Closed Bug 597669 Opened 14 years ago Closed 14 years ago

Chromium WebGL Aquarium demo crashes old display driver and screen goes black (ok with new driver)

Categories

(Core :: Graphics: CanvasWebGL, defect)

x86
Windows 7
defect
Not set
critical

Tracking

()

RESOLVED WORKSFORME
Tracking Status
blocking2.0 --- -

People

(Reporter: bugzilla, Assigned: bjacob)

References

(Blocks 1 open bug, )

Details

(Keywords: dataloss, Whiteboard: webglsamples, blacklist ati <= 2009-08 ?, fixed by ATI 8.752.4.0)

Using Windows 7 32-bit of firefox trunk. Build 09-17-2010

Steps to reproduce:
1. Enable HA and webgl
2. Visit http://webglsamples.googlecode.com/hg/aquarium/aquarium.html
3. Screen goes black

The problem with this issue is that there is no way to get back to Windows. Screen is black (display driver crashed?) and the only way to get Windows back is to shut down the computer. This causes serious data-loss.

Computer is a Lenovo T400 and I am using the latest graphic drivers from Lenovo.

Additional information:
Adapter Description	ATI Mobility Radeon HD 3400 Series
Vendor ID		1002
Device ID		95c4
Adapter RAM		256A
dapter Drivers		atiumd64 atidxx64 atiumdag atidxx32 atiumdva atiumd6a atitmm64
Driver Version		8.641.1.1000
Driver Date		8-24-2009
Direct2D Enabled	true
DirectWrite Enabled	true
GPU Accelerated Windows	1/1 Direct3D 9
Note that there is no crash report in about:crashes
Summary: Chromium Aquarium demo crashes display driver and screen goes black → Chromium WebGL Aquarium demo crashes display driver and screen goes black
There is no crash report because the display driver crashed and not Firefox.
That is of course a fault of the display driver but the screen. I had crashes of my nvidia driver but only the Seamonkey window is black in that case.
Can you test with the new generic ATI driver version : 8.762 (2010-08-25) :
http://support.amd.com/fr/gpudownload/windows/Pages/radeonmob_win7-32.aspx
that supports full opengl ES 2.0 since 8.762 :
https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/Catalyst_108_release_notes.pdf
Re: Comment 2:

Agree. We need a ATI driver blocklist just like we already have with Intel drivers.

Re: Comment 3:

Upgrading drivers is always a good idea in this case but I just wanted to say that we're not yet picking up GL ES 2.0 drivers on Windows. Right now the only platforms where we're using ES 2.0 drivers are on Android and Maemo. Of course we need to change this, it's just that until recently there were no ES drivers on desktop OS's. Another thing that we have yet to do, that could help with Windows crashes, is to use Direct3D via ANGLE for WebGL rendering.
As I mentioned in comment 0, I am using the latest official Lenovo drivers. The driver suggested in comment 3 results in the following message:
"ATI Catalyst Mobile cannot be downloaded due to incompatible hardware/software on your computer. 

The version of your graphics adapter is not supported.

Please contact your notebook manufacturer for a graphic driver update"
Note that this does not crash the Chrome canary build. I dont seem get many fps on that demo, but at least there is no black screen. 

See bug 596736, another failure on that test on another computer that Chrome handles without problems.
OK. Before we consider reporting a bug to ATI, it would help to know a few more things. Here, Firefox is doing two different things that stress the graphics driver:  - WebGL rendering, using OpenGL  - accelerated compositing, using Direct3D 9It would help a lot if we could know which of these two things is related to the crash. So could you go to about:config and set:  layers.accelerate-all   --> false  layers.accelerate-none  --> trueThen retry, please.
(bugzilla killed formatting)

OK. Before we consider reporting a bug to ATI, it would help to know a few more
things. Here, Firefox is doing two different things that stress the graphics
driver:
  - WebGL rendering, using OpenGL
  - accelerated compositing, using Direct3D 9

It would help a lot if we could know which of these two things is related to
the crash. So could you go to about:config and set:
  layers.accelerate-all   --> false
  layers.accelerate-none  --> true
Then retry, please.
Chrome is not using D3D at all. So if it's a D3D driver crash, they're not hitting it. Even otherwise, different programs do different things making them hit different bugs in drivers.
Oops, they are actually using D3D via angle. But, the point remains.
(In reply to comment #8)
> It would help a lot if we could know which of these two things is related to
> the crash. So could you go to about:config and set:
>   layers.accelerate-all   --> false
>   layers.accelerate-none  --> true
> Then retry, please.

