838731 - Firefox crashes on nVidia 555m when using ANGLE for WebGL

Status: RESOLVED INCOMPLETE

(Core :: Graphics, defect)

Description

[gordonrankin82]

User Agent: Mozilla/5.0 (Windows NT 6.2; WOW64; rv:21.0) Gecko/20130206 Firefox/21.0
Build ID: 20130206031027

Steps to reproduce:

Tried to open any site which initiates a webGL context on a Dell L702x XPS laptop with a nVidia Geforce 555m GPU


Actual results:

Firefox crashes instantly on any WebGL website unless I turn on webgl.prefer-native-gl and webgl-force-enabled.

The WebGL performance increase when using ANGLE is far superior to Firefox's native implementation on this GPU.

Things to note:

There are two GPU's on this laptop.  If I force it to use only the low performance Intel 3000 chipset, all works fine.

The nVidia driver is the latest stable release.  I recall this issue appearing upon release of the 300+ drivers, with version 296.x being the last that worked.

I can confirm that the problem occurs on version of firefox from around 14-15 and probably earlier.  It also occurs on the beta and nightly versions.

Although given the fact that the problem did not occur with much older drivers one would expect the issue to be related to nVidias code, the same drivers work perfectly in Chrome which also uses ANGLE by default.


The crash report is as follows:


AdapterDeviceID: 0x0116
AdapterVendorID: 0x8086
Add-ons: firepicker%40thedarkone:1.4.3,%7B972ce4c6-7e08-4474-a285-3208198ce6fd%7D:21.0a1,firebug%40software.joehewitt.com:1.11.1
AvailablePageFile: 31065550848
AvailablePhysicalMemory: 14329487360
AvailableVirtualMemory: 3350990848
BuildID: 20130206031027
CrashTime: 1360174652
EMCheckCompatibility: true
FramePoisonBase: 00000000f0de0000
FramePoisonSize: 65536
InstallTime: 1360170089
Notes: AdapterVendorID: 0x8086, AdapterDeviceID: 0x0116, AdapterSubsysID: 04b81028, AdapterDriverVersion: 9.17.10.2932
Has dual GPUs. GPU #2: AdapterVendorID2: 0x10de, AdapterDeviceID2: 0x0dcd, AdapterSubsysID2: 04b81028, AdapterDriverVersion2: 9.18.13.1396D2D? D2D+ DWrite? DWrite+ D3D10 Layers? D3D10 Layers+ WebGL? EGL? EGL+ GL Context? GL Context+ WebGL+ 
ProductID: {ec8030f7-c20a-464f-9b0e-13a3a9e97384}
ProductName: Firefox
ReleaseChannel: nightly
SecondsSinceLastCrash: 53
StartupTime: 1360174617
SystemMemoryUsePercentage: 16
Theme: classic/1.0
Throttleable: 1
TotalVirtualMemory: 4294836224
URL: http://www.songbanc.com/profile
Vendor: Mozilla
Version: 21.0a1
Winsock_LSP: MSAFD Tcpip [TCP/IP] : 2 : 1 :  
 MSAFD Tcpip [UDP/IP] : 2 : 2 : %SystemRoot%\system32\mswsock.dll 
 MSAFD Tcpip [RAW/IP] : 2 : 3 :  
 MSAFD Tcpip [TCP/IPv6] : 2 : 1 : %SystemRoot%\system32\mswsock.dll 
 MSAFD Tcpip [UDP/IPv6] : 2 : 2 :  
 MSAFD Tcpip [RAW/IPv6] : 2 : 3 : %SystemRoot%\system32\mswsock.dll 
 RSVP TCPv6 Service Provider : 2 : 1 :  
 RSVP TCP Service Provider : 2 : 1 : %SystemRoot%\system32\mswsock.dll 
 RSVP UDPv6 Service Provider : 2 : 2 :  
 RSVP UDP Service Provider : 2 : 2 : %SystemRoot%\system32\mswsock.dll 
 MSAFD RfComm [Bluetooth] : 2 : 1 : 



And the Graphics section from about:support gives the following information:



  Graphics

        Adapter Description
        Intel(R) HD Graphics 3000

        Adapter Description (GPU #2)
        NVIDIA GeForce GT 555M

        Adapter Drivers
        igdumd64 igd10umd64 igd10umd64 igdumd32 igd10umd32 igd10umd32

        Adapter Drivers (GPU #2)
        nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um

        Adapter RAM
        Unknown

        Adapter RAM (GPU #2)
        3071

        Device ID
        0x0116

        Device ID (GPU #2)
        0x0dcd

        Direct2D Enabled
        true

        DirectWrite Enabled
        true (6.2.9200.16433)

        Driver Date
        12-12-2012

        Driver Date (GPU #2)
        1-25-2013

        Driver Version
        9.17.10.2932

        Driver Version (GPU #2)
        9.18.13.1396

        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 3000)

        AzureCanvasBackend
        direct2d

        AzureContentBackend
        direct2d

        AzureFallbackCanvasBackend
        cairo



This is with no add-ons/extensions installed and still occurs following a reset.


