Closed Bug 1111808 Opened 10 years ago Closed 5 years ago

kernel panic with webgl on osx 10.6

Categories

(Core :: Graphics: CanvasWebGL, defect)

Product:
Core
Component:
Graphics: CanvasWebGL
Version:
34 Branch
Platform:
x86
macOS
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED WONTFIX

People

(Reporter: spamme.bugzilla, Unassigned)

References

Details

Attachments

(1 file)

Kernel_2014-12-15-222154_six.panic
5.85 KB, text/plain
Details 
User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:34.0) Gecko/20100101 Firefox/34.0
Build ID: 20141126041045

Steps to reproduce:

go to http://threejs.org/examples/webgl_animation_cloth.html with a computer running osx 10.6. It also happens with a lot of webGL pages


Actual results:

kernel panic


Expected results:

no kernel panic. I would prefer to disable the faulty acceleration...
same with webgl.disable-extensions = true
does not crash with webgl.disabled = true

but page does not render
The same happens with an an other computer (same graphics Radeon X1600) running the same system

Is there a way to incrementally test the webGL components?
Next crash, I tried the following settings, but it is still the same:
webgl.can-lose-context-in-foreground = false
webgl.default-no-alpha = true
webgl.disable-angle = true
webgl.disable-extensions = true
webgl.min_capability_mode = true

the other are default, except gfx.blacklist.webgl.msaa = 4 (that was set beforehand)

For now, I will keep with
webgl.disabled = true
Component: Untriaged → Canvas: WebGL
Product: Firefox → Core
Markus/Steven, can either of you help route this to someone who knows how to get this addressed? Kernel panics are pretty bad, as issues go. :-\
Flags: needinfo?(smichaud)
Flags: needinfo?(mstange)
I don't know anyone in Mozilla, including myself, who has experience dealing with (i.e. working around) kernel panics -- on OS X or any other platform.  That's to be expected, since *nothing* you do in a user-level app should be able to cause one.  And if it does, that's almost certainly either a hardware problem or an Apple bug (or both).

That said, if this bug is actually 100% reproducible by everyone who has OS X 10.6 set up in a certain way, it's conceivable we (Markus or I) might be able to figure something out.  But I think that's *highly* unlikely.
Flags: needinfo?(smichaud)
Randell, does this ring any bells?
Flags: needinfo?(rjesup)
(In reply to comment #3)

I didn't notice this comment until I reread the bug.  If you can reproduce these kernel panics on more than one machine, then hardware trouble is unlikely to be the cause.

Please give as much detail as you can about each computer on which you see these kernel panics.

And by the way, please describe in detail what happens when you see a kernel panic.  I want to make sure this is actually what you're seeing.  I've only seen two or three OS X kernel panics in the last 15 years.
> And by the way, please describe in detail what happens when you see a kernel panic.

Never mind.  Now I see your attachment.
Flags: needinfo?(spamme.bugzilla)
The HW driver may well be the cause (as opposed to the HW itself).

Does it work with HW acceleration disabled?  As with OMTC bad drivers may require blacklisting acceleration.

If so, we may need a GFX person to make the call
Flags: needinfo?(rjesup)
Of course, it should not happen due to an operation in the user space, so the driver is eventually faulty. But given that it is old an not updated any more (com.apple.ATIRadeonX1000), if Firefox consistently crashes with it, it should be blacklisted. (By the way, one very big advantage of Firefox, over the Apple browser, is that Firefox is still upgraded years after Apple dropped any support.)

How could I test the hardware acceleration?
Flags: needinfo?(spamme.bugzilla)
(In reply to jip from comment #11)
> Of course, it should not happen due to an operation in the user space, so
> the driver is eventually faulty. But given that it is old an not updated any
> more (com.apple.ATIRadeonX1000), if Firefox consistently crashes with it, it
> should be blacklisted. (By the way, one very big advantage of Firefox, over
> the Apple browser, is that Firefox is still upgraded years after Apple
> dropped any support.)
> 
> How could I test the hardware acceleration?

In the preferences, under the "Advanced" section, in the General subsection, you can turn it off.
jip, please do provide detailed information on the computers where you've seen this bug (as I asked above).

