crashes in atidxx32.dll/atidxx64.dll (more prevalent in windows 10)

RESOLVED INCOMPLETE

Status

()

Core
Graphics
--
critical
RESOLVED INCOMPLETE
3 years ago
2 years ago

People

(Reporter: philipp, Assigned: bas)

Tracking

(Blocks: 1 bug, {crash})

39 Branch
x86
Windows
crash
Points:
---

Firefox Tracking Flags

(firefox39 wontfix, firefox40+ fixed, firefox41 wontfix, firefox42+ wontfix, firefox43+ wontfix, firefox44- wontfix, firefox45- wontfix, firefox46 affected)

Details

(crash signature)

Attachments

(1 attachment)

(Reporter)

Description

3 years ago
This bug was filed from the Socorro interface and is 
report bp-ca6925e5-3866-4d4e-ae91-b25e22150726.
=============================================================

Crashing Thread
Frame 	Module 	Signature 	Source
Ø 0 	atidxx32.dll 	atidxx32.dll@0x55ed9f 	
Ø 1 	atidxx32.dll 	atidxx32.dll@0x1891eb 	
Ø 2 	atidxx32.dll 	atidxx32.dll@0x18995f 	
Ø 3 	atidxx32.dll 	atidxx32.dll@0xce43e 	
Ø 4 	atidxx32.dll 	atidxx32.dll@0xce680 	
Ø 5 	atidxx32.dll 	atidxx32.dll@0xcb494 	
Ø 6 	atidxx32.dll 	atidxx32.dll@0xca103 	
Ø 7 	atidxx32.dll 	atidxx32.dll@0xc8ed8 	
Ø 8 	atidxx32.dll 	atidxx32.dll@0x5e6eb 	
Ø 9 	atidxx32.dll 	atidxx32.dll@0x50f44d 	
Ø 10 	atidxx32.dll 	atidxx32.dll@0x50f2bd 	
Ø 11 	atidxx32.dll 	atidxx32.dll@0x4f0dd8 	
Ø 12 	atidxx32.dll 	atidxx32.dll@0x73e37 	
Ø 13 	atidxx32.dll 	atidxx32.dll@0x55cf9e 	
Ø 14 	atidxx32.dll 	atidxx32.dll@0x55d0c9 	
Ø 15 	kernel32.dll 	kernel32.dll@0x13743 	
Ø 16 	ntdll.dll 	ntdll.dll@0x5a063 	
Ø 17 	ntdll.dll 	ntdll.dll@0x5a02e

variants of atidxx32.dll/atidxx64.dll crashing may currently be a topcrash on windows 10 - responsible for 7,5% of crashes in the 10240 build so far. as far as i can tell, this is also happening with the most current drivers available:
https://crash-stats.mozilla.com/search/?product=Firefox&platform=Windows&platform_version=%3D10.0.10240&signature=%24atidxx&_facets=signature&_facets=platform_version&_facets=adapter_device_id&_facets=adapter_driver_version&_facets=version&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-signature

these crashes are happening in other windows versions as well though
(Reporter)

Comment 1

3 years ago
[Tracking Requested - why for this release]:
to get this on the radar of relman
status-firefox39: --- → affected
status-firefox40: --- → affected
status-firefox41: --- → affected
status-firefox42: --- → affected
tracking-firefox40: --- → ?

Comment 2

3 years ago
Not much to go on in this stack, unfortunately. Maybe Milan's team is aware of issues on this driver/device?

Adapter device id facet Rank 	Adapter device id 	Count 	%
1 	0x6939 	385 	90.38 %
2 	0x7300 	30 	7.04 %
3 	0x9874 	8 	1.88 %
4 	0x6921 	3 	0.70 %

Rank 	Adapter driver version 	Count 	%
1 	15.200.1046.2 	227 	53.29 %
2 	15.200.1062.1002 	99 	23.24 %
3 	15.200.1062.0 	89 	20.89 %
4 	15.200.0.0 	6 	1.41 %
5 	14.502.1014.1001 	2 	0.47 %
6 	15.200.1046.0 	2 	0.47 %
7 	15.200.1023.7 	1 	0.23 %
Flags: needinfo?(milan)
I did some supersearch correlations to see if anything interesting jumps out.

