1606549 - Intermittent Android crashtest <test-name> | application crashed [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext]

Reporter

Description

•

5 years ago

treeherder

Filed by: aciure [at] mozilla.com
Parsed log: https://treeherder.mozilla.org/logviewer.html#?job_id=283140202&repo=autoland
Full log: https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/SzdXI02KRNaQaG_Mqrgpew/runs/0/artifacts/public/logs/live_backing.log
Reftest URL: https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/SzdXI02KRNaQaG_Mqrgpew/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1

[task 2020-01-01T13:26:57.941Z] 13:26:57 [task 2020-01-01T13:26:57.941Z] 13:26:57 [task 2020-01-01T13:27:18.673Z] 13:27:18 [task 2020-01-01T13:27:18.775Z] 13:27:18 [task 2020-01-01T13:27:19.535Z] 13:27:19 [task 2020-01-01T13:27:24.178Z] 13:27:24 [task 2020-01-01T13:27:24.179Z] 13:27:24 [task 2020-01-01T13:27:24.185Z] 13:27:24 [task 2020-01-01T13:27:24.185Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.187Z] 13:27:24 [task 2020-01-01T13:27:24.188Z] 13:27:24 [task 2020-01-01T13:27:24.188Z] 13:27:24 [task 2020-01-01T13:27:24.188Z] 13:27:24 [task 2020-01-01T13:27:24.188Z] 13:27:24 [task 2020-01-01T13:27:24.188Z] 13:27:24 [task 2020-01-01T13:27:24.188Z] 13:27:24 [task 2020-01-01T13:27:24.188Z] 13:27:24 INFO - REFTEST TEST-START | http://10.0.2.2:8854/tests/dom/media/test/crashtests/1270303.html
INFO - REFTEST TEST-LOAD | http://10.0.2.2:8854/tests/dom/media/test/crashtests/1270303.html | 623 / 3748 (16%)
INFO - wait for org.mozilla.geckoview.test complete; top activity=com.android.launcher3
INFO - remoteautomation.py | Application ran for: 0:02:10.018884
INFO - REFTEST INFO | Copy/paste: /builds/worker/workspace/build/linux64-minidump_stackwalk /tmp/tmpEjVO4b/40c1ab12-8c29-9265-6955-1a70797e5105.dmp /builds/worker/workspace/build/symbols
INFO - REFTEST INFO | Saved minidump as /builds/worker/workspace/build/blobber_upload_dir/40c1ab12-8c29-9265-6955-1a70797e5105.dmp
INFO - REFTEST INFO | Saved app info as /builds/worker/workspace/build/blobber_upload_dir/40c1ab12-8c29-9265-6955-1a70797e5105.extra
WARNING - REFTEST PROCESS-CRASH | http://10.0.2.2:8854/tests/dom/media/test/crashtests/1270303.html | application crashed [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()]
