1570046 - Opening simultaneously more than 3 tabs with video causes massive amount of WMF VPX errors and warnings in about:support/Troubleshooting Information in Graphics section in Failure Log sub-section after landing patch from bug #1521370

Status: NEW

(Core :: Graphics, defect, P2)

Description

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

STR:

1. Open movie website page (e.g. YouTube - https://www.youtube.com/ )
2. Open simultaneously more then 3 movies as background tabs (or more [like 10 for example]) if you don't see STR #3)
3. Open about:support/Troubleshooting Information
   and see in Graphics section in Failure Log sub-section many errors like these:

• (#0) Error: WMF VPX decoder just crashed; hardware video will be disabled.
• (#1): CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.

Oddly that is shows like this in:
Important Modified Preferences

• media.hardware-video-decoding.failed: false

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

:Virtual, could you try to find a regression range in using for example mozregression?

Comment 2

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Found commit message:
Bug 1521370 - Ensure WMF PDM is always initialized on the right thread. r=mattwoodrow,gerald.
And only do it once.
Depends on D22874

Differential Revision:
https://phabricator.services.mozilla.com/D23247

Comment 3

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Attachment: mozregression-gui log.txt

There are massive amount of warning errors in bugged builds like:

• 2019-07-30T19:11:33: INFO : [Child 5296, Chrome_ChildThread] WARNING: pipe error: 109: file z:/task_1552687340/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 341
• 2019-07-30T19:11:34: INFO : [Child 3256, MediaPlayback #2] WARNING: Decoder=115ace00 Decode error: NS_ERROR_DOM_MEDIA_FATAL_ERR (0x806e0005) - RefPtr<MediaSourceTrackDemuxer::SamplesPromise> __cdecl mozilla::MediaSourceTrackDemuxer::DoGetSamples(int32_t): manager is detached.: file z:/task_1552687340/build/src/dom/media/MediaDecoderStateMachine.cpp, line 3296

in mozregresion-gui log, see attachment for more details.

Comment 4

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Attachment: Graphics section from about:support.txt

Comment 5

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Attachment: Graphics section from about:support (before issue).txt

Comment 6

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Attachment: Graphics section from about;support (after issue).txt

Comment 7

[Jean-Yves Avenard [:jya]]

I don't see how that patch could cause the issue mentioned.

A Gtx 750 Ti doesn't have a VP9 hardware accelerator, WMF VPX will only be used if your machine supports HW decoding. And yours doesn't.
So that WMF VPX gets disabled because attempting to load a decoder failed is of no consequence.

I've just tested on a machine with a nvidia 1080ti and I can open over 10 videos, all HW decoded.

Comment 8

[Jim Jeffery not reading bug-mail 1/2/11]

Attachment: graphics data.txt

I am also seeing these failure, attached is my graphics data from about:support

Comment 9

[Jim Jeffery not reading bug-mail 1/2/11]

I also see these failures , see attachment above this post..
Testing using latest Nightly:
Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:70.0) Gecko/20100101 Firefox/70.0

EDIT: after 'resetting' the Webrender and restarting the browser a quick check is not showing this error:
Failure Log
(#0) Error Use Premul SwapChain for Intel GPU

Comment 10

[Jeny]

(In reply to Jean-Yves Avenard [:jya] from comment #7)

I don't see how that patch could cause the issue mentioned.

A Gtx 750 Ti doesn't have a VP9 hardware accelerator, WMF VPX will only be used if your machine supports HW decoding. And yours doesn't.
So that WMF VPX gets disabled because attempting to load a decoder failed is of no consequence.

I've just tested on a machine with a nvidia 1080ti and I can open over 10 videos, all HW decoded.

I have 1080 SLI with 64 GB RAM, I can confirm the bug, except you need to open a lot of tabs with lots of video,

Every now and then I got media.hardware-video-decoding.failed: false even I reset it,

I think this is cause by a memory leak in GPU process

Comment 11

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

I redone regression range hunting with updated STR from Comment #0 and mozregression-gui shows that bug #1521370 is for sure the cause of this bug.

In about:support/Troubleshooting Information in Graphics section in Failure Log sub-section there are many errors like these:

• (#0) Error: WMF VPX decoder just crashed; hardware video will be disabled.
• (#1): CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.

Also in mozregresion-gui log there are massive amount of warning errors in bugged builds like:

• 2019-07-30T19:11:33: INFO : [Child 5296, Chrome_ChildThread] WARNING: pipe error: 109: file z:/task_1552687340/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 341
• 2019-07-30T19:11:34: INFO : [Child 3256, MediaPlayback #2] WARNING: Decoder=115ace00 Decode error: NS_ERROR_DOM_MEDIA_FATAL_ERR (0x806e0005) - RefPtr<MediaSourceTrackDemuxer::SamplesPromise> __cdecl mozilla::MediaSourceTrackDemuxer::DoGetSamples(int32_t): manager is detached.: file z:/task_1552687340/build/src/dom/media/MediaDecoderStateMachine.cpp, line 3296

So I'm reopening this bug, as it's not WORKSFORME for sure.

