It first appeared on Windows 8 in 17.0a1/20120815 and 16.0a2/20120815. The regression windows are: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=22288130fea2&tochange=86ee4deea55b http://hg.mozilla.org/releases/mozilla-aurora/pushloghtml?fromchange=1b8af79d5194&tochange=ecff9596cfba Bug 780035 is the only desktop bug that belongs to the two regression windows. Signature atidxx32.dll@0x624f2 More Reports Search UUID a1a43efb-b90d-4b75-8d71-842a02120817 Date Processed 2012-08-17 07:42:12 Uptime 22 Install Age 12.8 hours since version was first installed. Install Time 2012-08-16 18:53:36 Product Firefox Version 17.0a1 Build ID 20120813030532 Release Channel nightly OS Windows NT OS Version 6.2.9200 Build Architecture x86 Build Architecture Info GenuineIntel family 6 model 15 stepping 11 Crash Reason EXCEPTION_ACCESS_VIOLATION_READ Crash Address 0x501fffc App Notes AdapterVendorID: 0x1002, AdapterDeviceID: 0x68b8, AdapterSubsysID: 29991682, AdapterDriverVersion: 8.982.0.0 D2D? D2D+ DWrite? DWrite+ D3D10 Layers? D3D10 Layers+ Processor Notes EMCheckCompatibility True Adapter Vendor ID 0x1002 Adapter Device ID 0x68b8 Total Virtual Memory 4294836224 Available Virtual Memory 3891982336 System Memory Use Percentage 27 Available Page File 6353928192 Available Physical Memory 3097853952 Frame Module Signature Source 0 atidxx32.dll atidxx32.dll@0x624f2 1 atidxx32.dll atidxx32.dll@0xfdbe 2 atidxx32.dll atidxx32.dll@0x67b8 3 atidxx32.dll atidxx32.dll@0x2478e 4 atiuxpag.dll atiuxpag.dll@0x4b39 5 d3d11.dll d3d11.dll@0x8187 6 d3d11.dll d3d11.dll@0x8174 7 d2d1.dll d2d1.dll@0x6115 8 d2d1.dll d2d1.dll@0x5c31 9 d2d1.dll d2d1.dll@0x16ebd8 10 gkmedias.dll mozilla::gfx::DrawTargetD2D::~DrawTargetD2D gfx/2d/DrawTargetD2D.cpp:163 11 gkmedias.dll mozilla::gfx::DrawTargetD2D::`vector deleting destructor' 12 gkmedias.dll mozilla::RefCounted<mozilla::gfx::PathSink>::Release obj-firefox/dist/include/mozilla/RefPtr.h:64 13 gkmedias.dll mozilla::gfx::DrawTargetDual::~DrawTargetDual 14 gkmedias.dll mozilla::gfx::DrawTargetDual::`scalar deleting destructor' 15 xul.dll mozilla::RefCounted<mozilla::gfx::Path>::Release obj-firefox/dist/include/mozilla/RefPtr.h:64 16 xul.dll mozilla::RefPtr<mozilla::gfx::DrawTarget>::operator= obj-firefox/dist/include/mozilla/RefPtr.h:120 17 xul.dll mozilla::layers::ThebesLayerD3D10::CreateNewTextures gfx/layers/d3d10/ThebesLayerD3D10.cpp:499 18 xul.dll mozilla::layers::ThebesLayerD3D10::Validate gfx/layers/d3d10/ThebesLayerD3D10.cpp:212 19 xul.dll mozilla::layers::ContainerLayerD3D10::Validate gfx/layers/d3d10/ContainerLayerD3D10.cpp:356 20 xul.dll mozilla::layers::ContainerLayerD3D10::Validate gfx/layers/d3d10/ContainerLayerD3D10.cpp:358 21 xul.dll mozilla::layers::LayerManagerD3D10::Render gfx/layers/d3d10/LayerManagerD3D10.cpp:708 22 xul.dll mozilla::layers::LayerManagerD3D10::EndTransaction gfx/layers/d3d10/LayerManagerD3D10.cpp:365 23 xul.dll xul.dll@0x1047bf 24 xul.dll nsDisplayList::PaintForFrame layout/base/nsDisplayList.cpp:1040 More reports at: https://crash-stats.mozilla.com/report/list?signature=atidxx32.dll%400x624f2

The stack has absolutely nothing to do with what bug 780035 touches, so I'm almost certain the connection is a fluke.

In fact, it's not a regression in the Firefox code. It has appeared across all versions since August 15 at 18:16 UTC when Catalyst 12.8 was released.

This only happens on Win8 and is spiking in yesterday's data, probably due to this new consumer preview of Win8 that shipped.

