Open Bug 1479795 Opened Last year Updated 2 months ago

Crash in igd10iumd64.dll with Intel driver 23.20.16.4973

Categories

(Core :: Graphics, defect, P3, critical)

All
Windows 10
defect

Tracking

()

ASSIGNED
Tracking Status
firefox-esr52 --- wontfix
firefox-esr60 --- wontfix
firefox-esr68 --- affected
firefox61 --- wontfix
firefox62 --- wontfix
firefox63 --- wontfix
firefox67 --- wontfix
firefox68 --- affected
firefox69 --- affected

People

(Reporter: philipp, Assigned: aosmond)

References

Details

(Keywords: crash, Whiteboard: [gfx-noted])

Crash Data

Attachments

(1 file)

This bug was filed from the Socorro interface and is
report bp-f11ffa7b-64e5-4fb2-bda9-ecf350180720.
=============================================================

Top 10 frames of crashing thread:

0 igd10iumd64.dll igd10iumd64.dll@0x903f6f 
1 d3d11.dll CContext::UMQueryOM_DepthState_ 
2 igd10iumd64.dll igd10iumd64.dll@0xbf705 
3 igd10iumd64.dll igd10iumd64.dll@0x114de 
4 d3d11.dll d3d11.dll@0x17af3f 
5 igd10iumd64.dll igd10iumd64.dll@0xbf5e0 
6 igd10iumd64.dll igd10iumd64.dll@0x8ebad3 
7 d3d11.dll CDevice::CondObjectLock::CondObjectLock 
8 d3d11.dll CContext::TID3D11DeviceContext_Draw_<2> 
9 d2d1.dll CD3DDeviceLevel1::DrawDynamicVB 

=============================================================

these content crashes are on the increase for users on windows 10 with an intel gpu. the graphics driver version heavily skews towards the recently released v23.20.16.4973, so perhaps we can attempt some workaround through blocklisting it or else reach out to intel about this...

Adapter device id facet
1 	0x5912 	2848 	52.30 %
2 	0x5916 	1397 	25.66 %
3 	0x5917 	455 	8.36 %
4 	0x591b 	416 	7.64 %
5 	0x5902 	190 	3.49 %
User reported this for Intel HD Graphics 610:

https://downloadcenter.intel.com/product/98910/Intel-HD-Graphics-610

Latest version is 24.20.100.6194. There doesn't seem to be any issues with that driver and Firefox 60+ (although there are some crashes with 52 ESR / 32-bit). Probably safe just to block it and let them upgrade.
Priority: -- → P3
Update crash signatures to take the report sets which are dominated (90+%) by users with adapter version 23.20.16.4973.
Crash Signature: [@ igd10iumd64.dll | CContext::UMQueryOM_DepthState_] [@ igd10iumd32.dll | CContext::UMQueryOM_DepthState_] → [@ igd10iumd64.dll | CContext::UMQueryOM_DepthState_] [@ igd10iumd32.dll | CContext::UMQueryOM_DepthState_] [@ igd10iumd64.dll | DDIContextST::DDIContextST ] [@ @0x0 | igd11dxva64.dll | CSampler::GetDesc] [@ igd10iumd64.dll | CSampler::GetDesc] [@ ig…
Duplicate of this bug: 1454318
Summary: Crash in igd10iumd64.dll | CContext::UMQueryOM_DepthState_ with Intel driver 23.20.16.4973 → Crash in igd10iumd64.dll with Intel driver 23.20.16.4973
Duplicate of this bug: 1479778
Adapter device ID graphics from crash reports suggest at least 5 different adapter device IDs are seeing significant crash rates, so it is likely the driver itself than some specific combination of driver and device.
Assignee: nobody → aosmond
Status: NEW → ASSIGNED
Whiteboard: [gfx-noted]
How common is this driver?
Flags: needinfo?(aosmond)
(In reply to Andrew Osmond [:aosmond] from comment #6)
> Created attachment 8997036 [details] [diff] [review]
> 0001-Bug-1479795-Block-Intel-graphics-driver-23.20.16.497.patch, v1
> 
> try:
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=3566e7470df99dd1b27779e450d3c14c34eb0484

1.93% from gfx telemetry
Flags: needinfo?(aosmond)
Of all users? Are all the crashes Direct2D related?
Flags: needinfo?(aosmond)
(In reply to Jeff Muizelaar [:jrmuizel] from comment #9)
> Of all users? Are all the crashes Direct2D related?

