623107 - WebGL crash [@ ig4icd32.dll@0x2095ed ] on Intel 945GM

Status: RESOLVED DUPLICATE of bug 625118

(Core :: Graphics: CanvasWebGL, defect)

Description

[Makoto Kato [:m_kato]]

Should we add Intel 945GM on Win7 to blacklist of WebGL?

This crash occurs on latest driver too.


bp-f7adc517-2be7-4b35-b71f-4809b2110104

0 	ig4icd32.dll 	ig4icd32.dll@0x2095ed 	
1 	ig4icd32.dll 	ig4icd32.dll@0x34790a 	
2 	ig4icd32.dll 	ig4icd32.dll@0x26a548 	
3 	KERNELBASE.dll 	GlobalAlloc 	
4 	KERNELBASE.dll 	GlobalAlloc 	
5 	ntdll.dll 	RtlpLowFragHeapFree 	
6 	KERNELBASE.dll 	OutputDebugStringA 	
7 	KERNELBASE.dll 	GlobalAlloc 	
8 	ig4icd32.dll 	ig4icd32.dll@0x2978c9 	
9 	ig4icd32.dll 	ig4icd32.dll@0x34b987 	
10 	ig4icd32.dll 	ig4icd32.dll@0x34b7ff 	
11 	ig4icd32.dll 	ig4icd32.dll@0x3499cc 	
12 	ig4icd32.dll 	ig4icd32.dll@0x284c77 	
13 	ig4icd32.dll 	ig4icd32.dll@0x282c09 	
14 	xul.dll 	mozilla::WebGLProgram::QueryInterface 	content/canvas/src/WebGLContext.cpp:654
15 	xul.dll 	nsQueryInterfaceWithError::operator 	obj-firefox/xpcom/build/nsCOMPtr.cpp:64
16 	xul.dll 	nsCOMPtr_base::assign_from_qi_with_error 	obj-firefox/xpcom/build/nsCOMPtr.cpp:107
17 	ig4icd32.dll 	ig4icd32.dll@0x34b747 	
18 	ig4icd32.dll 	ig4icd32.dll@0x28312d 	
19 	ig4icd32.dll 	ig4icd32.dll@0x34b970 	
20 	ig4icd32.dll 	ig4icd32.dll@0x284f04 	
21 	ig4icd32.dll 	ig4icd32.dll@0x284f2c 	
22 	xul.dll 	mozilla::WebGLContext::LinkProgram 	content/canvas/src/WebGLContextGL.cpp:2449
23 	xul.dll 	nsICanvasRenderingContextWebGL_LinkProgram 	obj-firefox/js/src/xpconnect/src/dom_quickstubs.cpp:31571
24 	mozjs.dll 	CallCompiler::generateNativeStub 	js/src/methodjit/MonoIC.cpp:627
25 	mozjs.dll 	js::mjit::ic::NativeCall 	js/src/methodjit/MonoIC.cpp:851
26 		@0xbc567c8 	
27 	mozjs.dll 	js::mjit::EnterMethodJIT 	js/src/methodjit/MethodJIT.cpp:739
28 	mozjs.dll 	CheckStackAndEnterMethodJIT 	js/src/methodjit/MethodJIT.cpp:764
29 	mozjs.dll 	js::mjit::JaegerShot 	js/src/methodjit/MethodJIT.cpp:781
30 	mozjs.dll 	js::RunScript 	js/src/jsinterp.cpp:662
31 	mozjs.dll 	js::Invoke 	js/src/jsinterp.cpp:768
32 	mozjs.dll 	js_fun_call 	js/src/jsfun.cpp:2224
33 	mozjs.dll 	CallCompiler::generateNativeStub 	js/src/methodjit/MonoIC.cpp:627
34 	mozjs.dll 	js::mjit::ic::NativeCall 	js/src/methodjit/MonoIC.cpp:851
35 		@0xb4baa36 	
36 	mozjs.dll 	js::mjit::EnterMethodJIT 	js/src/methodjit/MethodJIT.cpp:739
37 	mozjs.dll 	CheckStackAndEnterMethodJIT 	js/src/methodjit/MethodJIT.cpp:764
38 	mozjs.dll 	js::mjit::JaegerShot 	js/src/methodjit/MethodJIT.cpp:781
39 	mozjs.dll 	js::RunScript 	js/src/jsinterp.cpp:662
40 	mozjs.dll 	js::Invoke 	js/src/jsinterp.cpp:768
41 	mozjs.dll 	js::ExternalInvoke 	js/src/jsinterp.cpp:881
42 	mozjs.dll 	JS_CallFunctionValue 	js/src/jsapi.cpp:4898
43 	xul.dll 	nsXPCWrappedJSClass::CallMethod 	js/src/xpconnect/src/xpcwrappedjsclass.cpp:1694
44 	xul.dll 	nsXPCWrappedJS::CallMethod 	js/src/xpconnect/src/xpcwrappedjs.cpp:577
45 	xul.dll 	PrepareAndDispatch 	xpcom/reflect/xptcall/src/md/win32/xptcstubs.cpp:114
46 	xul.dll 	SharedStub 	xpcom/reflect/xptcall/src/md/win32/xptcstubs.cpp:141
47 	xul.dll 	nsEventListenerManager::HandleEventSubType 	content/events/src/nsEventListenerManager.cpp:1112

http://crash-stats.mozilla.com/report/list?product=Firefox&query_search=signature&query_type=exact&query=ig4icd32.dll%400x2095ed&date=01%2F04%2F2011%2018%3A20%3A18&range_value=1&range_unit=weeks&hang_type=any&process_type=any&plugin_field=&plugin_query_type=&plugin_query=&do_query=1&admin=&signature=ig4icd32.dll%400x2095ed