"Hardware acceleration" can be turned off or on at "Firefox : Preferences : Advanced : General : Use hardware acceleration when available".
Flags: needinfo?(mstange)
Same kernel panic with hardware acceleration disabled. (The only way I found to avoid the panic is to disable WebGL entirely. Of course, it would be great to be able to test incrementally...)

I am experiencing this panic with 2 different machines, with the same configuration:
- MacBookPro 2,2 (Intel Core2Duo 2.33 GHz, 3GB RAM)
- ATI RadeonX1600 (256 MB VRAM)
- MacOS 10.6.8
Opening this pictures [1] with Firefox (37.0.2) on Google seems cause a kernel panic on Mac OS X 10.6.8 (iMac5,1 - Intel Core 2 Duo).
[1]: https://www.google.it/maps/uv?hl=it&pb=!1s0x880efc127a7177d5:0xf99c85f2fe837520!2m5!2m2!1i80!2i80!3m1!2i100!3m1!7e1!4shttp://www.panoramio.com/photo/68371202!5sfermi+lab+chicago+-+Cerca+con+Google&sa=X&ei=hLZEVZaeFsH5UsmmgbAM&ved=0CIcBEKIqMAo

(I'm not sure that it is related to this bug, but I repeat it twice and kernel panic happens).
I've been struggling with this issue for the past several weeks.  Took me quite a bit of time to get to this Bugzilla Report.  Initially, I thought it might be a virus, but pretty much ruled that out through trial-error and a half dozen anti-virus scans (different apps).  Then I was pretty much ready to download and install Chrome as a workaround (and would have except the Google Chrome download page caused a WebGL kernel crash on my machine!!), when I found this thread.

My System Info (kinda old these days, I know):

iMac -- 2.16 GHz Intel Core 2 Duo, 2.5 GB SDRAM
MacOS X 10.6.8 (last stable released of Snow Leopard)
Graphics: ATI Radeon X1600 driving two (2) monitors: 17" built-in Apple display, external 17" Dell 1703FP, both 32-bit, 60Hz.

After reviewing this bug report and the comments, I web into Firefox Configuration Panel and set webgl.disabled = true.  Since I did that about 24 hours ago, I haven't encountered a single website causing a kernel crash.  Nor have I encountered a webpage I was unable to use because of missing graphics, but I'm not betting there aren't some pages out there which are unusable without WebGL.

This is a fairly MAJOR BUG, IMO.  Any software program that is in release level 37.x which causes a kernel crash has something terribly broken.  I can live with WebGL turned off for a while, but I really would expect a fix for this issue to come sooner rather than later.  Trying to stay true to Firefox...
> Please give as much detail as you can about each computer on which you see
> these kernel panics.

In reply to your request for specifics about what you see.  The problem seems to appear most frequently, perhaps exclusively, related to login pages for websites.  A couple of times I first got a 5xx HTTP error, usually 504, because of a failed redirect following a login.  In those cases, I attempted to reload the root index of the website to see if the login had taken, navigate without login, etc., and the machine crashed.

When the crash happens, the main display (I have two 17" displays) has the grey curtain come down over the display, and the multi-language dialogue box appears which says to hold the power button down until the machine shuts down, then depress the power button a second time to reboot.  (Pretty much standard mach kernel panic scenario for Mac OS X.)

One thing I can tell you is that cookies do not appear to be responsible for the problem.  I tried several websites which caused problems with cookies disabled, flushed the cache, logins, cookies, and everything else in Firefox preferences.  Aside from the fact Firefox ran a lot faster on restart after flushing the cache, there was no change in terms of the consistency of kernel crashes on certain pages.
> This is a fairly MAJOR BUG, IMO.

Yes it is, but it's not a bug in Firefox.

No user level app can cause a kernel panic.  They can only be caused by hardware problems or buggy drivers (or kernel extensions).  Since all your drivers and kernel extensions probably came from Apple, this is probably an Apple bug.

