1265812 - [e10s] Firefox crashes [@ IPCError-browser | (msgtype=0x7,name=???) Route error: message sent to unknown actor ID ]

Status: VERIFIED FIXED

(Core :: IPC, defect, P4)

Description

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Some crashlog reports appeared in about:crashes after long browsing with e10s enabled.

Crashlog reports:
https://crash-stats.mozilla.com/report/index/a550cd66-1329-4706-b918-de0b12160418
https://crash-stats.mozilla.com/report/index/4e9a73d2-8c70-4878-8949-63e4c2160418

Comment 1

[Andrew McCreight [:mccr8]]

These two are happening inside IMEStateManager::SetIMEState(), when it is doing some kind of IPC send.

Comment 2

[Jim Mathies [:jimm]]

https://crash-stats.mozilla.com/report/index/a550cd66-1329-4706-b918-de0b12160418
https://crash-stats.mozilla.com/report/index/4e9a73d2-8c70-4878-8949-63e4c2160418

There's an ipc error here on the parent side, which triggers an abort - 

ipc_channel_error 	

(msgtype=0x7,name=???) Route error: message sent to unknown actor ID

These happen in various protocols and are currently low volume. See bug 1116884.

Comment 3

[Tobias B. Besemer [:BesTo] (QA)]

(In reply to Jim Mathies [:jimm] from comment #2)
> There's an ipc error here on the parent side, which triggers an abort - 
> 
> ipc_channel_error 	
> 
> (msgtype=0x7,name=???) Route error: message sent to unknown actor ID
> 
> These happen in various protocols and are currently low volume. See bug
> 1116884.

ipc_channel_error = ShutDownKill -> It's this bug:
Bug 1219672 - (ShutDownKills-Win) [meta] e10s related ShutDownKill parent side abort of the content process

Operating System 	Percentage 	Number Of Crashes
Windows 7 		100.00% 	105

Product 	Version 	Percentage 	Number Of Crashes
Firefox 	48.0a1 		56.19% 		59
Firefox 	38.0a1 		30.48% 		32
Firefox 	47.0a2 		12.38% 		13
Firefox 	42.0a1 		 0.95% 		 1

Results within 7 days of 2016-04-26 and the crashing process was of any type (including unofficial release channels).

Comment 5

[Tobias B. Besemer [:BesTo] (QA)]

Seems to be a other crash...
...sig still there! (2016-05-03)


https://crash-stats.mozilla.com/report/index/9510ea53-2252-4004-8f3e-731d92160503

Crashing Thread (0)
Frame 	Module 	Signature 	Source
Ø 0 	ntdll.dll 	ntdll.dll@0x4bb7a 	
Ø 1 	kernelbase.dll 	kernelbase.dll@0x10ab 	
2 	xul.dll 	std::_Deque_iterator<std::_Deque_val<std::_Deque_simple_types<IPC::Message> > >::operator++(int) 	vs2015u2/VC/include/deque:621
3 	xul.dll 	mozilla::ipc::MessageChannel::ProcessPendingRequests(mozilla::ipc::AutoEnterTransaction&) 	ipc/glue/MessageChannel.cpp:1055
4 	xul.dll 	mozilla::ipc::MessageChannel::Send(IPC::Message*, IPC::Message*) 	ipc/glue/MessageChannel.cpp:1179
5 	xul.dll 	mozilla::dom::PContentChild::SendGetGraphicsDeviceInitData(mozilla::gfx::DeviceInitData*) 	obj-firefox/ipc/ipdl/PContentChild.cpp:4783
6 	xul.dll 	gfxPlatform::UpdateDeviceInitData() 	gfx/thebes/gfxPlatform.cpp:2447
7 	xul.dll 	gfxWindowsPlatform::UpdateDeviceInitData() 	gfx/thebes/gfxWindowsPlatform.cpp:1994
8 	xul.dll 	gfxWindowsPlatform::InitializeDevices() 	gfx/thebes/gfxWindowsPlatform.cpp:2297
9 	xul.dll 	gfxWindowsPlatform::InitAcceleration() 	gfx/thebes/gfxWindowsPlatform.cpp:418
10 	xul.dll 	gfxPlatform::Init() 	gfx/thebes/gfxPlatform.cpp:664
11 	xul.dll 	gfxPlatform::GetPlatform() 	gfx/thebes/gfxPlatform.cpp:514
12 	xul.dll 	mozilla::dom::ContentProcess::Init() 	dom/ipc/ContentProcess.cpp:125
13 	xul.dll 	XRE_InitChildProcess 	toolkit/xre/nsEmbedFunctions.cpp:609
14 	plugin-container.exe 	wmain 	toolkit/xre/nsWindowsWMain.cpp:131
15 	plugin-container.exe 	__scrt_common_main_seh 	f:/dd/vctools/crt/vcstartup/src/startup/exe_common.inl:255
Ø 16 	kernel32.dll 	kernel32.dll@0x159bc 	
Ø 17 	ntdll.dll 	ntdll.dll@0x2a2e0


Seems to be a Win7-only, only-64bit topcrash-win:
https://crash-stats.mozilla.com/signature/?product=Firefox&signature=ntdll.dll%400x4bb7a

Operating System
Operating System 	Count 	Percentage
Windows 7 	59 	100.0%

Architecture
Architecture 	Count 	Percentage
amd64 	59 	100.0%

Product
Product 	Version 	Count 	Percentage 	Installations
Firefox 	38.0a1 	23 	39.0% 	25
Firefox 	49.0a1 	16 	27.1% 	6
Firefox 	48.0a2 	11 	18.6% 	9
Firefox 	47.0a2 	4 	6.8% 	5
Firefox 	48.0a1 	2 	3.4% 	2
Firefox 	45.0a1 	1 	1.7% 	1
Firefox 	47.0a1 	1 	1.7% 	1
Firefox 	47.0b1 	1 	1.7% 	1

Comment 6

[Andrew McCreight [:mccr8]]

This signature, ntdll.dll@0x4bb7a, just seems to indicate a crash while in a CondVar on Windows. Bug 1269471 should add them to the irrelevant signature list, so we'll separate these issues out a bit.

Comment 7

[Andrew McCreight [:mccr8]]

This is not happening during shutdown, so I'm removing the ShutDownKill references.

Comment 8

[(not currently active) Ted Mielczarek]

FYI we fixed bug 1140358 so these signatures should be more useful now. I tried reprocessing one of the crashes from comment 0 but I may not have done it right. I'll try to get that to work so we can know what the new signature would be.

Comment 9

[(not currently active) Ted Mielczarek]

After reprocessing, one of the crashes from comment 0 winds up with the signature [@ IPCError-browser | (msgtype=0x7,name=???) Route error: message sent to unknown actor ID ]:
https://crash-stats.mozilla.com/report/index/a550cd66-1329-4706-b918-de0b12160418

(This may have also been impacted by other signature changes in the interim.)

Comment 10

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

This should be fixed by patch from bug #1255823.