Harvested from chofmann's diff report, and new to B5. http://tinyurl.com/24g7oaf to the crashes. This is currently the #29 top crash in Beta 5. All the crash reports are from Win 7 users. One person crashed on Win 7 on this site: http://ie.microsoft.com/testdrive/Performance/01FlyingImages/Default.html Frame Module Signature [Expand] Source 0 igd10umd32.dll igd10umd32.dll@0x1195c

File Name: igd10umd32.dll Product Name: Intel Graphics Accelerator Drivers for Windows Vista(R)

wonder if we should open up a new component for graphic drivers like we have for plugins...

After analyzing some crashes : DeviceID : 2A02 or 2A12 = GL960/GLE960 or GM965/GME965 (GMA X3100) Module version : 8.14.10.1930 or 8.15.10.1867 mainly The up-to-date version for GMA X3100 is 8.15.10.1930. It will be fixed by bug 594877 : no D2D if Intel driver is not up-to-date

In b7pre, even with the Intel driver Blacklisting, it happens. OS : Windows 7 DeviceID : 2A02 (GL960 < GMA X3100), 2A12 (GM965 < GMA X3100), 2982 (GMA X3500) Driver version : 8.15.10.1930 igd10umd32.dll version : 8.14.10.1930 igd10umd32.dll : LDDM user mode driver for Intel Graphic Technology

(In reply to comment #4) > In b7pre, even with the Intel driver Blacklisting, it happens. > OS : Windows 7 > DeviceID : 2A02 (GL960 < GMA X3100), 2A12 (GM965 < GMA X3100), 2982 (GMA X3500) > Driver version : 8.15.10.1930 > igd10umd32.dll version : 8.14.10.1930 > > igd10umd32.dll : LDDM user mode driver for Intel Graphic Technology Is this with D2D or just D3D9?

Crashes occur before 20100905 builds and after 20100916 builds. D2D is OFF for every Intel cards from 20100906 to 20100915 builds. So it is a D2D issue.

Should this be a blocker? The build is unusable unless this bug is fixed.

(In reply to comment #7) > Should this be a blocker? The build is unusable unless this bug is fixed. Which driver version is this? And shouldn't it be getting blacklisted looking at the version number?

(In reply to comment #4) > In b7pre, even with the Intel driver Blacklisting, it happens. > OS : Windows 7 > DeviceID : 2A02 (GL960 < GMA X3100), 2A12 (GM965 < GMA X3100), 2982 (GMA X3500) > Driver version : 8.15.10.1930 > igd10umd32.dll version : 8.14.10.1930 Wait... in order to work, I need to understand this: is your driver version 8.15.10.1930 or 8.14.10.1930 ?? Or is there an inconsistency between the driver installed on your system (igd10umd32.dll) and what the Windows registry believes ("driver version") ??

> Wait... in order to work, I need to understand this: is your driver version > 8.15.10.1930 or 8.14.10.1930 ?? > Or is there an inconsistency between the driver installed on your system > (igd10umd32.dll) and what the Windows registry believes ("driver version") ?? This is results of the analysis of crash stats. For me, my graphic driver and igd10umd32.dll versions are the same : 8.15.10.2202. An Intel graphic card driver for Windows 7 32-bit contains several dll : * igd10umd32.dll : LDDM user mode driver for Intel Graphics Technology (DX10) * igdumd32.dll : LDDM user mode driver for Intel Graphics Technology (DX9) * igdumdx32.dll : LDDM user mode driver for Intel Graphics Technology * ig4icd32.dll : Open GL driver for Intel Graphics Accelerator * ig4dev32.dll : Open GL device driver for Intel Graphics Accelerator These 3 cards are DX10 cards of first generation because they can not use 8.15.10.2202. I see two possible reasons for the 8.14 pattern in dll version : 1. related to first generation of DX10 cards 2. related to some users that upgraded from Vista to 7 According to me, reason #1 is the more plausible because others cards should have been implicated with reason #2. So, a D2D blacklisting of the following deviceID must be done : 2982, 2983, 2A02, 2A03, 2A12, 2A13

(In reply to comment #10) > So, a D2D blacklisting of the following deviceID must be done : > 2982, 2983, 2A02, 2A03, 2A12, 2A13 So, I understand that you know that first-hand for 2A02, 2A03, 2A12, 2A13. How do you know about 2982, 2983 ? How can we be sure that there are no other device IDs to block in the GMA X3000 family? Why don't we just block the whole GMA X3000 family an earlier for Direct2D, making Direct2D on Intel happen only with the X4500/HD generation ?