That said, if we can reproduce this bug we may be able to find a way to work around it.  But I'm afraid I'm not very optimistic about that :-(
MIDI Maven, here's a thought:

Try disconnecting as much equipment as you can from your computer, including the second monitor.  Then see if you can still reproduce these panics with webgl turned on.  If you can't, then try reconnecting that equipment, piece by piece, until you start seeing the panics again.
I tried disconnecting the Dell 1703FP monitor, turned WebGL back on, and was able to reproduce the error.  

I do have a bunch of additional Firewire and USB equipment (Pro audio) connected to this system, but anytime it is powered up and I am using it... I am not using Firefox.  (As I previously said, the system is getting a little old these days, and generally I shut down all non-essentials when I am working with the sequencing software and A/D audio processors.)  Just to be certain, however; I removed all Firewire and USB devices, rebooted the system with webgl.disabled = false, and was able to reproduce the error on the basic Apple equipment.

I can certainly see how the kernel panic could be caused by the Radeon video driver rather than the application.

If, however, this is a driver issue, I fail to see why this error only begun to occur in the last two weeks on my system.  WebGL has been enabled in Firefox since version 4.x as I recall, and I've been running Firefox on this system since version 4.x.  I have had MacOS X 10.6.8 installed on it since 4Q 2011 a few months after it was released.  Never has Firefox caused a kernel panic until the last two weeks or so.  And frankly, Apple hasn't issued an relevant update for MacOS X 10.6.8 in YEARS, so I very seriously doubt the error crept in there.

The problem began shortly after I upgraded to Firefox 37.0.2 -- on April 28 according to Firefox Update History.  (I had skipped several previous updates.  Pretty sure the release installed prior to this one was 35.x, but the Firefox Update History doesn't show anything before 37.0.2.  Perhaps the data was purged when I flushed Everything?)  I saw in the release notes for 37.0.1 there were some updates related to video drivers:  "Stability fixes for select graphics hardware and feature sets."  So, I suppose it is possible the bug crept in in version 37.0.1.
Once again, Firefox *can't* cause a kernel panic -- it's just not possible.  If this isn't a hardware problem (which it might be, especially on old equipment), it's a bug in a kernel extension.

If it's a bug in a kernel extension, it's possible that something changed recently in Firefox's behavior to trigger the bug (more often or at all).  But for us to pinpoint the change, we'd need to know the regression range in mozilla-central (trunk) nightlies.  More on this later, in another comment.

First try the following, to rule out the possibility that this is a hardware problem:

Run your old version of Firefox for a while (the one you called 35.X above), to see if these panics now also happen with it.  If they do, then it's likely you have a hardware problem.
This confirms that the bug lies in the ATI Radeon X1600 driver for Mac OS 10.6.8. As for me, Firefox triggers the bug at least with Firefox 34 to 37, with WebGL. The other change is that more web pages are using WebGL now, so the behaviour could have been there for a while, albeit not revealed.
> This confirms that the bug lies in the ATI Radeon X1600 driver for Mac OS 10.6.8.

I don't know about that.  But it *is* true that the only two people who see this bug and who've reported their graphics hardware have ATI Radeon X1600.

I have an ancient MacBookPro1,1 -- yes, the original MacBook Pro.  I just dug it out of my closet and discovered that it still works (or at least it still turns on).  It has ATI Radeon X1600 hardware.

It has a tiny hard drive and currently has OS X 10.4.11.  So it will be a lot of work to get another bootable partition on it and install OS X 10.6.8 there ... if that's even possible.  But sometime in the next few weeks I hope to find the time to do this.  Then, if I can also reproduce these kernel panics, it'll be a matter of a few days to find out whether or not it's feasible for us to try to work around the presumed bug in Apple's ATI Radeon X1600 driver.
I also saw that on a third machine, with the same hardware, and system.

Thank you for your time. By the way, if you want to use your MacBookPro1,1 for testing, it would be easier to plug an external firewire hard drive, and boot from it (by holding the option key at start-up).