Comment 12

[Jean-Yves Avenard [:jya]]

(In reply to Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.) from comment #11)

In about:support/Troubleshooting Information in Graphics section in Failure Log sub-section there are many errors like these:

• (#0) Error: WMF VPX decoder just crashed; hardware video will be disabled.
• (#1): CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.

Also in mozregresion-gui log there are massive amount of warning errors in bugged builds like:

• 2019-07-30T19:11:33: INFO : [Child 5296, Chrome_ChildThread] WARNING: pipe error: 109: file z:/task_1552687340/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 341

So I'm reopening this bug, as it's not WORKSFORME for sure.

What problem are you exactly having? That after 3 videos it will no longer attempt to use WMF VPX?
Why would you want it to use WMF VPX here, it will never do so even before bug 1521370. WMF VPX is only used if hardware accelerated vp9 decoding is supported.
The logic is as follow:
1- Load WMF VPX framework once during the entire lifetime of the process.
2- When playing a new VP9 video, instanciate a new WMF VP9 decoder
3- Check if the newly created decoder is hardware accelerated. If Yes, use it. If not fallback to using FFmpeg ffvp9 software decoder.

Your graphic card doesn't support VP9, only nvidia > 10xx do so. So WMF VPX will never be used ever.

If "Error: WMF VPX decoder just crashed; hardware video will be disabled." appearing inside about:support is what worries you, it shouldn't.

Before bug 1521370 it would have crashed just the same. Bug 1521370 detects when attempting to load the WMF VPX actually crashed. Under most circumstances, such crash would have been recoverable, but not always. If it had crashed at startup, it would have prevented Firefox from even loading.
So bug 1521370 added a crashguard so that if loading the WMF VPX framework ever crashed, it wouldn't attempt to load it again.

So your regression range just finds where the crash would have been detected and prevented from crashing again. This is working as designed.
Of course your run of mozregression will find that commit, your machine would have crashed even before 1521370 . Bug 1521370 detects it and adds it to the history of events as shown in about:support.

This is not a bug nor a regression.

• 2019-07-30T19:11:34: INFO : [Child 3256, MediaPlayback #2] WARNING: Decoder=115ace00 Decode error: NS_ERROR_DOM_MEDIA_FATAL_ERR (0x806e0005) - RefPtr<MediaSourceTrackDemuxer::SamplesPromise> __cdecl mozilla::MediaSourceTrackDemuxer::DoGetSamples(int32_t): manager is detached.: file z:/task_1552687340/build/src/dom/media/MediaDecoderStateMachine.cpp, line 3296

This will show if you start playback, and close the window in the middle of it. It's just a warning that the decoding stopped because the video element got deleted from under us. It is not an actual decoding error, and it will not be observable by javascript.

There is no bug here, nor a regression. Detecting that loading the framework caused a crash is a feature.

Comment 13

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Thank you very much for detailed information.

This bug and regression is that after landing patch from bug #1521370 and opening simultaneously more then 3 movies as background tabs, I'm getting spammed with massive amount of errors in about:support/Troubleshooting Information in Graphics section in Failure Log sub-section, like these:

• (#0) Error: WMF VPX decoder just crashed; hardware video will be disabled.
• (#1): CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.

It happens every time on every video after doing STR. Without doing STR, like opening simultaneously only 2 movies or one by one, this bug isn't reproducible.

It's noisy annoying nuisance masking real errors and warnings when I'm debugging other things. It should be disabled, on non supporting GPUs like mine as you wrote, which doesn't even support WMF VPX decoder and VP8 and VP9 hardware accelerated decoding.

What's more I'm still having "media.hardware-video-decoding.failed" preference set as "false" value, which is illogical per above statements.
But if it controls only H.264/MPEG-4 AVC hardware accelerated decoding, it should change name to for example "media.hardware-video-decoding-h264.failed",
and new preference for example "media.hardware-video-decoding-vp.failed" should be added to control only VP 8 and VP9 hardware accelerated decoding and to control ["wmfvpxvideoCrashGuard" - "Reset on Next Restart"] feature in about:support/Troubleshooting Information in Graphics section in "Crash Guard Disabled Features" sub-section.

What do you think about this?

Comment 14

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

(In reply to Jeny from comment #10)

(In reply to Jean-Yves Avenard [:jya] from comment #7)

I don't see how that patch could cause the issue mentioned.

A Gtx 750 Ti doesn't have a VP9 hardware accelerator, WMF VPX will only be used if your machine supports HW decoding. And yours doesn't.
So that WMF VPX gets disabled because attempting to load a decoder failed is of no consequence.

I've just tested on a machine with a nvidia 1080ti and I can open over 10 videos, all HW decoded.

I have 1080 SLI with 64 GB RAM, I can confirm the bug, except you need to open a lot of tabs with lots of video,

Every now and then I got media.hardware-video-decoding.failed: false even I reset it,

I think this is cause by a memory leak in GPU process

Could please you use ["wmfvpxvideoCrashGuard" - "Reset on Next Restart"] feature in about:support/Troubleshooting Information in Graphics section in "Crash Guard Disabled Features" sub-section, next redo STR and in end attach your about:support here?