In combined signatures:
By platform, 85% are on Windows 10 while 15% are on Windows 8.1.
By version, 57% are on Firefox 39 while 26% are on Firefox 42, remainder go back to 34esr.
By device, 82% are on 0x6939 while 16% are on 0x7300, remainder are on various other devices.
By device family, 83% are on VI Tonga series, 16% are on Fiji XT series.
By driver, 56% are on 15.200.1046.2 [Catalyst 15.7] while 26% are on 15.200.1062.1002 [Catalyst 15.7.1]

In terms of topcrashiness this would rank 141st in release, 125th if we're looking at strictly Windows crashes, and #1 if we're strictly looking at Windows 10 crashes. That said, we're still extremely early in the Windows 10 roll out so I think it's premature to conclude anything from the Windows 10 rank.

It's also worth noting that these all might be separate issues. For example, the atidxx32 crashes show up with earlier versions of Firefox while the atidxx64 crashes only start with Firefox 39. I'd be cautious to make any conclusions about these having the same root cause.

Based on this our best lead might be trying to get a system with a Radeon R9 285 (0x6939) but I'd be hesitant to call this a topcrash. I checked the hardware inventory and we don't have one so this would need to be purchased. Milan can we get this in the lab for testing?

Comment 4

3 years ago
> In terms of topcrashiness this would rank 141st in release, 125th if we're
> looking at strictly Windows crashes, and #1 if we're strictly looking at
> Windows 10 crashes. That said, we're still extremely early in the Windows 10
> roll out so I think it's premature to conclude anything from the Windows 10
> rank.
It's worth getting a head start on this one. We need to pay extra attention to Win10 crashes in the coming days, as more and more users will be upgrading. If the crash profile on Win10 is dramatically different from previous Windows then we could be in for an unpleasant surprise.
(In reply to David Major [:dmajor] from comment #4)
> It's worth getting a head start on this one. We need to pay extra attention
> to Win10 crashes in the coming days, as more and more users will be
> upgrading. If the crash profile on Win10 is dramatically different from
> previous Windows then we could be in for an unpleasant surprise.

Definitely, which is why I asked Milan if we can get a card for testing. I just want to be cautious about making any decisions or conclusions about this until we've had the opportunity to collect more data and look at it more closely.
We don't currently blocklist any of these cards.  As far as the drivers go, we wholesale blocklist Windows 10 and 15.200.1006.0 for hardware video, but that driver version doesn't show up in the list above.

I'm not sure where we can go from here, but I'll add a few people so that it's not a surprise if it blooms.  Note that we don't have the ability in 39 to use the downloadable blocklist for Windows 10 (bug 1183725, fixed in 40 and up.)
Flags: needinfo?(milan)
Flags: needinfo?(matt.woodrow)
Flags: needinfo?(jmuizelaar)
Flags: needinfo?(dvander)
Flags: needinfo?(bas)
Some telemetry data. For Windows 10 sessions across beta, aurora, nightly, selecting driver versions "15.200.1046.2", "15.200.1062.1002", and "15.200.1062.0":

   5313 out of 49991 sessions matched. (10.63%)

On Windows 10, the vendor makeup is skewed differently than our overall population. Intel is ~50% of sessions, AMD and nVidia are both around 24%. So these versions are about 50% of AMD driver versions on Windows 10.

Across all Windows sessions:

   6466 out of 2897764 sessions matched. (0.22%)

This translates to about 1.5% of AMD drivers.

So for whatever reason these driver versions are only very common on Windows 10.
Flags: needinfo?(dvander)

Comment 8

3 years ago
> So for whatever reason these driver versions are only very common on Windows 10.