Expected results:

Firefox should not have crashed.

Comment 1

[gordonrankin82]

I have just noticed that about:support shows that the webGL renderer is the intel HD Graphics 3000 GPU despite the fact that in nVidias control panel I am forcing Firefox to use the nVidia GT 555m GPU instead.

Could it be that the GPU's are not switching correctly?  Despite it working perfectly in every other application, including Chrome?

Comment 2

[gordonrankin82]

Also my apologies for not stating before.  This occurs in both windows 7 and windows 8,  The only OSes I have tried.

Comment 3

[Matthias Versen [:Matti]]

That this doesn't happen with other browser doesn't matter if the nvidia code crashes.
The technical details of a crash report are useless for us.
We need the crash ID from about:crashes ->
https://developer.mozilla.org/en/How_to_get_a_stacktrace_for_a_bug_report

Comment 4

[gordonrankin82]

Thanks Matthias.  This is the first bug (?) I have ever reported so my apologies for not knowing the correct etiquette.

Here is the link to the crash report:

https://crash-stats.mozilla.com/report/index/bp-e1a793aa-825a-4aca-946a-138422130206

Comment 5

[Matthias Versen [:Matti]]

> I have ever reported so my apologies for not knowing the correct etiquette.
No problem !
Your bug report is good and I tried to explain what we need.


The crash is deep in the Nvidia driver.
Frame #13 (xul.dll) is the last code from Firefox.

Comment 6

[Scoobidiver (away)]

(In reply to gordonrankin82 from comment #0)
> Although given the fact that the problem did not occur with much older
> drivers one would expect the issue to be related to nVidias code
There are no Firefox components in the stack trace only graphics driver trace so it's indeed a driver bug.
Does it happen in Safe Mode (see https://support.mozilla.org/kb/troubleshoot-firefox-issues-using-safe-mode) that turns off HW acceleration?

Comment 7

[gordonrankin82]

(In reply to Scoobidiver from comment #6)
> (In reply to gordonrankin82 from comment #0)
> > Although given the fact that the problem did not occur with much older
> > drivers one would expect the issue to be related to nVidias code
> There are no Firefox components in the stack trace only graphics driver
> trace so it's indeed a driver bug.
> Does it happen in Safe Mode (see
> https://support.mozilla.org/kb/troubleshoot-firefox-issues-using-safe-mode)
> that turns off HW acceleration?

I've just tried in safe mode and it works... albeit without 2D hardware acceleration.  Although don't take this as gospel yet as I have had brief periods where WebGL didn't insta-crash when using ANGLE anyway, so it may be a fluke.

The second i exit safe mode and turn HW acceleration back on, it crashes.

Comment 8

[gordonrankin82]

(In reply to Matthias Versen [:Matti] from comment #5)
> > I have ever reported so my apologies for not knowing the correct etiquette.
> No problem !
> Your bug report is good and I tried to explain what we need.
> 
> 
> The crash is deep in the Nvidia driver.
> Frame #13 (xul.dll) is the last code from Firefox.

I see, thanks for clearing this up.  Would the correct course of action be for me to file a bug report with nVidia linking to the stack trace?

Comment 9

[Matthias Versen [:Matti]]

Only nvidia can fix the bug and it can be useful for the Nvidia guys if you mention this bug report and the stacktrace in your report. I don't know if the stacktrace helps because we have no debug symbols for nvidia drivers (closed source) in the stacktrace.

You should include an example page that is causing this crash in your report. 
Every bug report needs exact steps to reproduce and that includes an example URL.

Scoobidiver: Do you know if we have nvidia contacts ?

Comment 10

[gordonrankin82]

> You should include an example page that is causing this crash in your
> report. 

I have filed a bug report with Nvidia complete with a testcase URL as you suggested.  Though this is simply with their customer feedback/issue report system.

Any contacts you guys have would likely push the issue further down the chain.

Comment 11

[gordonrankin82]

Just an interesting note.  When forced into native webGL about:support reports NVIDIA Corporation -- GeForce GT 555M/PCIe/SSE2 as the WebGL renderer, but with it set to default/ANGLE it reports the intel 3000 GPU as the renderer.

Comment 12

[Wayne Mery (:wsmwk)]

Gordon, did you get a response from NVIDIA and do you still see this problem?

Comment 13

[gordonrankin82]

My apologies for not getting back sooner. I'm not sure how I missed this. Yes the issue appears to be fixed. I have no idea how or why. But it no longer occurs in recent builds. If my recollection serves me correctly the improvement directly followed a new driver release.

Comment 14

[u279076]

Thanks Gordon, let's call this incomplete. Please reopen if it happens again.