Fix a crasher in the GPU process caused by the DXVA code attempting to create an instance of the Vi…

Fix a crasher in the GPU process caused by the DXVA code attempting to create an instance of the Video processor MFT

The DXVA decoder object on Windows 8+ uses DX11 if available and in this mode it uses the VideoProcessor MFT for colorspace
conversion. This object exposes the IMFTransform interface and can be created like a traditional COM object via the
CoCreateInstance API. However this API expects CoInitialize to be called which is not the case on the main GPU thread and
fails with an error indicating that the COM apartment is not initialized. It is not clear as to why it is not failing in all
cases.

Proposed fix is to use DllGetClassObject and IClassFactory::CreateInstance to create the object directly and avoid using
CoCreateInstance. For this purpose added a helper function CreateCOMObjectFromDll in the dxva code.

BUG=493894
R=dalecurtis

Committed: https://crrev.com/f74c38420401d72fc8fdae2b8d03aece1279f7e7
Cr-Commit-Position: refs/heads/master@{#338975}

[DaleCurtis, 2015-07-15 23:11:56]

I don't know enough to say if that's the right dll to use, but if it works, on
the surface this lgtm % nits.

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
File content/common/gpu/media/dxva_video_decode_accelerator.cc (right):

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
content/common/gpu/media/dxva_video_decode_accelerator.cc:254: typedef
HRESULT(WINAPI * GetClassObject)(
No space before & and * on all lines? If you want, you can use "using" instead
here too for C++11 goodness.

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
content/common/gpu/media/dxva_video_decode_accelerator.cc:266:
reinterpret_cast<void**>(factory.Receive()));
ReceiveVoid to avoid cast?

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
content/common/gpu/media/dxva_video_decode_accelerator.cc:1029: CLSID clsid =
{};
Seems you could just roll this into the conditional above where decoder_dll is
set?

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
content/common/gpu/media/dxva_video_decode_accelerator.cc:1044:
decoder_.Receive()));
ReceiveVoid

[ananta, 2015-07-15 23:15:01]

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
File content/common/gpu/media/dxva_video_decode_accelerator.cc (right):

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
content/common/gpu/media/dxva_video_decode_accelerator.cc:254: typedef
HRESULT(WINAPI * GetClassObject)(
On 2015/07/15 23:11:56, DaleCurtis wrote:
> No space before & and * on all lines? If you want, you can use "using" instead
> here too for C++11 goodness.

eek. i dunno how i uploaded with this :(
Will upload a new patch

[ananta, 2015-07-15 23:23:25]

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
File content/common/gpu/media/dxva_video_decode_accelerator.cc (right):

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
content/common/gpu/media/dxva_video_decode_accelerator.cc:254: typedef
HRESULT(WINAPI * GetClassObject)(
On 2015/07/15 23:11:56, DaleCurtis wrote:
> No space before & and * on all lines? If you want, you can use "using" instead
> here too for C++11 goodness.

Done.

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
content/common/gpu/media/dxva_video_decode_accelerator.cc:266:
reinterpret_cast<void**>(factory.Receive()));
On 2015/07/15 23:11:56, DaleCurtis wrote:
> ReceiveVoid to avoid cast?

Done.

https://codereview.chromium.org/1235373003/diff/1/content/common/gpu/media/dx...
content/common/gpu/media/dxva_video_decode_accelerator.cc:1029: CLSID clsid =
{};
On 2015/07/15 23:11:56, DaleCurtis wrote:
> Seems you could just roll this into the conditional above where decoder_dll is
> set?

Yes. Done.

[DaleCurtis, 2015-07-16 00:22:41]

https://codereview.chromium.org/1235373003/diff/20001/content/common/gpu/medi...
File content/common/gpu/media/dxva_video_decode_accelerator.cc (right):

https://codereview.chromium.org/1235373003/diff/20001/content/common/gpu/medi...
content/common/gpu/media/dxva_video_decode_accelerator.cc:255: const CLSID
&clsid, const IID& iid, void **object);
& and ** are still not mapped right, should be CLSID& and void** :)

[ananta, 2015-07-16 00:24:13]

https://codereview.chromium.org/1235373003/diff/20001/content/common/gpu/medi...
File content/common/gpu/media/dxva_video_decode_accelerator.cc (right):

https://codereview.chromium.org/1235373003/diff/20001/content/common/gpu/medi...
content/common/gpu/media/dxva_video_decode_accelerator.cc:255: const CLSID
&clsid, const IID& iid, void **object);
On 2015/07/16 00:22:41, DaleCurtis wrote:
> & and ** are still not mapped right, should be CLSID& and void** :)

Fixed. Old age :(

[DaleCurtis, 2015-07-16 00:30:29]

lgtm

[ananta, 2015-07-16 01:38:33]

The CQ bit was checked by ananta@chromium.org

[commit-bot: I haz the power, 2015-07-16 01:38:52]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1235373003/40001

[commit-bot: I haz the power, 2015-07-16 02:41:29]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-07-16 02:42:36]

Patchset 3 (id:??) landed as
https://crrev.com/f74c38420401d72fc8fdae2b8d03aece1279f7e7
Cr-Commit-Position: refs/heads/master@{#338975}

[brucedawson, 2015-07-17 01:08:16]

Small bug hidden in the macros:

https://code.google.com/p/chromium/issues/detail?id=511044