INFO - Crash dump filename: /tmp/tmpEjVO4b/40c1ab12-8c29-9265-6955-1a70797e5105.dmp
INFO - Operating system: Android
INFO - 0.0.0 Linux 3.10.0+ #260 SMP PREEMPT Fri May 19 12:48:14 PDT 2017 x86_64
INFO - CPU: amd64
INFO - family 6 model 6 stepping 3
INFO - 4 CPUs
INFO - GPU: UNKNOWN
INFO - Crash reason: SIGSEGV /SEGV_MAPERR
INFO - Crash address: 0x0
INFO - Process uptime: not available
INFO - Thread 40 (crashed)
INFO - 0 libxul.so!mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext() [RenderAndroidSurfaceTextureHostOGL.cpp:a748a5149bda383173368cd3a8df84c8423f9f7b : 131 + 0x29]
INFO - rax = 0x00007c64ebd733f4 rdx = 0x0000000000000001
INFO - rcx = 0x00007c64efa3ca80 rbx = 0x0000000000000000
INFO - rsi = 0x00007c64dbbb5ea0 rdi = 0x00007c64dbbb5bf0
INFO - rbp = 0x00007c64dbbb5f30 rsp = 0x00007c64dbbb5ee0
INFO - r8 = 0x0000000000000b2c r9 = 0x00007c64dbbb6450
INFO - r10 = 0x00007c64e75d5a3b r11 = 0x0000000000000000
INFO - r12 = 0x00007c64d5d8fc00 r13 = 0x00007c64dbbb6080
INFO - r14 = 0x00007c64d5d8fc38 r15 = 0x00007c64dbbb5eec
INFO - rip = 0x00007c64e75d5a4d
INFO - Found by: given as instruction pointer in context
INFO - 1 libxul.so!mozilla::wr::RenderAndroidSurfaceTextureHostOGL::PrepareForUse() [RenderAndroidSurfaceTextureHostOGL.cpp:a748a5149bda383173368cd3a8df84c8423f9f7b : 150 + 0x8]
INFO - rbp = 0x00007c64dbbb5f70 rsp = 0x00007c64dbbb5f40
INFO - rip = 0x00007c64e75d5c12
INFO - Found by: previous frame's frame pointer
INFO - 2 libxul.so!mozilla::wr::RenderThread::HandlePrepareForUse() [RenderThread.cpp:a748a5149bda383173368cd3a8df84c8423f9f7b : 728 + 0xe]
INFO - rbp = 0x00007c64dbbb5fc0 rsp = 0x00007c64dbbb5f80
INFO - rip = 0x00007c64e75d9273
INFO - Found by: previous frame's frame pointer
INFO - 3 libxul.so!mozilla::detail::RunnableMethodImpl<mozilla::wr::RenderThread*, void (mozilla::wr::RenderThread::*)(), true, (mozilla::RunnableKind)0>::Run() [nsThreadUtils.h:a748a5149bda383173368cd3a8df84c8423f9f7b : 1217 + 0x17]
INFO - rbp = 0x00007c64dbbb5fd0 rsp = 0x00007c64dbbb5fd0
INFO - rip = 0x00007c64e75e8019
INFO - Found by: previous frame's frame pointer
INFO - 4 libxul.so!MessageLoop::RunTask(already_AddRefed<nsIRunnable>) [message_loop.cc:a748a5149bda383173368cd3a8df84c8423f9f7b : 442 + 0x11]
INFO - rbp = 0x00007c64dbbb6030 rsp = 0x00007c64dbbb5fe0
INFO - rip = 0x00007c64e6cb2cde
INFO - Found by: previous frame's frame pointer