And If we can run some specific tests, please let us know.
Another report on the French forum: https://forums.mozfr.org/viewtopic.php?f=5&t=125150
See Also: → 1186878
I have a 2006 MacBook Pro with Firefox 35.0.1 and the ATI Radeon X1600 graphics hardware and the kernel panic is repeatable. I too thought it might be a hardware issue, but it only ever happens with certain web sites. Midi Maven's comments about it happening when you try and access the Google Chrome web site are rather prescient - I too figured I would download Chrome to see if I could reproduce the problem in that browser (which may have helped me to confirm whether or not it was actually a hardware problem) but the kernel panic happens as soon as you hit the site, an irony which was not lost on me. If you try to access any WebGL demonstration web sites, a kernel panic is also very likely.

After a kernel panic I've checked the Console to see what error messages were generated but I had no idea what I was looking for and didn't find anything obvious, so it would be helpful if somebody could give some input on that please. Which log files do we need to be looking at and what messages are we looking for? I've just set webgl.disabled to true so I'll see whether that works out or not.
Web sites crash Firefox 39.0.3 and cause kernel panics. I've been having this problem repeatedly for the past two years. Terrible. It's gotten worse over time. The more time that passes, the worse it gets. Up until then, I didn't even know what a kernel panic was. Now I get about one a week, no more only because I try to observe (and remember) what site caused the crash. I always used tabbed browsing, so I'm not always sure. I don't like that there's a growing list of sites I can't use.

Last night a Google search caused a kernel panic! That's new. I even had trouble restarting the Mac by holding down the power button. Thought I would have to unplug it. On restart, I had to use Disk Utility to repair the volume.

One site I'm sure of is spokeo.com.

I'm using 39.0.3 because that's the last version that would work with 1Password. I can't upgrade 1Password without upgrading my OS. I can't upgrade my OS without upgrading my computer.

Nope.

imac 5,1; intel core2duo, 2ghz, 17", RadeonX1600

I've tried everything I know, and I know a fair amount. Obviously not as much as you guys. A bad graphic driver makes sense.
Sorry, that's OS 10.6.7, because 10.6.8 always made fonts jaggy. I spent years on that one too and never solved it.
One site I'm sure of now is B&H. 

Someone else on the web suggest the problem was heat. I installed a fan control program, and things seemed to be OK, but then B&H caused the kernel panic again. I do not have this problem with Safari 5 or Opera 25.
Re 28: You probably won't learn much on the technical side unless you're a techie to begin with. Here's an article about how to read to troubleshoot a kernel panic. http://www.macworld.com/article/2027201/how-to-troubleshoot-a-kernel-panic.html

Start up Console. On the left column you've got several logs. One is Crash Reporter. One is System Diagnostic Reports. Click on the any the start with "kernel" and you'll see a long list of kernel extensions. You should be able to tell which program was running or caused the panic. 

Some interesting things on my latest report.

 Kernel Extensions in backtrace (with dependencies):
         com.apple.ATIRadeonX1000(6.2.6)@0x948000->0x9a2fff
            dependency: com.apple.iokit.IOPCIFamily(2.6)@0x908000
            dependency: com.apple.iokit.IONDRVSupport(2.2)@0x93b000
            dependency: com.apple.iokit.IOGraphicsFamily(2.2)@0x919000

BSD process name corresponding to current thread: firefox

And

panic(cpu 1 caller 0x2aab59): Kernel trap at 0x0097b1a1, type 14=page fault, registers:
CR0: 0x80010033, CR2: 0x0000002c, CR3: 0x00100000, CR4: 0x00000660
EAX: 0x35f1c000, EBX: 0x00000027, ECX: 0x35f1d088, EDX: 0x00000000
CR2: 0x0000002c, EBP: 0x48e92b88, ESI: 0x00000000, EDI: 0x083a9000
EFL: 0x00010206, EIP: 0x0097b1a1, CS:  0x00000008, DS:  0x00000010
Error code: 0x00000000

You can also use Etre Check to view the panic report.
This guy thought heat might be the problem, so he used a fan control program, but that didn't solve the problem for me. Not on B&H.com anyway. It's still worth a look.

https://superuser.com/questions/73518/ati-x1600-driver-problem-on-mac

10.6 is WONTFIX

Status: UNCONFIRMED → RESOLVED
Closed: 5 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: