1112331 - No video displayed on Acer Iconia Win8 tablet with OMTC, Cairo, non-E10S

Status: RESOLVED FIXED

(Core :: Graphics: Layers, defect, P1)

Description

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

I have a low end Acer Iconia Windows 8.0 32bit tablet. In nightly I don't get video frames rendering unless I disable layers.offmainthreadcomposition.enabled.

The graphics driver is already blacklisted.

CPU: Atom Z2760 @ 1.8Ghz

Application Basics
------------------

Name: Firefox
Version: 37.0a1
User Agent: Mozilla/5.0 (Windows NT 6.2; rv:37.0) Gecko/20100101 Firefox/37.0
Multiprocess Windows: 0/1

Crash Reports for the Last 3 Days
---------------------------------

Report ID: bp-cc2f20e0-3421-4ca7-bd56-6f5002141216
Submitted: 21 minutes ago

Report ID: bp-ca904261-55c0-4536-ad69-b05ab2141216
Submitted: 22 minutes ago

Report ID: bp-a3badad9-17ee-46ed-8026-70d952141216
Submitted: 22 minutes ago

Report ID: bp-4a4f4d63-c9d4-43c0-9d73-b49ec2141216
Submitted: 22 minutes ago

All Crash Reports

Extensions
----------

Graphics
--------

Adapter Description: Intel(R) Graphics Media Accelerator
Adapter Drivers: igdumd32
Adapter RAM: 0
Device ID: 0x08cf
Direct2D Enabled: Blocked for your graphics driver version.
DirectWrite Enabled: false (6.2.9200.16581)
Driver Date: 11-2-2012
Driver Version: 9.14.3.1102
GPU #2 Active: false
GPU Accelerated Windows: 0/1 Basic Blocked for your graphics driver version.
Subsys ID: 074b1025
Vendor ID: 0x8086
WebGL Renderer: Blocked for your graphics driver version.
windowLayerManagerRemote: false
AzureCanvasBackend: skia
AzureContentBackend: cairo
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0

Important Modified Preferences
------------------------------

browser.cache.disk.capacity: 358400
browser.cache.disk.smart_size.first_run: false
browser.cache.frecency_experiment: 2
browser.places.smartBookmarksVersion: 7
browser.sessionstore.upgradeBackup.latestBuildID: 20141216030203
browser.startup.homepage_override.buildID: 20141216030203
browser.startup.homepage_override.mstone: 37.0a1
browser.tabs.remote.autostart.disabled-because-using-a11y: true
dom.mozApps.used: true
extensions.lastAppVersion: 37.0a1
layers.offmainthreadcomposition.enabled: false
media.gmp-gmpopenh264.lastUpdate: 1418763555
media.gmp-gmpopenh264.path: C:\Users\Chris\AppData\Roaming\Mozilla\Firefox\Profiles\lmzaxrb7.firetruck\gmp-gmpopenh264
media.gmp-gmpopenh264.version: 1.1
media.gmp-manager.lastCheck: 1418763552
network.cookie.prefsMigrated: true
places.database.lastMaintenance: 1418763685
places.history.expiration.transient_current_max_pages: 52624
plugin.disable_full_page_plugin_for_types: application/pdf
plugin.importedState: true
privacy.sanitize.migrateFx3Prefs: true
storage.vacuum.last.index: 0
storage.vacuum.last.places.sqlite: 1418763685

Important Locked Preferences
----------------------------

JavaScript
----------

Incremental GC: true

Accessibility
-------------

Activated: true
Prevent Accessibility: 0

Library Versions
----------------

NSPR
Expected minimum version: 4.10.7
Version in use: 4.10.7

NSS
Expected minimum version: 3.17.3 Basic ECC
Version in use: 3.17.3 Basic ECC

NSSSMIME
Expected minimum version: 3.17.3 Basic ECC
Version in use: 3.17.3 Basic ECC

NSSSSL
Expected minimum version: 3.17.3 Basic ECC
Version in use: 3.17.3 Basic ECC

NSSUTIL
Expected minimum version: 3.17.3
Version in use: 3.17.3

Experimental Features
---------------------

Comment 1

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

Milan - can you get someone to look into this? A year ago this graphics hardware wasn't blacklisted and it played full screen video beautifully.