Comment 15

[Jean-Yves Avenard [:jya]]

"media.hardware-video-decoding.failed" pref has nothing to do with bug 1521370. This pref is set on first startup ; it will attempt to decode a very short videos and see if it gets the proper image. If not HW video decoding will be disabled until either one of the following change:

• FF version
• Graphic cards
• Drivers

This pref will always be modified and will always be listed in about:support modified preferences.

It's noisy annoying nuisance masking real errors and warnings when I'm debugging other things. It should be disabled, on non supporting GPUs like mine as you wrote, which doesn't even support WMF VPX decoder and VP8 and VP9 hardware accelerated decoding.

The only way to know if your video card doesnt' support VP9 HW decoding is by loading the WMF VPX framework and creating a decoder.
Now why the entry about the crashguard would be added multiple times is a gfx issue. Re-assigning to the right component.

Comment 16

[Ryan VanderMeulen [:RyanVM]]

Too late for 69 at this point with RC coming next week.

Comment 17

[Sotaro Ikeda [:sotaro]]

I could reproduce the problem on 70.

Comment 18

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

@ Jean-Yves Avenard [:jya] - Thank you yet again very much for very detailed information.

Updating summary to better corresponding about what this bug is about after very detailed explanation by Jean-Yves Avenard [:jya].

Comment 19

[Pascal Chevrel:pascalc (PTO until April 26)]

Sotaro, could you assign a priority to this bug please? Thanks.

Comment 20

[Tascorla]

Attachment: NVIDIA GeForce RTX 2070.txt

It happens on my RTX 2070 as well. Either just browsing the internet or while watching a video, H264 or VP9, it doesn't seem to matter.

The issue started when I changed my GFX card to the RTX2070. I had never had this issue with my previous graphic card, it's a RX 480.

Comment 21

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

Too late for a fix in 70 but we could still take a patch for 71/72.

Comment 22

[Jeny]

Here what I noticed, in my case anyway.

the crash will always happen when I open multiple tabs fast, for example tabs saved in a bookmark folders, or click with control multiple links fast, regardless the tabs have video or not.

The crash has something to to with tab openings even if no video present, I hope this will help.

Usually they are up to to 11 errors, no they were only 3, I will paste them anyway

(#0) Error: WMF VPX decoder just crashed; hardware video will be disabled.
(#1): CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.
(#2) Error: WMF VPX video decoding is disabled due to a previous crash.
(#3): CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.

Comment 23

[Andrew]

(In reply to Jeny from comment #22)

Here what I noticed, in my case anyway.

the crash will always happen when I open multiple tabs fast, for example tabs saved in a bookmark folders, or click with control multiple links fast, regardless the tabs have video or not.

The crash has something to to with tab openings even if no video present, I hope this will help.

Usually they are up to to 11 errors, no they were only 3, I will paste them anyway

(#0) Error: WMF VPX decoder just crashed; hardware video will be disabled.
(#1): CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.
(#2) Error: WMF VPX video decoding is disabled due to a previous crash.
(#3): CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.

I have exactly how you wrote. This bug is almost a year old.

Comment 24

[Andrew]

Developers if you can’t fix the error, then at least give users the option to disable crash guard, we don’t have any errors from which crash guard would save us, we are inconvenient because you created the errors yourself.

Comment 25

[Jeff Muizelaar [:jrmuizel]]

This seems to be a media playback bug and not a graphics one.

Comment 26

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #25)

This seems to be a media playback bug and not a graphics one.

FYI - It was changed from " Audio/Video: Playback" to "Graphics" by:

(Jean-Yves Avenard [:jya] in comment #15)

[...] Now why the entry about the crashguard would be added multiple times is a gfx issue. Re-assigning to the right component.

Comment 27

[Jeff Muizelaar [:jrmuizel]]

Jean-Yves, can you elaborate on what you think needs fixing here?

Comment 28

[Jean-Yves Avenard [:jya]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #27)

Jean-Yves, can you elaborate on what you think needs fixing here?

It's not a media one, the issue is in the CrashGuard class which logs everything on the about:support page.

CrashGuard is a gfx component.

It shouldn't log every time, maybe once per session only

Comment 29

[Jeny]

today for the first time I got a new error message, i wonder if it has anything with this bog? ff 70.0.1

(#0) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0
(#1) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(612,21117) Code: 0x80070057 format 0
(#2) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0
(#3) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(612,21117) Code: 0x80070057 format 0
(#4) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0
(#5) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(612,21117) Code: 0x80070057 format 0
(#6) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0
(#7) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(612,21117) Code: 0x80070057 format 0
(#8) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0
(#9) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(612,21117) Code: 0x80070057 format 0
(#10) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0
(#11) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(612,21117) Code: 0x80070057 format 0
(#12) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0
(#13) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(612,21117) Code: 0x80070057 format 0
(#14) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0
(#15) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(612,21117) Code: 0x80070057 format 0

Comment 30

[Pascal Chevrel:pascalc (PTO until April 26)]

We have shipped our last beta for 71 and this is an unfixed P3, so this is wontfix for 71.

Comment 31

[Sotaro Ikeda [:sotaro]]

Attachment: temporal patch - Add log

The patch was used to know the place of the crash.