Perhaps those driver versions are specifically their Win10 driver. I wonder if the (6466 minus 5313) sessions were on prerelease build numbers of Win10?

Updated

3 years ago
Blocks: 1148406
Do we have enough information to just blocklist some drivers on Windows 10 for AMD, the equivalent of https://bugzilla.mozilla.org/show_bug.cgi?id=1189940#c4 for Nvidia?
Trying to get one of these cards locally.
I've reproduced this locally
Flags: needinfo?(jmuizelaar)
Locally crashes tend to end up on the main thread waiting for another thread (perhaps the crashing one) while calling OfferResources() Video does not seem required to trigger the crash. It tends to happen when creating or switching tabs shortly after startup.
Need this added to the downlodable blocklist:

<gfxBlacklistEntry>
  <os>WINNT 10.0</os>
  <vendor>0x1002</vendor>
  <feature>DIRECT3D_11_LAYERS</feature>
  <featureStatus>BLOCKED_DEVICE</featureStatus>
  <devices>
    <device>0x6939</device>
  </devices>
</gfxBlacklistEntry>

<gfxBlacklistEntry>
  <os>WINNT 10.0</os>
  <vendor>0x1002</vendor>
  <feature>DIRECT2D</feature>
  <featureStatus>BLOCKED_DEVICE</featureStatus>
  <devices>
    <device>0x6939</device>
  </devices>
</gfxBlacklistEntry>
Flags: needinfo?(jorge)
Flags: needinfo?(jmuizelaar)
(In reply to Milan Sreckovic [:milan] from comment #13)
> Need this added to the downlodable blocklist:
> 
> <gfxBlacklistEntry>
>   <os>WINNT 10.0</os>
>   <vendor>0x1002</vendor>
>   <feature>DIRECT3D_11_LAYERS</feature>
>   <featureStatus>BLOCKED_DEVICE</featureStatus>
>   <devices>
>     <device>0x6939</device>
>   </devices>
> </gfxBlacklistEntry>
> 
> <gfxBlacklistEntry>
>   <os>WINNT 10.0</os>
>   <vendor>0x1002</vendor>
>   <feature>DIRECT2D</feature>
>   <featureStatus>BLOCKED_DEVICE</featureStatus>
>   <devices>
>     <device>0x6939</device>
>   </devices>
> </gfxBlacklistEntry>

I think just the DIRECT2D one should be sufficient.
Flags: needinfo?(jmuizelaar)
FWIW, I was able to reproduce this on the driver that Windows 10 installs by default. I was not able to reproduce this on a R7 240 with the default drivers.
We should probably include all of these ids unless we have evidence to suggest otherwise.
            "TONGA": [
                "6920",
                "6921",
                "6928",
                "6929",
                "692b",
                "692f",
                "6930",
                "6938",
                "6939"
            ],
            "ICELAND": [
                "6900",
                "6901",
                "6902",
                "6903",
                "6907"
            ],
            "FIJI": [
                "7300"
            ]
This is the up to date entry for the downloadable blocklist:

<gfxBlacklistEntry>
  <os>WINNT 10.0</os>
  <vendor>0x1002</vendor>
  <feature>DIRECT2D</feature>
  <featureStatus>BLOCKED_DEVICE</featureStatus>
  <devices>
    <device>0x6920</device>
    <device>0x6921</device>
    <device>0x6928</device>
    <device>0x6929</device>
    <device>0x692b</device>
    <device>0x692f</device>
    <device>0x6930</device>
    <device>0x6938</device>
    <device>0x6939</device>
    <device>0x6900</device>
    <device>0x6901</device>
    <device>0x6902</device>
    <device>0x6903</device>
    <device>0x6907</device>
    <device>0x7300</device>

  </devices>
</gfxBlacklistEntry>
Note that 39 does not know about Windows 10, the downloadable blocklist will only work for 40 and later (bug 1183725), so I wouldn't expect 39 crashes to change, other than as people update.
The block has been added, per comment #17.
Flags: needinfo?(jorge)
(In reply to Jorge Villalobos [:jorgev] from comment #19)
> The block has been added, per comment #17.