Comment 2

[Milan Sreckovic [:milan] (needinfo for best results)]

Don't have the device, but maybe we can do something just based on the regression range - can somebody get that?

Comment 3

[Marcia Knous [:marcia]]

ni on marcia to check which Acer Iconia tablet she is in the QA lab.

Comment 4

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

The one I repro this on is a W510, with atom z2760 CPU.

Comment 5

[Marcia Knous [:marcia]]

(In reply to Marcia Knous [:marcia - use needinfo] from comment #3)
> ni on marcia to check which Acer Iconia tablet she is in the QA lab.

My machine in the lab is Acer Iconia W700 which has an Intel based processor.

Comment 6

[Matt Woodrow (:mattwoodrow)]

(In reply to Milan Sreckovic [:milan] from comment #2)
> Don't have the device, but maybe we can do something just based on the
> regression range - can somebody get that?

This will almost certainly just show the changeset where OMTC was enabled.

Comment 7

[Milan Sreckovic [:milan] (needinfo for best results)]

(In reply to Matt Woodrow (:mattwoodrow) from comment #6)
> (In reply to Milan Sreckovic [:milan] from comment #2)
> > Don't have the device, but maybe we can do something just based on the
> > regression range - can somebody get that?
> 
> This will almost certainly just show the changeset where OMTC was enabled.

I imagine you're right, we should just try chasing it on the solution side.

David, can you borrow this device from Marcia and see if you can trace the problem down? We're on a short timeline for MSE, so it'd be good to find out what's going on quickly.

Comment 8

[David Anderson [:dvander] - inactive, e-mail if emergency]

Can repro pretty easily on the device. Will take a look.

Comment 9

[David Anderson [:dvander] - inactive, e-mail if emergency]

FWIW, seems to work fine if I disable D3D11 and switch to the software compositor, but that doesn't match up with the original report.

Comment 10

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

The tablet is an Acer Iconia W520.

Comment 11

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

(In reply to Anthony Jones (:kentuckyfriedtakahe, :k17e) from comment #10)
> The tablet is an Acer Iconia W520.

No, our tablet on which the bug was reported is a W510. It has SNID 24700724865 on the side, and Acer's website lists that as a W510.
http://us.acer.com/ac/en/US/content/drivers/-;24700724865;-

Since dvander says he can repro on Marcia's W700, I bet the bug is not limited to only the W510.

Comment 12

[David Anderson [:dvander] - inactive, e-mail if emergency]

My drivers don't seem to be blacklisted though (at least on nightly).

Comment 13

[Milan Sreckovic [:milan] (needinfo for best results)]

David, how are you disabling D3D11?  With layers.prefer-d3d9 pref?

As for the original device being blacklisted - I'm not sure what actually catches it, and I wonder if it's run time test.

Comment 14

[David Anderson [:dvander] - inactive, e-mail if emergency]

(In reply to Milan Sreckovic [:milan] from comment #13)
> David, how are you disabling D3D11?  With layers.prefer-d3d9 pref?
> 
> As for the original device being blacklisted - I'm not sure what actually
> catches it, and I wonder if it's run time test.

I turned layers acceleration off - but turning prefer-d3d9 to true seems to work as well, actually.

Comment 15

[David Anderson [:dvander] - inactive, e-mail if emergency]

And it also looks like on Release/Beta/Aurora these drivers are blacklisted, but not on nightly.

Comment 16

[Milan Sreckovic [:milan] (needinfo for best results)]

(In reply to David Anderson [:dvander] from comment #15)
> And it also looks like on Release/Beta/Aurora these drivers are blacklisted,
> but not on nightly.

Could be related to bug 1097321.

Comment 17

[David Anderson [:dvander] - inactive, e-mail if emergency]

It looks like it wasn't a blacklist change but the fact that we use WARP now. At least, that is the reason for the nightly D3D11 bustage.

Unfortunately, that is going to be a different problem from cpearce's. It looks like the tablet I have has newer drivers (from 2013). Maybe there's a way to downgrade them, but failing that, do we have any more of these lying around?

Comment 18

[David Anderson [:dvander] - inactive, e-mail if emergency]

According to the model number the tablet I've got is a W3-810, and it has Windows 8.1 (not 8.0). Downgrading drivers doesn't appear to be an option.