(In reply to Robert Kaiser (:kairo@mozilla.com) from comment #3) > This only happens on Win8 and is spiking in yesterday's data, probably due > to this new consumer preview of Win8 that shipped. Just for the record, this was the RTM that shipped to developers, not a consumer preview :).

It's #9 top browser crasher in 15.0b5, #34 in 16.0a2, #22 in 17.0a1.

I was just scrolling up on a Bing search results page when I crashed.

Blocklisting this driver version for D2D on Windows 8 should stop it.

(In reply to Scoobidiver from comment #7) > Blocklisting this driver version for D2D on Windows 8 should stop it. Hrm, that's not great. We should definitely get AMD on this! Do we have someone yet who we can contact there?

(In reply to Bas Schouten (:bas) from comment #8) > (In reply to Scoobidiver from comment #7) > > Blocklisting this driver version for D2D on Windows 8 should stop it. > > Hrm, that's not great. We should definitely get AMD on this! Do we have > someone yet who we can contact there? We'll find a contact - just pinged Kev and I'll follow up.

(In reply to Bas Schouten (:bas) from comment #8) > Hrm, that's not great. We should definitely get AMD on this! I don't think AMD will do something as the latest versions don't crash (see http://developer.amd.com/download/ccc/pages/default.aspx). Bug 774672 is the right place to blocklist.

(In reply to Scoobidiver from comment #10) > Bug 774672 is the right place to blocklist. I meant bug 744672.

> the latest versions don't crash Wrong! Investigating further, I found that the atidxx32.dll version in crash correlations is the latest one: atidxx32.dll@0x624f2|EXCEPTION_ACCESS_VIOLATION_READ (344 crashes) 100% (344/344) vs. 1% (524/64449) 8.17.10.451 atidxx32.dll@0x3bf6ee|EXCEPTION_ACCESS_VIOLATION_READ (16 crashes) 100% (16/16) vs. 1% (524/64449) 8.17.10.451 atidxx32.dll@0x624f8|EXCEPTION_ACCESS_VIOLATION_WRITE (15 crashes) 100% (15/15) vs. 1% (524/64449) 8.17.10.451 atidxx64.dll@0x747db|EXCEPTION_ACCESS_VIOLATION_WRITE (52 crashes) 100% (52/52) vs. 1% (129/10679) 8.17.10.451

(In reply to Alex Keybl [:akeybl] from comment #9) > (In reply to Bas Schouten (:bas) from comment #8) > > (In reply to Scoobidiver from comment #7) > > > Blocklisting this driver version for D2D on Windows 8 should stop it. > > > > Hrm, that's not great. We should definitely get AMD on this! Do we have > > someone yet who we can contact there? > > We'll find a contact - just pinged Kev and I'll follow up. Generally, we have specific gfx driver contacts, no need to ping Kev about that, except for rare GPU vendors for which we don't have contacts yet. Are we certain that the crashes are specific to 8.982.0.0 ?

(In reply to Benoit Jacob [:bjacob] from comment #13) > Are we certain that the crashes are specific to 8.982.0.0 ? Yes, indirectly by the atidxx32.dll version in crash correlations (see comment 12) and directly by App Notes in crash reports (no AdapterDriverVersion field on Socorro in order to have correlations per driver version).

#17 top crash in 15.0b6, and there are crashes in the candidate build for Firefox 15. Are we planning to try to do something about this before we ship 15?

(In reply to Marcia Knous [:marcia] from comment #15) > Are we planning to try to do something about this before we ship 15? Windows 8 is not yet supported: http://www.mozilla.org/en-US/firefox/system-requirements.html

Official Win8 release however, will occur before 16 is shipped so it would be good to do something about this!

(In reply to Bas Schouten (:bas) from comment #17) > Official Win8 release however, will occur before 16 is shipped so it would > be good to do something about this! October 28th (see http://windowsteamblog.com/windows/b/bloggingwindows/archive/2012/07/18/windows-8-will-be-available-on.aspx) is after the release of Firefox 16. To get it fixed, the patch of bug 706908 and a new patch in bug 744672 are required along with a new AMD driver version to know the min driver version.

I realize Windows 8 is not yet a supported platform, and that General Availability (GA) is not until 26 October (see links to the windows team blog in previous posts), HOWEVER it is already released both to Volume Licensing (VL) customers, Enterprise Agreement(EA) customers, Tech Net Subscribers AND MSDN. That is quite a large swath particularly of Enterprise and Developers. Realistically it should be treated as if the official release HAS ALREADY OCCURRED as from 16 August onwards the install base of Windows 8 is only going to climb significantly. I do however agree that this is not a priority for 15, but it should be a priority for 16, which will be the available version at the time of GA.