To be clear, the other reason why I'm inclined to block all Intel cards inferior to GMA X3000 anyway, is that we have had lots of reports of people complaining that with Intel graphics, D2D actually makes Firefox slower.

Well, I mean they complain that "hardware acceleration" makes Firefox slower, I don't know if it's specifically D2D or if it's layers, anyway. But I was supposing that older Intel cards either wouldn't support D2D at all or would support it poorly, since it seems to require a modern GPU.

> So, I understand that you know that first-hand for 2A02, 2A03, 2A12, 2A13. > How do you know about 2982, 2983 ? see comment 4 > How can we be sure that there are no other device IDs to block in the GMA > X3000 family? Why don't we just block the whole GMA X3000 family an earlier > for Direct2D, making Direct2D on Intel happen only with the X4500/HD > generation ? We have enough feedback from 4.0b5/b6 crash stats. > we have had lots of reports of people complaining that with Intel graphics, > D2D actually makes Firefox slower we have also a lot of people that complained they didn't have D2D with Intel cards (see bug 594025). Let people have choice to disable HW acceleration by their own. Blacklisting must be only done to prevent driver crashes.

(In reply to comment #14) > > So, I understand that you know that first-hand for 2A02, 2A03, 2A12, 2A13. > > How do you know about 2982, 2983 ? > see comment 4 ah ok. > > > How can we be sure that there are no other device IDs to block in the GMA > > X3000 family? Why don't we just block the whole GMA X3000 family an earlier > > for Direct2D, making Direct2D on Intel happen only with the X4500/HD > > generation ? > We have enough feedback from 4.0b5/b6 crash stats. ok, makes sense. > > > we have had lots of reports of people complaining that with Intel graphics, > > D2D actually makes Firefox slower > we have also a lot of people that complained they didn't have D2D with Intel > cards (see bug 594025). Let people have choice to disable HW acceleration by > their own. > Blacklisting must be only done to prevent driver crashes. I don't know if I agree with that: most users won't bother trying that and first impressions matter; and one could also say "Let people have choice to *enable* HW acceleration by their own." But until we reach a decision on this, I'll only blacklist the crashy cases.

We have crashes on these DeviceIDs ---> blocking. Bas, what's your opinion on the above debate between blocking only what's crashy, and blocking all "old" Intel cards so we avoid complaints from users about slowness? Do you know if on old Intel cards we are actually running Direct2D or if it's detecting that the card is too old and just not using it?

(In reply to comment #8) > (In reply to comment #7) > > Should this be a blocker? The build is unusable unless this bug is fixed. > > Which driver version is this? And shouldn't it be getting blacklisted looking > at the version number? I have crashes on start up with my laptop running Windows 7 Home Premium 64-bit Graphics Adapter Description Intel(R) HD Graphics Vendor ID 8086 Device ID 0046 Adapter RAM Unknown Adapter Drivers igdumd64 igd10umd64 igdumdx32 igd10umd32 Driver Version 8.15.10.2189 Driver Date 7-28-2010 Direct2D Enabled false DirectWrite Enabled false GPU Accelerated Windows

Michael, Could you provide the crash id to know if your crash is related to this bug : https://developer.mozilla.org/En/How_to_get_a_stacktrace_for_a_bug_report