Comment 1

[Makoto Kato [:m_kato]]

Also, about:support is

Adapter Description Mobile Intel(R) 945 Express Chipset Family
Vendor ID           8086
Device ID           27a2
Adapter RAM         Unknown
Adapter Drivers     igdumd64
Driver Version      8.15.10.1930
Driver Date         9-23-2009

Comment 2

[Benoit Jacob [:bjacob] (mostly away)]

Does the crash persist if you install the DirectX runtime on this computer?

This allows us to use ANGLE instead of OpenGL for WebGL, which avoids using buggy OpenGL implementations on Windows.

We hope to do this automatically (ship relevant DLLs) very soon.

Comment 3

[Makoto Kato [:m_kato]]

(In reply to comment #2)
> Does the crash persist if you install the DirectX runtime on this computer?

This is Windows 7.

> This allows us to use ANGLE instead of OpenGL for WebGL, which avoids using
> buggy OpenGL implementations on Windows.

ANGLE doesn't work on 945GM.
 
> We hope to do this automatically (ship relevant DLLs) very soon.

Comment 4

[Benoit Jacob [:bjacob] (mostly away)]

Oh sorry, I had missed the fact that it's a 945 GM. This is not OpenGL 2 hardware, so WebGL should never have been able to initialize. The first bug here is that it did.

Comment 5

[Benoit Jacob [:bjacob] (mostly away)]

Yep, we should just blacklist this hardware. There's nothing useful we can do with it anyway, it's too old.

Comment 6

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Can you get the GL info from something like http://www.realtech-vr.com/glview/download.html and post it here?  I think we need to be more aggressive about validating OpenGL for WebGL; we kind of fudge it and hope for the best currently.

InitAndValidateGL needs to check for a GL version and some extensions; I think checking for 2.1 plus ARB_framebuffer_object should be enough, but I'm not sure.

Comment 7

[Benoit Jacob [:bjacob] (mostly away)]

Attachment: block old intel cards

This is explained in a comment in the patch.

Comment 8

[Scoobidiver (away)]

Are you sure this D2D feature patch is for this WebGL/OpenGL bug?
Moreover, I don't have any information of crashes with D2D on old Intel GPUs. Why blocklist them for D2D features?

Comment 9

[Scoobidiver (away)]

In addition to comment 8, see bug 625118 comment 6.

Comment 10

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 503282 [details] [diff] [review]
block old intel cards

It looks like it's just the vertex shaders that are done in software. These cards do have proper pixel shaders. If the software fallback for vertex shaders is a problem then ANGLE should probably be fixed to check for it. However, I don't necessarily think using software for vertex shading is even that big of a problem, especially in a shared memory architecture.

Comment 11

[Jeff Muizelaar [:jrmuizel]]

It seems more likely that this is a opengl driver bug or incompatibility.

Comment 12

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

We don't want to touch OpenGL on Intel cards right now for WebGL (or for layers); it should be blocked.  The patch isn't right though and neither are the comments :-)

Comment 13

[Benoit Jacob [:bjacob] (mostly away)]

Jeff --- just FYI, bug 626618 is about GeForce FX 5200 being too slow for D3D9 layers. This makes me think again that we should just blacklist old (pre-X3000) intel chipsets just to 'cut our risks' as we say.

Comment 14

[Benoit Jacob [:bjacob] (mostly away)]

Oh and on Mac we already are blacklisting all those cards from the early 2000s that have software shader fallbacks, just to be safe. I think that hw accel should only be on for known fast hardware (intel >= gma x3000 from 2006).

Comment 15

[Scoobidiver (away)]

> I think that hw accel should only be on for known fast hardware (intel >= gma
> x3000 from 2006).
There is no crash issue with these cards, but may be there is no performance improvement with HW acceleration. Do we have feedback? In that case, D2D might be disabled on 900, 950, and 3150 families and still enabled on 500, X3000, X4500 and HD families. It can be discussed in a dedicated bug.

Indeed it won't fix this bug because OpenGL is going to be used with WebGL whatever the D2D/D3D status.

Comment 16

[Benoit Jacob [:bjacob] (mostly away)]

(In reply to comment #15)
> > I think that hw accel should only be on for known fast hardware (intel >= gma
> > x3000 from 2006).
> There is no crash issue with these cards, but may be there is no performance
> improvement with HW acceleration. Do we have feedback? In that case, D2D might
> be disabled on 900, 950, and 3150 families and still enabled on 500, X3000,
> X4500 and HD families. It can be discussed in a dedicated bug.

I was really talking about D3D9 layers: these old cards don't support D3D10 and D2D anyway.

> 
> Indeed it won't fix this bug because OpenGL is going to be used with WebGL
> whatever the D2D/D3D status.

... which my patch was blocking too. But whatever, ok, let's not blacklist for now.

Comment 17

[Benoit Jacob [:bjacob] (mostly away)]

OK, the patch in bug 625118 will take care of this crash.