(In reply to blackmage71 from comment #19) > it should be a priority for 16 See the + in tracking flags. It can't be fixed as long as AMD hasn't released a non buggy driver (currently Catalyst 12.8) unless HW acceleration is disabled for AMD GPUs whatever their driver version. See comment 18.

This should be handled with priority. You should also make 15.0.1 for w8 users. Next AMD driver is at least a month away so everybody who installs windows 8 on amd GPU till at least end of September will have unusable Firefox. I am experiencing 5-20 crashes per day.

(In reply to Boris Prpic from comment #21) > This should be handled with priority. You should also make 15.0.1 for w8 > users. Windows 8 will be officially released on October 26 after Firefox 16. To workaround those crashes, disable HW acceleration.

(In reply to Scoobidiver from comment #22) > (In reply to Boris Prpic from comment #21) > > This should be handled with priority. You should also make 15.0.1 for w8 > > users. > Windows 8 will be officially released on October 26 after Firefox 16. > To workaround those crashes, disable HW acceleration. Officially in stores. But its already released for Volume licensing, TechNet, MSDN, Students and it is available on torrent trackers. I think it already have 0.5% market share.

The Windows 8 market share is currently 0.2% on July (see http://marketshare.hitslink.com/operating-system-market-share.aspx?qprid=10&qpcustomd=0) and the AMD GPU one is about 33% so only 0.07% of Firefox users are currently concerned, so there's nothing to worry about. The patch will disable HW acceleration for this AMD driver version on Windows 8, hoping AMD will provide a non buggy driver soon.

Ted's looking into getting symbols so that Bas can more easily investigate this crash.

I fixed a quirk of the Microsoft symbol fetching process and re-ran it, and we should now have symbols for most of this stack. I don't think we'll have symbols for the driver at the top of the stack, although we appear to have symbols for other versions of that driver that must have come from the Microsoft symbol server.

We're starting to get new reports that are properly symbolized (except for the drivers, as I suspected): https://crash-stats.mozilla.com/report/index/3e1a646e-0cc5-49ec-bf90-c95852120905

Hrm, only the top of the stack looks similar to me :s and pretty generic. That makes a workaround less likely, although not impossible, I'll have to look at some more stacks.

Can someone at this point a)make a hardware recommendation for us to reproduce the issue or b)help find a microsoft contact who can help us check the symbol issues with newer ATI drivers

I had the crash once with a Radeon HD 6870, running with the 8.982.0.0 driver

I am crashing on 6950. Windows 8 x64 enterprise 3 monitors. Crash time pretty random. I have not found any patterns. Maybe I can say that it happens mostly when changing tabs but I am not sure. I am willing to help if there is anything I can do.

Crashing with Windows 8, Radeon 6950 8.982-120727a-144945C-ATI https://crash-stats.mozilla.com/report/index/bp-758362f3-e3fd-47c8-a29a-753dc2120910 https://crash-stats.mozilla.com/report/index/bp-5d8bf3be-7069-4963-92fc-d3f842120907

I'm running on Firefox(v15.0.1) and it keeps crashing. I have AMD Radeon HD 6670 and Windows 8 Pro installed. https://crash-stats.mozilla.com/report/index/bp-88b85703-2954-4f84-9fef-31dc12120913 https://crash-stats.mozilla.com/report/index/bp-025c2f36-e4ce-4901-9b7a-c04f62120913

Let's just block everything on AMD on Windows 8 for now. We'll presume that AMD will fix it in an update; if not, we'll have to update this blacklist entry.

Try push: https://tbpl.mozilla.org/?tree=Try&rev=8397a69ce405

Does this mean i won't have HWA on win8 with cat 12.8?

Comment on attachment 661023 [details] [diff] [review] block up to and including catalyst 12.8 on windows 8 Review of attachment 661023 [details] [diff] [review]: ----------------------------------------------------------------- Yup, checked the 8.982 <-> 12.8 number, seems correct.

Comment on attachment 661023 [details] [diff] [review] block up to and including catalyst 12.8 on windows 8 [Approval Request Comment] Bug caused by (feature/regressing bug #): no bug User impact if declined: crashes on AMD drivers on Windows 8 Testing completed (on m-c, etc.): minimal Risk to taking this patch (and alternatives if risky): No hardware acceleration for AMD drivers on Windows 8. Could accidentally block hardware acceleration on all Windows systems. String or UUID changes made by this patch: None

Try is green, pushed to inbound: https://hg.mozilla.org/integration/mozilla-inbound/rev/eb3bb2153ade

Oh! One other thing to be aware of when approving: This only blocks *known* AMD drivers on Windows 8. Even if they don't fix their crashiness with the next public release, hardware acceleration will still be allowed. (I did this on purpose, because if they do fix their crashiness, it'd be nice for hardware acceleration to start working without further work on our part. We can also deploy graphics blocklist entries via the downloaded AMO blocklist.)