Comment hidden (Intermittent Failures Robot)

BugBot [:suhaib / :marco/ :calixte]

Comment 3

•

5 years ago

The priority flag is not set for this bug.
:bryce, could you have a look please?

For more information, please visit auto_nag documentation.

Flags: needinfo?(bvandyk)

Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail

Comment 4

•

5 years ago

Looks like a crash in WebRender code. Possibly an assertion failure given the stack and that all reports are in debug builds. Not clear to me if the webm in the test does something to trigger this.

Moving to Web Render.

Component: Audio/Video: Playback → Graphics: WebRender

Flags: needinfo?(bvandyk)

Comment hidden (Intermittent Failures Robot)

BugBot [:suhaib / :marco/ :calixte]

Comment 6

•

5 years ago

The priority flag is not set for this bug.
:jbonisteel, could you have a look please?

For more information, please visit auto_nag documentation.

Flags: needinfo?(jbonisteel)

Jessie [:jbonisteel] pls NI

Updated

•

5 years ago

Blocks: wr-android

Flags: needinfo?(jbonisteel) → needinfo?(ktaeleman)

Priority: -- → P3

Kris Taeleman (:ktaeleman)

Comment 7

•

5 years ago

Hey Sotaro, any idea what's going on here?

Flags: needinfo?(ktaeleman) → needinfo?(sotaro.ikeda.g)

Comment hidden (Intermittent Failures Robot)

Sotaro Ikeda [:sotaro]

Comment 9

•

5 years ago

It seems that GeckoSurfaceTexture::AttachToGLContext() call in RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext() seemed to be failed. It seems that SurfaceTexture state was not valid. I am going to take a look.

https://searchfox.org/mozilla-central/rev/2e355fa82aaa87e8424a9927c8136be184eeb6c7/gfx/webrender_bindings/RenderAndroidSurfaceTextureHostOGL.cpp#123

Assignee: nobody → sotaro.ikeda.g

Flags: needinfo?(sotaro.ikeda.g)

Andreea Pavel [:apavel]

Updated

•

5 years ago

Crash Signature: [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()] → [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::Lock(unsigned char, mozilla::gl::GLContext*, mozilla::wr::ImageRendering)]

Comment hidden (Intermittent Failures Robot)

Cosmin Sabou [:CosminS]

Comment 14

•

5 years ago

Worth mentioning that all these crashes have this assertion in the log:
Assertion failure: 0, at /builds/worker/workspace/build/src/gfx/webrender_bindings/RenderAndroidSurfaceTextureHostOGL.cpp:131

[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.484 E/ResourceManagerService( 1357): Rejected removeResource call with invalid pid.
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.485 I/GeckoMediaManager( 3063): Media service has been unbound. Stopping.
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.487 I/Gecko   ( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.487 I/Gecko   ( 2802): {"action":"log","time":1582286312486,"thread":null,"pid":null,"source":"reftest","level":"DEBUG","message":"[CONTENT] MakeProgress: Completed"}
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.488 I/Gecko   ( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.488 I/Gecko   ( 2802): {"action":"log","time":1582286312488,"thread":null,"pid":null,"source":"reftest","level":"DEBUG","message":"[CONTENT] RecordResult fired"}
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.489 I/Gecko   ( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.489 I/Gecko   ( 2802): {"action":"log","time":1582286312489,"thread":null,"pid":null,"source":"reftest","level":"DEBUG","message":"RecordResult fired"}
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 I/Gecko   ( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 I/Gecko   ( 2802): {"action":"test_status","time":1582286312489,"thread":null,"pid":null,"source":"reftest","test":"dom/media/test/crashtests/1267263.html","subtest":"(LOAD ONLY)","status":"PASS"}
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 E/GLConsumer( 2802): [SurfaceTexture-0-2802-6] attachToContext: abandoned GLConsumer
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 W/System.err( 2802): java.lang.RuntimeException: Error during attachToGLContext (see logcat for details)
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 I/Gecko   ( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 I/Gecko   ( 2802): {"action":"test_end","time":1582286312490,"thread":null,"pid":null,"source":"reftest","test":"dom/media/test/crashtests/1267263.html","status":"OK"}
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 W/System.err( 2802): 	at android.graphics.SurfaceTexture.attachToGLContext(SurfaceTexture.java:286)
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 W/System.err( 2802): 	at org.mozilla.gecko.gfx.GeckoSurfaceTexture.attachToGLContext(GeckoSurfaceTexture.java:90)
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 F/MOZ_Assert( 2802): Assertion failure: 0, at /builds/worker/workspace/build/src/gfx/webrender_bindings/RenderAndroidSurfaceTextureHostOGL.cpp:131
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 I/Gecko   ( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.490 I/Gecko   ( 2802): {"action":"log","time":1582286312490,"thread":null,"pid":null,"source":"reftest","level":"DEBUG","message":"Loading a blank page"}
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.495 W/google-breakpad( 2802): ExceptionHandler::GenerateDump cloned child
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.495 W/google-breakpad( 2802): 3315
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.495 W/google-breakpad( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.495 W/google-breakpad( 2802): ExceptionHandler::SendContinueSignalToChild sent continue signal to child
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:32.495 W/google-breakpad( 3315): ExceptionHandler::WaitForContinueSignal waiting for continue signal...
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:33.198 I/Choreographer( 2802): Skipped 41 frames!  The application may be doing too much work on its main thread.
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:33.206 I/Gecko   ( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:33.206 I/Gecko   ( 2802): {"action":"assertion_count","time":1582286313206,"thread":null,"pid":null,"source":"reftest","test":"dom/media/test/crashtests/1267263.html","min_expected":0,"max_expected":0,"count":0}
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:33.207 I/Gecko   ( 2802):
[task 2020-02-21T11:59:07.074Z] 11:59:07     INFO -  02-21 11:58:33.207 I/Gecko   ( 2802): {"action":"test_start","time":1582286313207,"thread":null,"pid":null,"source":"reftest","test":"dom/media/test/crashtests/1270303.html"}