(In reply to comment #17) > (In reply to comment #8) > > (In reply to comment #7) > > > Should this be a blocker? The build is unusable unless this bug is fixed. > > > > Which driver version is this? And shouldn't it be getting blacklisted looking > > at the version number? > > I have crashes on start up with my laptop running Windows 7 Home Premium 64-bit > Graphics > > Adapter Description > Intel(R) HD Graphics > > Vendor ID > 8086 > > Device ID > 0046 That's Intel Mobile HD Graphics, so a more recent generation than the GMA X3000 that are otherwise discussed here. > Driver Version > 8.15.10.2189 OK, recent driver version (we block < 2182) I wonder what we should do about this. First of all, could you please confirm that if you disable Direct2D, it doesn't crash anymore? To do that: 1. Run firefox in safe mode: from the windows command line, run firefox -safe-mode 2. Go to about:config, search for direct2d, disable it by flipping the relevant preferences. 3. Restart firefox (not in safe mode). Does it crash?

(In reply to comment #16) > Created attachment 478754 [details] [diff] [review] > block Direct2D for 6 deviceIDs from the GMA X3000 family > > We have crashes on these DeviceIDs ---> blocking. > > Bas, what's your opinion on the above debate between blocking only what's > crashy, and blocking all "old" Intel cards so we avoid complaints from users > about slowness? Do you know if on old Intel cards we are actually running > Direct2D or if it's detecting that the card is too old and just not using it? At this point in time DX10->D2D unless the card is blocklisted.

http://hg.mozilla.org/mozilla-central/rev/e619706a27d7 leaving open until the issues reported in comment 17 are resolved

Notice that in bug 593678 comment 60 we have someone reporting a crash on startup with this card (device 0x0046, Intel Mobile HD Graphics). Should we just blacklist it for now and get Intel to do their homework ? :-P

Sorry, I meant that in that bug we have someone reporting bad rendering (completely black window) with that card.

(In reply to comment #19) > > OK, recent driver version (we block < 2182) > > I wonder what we should do about this. First of all, could you please confirm > that if you disable Direct2D, it doesn't crash anymore? > > To do that: > > 1. Run firefox in safe mode: from the windows command line, run firefox > -safe-mode > > 2. Go to about:config, search for direct2d, disable it by flipping the relevant > preferences. > > 3. Restart firefox (not in safe mode). Does it crash? 1. Firefox starts in safe mode without crashes. 2. I have the 2 variables set to false: gfx.direct2d.disabled gfx.direct2d.force-enabled 3. When I restart Firefox normally (not in safe mode) it still crashes. Tested with the 0927 nightly build.

Sorry for the spam, just re-read my post: I have changed the setting for gfx.direct2d.disabled from false to true and Firefox now starts without crashes. Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b7pre) Gecko/20100927 Firefox/4.0b7pre ID:20100927041306

Do you have the hotfixes necessary for IE9 installed?

igd10umd32.dll@0x1e1ba0 crash signature happens also with the graphic driver version 8.14.10.1930 and for at least deviceID 2A02 (GL960 < GMA X3100). It is #266 top crasher in 4.0b6 for the last 2 weeks. Frame Module Signature [Expand] Source 0 igd10umd32.dll igd10umd32.dll@0x1e1ba0 1 igd10umd32.dll igd10umd32.dll@0x1f0cf 2 ntdll.dll ntdll.dll@0x703dc > Do you have the hotfixes necessary for IE9 installed? Crashes happen until now. So if some MS hotfixes can prevent this crash or others crashes, it is not in the automatic Windows 7 update.

I have IE9 public beta and Windows 7 Service Pack 1 Beta (v178) installed but still getting crashes until I disabled Direct2D.

(In reply to comment #24) > (In reply to comment #19) > > > > OK, recent driver version (we block < 2182) > > > > I wonder what we should do about this. First of all, could you please confirm > > that if you disable Direct2D, it doesn't crash anymore? > > > > To do that: > > > > 1. Run firefox in safe mode: from the windows command line, run firefox > > -safe-mode > > > > 2. Go to about:config, search for direct2d, disable it by flipping the relevant > > preferences. > > > > 3. Restart firefox (not in safe mode). Does it crash? > > 1. Firefox starts in safe mode without crashes. > > 2. I have the 2 variables set to false: > gfx.direct2d.disabled > gfx.direct2d.force-enabled > > 3. When I restart Firefox normally (not in safe mode) it still crashes. Thanks, that's interesting, proves that the crashes with device 0x0046 are not D2D-specific. Can you please then start in safe mode, set layers.accelerate-none=true. That will disable D3D compositing. Does that avoid the crash? If yes, can you then try reenabling D2D (do the opposite of comment 24)? The reason why I'm asking is that I would like to know if the crash is only with D3D compositing or if it's a generalized crash with any form of acceleration.