Try run for 8397a69ce405 is complete. Detailed breakdown of the results available here: https://tbpl.mozilla.org/?tree=Try&rev=8397a69ce405 Results (out of 7 total builds): success: 7 Builds (or logs if builds failed) available at: http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/jdrew@mozilla.com-8397a69ce405

Comment on attachment 661023 [details] [diff] [review] block up to and including catalyst 12.8 on windows 8 [Triage Comment] Let's land this on Aurora today, Marcia can look at it on Monday, and then we can land for Beta 4. I just want to make sure we don't disable HWA across the board (as the risk evaluation suggests).

FWIW... I was getting crashes every so often running Fx18 (hourlies) under Windows 8. I lowered the overclock on my AMD HD7970 and I haven't crashed since then. Might be coincidence or a cause and effect.

Win8 64bit RTM, Sapphire Radeon HD 6950, Catalyst 12.8, on 15.0.1 and latest nightly plus latest Flash, hwa enabled, two monitors, multiple tabs, multiple windows, no other add-ons, youtube, gmail, other media sites: I haven't been able to crash with common user operations after about a half hour of fiddling. If you have suggestions as to other things I should install or try, let me know.

I think I've been getting a similar regular crash on Windows 7 (atidxx32.dll@0x3bfa5a), and I think it's because I'm Bitcoin mining at the same time with DiabloMiner (so my GPU is under heavy stress). This could be a way to try and reproduce the crash.

I updated https://wiki.mozilla.org/Blocklisting/Blocked_Graphics_Drivers

I'm still crashing with the 2012-09-16 nightly, which includes the fix. I *believe* it's because of bug 789830.

Here's a crash report: https://crash-stats.mozilla.com/report/index/bp-3500cd45-d154-4248-a45b-03c1b2120916

Comment on attachment 661023 [details] [diff] [review] block up to and including catalyst 12.8 on windows 8 [Triage Comment] Now that we have bug 783517 and we haven't heard of major regressions, let's land on Beta as well.

Sid, I'm going to leave this as closed since we landed the patch. Can you open a follow-up bug if this persists? https://hg.mozilla.org/releases/mozilla-beta/rev/bb118ab06ae5

This block seems to have spread to Windows 7 as well. I get "Blocked for your graphics driver version." in about:support.

In 16b4 HW acceleration (direct2d) is also disabled in Windows 7 with my 5770. Was this intended to happen? gfx.blacklist.direct2d is always set to 3 even if I reset the value. Can I disable this ****? My firefox doesn't have any crashes related to drivers. Here is my about:support entry: Name Firefox Version 16.0 User-Agent Mozilla/5.0 (Windows NT 6.1; WOW64; rv:16.0) Gecko/20100101 Firefox/16.0 Grafik Karten-Beschreibung ATI Radeon HD 5700 Series Vendor-ID 0x1002 Geräte-ID 0x68b8 Karten-RAM 1024 Karten-Treiber aticfx64 aticfx64 aticfx64 aticfx32 aticfx32 aticfx32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64 Treiber-Version 8.982.0.0 Treiber-Datum 7-27-2012 Direct2D aktiviert Wurde auf Grund Ihrer Grafiktreiberversion blockiert. DirectWrite aktiviert false (6.1.7601.21935) ClearType-Parameter Gamma: 2200 Pixel Structure: RGB ClearType Level: 100 Enhanced Contrast: 300 WebGL-Renderer Google Inc. -- ANGLE (ATI Radeon HD 5700 Series) -- OpenGL ES 2.0 (ANGLE 1.0.0.1041) GPU-beschleunigte Fenster 1/1 Direct3D 9

For the block in Vista and 7, see https://wiki.mozilla.org/Blocklisting/Blocked_Graphics_Drivers

(In reply to Scoobidiver from comment #55) > For the block in Vista and 7, see > https://wiki.mozilla.org/Blocklisting/Blocked_Graphics_Drivers I know this ;) Why is Firefox blocking the AMD 12.8 driver on Windows 7? This is my question. When I reset the blacklisting in about:config it changes back again after a few minutes. How can I stop this?