Recent failure: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=289998025&repo=autoland&lineNumber=3669

Summary: Intermittent dom/media/test/crashtests/1270303.html | application crashed [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()] → Intermittent Android crashtest <test-name> | application crashed [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()]

Comment hidden (Intermittent Failures Robot)

Sotaro Ikeda [:sotaro]

Updated

•

5 years ago

Depends on: 1621836

Comment hidden (Intermittent Failures Robot)

Sotaro Ikeda [:sotaro]

Comment 23

•

5 years ago

This bug might be affected by Bug 1626142.

Comment hidden (Intermittent Failures Robot)

Jamie Nicol [:jnicol] out of office until 6th Jan

Assignee

Comment 27

•

5 years ago

I reproduced this with some logging: https://treeherder.mozilla.org/#/jobs?repo=try&revision=db140128b4a36f634a4aade80ffd80ec6c87451e&selectedTaskRun=bGm0D5mtTqKtr6O0f092uA-0

Here's the relevant output:

Creating AndroidSurfaceTextureData handle = 82
AndroidSurfaceTextureData::Serialize() handle=82
RemoteVideoDecoder::~RemoteVideoDecoder() handle=82
SurfaceAllocator.disposeSurface() handle=82
SurfaceAllocatorService.releaseSurface() handle=82
GeckoSurfaceTexture.decrementUse() handle=82 upstream=0 count=0
GST handle=82 count reached zero. releasing immediately
GeckoSurfaceTexture.release() handle=82
CreateTextureHostOGL() handle=82
SurfaceTextureHost::SurfaceTextureHost() 0x78d1584129d0 handle=82
GeckoSurfaceTexture.incrementUse() handle=82 upstream=0 count=1
RenderAndroidSurfaceTextureHostOGL::RenderAndroidSurfaceTextureHostOGL() 0x78d157c648c0 handle=82
GeckoSurfaceTexture.incrementUse() handle=82 upstream=0 count=2
SurfaceTextureHost::CreateRenderTexture() 0x78d157c648c0 id=30064771271 handle=82
RenderThread::PrepareForUse() 0x78d157c648c0
GeckoSurface.release() handle=82 82
GeckoSurfaceTexture.decrementUse() handle=82 upstream=82 count=0
GST handle=82 count reached zero. releasing immediately
GeckoSurfaceTexture.release() handle=82
RenderThread::HandlePrepareForUse()
RenderAndroidSurfaceTextureHostOGL::PrepareForUse() 0x78d157c648c0 handle=82
RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext() 0x78d157c648c0 handle=82
GeckoSurfaceTexture.attachToGLContext() handle=82
[SurfaceTexture-0-4874-185] attachToContext: abandoned GLConsumer
java.lang.RuntimeException: Error during attachToGLContext (see logcat for details)
	at android.graphics.SurfaceTexture.attachToGLContext(SurfaceTexture.java:286)
	at org.mozilla.gecko.gfx.GeckoSurfaceTexture.attachToGLContext(GeckoSurfaceTexture.java:92)
EnsureAttached 0x78d157c648c0 attach failed
MOZ_Assert: Assertion failure: 0, at /builds/worker/checkouts/gecko/gfx/webrender_bindings/RenderAndroidSurfaceTextureHostOGL.cpp:153