(In reply to comment #25) > Sorry for the spam, just re-read my post: > > I have changed the setting for gfx.direct2d.disabled from false to true and > Firefox now starts without crashes. Aaah ! OK, thanks! So, definitely a D2D crash with device 0x0046. Ignore my previous comment then.

Michael: Can you please also have a look at bug 593678: someone with the same card as you reports that everything is black for him with D2D disabled and D3D enabled, do you experience similar issues?

Michael, Can you provide a crash id for your crash with D2D enabled ? Benoit, One crash signature means one cause. Until now, we don't know the crash signature of the crash with DeviceID 0046. DeviceID 2A42 and 0046 crashes FF but not with this signature (see bug 600152)

Bas, I don't agree with HD graphics blacklisting as we don't know if it is a driver crash or not. In case of a driver crash, we need to have enough crash stats to know if it is caused by DeviceID or graphic driver version ?

(In reply to comment #34) > Bas, > I don't agree with HD graphics blacklisting as we don't know if it is a driver > crash or not. In case of a driver crash, we need to have enough crash stats to > know if it is caused by DeviceID or graphic driver version ? Hrm, we could delay this, but we can't have this crash ocuring in beta 7 I believe.

(In reply to comment #28) > I have IE9 public beta and Windows 7 Service Pack 1 Beta (v178) installed but > still getting crashes until I disabled Direct2D. from http://msdn.microsoft.com/en-us/ie/ff959805.aspx#_Known_Installation_Issues "Windows 7 SP1 Beta does not contain all the latest bug fixes for the Direct 2D (D2D) technology used by Internet Explorer 9. We recommend that you uninstall Windows 7 SP1 beta before installing Internet Explorer 9, to get the latest security updates."

Tough call: - we shouldn't ship b7 with such a crash - but if we blacklist we dont get user data anymore. Do we know if the crash still exists with the current driver version (.2202 ??) All I know is that it's still crashing with .2189. Notice that even if we blacklist, we have a force-enabled preference for direct2d, which ignores the blacklist (Bas: correct?), so we could still ask people to test.

(In reply to comment #34) > Bas, > I don't agree with HD graphics blacklisting as we don't know if it is a driver > crash or not. In case of a driver crash, we need to have enough crash stats to > know if it is caused by DeviceID or graphic driver version ? I'll let Bas and you discuss this --- as if it's not a driver crash then it would be a crash in Bas's code. But that would surprise me: why would such a crash happen only with 0x0046 ??

Let's see what the result of uninstalling the SP1 beta is. If it doesn't help I'll contact the intel guys. As far as I'm concerned this has to be a driver bug in any case, you shouldn't be able to really crash inside the driver from D2D, after which it goes through 2 levels of error checking (D2D and D3D) before it arrives at the UMD.

We should just get this into the tree, and take it up with Intel afterwards.

OK, here's a new version of this patch. The change is that instead of blacklisting Mobile HD Graphics, we require the latest driver .2202 for x4500/HD cards. The rationale comes from bug 600152. It's twofold: - MobileHD is not the only crashy card, half of the 4500/HD series turns out to be. So just blacklisting MobileHD is not a full solution anyhow. - We have yet to receive any such crash report with driver .2202. So either the bug is fixed in .2202 or users haven't updated yet. Either way, in the very short term (for beta 7), requiring .2202 will avoid crashing. ---> next step: ask people to upgrade to .2202 and see if crash persists...