Flagging myself to check crash-stats in a few days to see what difference this has made.
Flags: needinfo?(anthony.s.hughes)

Updated

3 years ago
See Also: → bug 1190519
Crashes for a bunch of people, tracking.
status-firefox39: affected → wontfix
tracking-firefox40: ? → +
We'll probably be visting AMD about this next week
Created attachment 8647634 [details]
atidxx-crashes.png

Over on bug 1189940 it was determined that we *may* have pushed our block out a bit too hastily. I wanted to make sure the same wasn't true for these AMD crashes. After crunching the numbers from crash-analysis I can definitely see that things got progressively worse once Windows 10 came up so there's some validity to blocking the AMD drivers (the same cannot be said for the NVidia drivers on bug 1189940).

The attached chart shows the crash data for atidxx**.dll crashes starting 10 days before Windows 10 was released with Windows 10 release date and the blocklist date annotated. 

We don't have enough data yet to say for sure if the blocklist is helping. I'm keeping my need-info flag set as a reminder to check this in a few more days.
It appears as though the block is having some impact on the crash volume but it's not eliminating this issue altogether. For the week before the block was staged we saw 2179 crashes (~311 crashes per day). Since pushing the block live we've seen 673 crashes (~134 crashes per day). Based on this the daily crash rate is down about 57%.
Flags: needinfo?(anthony.s.hughes)
(Assignee)

Comment 25

2 years ago
Let's see what AMD says.
Flags: needinfo?(bas)
Assignee: nobody → bas
(In reply to Bas Schouten (:bas.schouten) from comment #25)
> Let's see what AMD says.

Last work from AMD is - this is most likely Windows 8.1 and later, "all devices".  We may be able to mount an expedition and test with the "pre-release" driver that has a fix.
Flags: needinfo?(matt.woodrow)

Updated

2 years ago
Duplicate of this bug: 1195896

Comment 28

2 years ago
In bug 1195896, a user is able to reproduce the crash, you can contact him.
FWIW, we confirmed that this is fixed in the upcoming driver release.
Jorge,

Can you add this entry as well?

<gfxBlacklistEntry>
  <os>WINNT 8.1</os>
  <vendor>0x1002</vendor>
  <feature>DIRECT2D</feature>
  <featureStatus>BLOCKED_DEVICE</featureStatus>
  <devices>
    <device>0x6920</device>
    <device>0x6921</device>
    <device>0x6928</device>
    <device>0x6929</device>
    <device>0x692b</device>
    <device>0x692f</device>
    <device>0x6930</device>
    <device>0x6938</device>
    <device>0x6939</device>
    <device>0x6900</device>
    <device>0x6901</device>
    <device>0x6902</device>
    <device>0x6903</device>
    <device>0x6907</device>
    <device>0x7300</device>

  </devices>
</gfxBlacklistEntry>
Flags: needinfo?(jorge)
Added.
Flags: needinfo?(jorge)
Jeff, I will mark 40 as fixed as we blacklisted them. Please correct me if I am wrong! Thanks
status-firefox40: affected → fixed
Flags: needinfo?(jmuizelaar)

Comment 33

2 years ago
Forced direct2d on and I can confirm that I'm not getting crashes on the just released 15.8 Beta.
Jorge can you update the previous blocklist entries to:

<gfxBlacklistEntry>
  <os>WINNT 8.1</os>
  <vendor>0x1002</vendor>
  <feature>DIRECT2D</feature>
  <featureStatus>BLOCKED_DRIVER_VERSION </featureStatus>
  <driverVersion>15.201.1151.0</driverVersion>
  <driverVersionComparator>LESS_THAN</driverVersionComparator>
  <devices>
    <device>0x6920</device>
    <device>0x6921</device>
    <device>0x6928</device>
    <device>0x6929</device>
    <device>0x692b</device>
    <device>0x692f</device>
    <device>0x6930</device>
    <device>0x6938</device>
    <device>0x6939</device>
    <device>0x6900</device>
    <device>0x6901</device>
    <device>0x6902</device>
    <device>0x6903</device>
    <device>0x6907</device>
    <device>0x7300</device>
  </devices>
</gfxBlacklistEntry>
<gfxBlacklistEntry>
  <os>WINNT 10</os>
  <vendor>0x1002</vendor>
  <feature>DIRECT2D</feature>
  <featureStatus>BLOCKED_DRIVER_VERSION </featureStatus>
  <driverVersion>15.201.1151.0</driverVersion>
  <driverVersionComparator>LESS_THAN</driverVersionComparator>
  <devices>
    <device>0x6920</device>
    <device>0x6921</device>
    <device>0x6928</device>
    <device>0x6929</device>
    <device>0x692b</device>
    <device>0x692f</device>
    <device>0x6930</device>
    <device>0x6938</device>
    <device>0x6939</device>
    <device>0x6900</device>
    <device>0x6901</device>
    <device>0x6902</device>
    <device>0x6903</device>
    <device>0x6907</device>
    <device>0x7300</device>
  </devices>
</gfxBlacklistEntry>
Flags: needinfo?(jmuizelaar) → needinfo?(jorge)
Done.
Flags: needinfo?(jorge)
(Reporter)

Comment 36

2 years ago
revisiting the adapter device id facet from comment #0 shows that this crash is now dominated by:
>Rank 	Adapter device id 	Count 	%
>1 	0x9874 	929 	71.74 %
maybe this warrants putting the card on the blocklist as well?
Flags: needinfo?(jmuizelaar)
[Tracking Requested - why for this release]: We still have crashes but they're not in topcrash range. Do we still want to track this and if so what is left to do?
status-firefox41: affected → wontfix
status-firefox43: --- → affected
status-firefox44: --- → affected
status-firefox45: --- → affected
tracking-firefox42: --- → ?
[Tracking Requested - why for this release]:
This is too late for 42 but Liz might be interested for 43.
status-firefox42: affected → wontfix
tracking-firefox42: ? → +
tracking-firefox43: --- → ?
Sounds like maybe we should add the card Anthony mentions in comment 37. Jorge, what do you think?
Flags: needinfo?(jorge)
We should all of these ids:
                "9870",
                "9874",
                "9875",
                "9876",
                "9877"
Flags: needinfo?(jmuizelaar)
Jeff, so I should add the IDs in comment #40 to the blocks in comment #34?
Flags: needinfo?(jorge) → needinfo?(jmuizelaar)
(In reply to Jorge Villalobos [:jorgev] from comment #41)
> Jeff, so I should add the IDs in comment #40 to the blocks in comment #34?

Yes please.
Flags: needinfo?(jmuizelaar)
Okay, both blocks have been updated.
Tracking, as a reminder to come back and check if this has been fixed in a few days.
tracking-firefox43: ? → +
tracking-firefox44: --- → +
tracking-firefox45: --- → +
This crash is still happening, but at a lower volume, still in Windows 10. 
Comments here are all about the crash happening when opening a new tab. 
Here's a link to the remaining crashes across all versions:
https://crash-stats.mozilla.com/search/?product=Firefox&signature=%24atidxx32&_facets=signature&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-signature
I went through the crash signatures and in total I found only one instance occurring on FF44. Given that, this particular crash is not so significant for 44 stability and no use tracking it.
tracking-firefox44: + → -

Comment 47

2 years ago
I have what appear to be this crash in both 42.0 32-bit and developer edition 44.0a2 64-bit:

bp-a715a376-8b3b-41d0-92d3-717a32151121
bp-80325b8a-5084-4583-b60f-3493d2151204

Always happens on opening a new tab right after launch. Updating AMD drivers to latest beta from 15.7.1 doesn't change anything. Anything I can add to help fix this?
Could you try installing and running Firefox Nightly edition, and give us a link to a crash report with that version?