What I think is happening is:

We send a video frame from the decoder (content process) to compositor by using AndroidSurfaceTextureData::Serialize(). We don't appear to add any reference to the Surface here, which I think is the problem.
Then the RemoteVideoDecoder gets destroyed. This decrements the final reference to the Surface.
The Surface gets released, which decrements the reference count of the corresponding SurfaceTexture to zero.
Meanwhile in the Compositor thread, the video frame gets deserialized and a SurfaceTextureHost is constructed. It calls GeckoSurfaceTexture.lookup() and finds the SurfaceTexture. This increments the reference count back up to 1, but the SurfaceTexture is in the process of being released in GeckoSurfaceTexture.decrementUse().
Back in SurfaceTexture.decrementUse(), the SurfaceTexture is finally released, and removed from sSurfaceTextures, but it is too late.
Then later in RenderAndroidSurfaceTextureHost::PrepareForUse(), we call EnsureAttachedToGLContext(), which calls attachToGLContext(),
which fails because the SurfaceTexture has been released (or "abandoned" as it calls it).

Sometimes, the SurfaceTexture is in fact released and removed from sSurfaceTextures before the SurfaceTextureHost is deserialized. In this case, GeckoSurfaceTexture.lookup() fails, so the SurfaceTextureHost is constructed with nullptr as mSurfTex. This doesn't cause any crashes but will presumably means frames are dropped when rendering.

Comment hidden (Intermittent Failures Robot)

Sotaro Ikeda [:sotaro]

Comment 29

•

5 years ago

(In reply to Jamie Nicol [:jnicol] from comment #27)

What I think is happening is:

We send a video frame from the decoder (content process) to compositor by using AndroidSurfaceTextureData::Serialize(). We don't appear to add any reference to the Surface here, which I think is the problem.

Then the RemoteVideoDecoder gets destroyed. This decrements the final reference to the Surface.

The Surface gets released, which decrements the reference count of the corresponding SurfaceTexture to zero.

Thanks for the investigation. It seems that this problem could be addressed by using TextureFlags::WAIT_HOST_USAGE_END.

Sotaro Ikeda [:sotaro]

Comment 30

•

5 years ago

(In reply to Sotaro Ikeda [:sotaro] from comment #29)

Thanks for the investigation. It seems that this problem could be addressed by using TextureFlags::WAIT_HOST_USAGE_END

Created Bug 1636334 for it.

Sotaro Ikeda [:sotaro]

Updated

•

5 years ago

Depends on: 1636334

Sotaro Ikeda [:sotaro]

Comment 31

•

5 years ago

It seem that calling HandlePrepareForUse() before UpdateAndRender() is not enough. Emulator is very very slow. Then there might be a case that new RenderThread::PrepareForUse() and new WebRender transaction by WebRenderBridgeParent::MaybeGenerateFrame() achieve during calling UpdateAndRender().

Sotaro Ikeda [:sotaro]

Updated

•

5 years ago

Depends on: 1636352

Sotaro Ikeda [:sotaro]

Comment 32

•

5 years ago

(In reply to Sotaro Ikeda [:sotaro] from comment #31)

It seem that calling HandlePrepareForUse() before UpdateAndRender() is not enough. Emulator is very very slow. Then there might be a case that new RenderThread::PrepareForUse() and new WebRender transaction by WebRenderBridgeParent::MaybeGenerateFrame() achieve during calling UpdateAndRender().

Bug 1636352 was created for it.

Comment hidden (Intermittent Failures Robot)

Sotaro Ikeda [:sotaro]

Comment 34

•

5 years ago

Since Bug 1636352 fix, the failures did not happen.

Status: NEW → RESOLVED

Closed: 5 years ago

Resolution: --- → WORKSFORME

Dorel Luca [:dluca]

Comment 35

•

5 years ago

Still happening here:
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=302622761&repo=autoland&lineNumber=4057

Status: RESOLVED → REOPENED

Resolution: WORKSFORME → ---

Sotaro Ikeda [:sotaro]

Comment 36

•

5 years ago

Failure happened at mSurfTex->AttachToGLContext(). Form it, SurfaceTexture might be deallocated when AttachToGLContext() is called.

Jamie Nicol [:jnicol] out of office until 6th Jan

Assignee

Comment 37

•

5 years ago

I understand what the problem is and can reproduce lo locally so am happy to take this bug, sotaro.

Assignee: sotaro.ikeda.g → jnicol

Sotaro Ikeda [:sotaro]

Comment 38

•

5 years ago

Thanks.

Comment hidden (Intermittent Failures Robot)

BugBot [:suhaib / :marco/ :calixte]

Updated

•

4 years ago

Crash Signature: [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::Lock(unsigned char, mozilla::gl::GLContext*, mozilla::wr::ImageRendering)] → [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::Lock(unsigned char, mozilla::gl::GLContext*, mozilla::wr::ImageRendering)] [@ mozilla::wr::RenderAndroidSurfaceTextureHos…

Comment hidden (Intermittent Failures Robot)

Andreea Pavel [:apavel]

Comment 54

•

4 years ago

Retriggers for the recent spike: https://treeherder.mozilla.org/#/jobs?repo=autoland&group_state=expanded&searchStr=android%2C7.0%2Cwebrender&tochange=a24dbf67be695c274b18914a65a1c0b9746323eb&fromchange=0ae4f6d05bc52117a58b6f158584b3a6e2dd01cb&selectedTaskRun=c5zM_AXoTJeXQ44T7SoBiA.0

Crash Signature: , mozilla::wr::ImageRendering)] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::PrepareForUse()] → , mozilla::wr::ImageRendering)] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::PrepareForUse()]

Whiteboard: [stockwell needswork:owner] → [stockwell needswork:owner][retriggers]

Comment hidden (Intermittent Failures Robot)

Jamie Nicol [:jnicol] out of office until 6th Jan

Assignee

Comment 57

•

4 years ago

It's very confusing that this bug contains multiple crash signatures that are almost certainly different issues. I think I know how to fix the EnsureAttachedToGLContext crash. I will land a potential fix asap. But the Lock crash is likely unrelated. Is the recent spike predominantly in either one of the signatures? Is there a way to separate them?

Jamie Nicol [:jnicol] out of office until 6th Jan

Assignee

Comment 58

•

4 years ago

Attached file Bug 1606549 - Fix race condition between GeckoSurfaceTexture destruction and lookup. r?snorp — Details

Currently when a GeckoSurfaceTexture's refcount reaches zero, it calls
super.release() prior to removing itself from sSurfaceTextures. If a call to
GeckoSurfaceTexture.lookup() takes place in between these events, then it will
succesfully find the surface texture and return it. However, that surface
texture will be in a released state, which will cause problems.

To fix this, change lookup() to check the refcount of any surface texture that
it finds (within a synchronized block, of course), and return null if the
refcount is zero. I think this is preferable to locking sSurfaceTextures for
every decrementUse() call.

Change mRefCount from an AtomicInt to a plain int, as it is only accessed within
synchronized blocks or methods. The use of AtomicInt implied a false sense of
security, when in fact there is more work that must be done atomically than just
decrementing the count.

Additionally, remove some code which attempts to remove the surface texture from
sSurfaceTextures just after calling release(), as we have just done this in our
overridden release() method.

Andreea Pavel [:apavel]

Updated

•

4 years ago

Whiteboard: [stockwell needswork:owner][retriggers] → [stockwell needswork:owner][retriggered]

Comment hidden (Intermittent Failures Robot)

Jamie Nicol [:jnicol] out of office until 6th Jan

Assignee

Comment 60

•

4 years ago

The patch in comment 58 should hopefully fix the RenderAndroidSurfaceTextureHost::EnsureAttachedToGLContext() crash.

As far as I can tell, however, the spike is due to the crash in RenderAndroidSurfaceTextureHost::Lock(). (By quickly looking through the failures linked to by the bot the past week.) This bug is confusing enough already, I'd really like it if we could separate that out in to a different bug. Andreea, is it possible to do that? I obviously know how to file a new bug, but I have no idea how to make it so that treeherder matches the intermittent failures to the new bug instead.

Sotaro, the crash in Lock() occurs because mPrepareStatus == STATUS_MIGHT_BE_USED_BY_WR. This is because NofityForUse() has not been called before Lock(). Do you know why that might be happening?

Flags: needinfo?(sotaro.ikeda.g)

Flags: needinfo?(apavel)

Andreea Pavel [:apavel]

Comment 61

•

4 years ago

(In reply to Jamie Nicol [:jnicol] from comment #60)

The patch in comment 58 should hopefully fix the RenderAndroidSurfaceTextureHost::EnsureAttachedToGLContext() crash.

As far as I can tell, however, the spike is due to the crash in RenderAndroidSurfaceTextureHost::Lock(). (By quickly looking through the failures linked to by the bot the past week.) This bug is confusing enough already, I'd really like it if we could separate that out in to a different bug. Andreea, is it possible to do that? I obviously know how to file a new bug, but I have no idea how to make it so that treeherder matches the intermittent failures to the new bug instead.

Sotaro, the crash in Lock() occurs because mPrepareStatus == STATUS_MIGHT_BE_USED_BY_WR. This is because NofityForUse() has not been called before Lock(). Do you know why that might be happening?

I apologize for miscalssifying this in the first place. I filed bug 1658005.
TH usually shows the suggestions if they match exactly the failure line or part of the failure line. However, we'll keep in mind t make the distinction between the two crashes.

Thank you.

Flags: needinfo?(apavel)

Jamie Nicol [:jnicol] out of office until 6th Jan

Assignee

Comment 62

•

4 years ago

Thanks Andreea, much appreciated!

Andreea Pavel [:apavel]

Comment 63

•

4 years ago

(In reply to Jamie Nicol [:jnicol] from comment #62)

Thanks Andreea, much appreciated!

No problem, any time.

Comment hidden (Intermittent Failures Robot)

Jamie Nicol [:jnicol] out of office until 6th Jan

Assignee

Comment 65

•

4 years ago

Cancelling Sotaro's needinfo to transfer it to bug 1658005.

Flags: needinfo?(sotaro.ikeda.g)

Comment hidden (Intermittent Failures Robot)

BugBot [:suhaib / :marco/ :calixte]

Updated

•

4 years ago

Crash Signature: , mozilla::wr::ImageRendering)] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::PrepareForUse()] → , mozilla::wr::ImageRendering)] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::PrepareForUse()] [@ mozilla::wr::RenderAndroidSurfaceTextureHost::EnsureAttachedToGLContext()]

Comment hidden (Intermittent Failures Robot)

BMO Automation

Updated

•

2 years ago

Severity: normal → S3

Comment hidden (Intermittent Failures Robot)

BMO Automation

Updated

•

2 years ago

Crash Signature: [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::Lock(unsigned char, mozilla::gl::GLContext*, mozilla::wr::ImageRendering)] [@ mozilla::wr::RenderAndroidSurfaceTextureHo… → [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::Lock] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::PrepareForUse] [@ mozilla::wr::RenderAndroidSurfaceTextureHost::…

Summary: Intermittent Android crashtest <test-name> | application crashed [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext()] → Intermittent Android crashtest <test-name> | application crashed [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext]

Glenn Watson [:gw]

Updated

•

8 months ago

Status: REOPENED → RESOLVED

Crash Signature: [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::Lock] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::PrepareForUse] [@ → [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::EnsureAttachedToGLContext] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::Lock] [@ mozilla::wr::RenderAndroidSurfaceTextureHostOGL::PrepareForUse] [@

Closed: 5 years ago → 8 months ago

Resolution: --- → WORKSFORME