@ Michael: please upgrade driver to .2202 and tell us if it's still crashing!

http://hg.mozilla.org/mozilla-central/rev/b07c0925efe5 Michael: Please test tomorrow's nightly (so it has this fix) with Intel driver .2202 (or newer) and confirm it doesn't crash anymore. Leaving open for now.

I tag it as "closed" "fixed". Let's wait to change "closed" in "verified"

MS hotfix KB2028560 modified d2d1.dll, d3d10*.dll from version 6.1.7600.16385 to 6.1.7600.20781 May be the blacklisting of GMA X3000 series is no more useful. In that case, a new bug will be filed Chris, Is it possible to sort out by d2d1.dll version the crashes of 4.0b5 and b6 from the last 4 weeks ?

202 reports found in in 20100901-crashdata.csv 20100902-crashdata.csv 20100903-crashdata.csv 20100904-crashdata.csv 20100905-crashdata.csv 20100906-crashdata.csv 20100907-crashdata.csv 20100908-crashdata.csv 20100909-crashdata.csv 20100910-crashdata.csv 20100911-crashdata.csv 20100912-crashdata.csv 20100913-crashdata.csv 20100914-crashdata.csv 20100915-crashdata.csv 20100916-crashdata.csv 20100917-crashdata.csv 20100918-crashdata.csv 20100919-crashdata.csv 20100920-crashdata.csv 20100921-crashdata.csv 20100922-crashdata.csv 20100923-crashdata.csv 20100924-crashdata.csv 20100925-crashdata.csv 20100926-crashdata.csv 20100927-crashdata.csv 20100928-crashdata.csv 150 AdapterVendorID: 8086, AdapterDeviceID: 2a02 d2d1.dll 6.1.7600.16385 igd10umd32.dll 8.14.10.1930 35 AdapterVendorID: 8086, AdapterDeviceID: 2a02 d2d1.dll 6.1.7600.20781 igd10umd32.dll 8.14.10.1930 7 AdapterVendorID: 8086, AdapterDeviceID: 2982 d2d1.dll 6.1.7600.16385 igd10umd32.dll 8.14.10.1930 5 AdapterVendorID: 8086, AdapterDeviceID: 2a12 d2d1.dll 6.1.7600.20781 igd10umd32.dll 8.14.10.1930 2 AdapterVendorID: 8086, AdapterDeviceID: 2a02 d2d1.dll 6.1.7600.20710 igd10umd32.dll 8.14.10.1930 1 ___checking igd10umd32.dll@0x1e1ba0 1 AdapterVendorID: 8086, AdapterDeviceID: 2a12 d2d1.dll 6.1.7600.16385 igd10umd32.dll 8.14.10.1930 1 AdapterVendorID: 8086, AdapterDeviceID: 2a02 d2d1.dll 6.1.7601.16562 igd10umd32.dll 8.14.10.1930

According to comment 18, MS hotfix KB2028560 doesn't change anything. So we keep the solution of this bug : blocks GMA X3000 for D2D.

(In reply to comment #49) > According to comment 18, MS hotfix KB2028560 doesn't change anything. So we > keep the solution of this bug : blocks GMA X3000 for D2D. Any new deviceID in this family, that we should add to the blocklist? Currently blocking: // see bug 595364 comment 10 static const PRUint32 deviceFamilyIntelBlockDirect2D[] = { 0x2982, /* IntelG35_1 */ 0x2983, /* IntelG35_2 */ 0x2A02, /* IntelGL960_1 */ 0x2A03, /* IntelGL960_2 */ 0x2A12, /* IntelGM965_1 */ 0x2A13, /* IntelGM965_2 */ 0 };

(In reply to comment #9) > (In reply to comment #4) > > In b7pre, even with the Intel driver Blacklisting, it happens. > > OS : Windows 7 > > DeviceID : 2A02 (GL960 < GMA X3100), 2A12 (GM965 < GMA X3100), 2982 (GMA X3500) > > Driver version : 8.15.10.1930 > > igd10umd32.dll version : 8.14.10.1930 > > Wait... in order to work, I need to understand this: is your driver version > 8.15.10.1930 or 8.14.10.1930 ?? > > Or is there an inconsistency between the driver installed on your system > (igd10umd32.dll) and what the Windows registry believes ("driver version") ?? I still haven't understood this; could someone please explain me? Is this inconsistency between "driver version" and "igd10umd32.dll version" normal?