Comment 49

2 years ago
(In reply to T.Toft from comment #47)
> Always happens on opening a new tab right after launch. Updating AMD drivers
> to latest beta from 15.7.1 doesn't change anything. Anything I can add to
> help fix this?

Do the 15.11 or 15.11.1 Beta drivers help?

Comment 50

2 years ago
(In reply to Milan Sreckovic [:milan] from comment #48)
> Could you try installing and running Firefox Nightly edition, and give us a
> link to a crash report with that version?

Here you go:
bp-bc2c408f-db87-483c-ae76-248922151205

I installed Nightly, launched with a new profile and got the crash when closing the default tabs opening.

Comment 51

2 years ago
(In reply to Arthur K. from comment #49)
> (In reply to T.Toft from comment #47)
> > Always happens on opening a new tab right after launch. Updating AMD drivers
> > to latest beta from 15.7.1 doesn't change anything. Anything I can add to
> > help fix this?
> 
> Do the 15.11 or 15.11.1 Beta drivers help?

No, I tried both versions of the 15.11 driver* with no luck. I'm back to 15.7.1 at the moment as it works better for me in "Fallout 4" but I'll try to update again.


*- These two:
AMD-Catalyst-15.11Beta-64Bit-Win10-Win8.1-Win7-Nov3
AMD-Catalyst-15.11.1Beta-64Bit-Win10-Win8.1-Win7-Nov14
Too late for 43. Is this still an active issue in 44/45/46?
status-firefox43: affected → wontfix
Flags: needinfo?(kairo)
I just did a quick search and it looks like the issue still exists albeit very low in volume:
> Firefox 44.0b* has 20 reports
> Firefox 44.0a2 has 3 reports
> Firefox 45.0a2 has 4 reports
> Firefox 45.0a1 has 3 reports
> Firefox 46.0a1 has 9 reports
[based on 7 days worth of data]

I'll leave the need-info to KaiRo in case he wants to add anything.
Perhaps it's an ATI version of bug 1234616...
See Also: → bug 1234616

Comment 55

2 years ago
(In reply to Liz Henry (:lizzard) (needinfo? me) from comment #52)
> Too late for 43. Is this still an active issue in 44/45/46?

At least for me it is still happening, though less so after I switched to 45 and the latest AMD driver.

bp-bc2c408f-db87-483c-ae76-248922151205

Comment 56

2 years ago
https://crash-stats.mozilla.com/search/?product=Firefox&version=43.0.3&version=43.0.2&version=43.0.1&version=43.0&signature=%24atidxx at least says that this is still happening, but the volume esp. of single signatures is comparatively small.
Flags: needinfo?(kairo)
(Assignee)

Comment 57

2 years ago
I believe at the current volumes it may be best to untrack this bug.
OK, let's untrack it then.
Thanks
tracking-firefox45: + → -

Comment 59

2 years ago
I can absolutely reproduce this bug on Firefox 45. Constantly crashes my ati drivers. Any and all websites do this. I've refreshed Firefox, turned off all extensions and plugins, created a new profile, even went as far as reinstalling windows 10. Nothing has helped besides disabling hardware acceleration which makes Firefox unusable. If Firefox was even remotely usable in that state, I'd be satisfied. I'd go and debug my version of Firefox to let you know more, but the console doesn't spit out anything useful, and not to mention there is no documentation on how to build from source to debug. My card is the R7 370, and my driver is 16.15. I have also tried to see if beta drivers work at all, and they produce the exact same results. Firefox is the only application causing my drivers to crash. Anything I can do to help figure this one out, let me know.
(Assignee)