That's probably what the telemetry search result is yielding:

https://firefoxgraphics.github.io/telemetry/#view=hwsearch&vendor=&drivers=23.20.16.4973

It would be good to know how many of them are Windows 10, since there doesn't seem to be non-Windows 10 crashes. Maybe it is a smaller subset of the population (1.92% seems high to block, even for 200 crashes/day...).

As for examples from the crash reports, here are a few samples:

D2D:

https://crash-stats.mozilla.com/report/index/2aeec56f-fa99-4037-bdf7-2d4eb0180802
https://crash-stats.mozilla.com/report/index/ac3711e8-e3ec-4341-b239-2be1a0180802
https://crash-stats.mozilla.com/report/index/ab9fdc86-a12b-4d2d-9a67-5574f0180802

Video related (I think):

https://crash-stats.mozilla.com/report/index/0d3117fe-13db-4a61-bd35-c496b0180802

Even WebRender:

https://crash-stats.mozilla.com/report/index/a500870e-6f58-40dc-a820-7da8a0180731
Flags: needinfo?(aosmond)
(In reply to Jeff Muizelaar [:jrmuizel] from comment #7)
> How common is this driver?

This is a WHQL driver shipped via Windows Update so it's going to be pretty common.
Some additional information:
- 23.20.16.4973 is the most common driver on Gen9 hardware with 14.07% of Gen9 users.
- We don't have great evidence to suggest that 23.20.16.4973 is the only driver affected.
- It's possible to run this driver on the quantum reference hardware by manually installing it.
- It's not clear that black listing this driver will give an overall improvement in user experience.
https://downloadcenter.intel.com/download/27894/Intel-Graphics-Driver-for-Windows-10

Maybe you can suggest to Intel Gen9 users(Skylake) & Gen9.5(Kaby Lake/Coffee Lake) to update to the latest Intel 24.20.100.6229 driver. Intel has been regularly updating and fixing their driver for Gen9/9.5 GPUs.
Unfortunately, it's not practical to have users update their drivers manually.
The crash in comment 0 is from Win 10. Microsoft should distribute a stable driver through Windows Update.
the volume of this crash is steadily creeping up. user comments often refer to video playback (commonly on facebook, but also twitter, imgur and other common sites) and messages from windows like "Application firefox.exe has been blocked from accessing graphics hardware" at the time of the crash.
all the reports are happening when d2d1.1 is active and most of them also contain a gfx critical error section.
Duplicate of this bug: 1503187
I found another instance of this, most of the comments mention watching videos on YouTube, Facebook or just scrolling on Facebook (which must have triggered an auto-play video). Is there a way to at least prevent Firefox from crashing? While it's not our fault many of the users reporting crashes appear to be very annoyed.
Crash Signature: igd10iumd64.dll | CHwSurfaceRenderTarget::FlushQueuedOperations ] → igd10iumd64.dll | CHwSurfaceRenderTarget::FlushQueuedOperations ] [@ igd10iumd64.dll | d3d11.dll@0x189926]
And this one too, there must be even more.
Crash Signature: igd10iumd64.dll | CHwSurfaceRenderTarget::FlushQueuedOperations ] [@ igd10iumd64.dll | d3d11.dll@0x189926] → igd10iumd64.dll | CHwSurfaceRenderTarget::FlushQueuedOperations ] [@ igd10iumd64.dll | d3d11.dll@0x189926] [@ @0x0 | igd11dxva64.dll | d3d11.dll@0x2592d7]
I've reached out to Microsoft to see if there's anything they can do to help with the situation here.

did we receive a reply yet?

Just found another fairly high volume signature with similar stack. Not all the crashes are for the driver version specified in the title but over 90% of them are.

Crash Signature: [@ igd10iumd64.dll | CContext::UMQueryOM_DepthState_] [@ igd10iumd32.dll | CContext::UMQueryOM_DepthState_] [@ igd10iumd64.dll | DDIContextST::DDIContextST ] [@ @0x0 | igd11dxva64.dll | CSampler::GetDesc] [@ igd10iumd64.dll | CSampler::GetDesc] [@ ig… → [@ @0x0 | igd11dxva64.dll | CSampler::GetDesc] [@ @0x0 | igd11dxva64.dll | d3d11.dll@0x2592d7] [@ igd10iumd32.dll | CContext::UMQueryOM_DepthState_] [@ igd10iumd64.dll | CContext::UMQueryOM_DepthState_] [@ igd10iumd64.dll | CContext::UMSetError_] [@ …
Duplicate of this bug: 1560111
You need to log in before you can comment on or make changes to this bug.