So to move forward with the original bug we'll need an older device.

Comment 19

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

Matt Woodrow has "won" an Acer Iconia tablet at the Auckland office to take back to his home :-P

Comment 20

[Matt Woodrow (:mattwoodrow)]

With OTMC disabled, then all layers are blacklisted for this device and we end up with BasicLayers.

With OMTC enabled, we're not blacklisted, and get the D3D11 compositor.

In the latter configuration we're getting the hardware accelerated decoder, and we're failing in DXGITextureHostD3D11::OpenSharedShared.

ID3D11Device::OpenSharedResource is returning E_INVALIDARG.

Looks like switching to OMTC lost our blacklist entry, and we probably needed it.

Is there any way to blacklist the decoder only, or do we need to re-blacklist layers for this device?

Comment 21

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

If you set media.windows-media-foundation.use-dxva=false, do you still get no video?

We don't have a way to blacklist specific DXVA decoders; we've never needed it until now. In the past, if the HW decoder was bad we'd usually get a failure to initialize the DXVA path and we'd be able to fallback to software.

What if you force Direct3D9 compositing? Does that work?

Comment 22

[Matt Woodrow (:mattwoodrow)]

Yes, both of those changes (individually) result in working video.

Comment 23

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

What are the implications of us switching to prefer D3D9 on this device? Is it the least amount of work to get this working again?

Another idea would be to set media.windows-media-foundation.use-dxva=false when we detect this failure at runtime.

Or we could try to be clever and run enough of the pipeline at first start up in order to detect this failure and set the pref, so that the first video we try to play won't fail.

Comment 24

[Matt Woodrow (:mattwoodrow)]

(In reply to David Anderson [:dvander] from comment #17)
> It looks like it wasn't a blacklist change but the fact that we use WARP
> now. At least, that is the reason for the nightly D3D11 bustage.

Looks like this comment was entirely correct for this device too.

Disabing WARP also disables d3d11 (since we're blacklisted), and the problem goes away.

Comment 25

[Matt Woodrow (:mattwoodrow)]

Forcing WARP on my desktop seems to work fine though, so it's not a fundamental issue with DXVA and WARP.

Bas, any ideas why this combination might be broken on some hardware?

Comment 26

[Matt Woodrow (:mattwoodrow)]

Attachment: Disable DXVA when using WARP

We need something to uplift to beta, so I think the easiest thing to do here is to disable DXVA when using a WARP backed compositor.

WARP is a CPU implementation of D3D11, so it's probably faster to use CPU decoding instead of readback anyway.

Comment 27

[Bas Schouten (:bas.schouten)]

Comment on attachment 8552051 [details] [diff] [review]
Disable DXVA when using WARP

Review of attachment 8552051 [details] [diff] [review]:
-----------------------------------------------------------------

I feel this is a good idea.

Comment 28

[Matt Woodrow (:mattwoodrow)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/a98935079990

Comment 29

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/a98935079990

Comment 31

[Ralph Giles (:rillian)]

Comment on attachment 8552051 [details] [diff] [review]
Disable DXVA when using WARP

Approval Request Comment
[Feature/regressing bug #]: MSE
[User impact if declined]: Video fails to play on some Windows devices.
[Describe test coverage new/current, TreeHerder]: Landed on m-c.
[Risks and why]: This touches both compositing and video playback on Windows, but it just changes the code path selected for video in certain configurations. It's a straightforward change and easy to revert. The compositor changes have no effect outside video playback. I don't think there will be any regressions, but any which occur should be obvious.
[String/UUID change made/needed]: None.

Comment 32

[Ralph Giles (:rillian)]

Matt, this doesn't apply cleanly to beta, and I'm not sure how to resolve the conflicts. Would you mind doing a backport?

Comment 33

[Ralph Giles (:rillian)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/3670866ab6a9

Comment 34

[Matt Woodrow (:mattwoodrow)]

Sure thing!

Comment 35

[Matt Woodrow (:mattwoodrow)]

WARP support wasn't added until 37, so 36 isn't affected by this bug :)

Comment 36

[Ralph Giles (:rillian)]

oh! yay.