Crash in [@ mozilla::detail::MutexImpl::mutexTryLock | nsTimerImpl::nsTimerImpl]
Categories
(Core :: Graphics: WebRender, defect)
Tracking
()
People
(Reporter: RyanVM, Unassigned)
References
(Blocks 1 open bug)
Details
(Keywords: crash)
Crash Data
Crash report: https://crash-stats.mozilla.org/report/index/2cbfb2ae-bad9-444f-a646-834f90211004
New crash signature appearing in Fenix 93.
Reason: SIGSEGV / SEGV_MAPERR
Top 10 frames of crashing thread:
0 libandroid.so libandroid.so@0x155b9
1 libxul.so mozilla::wr::RenderCompositorOGLSWGL::Resume gfx/webrender_bindings/RenderCompositorOGLSWGL.cpp:236
2 libmozglue.so RedBlackTree<arena_chunk_map_t, ArenaRunTreeTrait>::RotateRight memory/build/rb.h:564
3 libmozglue.so RedBlackTree<arena_chunk_map_t, ArenaRunTreeTrait>::Insert memory/build/rb.h:377
4 u:object_r:debug_prop:s0 u:object_r:debug_prop:s0@0x107
5 libc.so libc.so@0x2a510
6 libc.so libc.so@0x2a84f
7 libc.so libc.so@0x92ead
8 libxul.so mozilla::wr::RenderThread::Resume gfx/webrender_bindings/RenderThread.cpp:577
9 libxul.so mozilla::wr::WebRenderAPI::Resume gfx/webrender_bindings/WebRenderAPI.cpp:622
|
Reporter | |
Comment 1•3 years ago
|
||
Interestingly, they started in beta.4. Here's a pushlog range for the Gecko changes between Fenix 93.0.0-beta.3 and beta.4:
https://hg.mozilla.org/releases/mozilla-beta/pushloghtml?fromchange=7bd77da50990b750c9b8681e6d530610341916fa&tochange=b7e99d71a046bdd3c135af97d9e900673ca7210f
|
|
Reporter | |
Comment 2•3 years ago
|
||
Jeff, is it possible that your commit is related?
https://hg.mozilla.org/releases/mozilla-beta/rev/d18009b6f24c72efd1a6461449826055d3153b92
|
||
Updated•3 years ago
|
|
||
Comment 3•3 years ago
|
||
It's unlikely to be the cause. These are all happening on Intel GPUs. Maybe Jamie has a guess as to what the cause might be.
|
||
Comment 4•3 years ago
|
||
I think this is probably the same as bug 1726792, which I did in fact notice yesterday was mysteriously low on 93 beta crashes, but is crashing in 92 release and 94 nightly.
The crash seems to occur when falling back to SWGL. I used to be able to reproduce on a chromebook, but wasn't able to yesterday when investigating. You can see in the crash reports' "graphics critical error" field that most (but not all) of the time we are falling back to SWGL due to a shader compilation error - that is bug 1733919 and the fix landed yesterday. That only fixes the cause of the majority of fallbacks to SWGL, but not the SWGL crash itself. However, the effect should be to significantly reduce the frequency of this crash. I'd feel pretty confident uplifting it to release, if we think the crash numbers justify it?
|
|
Reporter | |
Comment 5•3 years ago
|
||
Let's at least nominate it for Beta uplift and see how things look there.
|
||
Comment 6•3 years ago
|
||
The crash address looks like a null pointer deref in most of these reports. A good slice of those because mGL is null here: https://hg.mozilla.org/releases/mozilla-beta/file/b7e99d71a046bdd3c135af97d9e900673ca7210f/gfx/webrender_bindings/RenderAndroidSurfaceTextureHost.cpp#l180 as in crash report https://crash-stats.mozilla.org/report/index/268dad5f-e1e8-4d4b-8860-26f9f0211003
Others seem to be causing a separate null deref here, maybe because nativeWindow is null:
https://hg.mozilla.org/releases/mozilla-beta/file/b7e99d71a046bdd3c135af97d9e900673ca7210f/gfx/webrender_bindings/RenderCompositorOGLSWGL.cpp#l236 as in crash report https://crash-stats.mozilla.org/report/index/e1a52da9-b951-49f2-b930-3f18a0210926
|
|
||
Comment 7•3 years ago
|
||
I think this was dup of bug 1747116. certainly the latter signature Lee identified was. If the numbers spike again we can look in to the former signature
Description
•