> Is this inconsistency between "driver version" and "igd10umd32.dll version" > normal? I don't know. I think we have enough stats to know if it is a common thing or not. Chris, Is it possible for any crash (one or two weeks of crash-stats) whatever the signature to select those that have the following "App Notes" : AdapterVendorID: 8086, AdapterDeviceID: 2a02 AdapterVendorID: 8086, AdapterDeviceID: 2912 AdapterVendorID: 8086, AdapterDeviceID: 2a12 Then, sort them out by : * Windows version * igd10umd32.dll version

I think we need socorro 1.7 (tracked in Bug 600859) and an update to the .csv files to get at that kind of data. I've made the pitch for pushing that release tonight, but sounds like it will be thursday.

Filed Bug 601997 to start fresh discussion on the 8.14 vs 8.15 version mismatch between the dll and the DriverVersion registry key, and perhaps contact Intel about that...

I posted some of the data we started getting out of socorro last night over in https://bugzilla.mozilla.org/show_bug.cgi?id=601997#c14

There has been no crash with these 2 signatures (igd10umd32.dll@0x1195c and igd10umd32.dll@0x1e1ba0) since b7pre/20100928 build. So I change bug status to verified.

I have an Intel G965 (Device ID 29A2). Does this fix blocks my Minefield from using Direct2D irrespective to the block list implemented under this bug: https://bugzilla.mozilla.org/show_bug.cgi?id=594877 ?

Device 29A2 is not in the list of devices for which we block Direct2D in this fix. FYI the list is: 0x2982, /* IntelG35_1 */ 0x2983, /* IntelG35_2 */ 0x2A02, /* IntelGL960_1 */ 0x2A03, /* IntelGL960_2 */ 0x2A12, /* IntelGM965_1 */ 0x2A13, /* IntelGM965_2 */ So if you're not getting hardware acceleration the most likely reason is an old driver. If you're on win7, you need at least driver version 8.15.10.1930. On vista, you need 7.15.10.1666.

I do have the latest required drivers, i.e. the 8.15.10.1930. Direct3D 9 is enabled by default. DirectWrite can be manually enabled. Direct2D, however, refuses to work, even if forced manually. This happens on both Win7x86 and Minefieldx86 on Win7x64, on the same machine. Minefieldx64 crashes, so I cannot test that. I think I should file some new bugs then.

> This happens on both Win7x86 and Minefieldx86 on Win7x64, on the same > machine. Minefieldx64 crashes, so I cannot test that. I think I should file > some new bugs then. Minefieldx64 crash is bug 601355. Yes, you can file bugs about D2D feature that can not be forced and D2D/DW disabled with a compatible Intel driver.

I have just made a patch in bug 604771, that allows to spoof graphics information and circumvent the blocklist. Running with graphics info spoofed as yours, every graphics feature (D2D, DW, etc) is still enabled for me. So this regression is not caused by the blocklist.

(In reply to comment #60) > > I think I should file some new bugs then. > Yes, you can file bugs about D2D feature that can not be forced and D2D/DW > disabled with a compatible Intel driver. I've just reported bug 605053, which is somehow related with this. ;-)

Bug 604771 is now landed, you can now circumvent the blocklists. Instructions are given in that bug, but the short version is that just doing MOZ_GFX_SPOOF_VENDOR_ID=0 firefox in a terminal ("command line prompt"), i.e. running firefox with the environment variable MOZ_GFX_SPOOF_VENDOR_ID set to 0, will cause the whole blacklist to be ignored. I am especially interested in hearing from people with cards listed in comment 58. They currently are completely blacklisted because all Intel drivers were very buggy with them, as of .2202. But intel has recently published a newer driver version, and the only way we can learn if it fixes it, is by having people with these cards run with MOZ_GFX_SPOOF_VENDOR_ID=0 and report if it's still crashy.