(In reply to André Ziegler from comment #56) > I know this ;) Why is Firefox blocking the AMD 12.8 driver on Windows 7? It's explained in the bug of the downloaded blacklist section.

ok, I've overlocked that HW acceleartion is now blocked for all windows versions. using the force direct2D setting in about:config works for me to unlock it.

12.9 Beta driver is out and it looks like it is not crashing anymore. HWA is enabled.

I got one crash on 12.9: https://crash-stats.mozilla.com/report/index/bp-46ec9bac-b3fa-4be4-ad54-b4be22120929 Not sure if it is related to this but it crashed in driver.

(In reply to Boris Prpic from comment #60) > I got one crash on 12.9: > https://crash-stats.mozilla.com/report/index/bp-46ec9bac-b3fa-4be4-ad54- > b4be22120929 > > Not sure if it is related to this but it crashed in driver. Your crash report indicates you're running Firefox 17.0a2. Please update to the latest Nightly build then try again.

Oh never mind me. I guess 17.0a2 is Aurora. I thought it was an old nightly. :-/

Please verify by checking Socorro and also manually spotchecking that AMD drivers are blocked on Windows 8.

Mozilla/5.0 (Windows NT 6.2; WOW64; rv:17.0) Gecko/17.0 Firefox/17.0 Verified that the ATI driver version 8.982.0.0 is blocked on Win 8 and there are no crashes for this driver version in Socorro. Graphics Adapter Description AMD Radeon HD 6450 Vendor ID 0x1002 Device ID 0x6779 Adapter RAM 1024 Adapter Drivers aticfx64 aticfx64 aticfx64 aticfx32 aticfx32 aticfx32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64 Driver Version 8.982.0.0 Driver Date 7-27-2012 Direct2D Enabled Blocked for your graphics driver version. Try updating your graphics driver to version > 12.8 or newer. DirectWrite Enabled false (6.2.9200.16384) ClearType Parameters ClearType parameters not found WebGL Renderer Blocked for your graphics driver version. Try updating your graphics driver to version > 12.8 or newer. GPU Accelerated Windows 0. Blocked for your graphics driver version. Try updating your graphics driver to version > 12.8 or newer. AzureCanvasBackend cairo AzureFallbackCanvasBackend none AzureContentBackend none

Mozilla/5.0 (Windows NT 6.2; WOW64; rv:18.0) Gecko/20100101 Firefox/18.0 ATI driver version 8.982.0.0 is blocked on Win 8. Also, there are no crashes for Firefox 18 with this driver version in Socorro.

Looking at this crash: https://crash-stats.mozilla.com/report/index/d20cc5b9-28e8-49c3-aedd-009b42160520 46.0.1 Windows 8 AMD graphics 8.892.0.0 driver D3D11 Layers+ Yet, this should be blocked. Is it possible the user is forcing this? Jeff, we don't have the crash annotation with major preferences until 47, not sure if there is another way to check. Anthony, can you find if there are more of the crashes that should be blocked by this bug?

https://crash-stats.mozilla.com/search/?product=Firefox&adapter_driver_version=%3D8.892.0.0&_facets=signature&_facets=version&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-version There are 344 crashes reported in the last week with driver 8.892.0.0. 80.53% are with Firefox 46 or later. 4.65% appear to be graphics related. 0.58% appear to be related to atidxx32.dll. In terms of platform there are none on Windows 8. * 78.78% are on Windows 7 * 4.07% are on Windows XP * 1.16% are on Windows Vista Looking at just the atidxx32.dll crashes it is 100% on Windows Vista. It's not clear to me that blocklisting is supposed to block the type of crash you referenced above.

Sorry Anthony, I transposed the driver version in comment 66 - it's 8.982.0.0 we care about, rather than 8.892.0.0.

(In reply to Milan Sreckovic [:milan] from comment #68) > Sorry Anthony, I transposed the driver version in comment 66 - it's > 8.982.0.0 we care about, rather than 8.892.0.0. Here are all the crashes with Firefox 46.0.1 on Windows 8 with driver 8.982.0.0 with D3D11 Layers+ https://crash-stats.mozilla.com/search/?product=Firefox&adapter_driver_version=%3D8.982.0.0&app_notes=~Layers%2B&platform_pretty_version=%3DWindows%208&version=46.0.1&_facets=signature&_facets=version&_facets=platform_pretty_version&_facets=app_notes&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-signature All in all there are only 23 reports (~3/day) most of which are shutdown hangs. As far as I can tell none of these are graphics related and none involve atidxx32.dll. I'm not sure there is anything in this data which points to a problem related to this bug report.

So much has changed over the course of this bug I'm going to suggest that we drop it and open a new one for any new information or problems.