Comment 60

2 years ago
(In reply to Ceadeus Viktus from comment #59)
> I can absolutely reproduce this bug on Firefox 45. Constantly crashes my ati
> drivers. Any and all websites do this. I've refreshed Firefox, turned off
> all extensions and plugins, created a new profile, even went as far as
> reinstalling windows 10. Nothing has helped besides disabling hardware
> acceleration which makes Firefox unusable. If Firefox was even remotely
> usable in that state, I'd be satisfied. I'd go and debug my version of
> Firefox to let you know more, but the console doesn't spit out anything
> useful, and not to mention there is no documentation on how to build from
> source to debug. My card is the R7 370, and my driver is 16.15. I have also
> tried to see if beta drivers work at all, and they produce the exact same
> results. Firefox is the only application causing my drivers to crash.
> Anything I can do to help figure this one out, let me know.

Hi Ceadeus,

I'm a little hesitant to say 'great to hear' :-). But for us this might be an interesting opportunity to study the problem.

For starters, could you post the information from the graphics section of the 'about:support' page, as well as a couple of links to crashes you're experiencing as listed in 'about:crashes'?

Thanks!
Flags: needinfo?(fisch602)

Comment 61

2 years ago
about:support -->

Adapter Description	AMD Radeon (TM) R7 370 Series
Adapter Drivers	aticfx64 aticfx64 aticfx64 amdxc64 aticfx32 aticfx32 aticfx32 amdxc32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64
Adapter RAM	4095
Asynchronous Pan/Zoom	none
ClearType Parameters	D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 100 ] D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 100 ]
Device ID	0x6811
Direct2D Enabled	true
DirectWrite Enabled	true (10.0.10240.16430)
Driver Date	2-13-2016
Driver Version	16.150.1009.0
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
Subsys ID	20151462
Supports Hardware H264 Decoding	Yes
Vendor ID	0x1002
WebGL Renderer	Google Inc. -- ANGLE (AMD Radeon (TM) R7 370 Series Direct3D11 vs_5_0 ps_5_0)
windowLayerManagerRemote	true
AzureCanvasBackend	direct2d 1.1
AzureContentBackend	direct2d 1.1
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0

since it's my graphics cards crashing and not my browser, i don't have anything recent in my about:crashes
Flags: needinfo?(fisch602)
(Assignee)

Comment 62

2 years ago
(In reply to Ceadeus Viktus from comment #61)
> about:support -->
> 
> Adapter Description	AMD Radeon (TM) R7 370 Series
> Adapter Drivers	aticfx64 aticfx64 aticfx64 amdxc64 aticfx32 aticfx32
> aticfx32 amdxc32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32
> atiumdva atiumd6a atitmm64
> Adapter RAM	4095
> Asynchronous Pan/Zoom	none
> ClearType Parameters	D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100
> Enhanced Contrast: 100 ] D [ Gamma: 2200 Pixel Structure: R ClearType Level:
> 100 Enhanced Contrast: 100 ]
> Device ID	0x6811
> Direct2D Enabled	true
> DirectWrite Enabled	true (10.0.10240.16430)
> Driver Date	2-13-2016
> Driver Version	16.150.1009.0
> GPU #2 Active	false
> GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
> Subsys ID	20151462
> Supports Hardware H264 Decoding	Yes
> Vendor ID	0x1002
> WebGL Renderer	Google Inc. -- ANGLE (AMD Radeon (TM) R7 370 Series
> Direct3D11 vs_5_0 ps_5_0)
> windowLayerManagerRemote	true
> AzureCanvasBackend	direct2d 1.1
> AzureContentBackend	direct2d 1.1
> AzureFallbackCanvasBackend	cairo
> AzureSkiaAccelerated	0
> 
> since it's my graphics cards crashing and not my browser, i don't have
> anything recent in my about:crashes

Hrm, this bug is specifically about actual crashes occurring in the AMD user mode drivers that will crash firefox. So it sounds like what you're experiencing is a different issue.
(In reply to Ceadeus Viktus from comment #59)
> I can absolutely reproduce this bug on Firefox 45. Constantly crashes my ati
> drivers. Any and all websites do this. I've refreshed Firefox, turned off
> ...
> results. Firefox is the only application causing my drivers to crash.
> Anything I can do to help figure this one out, let me know.