Comment 32

[Sotaro Ikeda [:sotaro]]

Attachment: logout with Attachment 9112211

Took a log with the STR in comment 0 and with Attachment 9112211 [details] [diff]. From the log, it seemed that the crash happened at CanCreateMFTDecoder() in WMFDecoderModule::Init().

Comment 33

[Sotaro Ikeda [:sotaro]]

From Comment 32, the crash seems to related to media.

Comment 34

[Sotaro Ikeda [:sotaro]]

(In reply to Jeny from comment #29)

today for the first time I got a new error message, i wonder if it has anything with this bog? ff 70.0.1

(#0) CP+[GFX1-]: [D2D1.1] 4CreateBitmap failure Size(75,22017) Code: 0x80070057 format 0

I seems like a different bug that might be related to D2D1.1.

Comment 35

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

Jean-Yves, does comment 32 shed any light on this crash?

Comment 36

[master.engineer]

I have this issue since a long time, good it's worked on. I have an AMD 5700XT which is a Navi card and supports VP9 acceleration i can test your fix. Which build should i use?

Comment 37

[Jean-Yves Avenard [:jya]]

(In reply to Liz Henry (:lizzard) from comment #35)

Jean-Yves, does comment 32 shed any light on this crash?

the crash occurs when we can't load the WMF VP9 decoder. it crashes. There was an issue with some Windows N edition with an out of sync Windows Media Feature Pack.

If they are running windows N edition (used in Europe and Korea), I suggest that they update windows to the latest version and install the latest version of the Windows Media Feature Pack (for information about it: https://support.microsoft.com/en-us/help/4516397/media-feature-pack-for-windows-10-n-november-2019)

Comment 38

[Jean-Yves Avenard [:jya]]

(In reply to master.engineer from comment #36)

I have this issue since a long time, good it's worked on. I have an AMD 5700XT which is a Navi card and supports VP9 acceleration i can test your fix. Which build should i use?

Not according to AMD technical page: https://www.amd.com/en/products/graphics/amd-radeon-rx-5700-xt Only H264 and H265.
AFAIK, only AMD iGPU (used on laptop) have VP9 decoders.

AMD used to have some kind of hybrid VP HW accelerator using OpenCL, it was very inefficient and had a few bugs. It got removed in recent drivers.

Edit: confirmed by AMD that all Navi based card can do VP9, though only 24fps for 8K.

Comment 39

[master.engineer]

(In reply to Jean-Yves Avenard [:jya] from comment #38)

(In reply to master.engineer from comment #36)

I have this issue since a long time, good it's worked on. I have an AMD 5700XT which is a Navi card and supports VP9 acceleration i can test your fix. Which build should i use?

Not according to AMD technical page: https://www.amd.com/en/products/graphics/amd-radeon-rx-5700-xt Only H264 and H265.
AFAIK, only AMD iGPU (used on laptop) have VP9 decoders.

AMD used to have some kind of hybrid VP HW accelerator using OpenCL, it was very inefficient and had a few bugs. It got removed in recent drivers.

No you are definitely wrong:
https://imgur.com/a/l3cDzx6

Supported Technologies: Radeon Media Engine
https://www.techpowerup.com/256481/amd-navi-radeon-display-engine-and-multimedia-engine-detailed

Plus it's working i can see it under task manager video decode usage;)

Problem is this bug in Firefox.

Comment 40

[master.engineer]

(In reply to Jean-Yves Avenard [:jya] from comment #38)

(In reply to master.engineer from comment #36)

I have this issue since a long time, good it's worked on. I have an AMD 5700XT which is a Navi card and supports VP9 acceleration i can test your fix. Which build should i use?

Not according to AMD technical page: https://www.amd.com/en/products/graphics/amd-radeon-rx-5700-xt Only H264 and H265.
AFAIK, only AMD iGPU (used on laptop) have VP9 decoders.

AMD used to have some kind of hybrid VP HW accelerator using OpenCL, it was very inefficient and had a few bugs. It got removed in recent drivers.

Checked a bit deeper, you are half-wrong, VP9 was indeed dropped but on Polaris/Vega cards that used the hybrid decoder. Navi doesn't use that.

Comment 41

[Andrew]

I have 2070s and 1060 (they are 100% hw accelerated vp9) on 3 pc. I get this error and the disabled hardware acceleration on windows 10 pro is not "N". HW accererated always works on other browsers (edge/opera/chrome). The error occurred a year ago. Before that, everything worked fine.

Comment 42

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

(In reply to Jean-Yves Avenard [:jya] from comment #37)

(In reply to Liz Henry (:lizzard) from comment #35)

Jean-Yves, does comment 32 shed any light on this crash?

the crash occurs when we can't load the WMF VP9 decoder. it crashes. There was an issue with some Windows N edition with an out of sync Windows Media Feature Pack.

If they are running windows N edition (used in Europe and Korea), I suggest that they update windows to the latest version and install the latest version of the Windows Media Feature Pack (for information about it: https://support.microsoft.com/en-us/help/4516397/media-feature-pack-for-windows-10-n-november-2019)

FYI - I'm using Windows 7 Professional (no N edition) with nearly all latest patches (cautiously and carefully selected manually to not install all and any spyware "botnet" patches [aka diagnostics, telemetry and data gathering, etc.]).

Comment 43

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

This is log from "Enumerate database" in "Media Capabilities" sub-section in "Media" section in about:support:

video/av1:
ResolutionLevel2-FrameRateLevel1-8bit: 4100
ResolutionLevel2-FrameRateLevel2-8bit: 2100
ResolutionLevel3-FrameRateLevel2-8bit: 1100
video/vp8:
ResolutionLevel0-FrameRateLevel0-8bit: 2098
ResolutionLevel0-FrameRateLevel2-8bit: 20098
ResolutionLevel1-FrameRateLevel0-8bit: 19097
ResolutionLevel1-FrameRateLevel1-8bit: 20100
ResolutionLevel1-FrameRateLevel2-8bit: 20100
ResolutionLevel1-FrameRateLevel4-8bit: 3100
ResolutionLevel2-FrameRateLevel0-8bit: 20098
ResolutionLevel2-FrameRateLevel1-8bit: 20100
ResolutionLevel2-FrameRateLevel2-8bit: 20099
ResolutionLevel2-FrameRateLevel3-8bit: 2099
ResolutionLevel2-FrameRateLevel4-8bit: 12099
ResolutionLevel3-FrameRateLevel0-8bit: 8097
ResolutionLevel3-FrameRateLevel1-8bit: 20100
ResolutionLevel3-FrameRateLevel2-8bit: 20099
ResolutionLevel3-FrameRateLevel4-8bit: 20100
ResolutionLevel4-FrameRateLevel0-8bit: 20098
ResolutionLevel4-FrameRateLevel1-8bit: 20100
ResolutionLevel4-FrameRateLevel2-8bit: 20100
ResolutionLevel4-FrameRateLevel3-8bit: 7099
ResolutionLevel4-FrameRateLevel4-8bit: 20100
ResolutionLevel5-FrameRateLevel0-8bit: 12097
ResolutionLevel5-FrameRateLevel1-8bit: 20100
ResolutionLevel5-FrameRateLevel2-8bit: 20099
ResolutionLevel5-FrameRateLevel3-8bit: 20088
ResolutionLevel5-FrameRateLevel4-8bit: 20084
ResolutionLevel6-FrameRateLevel1-8bit: 2100
video/vp9:
ResolutionLevel0-FrameRateLevel0-8bit: 5100
ResolutionLevel0-FrameRateLevel1-8bit: 3093
ResolutionLevel0-FrameRateLevel2-8bit: 2100
ResolutionLevel1-FrameRateLevel0-8bit: 20100
ResolutionLevel1-FrameRateLevel1-8bit: 20100
ResolutionLevel1-FrameRateLevel2-8bit: 20100
ResolutionLevel1-FrameRateLevel4-8bit: 3098
ResolutionLevel2-FrameRateLevel1-8bit: 20100
ResolutionLevel2-FrameRateLevel2-8bit: 20100
ResolutionLevel2-FrameRateLevel4-8bit: 1100
ResolutionLevel3-FrameRateLevel0-8bit: 5100
ResolutionLevel3-FrameRateLevel1-8bit: 20100
ResolutionLevel3-FrameRateLevel2-8bit: 20100
ResolutionLevel4-FrameRateLevel0-8bit: 2100
ResolutionLevel4-FrameRateLevel1-8bit: 20100
ResolutionLevel4-FrameRateLevel2-8bit: 20100
ResolutionLevel4-FrameRateLevel3-8bit: 7099
ResolutionLevel4-FrameRateLevel4-8bit: 20100
ResolutionLevel5-FrameRateLevel0-8bit: 1100
ResolutionLevel5-FrameRateLevel1-8bit: 20100
ResolutionLevel5-FrameRateLevel2-8bit: 20100
ResolutionLevel5-FrameRateLevel3-8bit: 3099
ResolutionLevel5-FrameRateLevel4-8bit: 20094
video/avc:
ResolutionLevel0-FrameRateLevel0-8bit: 20100
ResolutionLevel0-FrameRateLevel1-8bit: 20098
ResolutionLevel0-FrameRateLevel2-8bit: 20100
ResolutionLevel0-FrameRateLevel3-8bit: 20100
ResolutionLevel0-FrameRateLevel4-8bit: 7096
ResolutionLevel1-FrameRateLevel0-8bit: 20100
ResolutionLevel1-FrameRateLevel1-8bit: 20100
ResolutionLevel1-FrameRateLevel2-8bit: 20100
ResolutionLevel1-FrameRateLevel3-8bit: 11100
ResolutionLevel1-FrameRateLevel4-8bit: 20094
ResolutionLevel2-FrameRateLevel0-8bit: 20100
ResolutionLevel2-FrameRateLevel1-8bit: 20100
ResolutionLevel2-FrameRateLevel2-8bit: 20100
ResolutionLevel2-FrameRateLevel3-8bit: 2100
ResolutionLevel2-FrameRateLevel4-8bit: 19095
ResolutionLevel3-FrameRateLevel0-8bit: 6100
ResolutionLevel3-FrameRateLevel1-8bit: 20100
ResolutionLevel3-FrameRateLevel2-8bit: 20100
ResolutionLevel3-FrameRateLevel3-8bit: 1100
ResolutionLevel3-FrameRateLevel4-8bit: 9096
ResolutionLevel4-FrameRateLevel0-8bit: 6100
ResolutionLevel4-FrameRateLevel1-8bit: 20100
ResolutionLevel4-FrameRateLevel2-8bit: 20100
ResolutionLevel4-FrameRateLevel3-8bit: 9098
ResolutionLevel4-FrameRateLevel4-8bit: 12099
ResolutionLevel5-FrameRateLevel0-8bit: 2100
ResolutionLevel5-FrameRateLevel1-8bit: 20100
ResolutionLevel5-FrameRateLevel2-8bit: 20099
ResolutionLevel5-FrameRateLevel3-8bit: 1100
ResolutionLevel5-FrameRateLevel4-8bit: 18088
video/theora:

Comment 44

[Jens Stutte [:jstutte]]

Jean Yves, do we have something actionable here? Thank you!

Comment 45

[master.engineer]

Is a fix coming along or should we wait until VP9 gets discontinued?

Comment 46

[Jeny]

I am not sure if I am allow to say that, but lately Firefox has become so slow, memory leaks (GPU process will reach 5GB and keep increasing) and high CPU usage when opening YouTube videos, I have to restart Firefox every now and then, I am not sure if it is this bug or not, but I installed Edge chromium just to avoid open YouTube videos on Firefox.

I have been using Edge for a while now only for YouTube, and it is super fast, no memory leaks, no lagging.

I have reached a point where I will compromise the privacy on the account of performance. Firefox still my main browser, except for the videos, and YouTube.

I hope if you can't find a solution soon, you try use chromium for it is super fast.

Comment 47

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

(In reply to Jeny from comment #46)

I am not sure if I am allow to say that, but lately Firefox has become so slow, memory leaks (GPU process will reach 5GB and keep increasing) and high CPU usage when opening YouTube videos, I have to restart Firefox every now and then, I am not sure if it is this bug or not, but I installed Edge chromium just to avoid open YouTube videos on Firefox.

I have been using Edge for a while now only for YouTube, and it is super fast, no memory leaks, no lagging.

I have reached a point where I will compromise the privacy on the account of performance. Firefox still my main browser, except for the videos, and YouTube.

I hope if you can't find a solution soon, you try use chromium for it is super fast.

It's probably very long standing issue in bug #1386177.

Comment 48

[Jeny]

(In reply to Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.) from comment #47)

(In reply to Jeny from comment #46)

I am not sure if I am allow to say that, but lately Firefox has become so slow, memory leaks (GPU process will reach 5GB and keep increasing) and high CPU usage when opening YouTube videos, I have to restart Firefox every now and then, I am not sure if it is this bug or not, but I installed Edge chromium just to avoid open YouTube videos on Firefox.

I have been using Edge for a while now only for YouTube, and it is super fast, no memory leaks, no lagging.

I have reached a point where I will compromise the privacy on the account of performance. Firefox still my main browser, except for the videos, and YouTube.

I hope if you can't find a solution soon, you try use chromium for it is super fast.

It's probably very long standing issue in bug #1386177.

it is for sure this https://bugzilla.mozilla.org/show_bug.cgi?id=1560728 but it wont open RDD process with high usage when restart, I think it maybe related to this bug,

I hope it ok to say that I think it is time to try some chromium codes, as it has been a long time now, and every updates it become just slower

Comment 49

[Jeny]

(In reply to Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.) from comment #47)

(In reply to Jeny from comment #46)

I am not sure if I am allow to say that, but lately Firefox has become so slow, memory leaks (GPU process will reach 5GB and keep increasing) and high CPU usage when opening YouTube videos, I have to restart Firefox every now and then, I am not sure if it is this bug or not, but I installed Edge chromium just to avoid open YouTube videos on Firefox.

I have been using Edge for a while now only for YouTube, and it is super fast, no memory leaks, no lagging.

I have reached a point where I will compromise the privacy on the account of performance. Firefox still my main browser, except for the videos, and YouTube.

I hope if you can't find a solution soon, you try use chromium for it is super fast.

It's probably very long standing issue in bug #1386177.

Now I started noticing this issue with HTML videos, I was talking about Youtube videos only, but now you mentioned it, I am suffering from this bug you mentioned ....

Comment 50

[Jean-Yves Avenard [:jya]]

not sure what I can do here to help.

Comment 51

[master.engineer]

Is a fix ever coming for this? This is like super basic end user functionality. Even APU's have VP9 acceleration since October 2017...

Comment 52

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

This bug is only about fixing massive amount of WMF VPX errors and warnings in about:support/Troubleshooting Information in Graphics section in Failure Log sub-section.

Also changing again Component to Graphics per:
(Jean-Yves Avenard [:jya] wrote in comment #15)

[...] Now why the entry about the crashguard would be added multiple times is a gfx issue. Re-assigning to the right component.

Comment 53

[master.engineer]

It's not that it doesn't work but it crashes with this error after a few tabs requiring reset on next restart which will crash again either way. If there is a different bug regarding why it crashes let me know.

Comment 54

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Oh, true, it happens also on GPUs with VP9 hardware accelerator support (comment #10, comment #20, comment #36, comment #41).

Comment 55

[master.engineer]

Guys, if you search VP9 it's a total cluster of bugs. Is there a specific one that is responsible for the actual reason VP9 acceleration crashes? We need to fix that IMMEDIATELY. There lot's of people on reddit complaining VP9 doesn't work when in fact it does work but crashes..then you wonder why people move to Chrome..

Comment 56

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

I'm trying to reproduce the crash that leads to the log spam. I'm interested in what is causing the initial Error: WMF VPX decoder just crashed; hardware video will be disabled. message, less so in WMF VPX video decoding is disabled due to a previous crash. I've tried opening many videos in different tabs and windows, but so far no luck.

For people encountering the issue I would also be interested in the values of your gfx.crash-guard.status.wmfvpxvideo and gfx.crash-guard.wmfvpxvideo.* prefs.

Edit: it would also be useful to know the Windows version reported under System Information as different versions of Windows may have different decoder implementations.

Comment 57

[Asif Youssuff]

Noticed this on my machine.

gfx.crash-guard.status.wmfvpxvideo 3
gfx.crash-guard.wmfvpxvideo.appVersion 81.0a1
gfx.crash-guard.wmfvpxvideo.deviceID 0x3185
gfx.crash-guard.wmfvpxvideo.driverVersion 23.20.16.4936

Oh, I guess this is not useful because I didn't see it crash.

Comment 58

[master.engineer]

Firefox v.200 -> Still has VP9 acceleration crashes. It's been 1 year already. RDNA 2 cards are coming out in 1 month. And it will break for those too.

Comment 59

[master.engineer]

(In reply to master.engineer from comment #58)

Firefox v.200 -> Still has VP9 acceleration crashes. It's been 1 year already. RDNA 2 cards are coming out in 1 month. And it will break for those too.

I think an option to force the acceleration ON continuously it's a band-aid fix that can used.

Comment 60

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

(In reply to master.engineer from comment #59)

(In reply to master.engineer from comment #58)

Firefox v.200 -> Still has VP9 acceleration crashes. It's been 1 year already. RDNA 2 cards are coming out in 1 month. And it will break for those too.

I think an option to force the acceleration ON continuously it's a band-aid fix that can used.

This option exist in some way, but it doesn't work properly (bug #1253923).

Comment 61

[master.engineer]

It's almost 2021 and this has yet to be fixed.
I'll say it again. How are you going to bring people away from Chrome when such basic stuff doesn't work?
I happen to have a 3950X so software decoding is not a big deal for me.
But for the majority of people with shitty laptops they need hardware decoding.
If they see the video plays without stuttering on Chrome of course they are going to stay there.

(In reply to Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.) from comment #60)

(In reply to master.engineer from comment #59)

(In reply to master.engineer from comment #58)

Firefox v.200 -> Still has VP9 acceleration crashes. It's been 1 year already. RDNA 2 cards are coming out in 1 month. And it will break for those too.

I think an option to force the acceleration ON continuously it's a band-aid fix that can used.

This option exist in some way, but it doesn't work properly (bug #1253923).

Thanks i didn't know that. It's almost comical how that doesn't work either...reported 5 years ago...WTF

Comment 62

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

(In reply to master.engineer from comment #61)

snip

If you're encountering this it would be useful if you could provide the information mentioned in comment 56, which would help further diagnose where the issue is taking place.

Comment 63

[master.engineer]

(In reply to Bryce Seager van Dyk (:bryce) from comment #62)

(In reply to master.engineer from comment #61)

snip

If you're encountering this it would be useful if you could provide the information mentioned in comment 56, which would help further diagnose where the issue is taking place.

gfx.crash-guard.status.wmfvpxvideo = 0 (note i had reset it manually from about:support before my last reply, so it's should have been a different value)
gfx.crash-guard.wmfvpxvideo.appVersion = 81.0.1
gfx.crash-guard.wmfvpxvideo.deviceID = 0x731f
gfx.crash-guard.wmfvpxvideo.driverVersion = 27.20.12031.1000

This is AMD 5700XT with W10 2004 and 20.9.2 drivers. But that's not the issue here since it's happening with older versions of the drivers and W10.
I don't know if the same issue happens with Nvidia cards supporting VP9 decoding but at least for AMD any GPU and APU using VCN should be affected: https://en.wikipedia.org/wiki/Video_Core_Next

As for reproducing, unless there is a way to put an active indicator to notify me of the crash i will have no way of knowing if something triggers it.
From what i understand you need to have lots of tabs and/or windows to trigger it. What many people say are a "lot" for me are a "few".

Still simplest temp solution is to have the WMF VPX reset on every start of Firefox. It's not rocket science.

Comment 64

[Andrew]

(In reply to master.engineer from comment #63)

(In reply to Bryce Seager van Dyk (:bryce) from comment #62)

(In reply to master.engineer from comment #61)

snip

If you're encountering this it would be useful if you could provide the information mentioned in comment 56, which would help further diagnose where the issue is taking place.

gfx.crash-guard.status.wmfvpxvideo = 0 (note i had reset it manually from about:support before my last reply, so it's should have been a different value)
gfx.crash-guard.wmfvpxvideo.appVersion = 81.0.1
gfx.crash-guard.wmfvpxvideo.deviceID = 0x731f
gfx.crash-guard.wmfvpxvideo.driverVersion = 27.20.12031.1000

This is AMD 5700XT with W10 2004 and 20.9.2 drivers. But that's not the issue here since it's happening with older versions of the drivers and W10.
I don't know if the same issue happens with Nvidia cards supporting VP9 decoding but at least for AMD any GPU and APU using VCN should be affected: https://en.wikipedia.org/wiki/Video_Core_Next

As for reproducing, unless there is a way to put an active indicator to notify me of the crash i will have no way of knowing if something triggers it.
From what i understand you need to have lots of tabs and/or windows to trigger it. What many people say are a "lot" for me are a "few".

Still simplest temp solution is to have the WMF VPX reset on every start of Firefox. It's not rocket science.

And nvidia is affected. 1060, 1080, 2700s

Comment 65

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

We've added markers to the Firefox profiler to help detect failures in Window's decoding pipeline. If you're encountering this issue it would be helpful if you could gather a profile using the 'Media' config on Firefox Nightly from after 2020-10-20, then share it and paste the link here.

Comment 66

[master.engineer]

(In reply to Bryce Seager van Dyk (:bryce) from comment #65)

We've added markers to the Firefox profiler to help detect failures in Window's decoding pipeline. If you're encountering this issue it would be helpful if you could gather a profile using the 'Media' config on Firefox Nightly from after 2020-10-20, then share it and paste the link here.

Since the issue happens randomly how do you recommend that i use the tool to catch the crash happening?

Comment 67

[master.engineer]

Also which settings from the dropdown menu to use?(In reply to Bryce Seager van Dyk (:bryce) from comment #65)

We've added markers to the Firefox profiler to help detect failures in Window's decoding pipeline. If you're encountering this issue it would be helpful if you could gather a profile using the 'Media' config on Firefox Nightly from after 2020-10-20, then share it and paste the link here.

Also which settings from the dropdown menu to use?

Comment 68

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

(In reply to master.engineer from comment #66)

Since the issue happens randomly how do you recommend that i use the tool to catch the crash happening?

I can't reproduce this myself. The bug suggests that opening several tabs containing video leads to the issue, is this not triggering the issue in your case? You could leave the profiler running and once you notice the issue taking place, then capture and share the profile. However, if it takes hours to repro then the profiles may grow to large to upload.

Also which settings from the dropdown menu to use?

There should be a 'Media' preset available in the profiler, that would be the best one to use for this.

Comment 69

[stanislav.khromov]

👋 I've been having this bug for as long as I can remember and it's still present on Firefox 86 and 100% reproducible. In my case it seems like opening 3-4 YouTube tabs simultaneously causes the following crash guard log in about:support.

WMF VPX Video Decoder	
Failure Log
(#0) Error	WMF VPX decoder just crashed; hardware video will be disabled.
(#1) 	CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.
(#2) Error	WMF VPX video decoding is disabled due to a previous crash.
(#3) 	CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.
(#4) Error	WMF VPX video decoding is disabled due to a previous crash.
(#5) 	CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.
(#6) Error	WMF VPX video decoding is disabled due to a previous crash.
(#7) 	CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.
(#8) Error	WMF VPX video decoding is disabled due to a previous crash.
(#9) 	CP+[GFX1-]: WMF VPX video decoding is disabled due to a previous crash.

The problem appears only with VP9 and AV1-encoded videos and disables GPU acceleration for these codecs. Curiously, h264 videos continue to be hardware accelerated. (I am using the "enhanced-h264ify" add-on to lock the codec to h264 on YouTube). I am using an Nvidia RTX 2070 GPU on Windows 10.

Is there any kind of log or dump I can submit to help with the debugging effort?

Comment 70

[BugBot [:suhaib / :marco/ :calixte]]

In the process of migrating remaining bugs to the new severity system, the severity for this bug cannot be automatically determined. Please retriage this bug using the new severity system.

Comment 71

[BugBot [:suhaib / :marco/ :calixte]]

The severity field is not set for this bug.
:bhood, could you have a look please?

For more information, please visit auto_nag documentation.

Comment 72

[Bob Hood [:bhood]]

Can you still see this behavior in the current release (Fx111)?

Comment 73

[Jason Salhab]

Attachment: same problem

Thanks from hontentbaber1982.

Comment 74

[master.engineer]

(In reply to Bob Hood [:bhood] from comment #72)

Can you still see this behavior in the current release (Fx111)?

I'll need to test and report back

Comment 75

[BugBot [:suhaib / :marco/ :calixte]]

Redirect a needinfo that is pending on an inactive user to the triage owner.
:bhood, since the bug has recent activity, could you please find another way to get the information or close the bug as INCOMPLETE if it is not actionable?

For more information, please visit BugBot documentation.