I can't find these values, I assume I have to create them?
(In reply to comment #11)
> I can't find these values, I assume I have to create them?

No, you really should have them. I checked, they should really be present on windows too. Can you try creating a fresh profile? Run firefox from the command line with the -P option, that should open the profile manager.

Have you really checked that you are using a recent firefox? What does Help->About say?
(In reply to comment #12)
> Have you really checked that you are using a recent firefox? What does
> Help->About say?

New about dialog does not provide any userAgent that you can copy, I got this one from about:support:
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b7pre) Gecko/20100918 Firefox/4.0b7pre

I tried a new profile and found the properties mentioned in comment 8, why are these not to be found with my old profile that I use for trunk builds?

Same result, Windows went black and I had to shut down the computer and restart Windows.
(In reply to comment #13)
> (In reply to comment #12)
> > Have you really checked that you are using a recent firefox? What does
> > Help->About say?
> 
> New about dialog does not provide any userAgent that you can copy, I got this
> one from about:support:
> Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b7pre) Gecko/20100918
> Firefox/4.0b7pre

OK.

> 
> I tried a new profile and found the properties mentioned in comment 8, why are
> these not to be found with my old profile that I use for trunk builds?

This is strange, I'll ask colleagues next week.

> 
> Same result, Windows went black and I had to shut down the computer and restart
> Windows.

OK, thanks for your patiences. Thanks to this, we now know that this is an OpenGL crash, not a Direct3D crash.

In any case the driver shouldn't crash like that, but that could either be triggered by faulty behavior from Firefox (so we need to fix this), or could be purely a driver bug (so we need to report to ATI).

I will soon make a change that will allow us to debug much better into OpenGL crashes. Stay tuned.
blocking2.0: --- → ?
Assignee: nobody → bjacob
blocking2.0: ? → final+
(In reply to comment #0)
> Driver Version        8.641.1.1000
> Driver Date        8-24-2009


OK so, this is a very old driver. The only thing to do is to blacklist it (require newer version) for OpenGL. Sorry if Lenovo doesn't offer a newer version, I can't do anything about that :-( Eventually your problem will be "fixed" when we use Direct3D via ANGLE.
(In reply to comment #15)
> (In reply to comment #0)
> OK so, this is a very old driver. The only thing to do is to blacklist it
> (require newer version) for OpenGL. Sorry if Lenovo doesn't offer a newer
> version, I can't do anything about that :-( Eventually your problem will be
> "fixed" when we use Direct3D via ANGLE.

I am OK with you blacklisting this for OpenGL, but please dont block this driver version for Direct2D, since all HA demos works quite fine.
Whiteboard: webglsamples
Actually, this driver version should be blocked completely due to bug 599661 (unless that bug gets fixed).
Blocks: 605780
Whiteboard: webglsamples → webglsamples, blacklist ati <= 2009-08 ?
We use ANGLE for WebGL now, so this shouldn't be an issue any more.
blocking2.0: final+ → -
Build ID: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b8pre) Gecko/20101212 Firefox/4.0b8pre

Display driver still crashes. Difference now is that I get a blue screen with the following message
*** Hardware malfunction
Call your hardware vendor for support
NMI: Parity check / memory parity error
*** The system has halted

This is fixed by bug 611322.
blocking2.0: - → ?
José, that seems to mean your hardware is physically broken. This is probably blocking- unless that's a false positive or others see this problem.
blocking2.0: ? → -
This is WFM now with the latest Lenovo drivers released this week:

  Graphics

        Adapter Description
        ATI Mobility Radeon HD 3400 Series

        Vendor ID
        1002

        Device ID
        95c4

        Adapter RAM
        256

        Adapter Drivers
        atiu9p64.dll atiuxp64 atiu9pag atiuxpag atiumdva atiumd6a atitmm64

        Driver Version
        8.752.4.0

        Driver Date

        Direct2D Enabled
        true

        DirectWrite Enabled
        true

      GPU Accelerated Windows
      1/1 Direct3D 10
Depends on: 611322
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → WORKSFORME
Summary: Chromium WebGL Aquarium demo crashes display driver and screen goes black → Chromium WebGL Aquarium demo crashes old display driver and screen goes black (ok with new driver)
Whiteboard: webglsamples, blacklist ati <= 2009-08 ? → webglsamples, blacklist ati <= 2009-08 ?, fixed by ATI 8.752.4.0
You need to log in before you can comment on or make changes to this bug.