User Agent: Mozilla/5.0 (Windows NT 6.3; Win64; x64; rv:89.0) Gecko/20100101 Firefox/89.0

Steps to reproduce:

I recently upgraded to Firefox 89, and have hardware acceleration ("Use hardware acceleration when available") turned on in its settings. I regularly use Firefox to watch videos on YouTube.

Actual results:

When watching YouTube videos while Firefox's hardware acceleration is enabled, I occasionally get blue screens (BugChecks) of type SYSTEM_THREAD_EXCEPTION_NOT_HANDLED (0x7e), pointing to the driver nvlddmkm.sys (NVIDIA Windows Kernel Mode Driver). I have so far not been able to reliably reproduce the crashes: sometimes I will get multiple in one day, other times it might work fine for a few days before crashing. The one constant is that they always occur while I am watching YouTube videos using Firefox. (This is reminiscent of Bug 1409949, but I am getting a different BugCheck code.)

These crashes do not occur when watching YouTube with Firefox's hardware acceleration disabled, nor when I am performing other graphically intensive tasks (e.g., playing video games). They also did not occur with FF's HW acceleration enabled prior to the update to version 89.

In an effort to resolve this issue, I have updated my NVIDIA drivers to the latest version (466.63), but I still encounter the same behavior. I have configured Firefox to clear all cookies/cache/history every time I close the browser, so those should not be a factor.

In case it's relevant, my PC's configuration is as follows.

OS: Windows 8.1 Pro (64-bit)
Motherboard: ASUS MAXIMUS VII HERO
CPU: Intel Core i7 4790K
RAM: Corsair Vengeance LP 32GB (4x8GB) DDR3 (PC3-12800)
GPU: EVGA GeForce GTX 1080 Ti SC2 GAMING

I have attached a kernel debugger analysis transcript for the latest BugCheck I encountered related to this issue. I can also provide the full memory dump from the crash if that would be helpful, but I would need a way to do so privately.

Expected results:

Watching YouTube videos using Firefox (with HW acceleration enabled) should not crash my computer.

The Bugbug bot thinks this bug should belong to the 'DevTools::Debugger' component, and is moving the bug to that component. Please revert this change in case you think the bot is wrong.

Moving to relevant component

Aaron Loessberg-Zahl, thank you for the report. Can you attach about:support to this bug?

about:support data attached. Thanks.

Thank you for the quick response. Attachment 9226050 [details] seemed to be get while hw acceleratio was off. But when hw acceleration was on, WebRender might be used.

Aaron Loessberg-Zahl, can you also attach about:support when hw acceleration is on? Thank you.

Ah, sorry about that. I had disabled it to stop it crashing. Here's the about:support with it turned on.

(In reply to Aaron Loessberg-Zahl from comment #6)

Created attachment 9226063 [details]

From about:support, WebRender was enabled.

Aleino, can you take a look at this nvidia kernel crash?

Failed to reproduce with a slightly different setup, but I only watched youtube for around 10 minutes.

Firefox: 90.0.2 (64-bit)
GPU: GTX 1080
Driver: Internal build. Rougly 470.xx.
Windows: 10 (version 1909)

I can file an internal Nvidia-bug, but I think we should first check if we can cut down the search space a bit, since the crash is so intermittent:

1. Do you have an example of a specific youtube video where the crash is observed?
2. Please try to reproduce with the latest drivers: 471.41.
3. Please try with Firefox 90.
4. Just in case you're set up to do it easily: can you try on windows 10?

(In reply to aleino from comment #10)

I can file an internal Nvidia-bug, but I think we should first check if we can cut down the search space a bit, since the crash is so intermittent:

1. Do you have an example of a specific youtube video where the crash is observed?
2. Please try to reproduce with the latest drivers: 471.41.
3. Please try with Firefox 90.
4. Just in case you're set up to do it easily: can you try on windows 10?

Unfortunately, I don't have a specific example of a video which reliably causes the crashes. I've just now installed the 471.41 driver and re-enabled HW acceleration on FF (currently on version 90.0.2). If it crashes again, I'll note down what video I was watching at the time.

As for Windows 10, I only have a laptop with Windows 10 installed, which has a completely different GPU, so I don't think it would be helpful in terms of testing this issue.

In any case, like I said, I'll run FF with HW accel turned on and using the latest Nvidia drivers and see what happens. I'll post an update after a week or so, or sooner if the crashes return.

I filed Nvidia bug 3353089 for this issue.

It seems difficult to reproduce with the current information, but at least we can take a look at the provided kernel log.

(In reply to Aaron Loessberg-Zahl from comment #11)

(In reply to aleino from comment #10)

I can file an internal Nvidia-bug, but I think we should first check if we can cut down the search space a bit, since the crash is so intermittent:

1. Do you have an example of a specific youtube video where the crash is observed?
2. Please try to reproduce with the latest drivers: 471.41.
3. Please try with Firefox 90.
4. Just in case you're set up to do it easily: can you try on windows 10?

Unfortunately, I don't have a specific example of a video which reliably causes the crashes. I've just now installed the 471.41 driver and re-enabled HW acceleration on FF (currently on version 90.0.2). If it crashes again, I'll note down what video I was watching at the time.

As for Windows 10, I only have a laptop with Windows 10 installed, which has a completely different GPU, so I don't think it would be helpful in terms of testing this issue.

In any case, like I said, I'll run FF with HW accel turned on and using the latest Nvidia drivers and see what happens. I'll post an update after a week or so, or sooner if the crashes return.

Thanks!
Please update the bug if you obtain any further information.

Also, if you catch the crash again it would be helpful if we could get a kernel memory dump.

https://docs.microsoft.com/en-us/windows-hardware/drivers/debugger/kernel-memory-dump

Well, I've been running FF 90.0.2 and Nvidia's 471.41 drivers for about a week, and I haven't had any crashes to speak of, so I think it's safe to say that something in one of those version bumps managed to inadvertently fix the problem.

Aleino, if you'd like a memory dump of this type of crash for academic purposes, I still have the original one which I used to generate the kd transcript I attached to this bug. If you can provide me a secure way to do so, I don't mind sending it over.

Otherwise, unless there's another reason to keep it open, this bug can probably be closed.

Thanks!