To summarize - drivers crash, but Firefox doesn't crash, and the problem goes away when hardware acceleration is disabled.

It may help us to know if it's content or composition.

Could you, in about:config, toggle the value of gfx.direct2d.disabled preference to true and restart Firefox.  Is the frequency of driver crashes any different?

Also, could you clarify if this is worse starting in 45, or comment 59 meant that you "still" see it in 45, but it was happening before as well?

Thanks!
Flags: needinfo?(fisch602)
Caedeus, we're still waiting on information from you re:comment 63. Currently this crash is extremely low volume with 38 reports over the last week against 46.0.1 (0.0075%). I nominate we close this bug as incomplete if we don't get any new information in the next month.
status-firefox44: affected → wontfix
status-firefox45: affected → wontfix
status-firefox46: --- → affected

Comment 65

2 years ago
Apologies for the wait. I had stopped using Firefox in the mean time after posting the original comment. I just started up Firefox again and I haven't actually gotten the drivers to crash again. I am now using the most recent version of Firefox stable. I went to roll20, and did the exact same process that would cause Firefox to crash. Firefox was completely stable. In fact it now seems to be more stable than Chrome. Chrome was freaking out with HBO NOW, while this is mostly fluff info that has nothing to do with this bug, I'm amazed how far Firefox has come recently. Would you still need info? I vote for closing the bug otherwise.

Comment 66

2 years ago
To be clear, I didn't change anything in about:config to get these results.
Ceadeus - for completeness, could you give us the most recent graphics about:support, I wouldn't mind comparing it to the one from comment 61.
Thanks for all your help!

Comment 68

2 years ago
Graphics
--------

Adapter Description: AMD Radeon (TM) R7 370 Series
Adapter Drivers: aticfx64 aticfx64 aticfx64 amdxc64 aticfx32 aticfx32 aticfx32 amdxc32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64
Adapter RAM: 4095
Asynchronous Pan/Zoom: none
ClearType Parameters: D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 100 ] D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 100 ]
Device ID: 0x6811
Direct2D Enabled: true
DirectWrite Enabled: true (10.0.10240.16430)
Driver Date: 4-29-2016
Driver Version: 16.150.2401.1002
GPU #2 Active: false
GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC)
Subsys ID: 20151462
Supports Hardware H264 Decoding: Yes
Vendor ID: 0x1002
WebGL Renderer: Google Inc. -- ANGLE (AMD Radeon (TM) R7 370 Series Direct3D11 vs_5_0 ps_5_0)
windowLayerManagerRemote: true
AzureCanvasBackend: direct2d 1.1
AzureContentBackend: direct2d 1.1
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0
Flags: needinfo?(fisch602)
OK, the driver version used to be 16.150.1009.0, and now is 16.150.2401.1002.  Anthony, can you check the telemetry for ATI drivers version 16.150.1009.0?  There are virtually no crashes with it - as Ceadeus points out, Firefox doesn't crash, so that's not surprising, but perhaps it's worth considering blocking it, as it's making the user experience worse.
Flags: needinfo?(anthony.s.hughes)
(In reply to Milan Sreckovic [:milan] from comment #69)
> Anthony, can you check the telemetry for ATI drivers version 16.150.1009.0?

7 out of 3,821,306 sessions report this driver.
Flags: needinfo?(anthony.s.hughes)
See Also: → bug 1267970
We've gone from seeing ~80 reports/week prior to June 23rd to ~14 reports/week after June 23rd and have not seen a single report since July 18, 2016. Based on the lack of reports I think this should be closed.
Status: NEW → RESOLVED
Last Resolved: 2 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.