1054651 - msmpeg2vdec.dll crashes on YouTube's HTML5 video

Status: RESOLVED INCOMPLETE

(Core :: Audio/Video: Playback, defect)

Description

[[e-mail]]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:33.0) Gecko/20100101 Firefox/33.0 (Beta/Release)
Build ID: 20140815004001

Steps to reproduce:

1. Start a completely clean profile to ensure it's not the fault of any add-ons
2. Go to https://www.youtube.com/html5 to verify that the HTML5 player is being used
3. Navigate to any video on YouTube


Actual results:

The video loading animation briefly plays, then Firefox crashes.
Crash report here: https://crash-stats.mozilla.com/report/index/642e20eb-3631-4421-8ce7-f63d92140816


Expected results:

The video should have played.

Comment 1

[[e-mail]]

GPUs are an Intel HD 4000 and NVIDIA GeForce 610M. Both have their drivers up-to-date. Disabling hardware acceleration beforehand still crashes the browser, although now Windows says "Firefox is not responding" instead of the Crash Reporter coming up.

Comment 2

[alex_mayorga]

Is this your crash too https://crash-stats.mozilla.com/report/index/da7183c6-1432-4172-ba6b-a650b2140816 ?

Do the videos at http://ie.microsoft.com/testdrive/Graphics/VideoFormatSupport/Default.html also cause the crash?

Comment 3

[[e-mail]]

(In reply to alex_mayorga from comment #2)
> Is this your crash too
> https://crash-stats.mozilla.com/report/index/da7183c6-1432-4172-ba6b-
> a650b2140816 ?
> 
> Do the videos at
> http://ie.microsoft.com/testdrive/Graphics/VideoFormatSupport/Default.html
> also cause the crash?
That UUID is not listed in my about:crashes, so I suppose not. 

Yes, (one of) those videos causes the crash as well.

I'm not sure how useful of a datapoint this is, but the crash does not happen when running Firefox inside a VM on the same machine.

Comment 4

[alex_mayorga]

Can you please tell us which exact video crashes and what's the new crash ID?

Comment 5

[alex_mayorga]

Also could you please check if https://support.mozilla.org/en-US/kb/troubleshoot-extensions-themes-to-fix-problems improves things?

Comment 6

[[e-mail]]

Safe mode does not improve things.

I'm not 100% sure which video causes the crash, since the entire browser just stops working, but part of the webm video appeared to have loaded and I have viewed other .webms without issue, so it's very likely one of the h.264 videos.

The crash ID of this latest crash is bp-67370869-9616-46d5-b00f-734262140818

Comment 7

[Liz Henry (:lizzard) (relman/hg->git project)]

More reports: https://crash-stats.mozilla.com/report/list?date=2014-08-18&range_unit=days&range_value=28&signature=msmpeg2vdec.dll%400x7348f   

Tracy, this isn't a high volume crash but since it mainly appears on 33 I thought you might want to take a look.

Comment 8

[[e-mail]]

I think the reason it happens mostly on 33 is that for this version and above YouTube offers html5 video by default, as opposed to flash-based videos. The issue is present in 31 as well, but its users are less likely to come across these problematic videos.

Comment 9

[Tracy Walker [:tracy]]

heh, with fx33, 18 crashes on 10 installations in 28 days isn't something I have the bandwidth to investigate.  If/when it becomes a top crasher, I'll certainly set my attention to it.

That said, nyubis, please keep trying to figure out why vm works but not on hardware. what's differently configured?

Comment 10

[[e-mail]]

(In reply to [:tracy] Tracy Walker - QA Mentor from comment #9)
> heh, with fx33, 18 crashes on 10 installations in 28 days isn't something I
> have the bandwidth to investigate.  If/when it becomes a top crasher, I'll
> certainly set my attention to it.
Unfortunate, but I understand.

> That said, nyubis, please keep trying to figure out why vm works but not on
> hardware. what's differently configured?
Nothing in the Firefox configuration itself seems to be different, and I uninstalled any programs that might interfere with it. My best bet is something relating to drivers, but I'm not entirely sure how VMWare handles those.

Comment 11

[alex_mayorga]

Is the OS fully updated? What about IE? Some patches for OS and IE fix things on Firefox strangely...

Please paste the contents of "Graphics" in about:support

Comment 12

[[e-mail]]

All important/security updates are applied, still got some optional updates that I will take care of now.

about:support Graphics:

Adapter Description	Intel(R) HD Graphics 4000
Adapter Description (GPU #2)	NVIDIA GeForce 610M
Adapter Drivers	igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32
Adapter Drivers (GPU #2)	nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
Adapter RAM	Unknown
Adapter RAM (GPU #2)	1023
Device ID	0x0166
Device ID (GPU #2)	0x1058
Direct2D Enabled	true
DirectWrite Enabled	true (6.2.9200.16571)
Driver Date	5-17-2014
Driver Date (GPU #2)	8-30-2012
Driver Version	10.18.10.3621
Driver Version (GPU #2)	9.18.13.623
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 10
Vendor ID	0x8086
Vendor ID (GPU #2)	0x10de
WebGL Renderer	Google Inc. -- ANGLE (Intel(R) HD Graphics 4000 Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote	false
AzureCanvasBackend	direct2d
AzureContentBackend	direct2d
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0

Comment 13

[alex_mayorga]

Does this driver http://www.geforce.com/drivers/results/77228 improve things?

Comment 14

[[e-mail]]

(In reply to alex_mayorga from comment #13)
> Does this driver http://www.geforce.com/drivers/results/77228 improve things?

No, same as usual. Windows reports the drivers of both my GPUs as being up-to-date.

Crash ID for the one with the new driver:
bp-63e4ff02-af3a-4d6a-9a05-ec9e92140819

Comment 15

[:Gijs (he/him)]

(In reply to nyubis from comment #14)
> (In reply to alex_mayorga from comment #13)
> > Does this driver http://www.geforce.com/drivers/results/77228 improve things?
> 
> No, same as usual. Windows reports the drivers of both my GPUs as being
> up-to-date.
> 
> Crash ID for the one with the new driver:
> bp-63e4ff02-af3a-4d6a-9a05-ec9e92140819

This dll file belongs to Windows' media foundation things, AFAICT, so I'm moving this to Core::Audio/Video.

If you toggle media.windows-media-foundation.enabled to false in about:config, does that fix the crash? (I actually think it might make it impossible to watch H264 media, and thereby "fixes" the issue, but it's worth checking, I suppose...)

Comment 16

[[e-mail]]

(In reply to :Gijs Kruitbosch from comment #15)
> (In reply to nyubis from comment #14)
> > (In reply to alex_mayorga from comment #13)
> > > Does this driver http://www.geforce.com/drivers/results/77228 improve things?
> > 
> > No, same as usual. Windows reports the drivers of both my GPUs as being
> > up-to-date.
> > 
> > Crash ID for the one with the new driver:
> > bp-63e4ff02-af3a-4d6a-9a05-ec9e92140819
> 
> This dll file belongs to Windows' media foundation things, AFAICT, so I'm
> moving this to Core::Audio/Video.
> 
> If you toggle media.windows-media-foundation.enabled to false in
> about:config, does that fix the crash? (I actually think it might make it
> impossible to watch H264 media, and thereby "fixes" the issue, but it's
> worth checking, I suppose...)

It's as you thought, H264 video no longer works, but the browser doesn't crash any more. 
So if the problem exists in the WMF code, does that mean it's something you have no control over?

Comment 17

[:Gijs (he/him)]

(In reply to nyubis from comment #16)
> (In reply to :Gijs Kruitbosch from comment #15)
> > (In reply to nyubis from comment #14)
> > > (In reply to alex_mayorga from comment #13)
> > > > Does this driver http://www.geforce.com/drivers/results/77228 improve things?
> > > 
> > > No, same as usual. Windows reports the drivers of both my GPUs as being
> > > up-to-date.
> > > 
> > > Crash ID for the one with the new driver:
> > > bp-63e4ff02-af3a-4d6a-9a05-ec9e92140819
> > 
> > This dll file belongs to Windows' media foundation things, AFAICT, so I'm
> > moving this to Core::Audio/Video.
> > 
> > If you toggle media.windows-media-foundation.enabled to false in
> > about:config, does that fix the crash? (I actually think it might make it
> > impossible to watch H264 media, and thereby "fixes" the issue, but it's
> > worth checking, I suppose...)
> 
> It's as you thought, H264 video no longer works, but the browser doesn't
> crash any more. 

Thanks for confirming that.

> So if the problem exists in the WMF code, does that mean it's something you
> have no control over?

I don't really know the video/audio code, so I can't knowledgeably answer that question, but it is at least possible that that is the case.

It's strange though, because I expect this isn't universally reproducible, or we would have noticed before now and/or there'd be more people crashing. If you can reproduce this with any HTML5 h264 video on youtube, that's pretty terrible. In other words, I'm wondering why this crashes reliably on your machine and doesn't on most other people's copies of win7 (at least, I'm assuming it doesn't). Perhaps Chris can help and/or knows of a way to gather more data? Maybe we have contacts with MS that can help?

Comment 18

[Chris Pearce [:cpearce (Not reading bugmail)]]

(In reply to nyubis from comment #16)

> > If you toggle media.windows-media-foundation.enabled to false in
> > about:config, does that fix the crash? (I actually think it might make it
> > impossible to watch H264 media, and thereby "fixes" the issue, but it's
> > worth checking, I suppose...)
> 
> It's as you thought, H264 video no longer works, but the browser doesn't
> crash any more. 

Try toggling the pref media.windows-media-foundation.use-dxva to false.

This means we won't try to use your GPU to accelerate the video decoding, which is often where these problems lie.

> So if the problem exists in the WMF code, does that mean it's something you
> have no control over?

Basically yes.

It looks like you're crashing right at the start of decoding. We'll be replacing the code that calls thing in a few months, so likely we'll see a change in behaviour here soon.

Comment 19

[[e-mail]]

(In reply to Chris Pearce (:cpearce) from comment #18)
> Try toggling the pref media.windows-media-foundation.use-dxva to false.
> 
> This means we won't try to use your GPU to accelerate the video decoding,
> which is often where these problems lie.

Setting media.windows-media-foundation.enabled back to true and media.windows-media-foundation.use-dxva to false causes Firefox to crash on the test page. Unfortunately it's one of those crashes where the Crash Reporter doesn't come up, so I can't report it.

Comment 20

[Sladky Vladimir]

msmpeg2vdec.dll crashed too on http://ie.microsoft.com/testdrive/Performance/FishBowl/
in SeaMonkey 2.32 (64 bit)

https://crash-stats.mozilla.com/report/index/cf82ca15-61e6-417c-929a-495bc2140904

SeaMonkey 2.31 crashed too, but older SeaMonkey 2.30 (20140722191524) is O.K. and NOT crashed !!!
(and latest FireFox on same PC too not crashed)

Comment 21

[Sladky Vladimir]

With media.windows-media-foundation.enabled set to false SM 2.32 NOT crashed !!!
But water is in rot color, and not in blue.

Comment 22

[Sladky Vladimir]

And with media.windows-media-foundation.use-dxva to false is O.K. (in blue color and NOT crashed),
but older SM 2.30 not crashed with both set to true.

Comment 23

[Chris Pearce [:cpearce (Not reading bugmail)]]

Does the crash still occur today's Nightly? Some major changes were made recently to our Windows video decoding code.

Comment 24

[Sladky Vladimir]

Build ID:20140914030209 CSet: d070787de8f7 is now O.K. and with both statements set to true not crashed.

Comment 25

[Chris Pearce [:cpearce (Not reading bugmail)]]

Excellent. Thanks for reporting the bug!

Comment 26

[Sladky Vladimir]

But my yesterday selfcompilled 64Bit SeaMonkey/2.32a1 ID:20140914171258 CSet: d070787de8f7 crashed always :-(

https://crash-stats.mozilla.com/report/index/7e651c4e-a3f2-4b12-b4f2-b4bf62140915


Please mus i create new bug in Bugzilla for SeaMonkey too, or recompile it later only ???

Thank You

Comment 27

[Chris Pearce [:cpearce (Not reading bugmail)]]

(In reply to Sladky Vladimir from comment #26)
> But my yesterday selfcompilled 64Bit SeaMonkey/2.32a1 ID:20140914171258
> CSet: d070787de8f7 crashed always :-(
> 
> https://crash-stats.mozilla.com/report/index/7e651c4e-a3f2-4b12-b4f2-
> b4bf62140915
> 
> 
> Please mus i create new bug in Bugzilla for SeaMonkey too, or recompile it
> later only ???

That's probably bug 1062277. Should be fixed in the next day or so.
> 
> Thank You

Comment 28

[Matthias Versen [:Matti]]

Reporter: Is the issue now fixed for you ?

Comment 29

[[e-mail]]

(In reply to Matthias Versen [:Matti] from comment #28)
> Reporter: Is the issue now fixed for you ?

No.
I tried it with Developer Edition 37.0a2, with media.windows-media-foundation.enabled set to true, and it still crashes exactly the way it used to. Crash report is here: bp-99173498-0aee-4f02-8e80-1a9992150127

Comment 30

[Anthony Jones (:ajones, :kentuckyfriedtakahe, :k17e)]

We've fixed a number of issues since last time we asked. Is it working now?

Comment 31

[[e-mail]]

(In reply to Anthony Jones (:kentuckyfriedtakahe, :k17e) from comment #30)
> We've fixed a number of issues since last time we asked. Is it working now?

I can't really say, I don't run Windows on this machine any more. For what it's worth, it works without problems on Debian.