service/gpu_memory_buffer_factory_ozone_native_pixmap.h:8: #include
"base/containers/hash_tables.h"
nit: this is deprecated. Can you use std::unordered_map?

[marcheu1, 2016-07-21 21:16:38]

On 2016/07/21 21:11:48, reveman wrote:
> On 2016/07/21 at 20:59:30, hshi wrote:
> > lgtm but this appears to be the right thing to do and not really a
workaround?
> > 
> > the behavior of kernel seems to be working as intended
> 
> I think it's appropriate for dma-buf shared buffers to survive when shared
with
> another process even if the original process died. This works as expected with
> other drivers.

I think it works by chance for other drivers because they have a special path
for self-reimport. But there is definitely no such guarantee generically...

[commit-bot: I haz the power, 2016-07-21 21:41:34]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-07-21 21:41:35]

Dry run: This issue passed the CQ dry run.

[reveman, 2016-07-21 21:53:16]

The CQ bit was checked by reveman@chromium.org

[reveman, 2016-07-21 21:53:17]

The patchset sent to the CQ was uploaded after l-g-t-m from hshi@chromium.org,
piman@chromium.org
Link to the patchset: https://codereview.chromium.org/2171713004/#ps20001
(title: "use std::unordered_map")

[commit-bot: I haz the power, 2016-07-21 21:53:52]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-21 22:43:18]

Description was changed from

==========
gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI
buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI
buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[commit-bot: I haz the power, 2016-07-21 22:43:19]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-07-21 22:46:10]

Description was changed from

==========
gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI
buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
gpu: Keep original GMB alive to workaround problem with importing of UDL/EVDI
buffers.

Destroying the original buffer invalidates kernel state such that
importing from the FD doesn't work as expected. By keeping the
original buffer alive we avoid this issue for scanout buffers used
by the browser process.

BUG=628334
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel

Committed: https://crrev.com/1ff18bae4738ee3d62c2310d9e03211054be0582
Cr-Commit-Position: refs/heads/master@{#406973}
==========

[commit-bot: I haz the power, 2016-07-21 22:46:11]

Patchset 2 (id:??) landed as
https://crrev.com/1ff18bae4738ee3d62c2310d9e03211054be0582
Cr-Commit-Position: refs/heads/master@{#406973}