1056893 - Crash in CompositorChild::ActorDestroy while receiving calls and SMS continuously

Status: RESOLVED WORKSFORME

(Core :: Graphics: Layers, defect)

Description

[Greg Grisco]

[Blocking Requested - why for this release]:

1. Enable auto answer.
2. Receive MT call and MT SMS continuously.
3. After few hours of run, mini dumps are generated in the phone.

This is really a duplicate of bug 945992, but opening it on 2.0

[@ mozalloc_abort(char const*) | NS_DebugBreak | mozilla::layers::CompositorChild::ActorDestroy(mozilla::ipc::IProtocolManager::ActorDestroyReason) | mozilla::layers::PCompositorChild::DestroySubtree(mozilla::ipc::IProtocolManager::ActorDestroyReason) ]

Comment 1

[Sotaro Ikeda [:sotaro]]

Greg, why does this bug flagged to 2.0? Does the crash happen on b2g-2.0? If it is a dup of bug 945992, the crash seems to happen on b2g-v1.3.

Comment 2

[Greg Grisco]

(In reply to Sotaro Ikeda [:sotaro] from comment #1)
> Greg, why does this bug flagged to 2.0? Does the crash happen on b2g-2.0? If
> it is a dup of bug 945992, the crash seems to happen on b2g-v1.3.

Hi Sotaro.  I'll upload logs in a second, but this bug is crashing on 2.0 now.

Comment 3

[cafbot (PoC: ggrisco)]

Attachment: EXTRA file attachment

Comment 4

[cafbot (PoC: ggrisco)]

Attachment: decoded minidump

Comment 5

[Milan Sreckovic [:milan] (needinfo for best results)]

Can we confirm why this is a CAF blocker?

Comment 6

[Sotaro Ikeda [:sotaro]]

>AbortMessage=[Child 3653] ###!!! ABORT: ActorDestroy by IPC channel failure at CompositorChild: file ../../../../../../../../gecko/gfx/layers/ipc/CompositorChild.cpp, line 169

The above log was triggered by the following code. It just say that CompositorChild is going to be closed because of IPC error. The log does not have an inforation about why IPC error happens. 

void
CompositorChild::ActorDestroy(ActorDestroyReason aWhy)
{
  MOZ_ASSERT(sCompositor == this);

#ifdef MOZ_B2G
  // Due to poor lifetime management of gralloc (and possibly shmems) we will
  // crash at some point in the future when we get destroyed due to abnormal
  // shutdown. Its better just to crash here. On desktop though, we have a chance
  // of recovering.
  if (aWhy == AbnormalShutdown) {
    NS_RUNTIMEABORT("ActorDestroy by IPC channel failure at CompositorChild");
  }
#endif

Comment 7

[Sotaro Ikeda [:sotaro]]

Greg, can you re-test again with IPC log enabled?

Comment 8

[Sotaro Ikeda [:sotaro]]

To analyze the problem, IPC enabled log with crash log becomes necessary.

Comment 9

[Greg Grisco]

(In reply to Milan Sreckovic [:milan] from comment #5)
> Can we confirm why this is a CAF blocker?

Hi Milan, this is affecting our stability milestone for this release.  We've seen this crash 3 times on latest AU.

Comment 10

[Sotaro Ikeda [:sotaro]]

(In reply to Sotaro Ikeda [:sotaro] from comment #7)
> Greg, can you re-test again with IPC log enabled?

Greg, is it possible?
Without the IPC log enabled log, it is not possible to investigate about the problem.

Comment 11

[cafbot (PoC: ggrisco)]

Observed on: 

Device: 
Gonk Version: AU_LINUX_GECKO_B2G_JB_3.2.01.03.00.112.255
Moz BuildID: 20140226004002
B2G Version: 2.0
Gecko Version: 28.0
Gaia:  http://git.mozilla.org/?p=releases/gaia.git;a=commit;h=8039a5cb7519adfa81677df577f494c6a4de6140
Gecko: http://git.mozilla.org/?p=releases/gecko.git;a=commit;h=7599b5ccc906f556076798fc062a9e51e2c0eece

Comment 12

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

The bot appears to be reporting this on 1.3 ...

Comment 13

[Greg Grisco]

(In reply to Kyle Huey [:khuey] ([e-mail]) from comment #12)
> The bot appears to be reporting this on 1.3 ...

Yeah, right now the bot doesn't discriminate between versions.  We're tracking this locally as a single issue, so it will log here whenever it witnesses the crash on any version.  I'll see if we can improve this but it might take a little time.  In the meantime, I'll try to force the bot to comment on 2.0 version here.

Comment 14

[cafbot (PoC: ggrisco)]

Observed on: 

Device: msm8226
Gonk Version: AU_LINUX_GECKO_B2G_KK_3.6.01.04.00.000.060
Moz BuildID: 20140810000201
B2G Version: 2.0
Gecko Version: 32.0
Gaia:  http://git.mozilla.org/?p=releases/gaia.git;a=commit;h=de28796a8956a48bb98ca67df6a33e0622d642d1
Gecko: http://git.mozilla.org/?p=releases/gecko.git;a=commit;h=2b27becae85092d46bfadcd4fb5605e82e1e1093

Comment 15

[Sotaro Ikeda [:sotaro]]

(In reply to cafbot (PoC: ggrisco) from comment #14)
> Observed on: 
> 
> Device: msm8226
> Gonk Version: AU_LINUX_GECKO_B2G_KK_3.6.01.04.00.000.060
> Moz BuildID: 20140810000201
> B2G Version: 2.0
> Gecko Version: 32.0
> Gaia: 
> http://git.mozilla.org/?p=releases/gaia.git;a=commit;
> h=de28796a8956a48bb98ca67df6a33e0622d642d1
> Gecko:
> http://git.mozilla.org/?p=releases/gecko.git;a=commit;
> h=2b27becae85092d46bfadcd4fb5605e82e1e1093

This gecko does not include Bug 1053204 fix.

Comment 16

[Greg Grisco]

(In reply to Sotaro Ikeda [:sotaro] from comment #15)
> This gecko does not include Bug 1053204 fix.

Right.  According to bug 1053204, that fix should be built into AU63 and later.

Comment 17

[Sotaro Ikeda [:sotaro]]

(In reply to Sotaro Ikeda [:sotaro] from comment #10)
> (In reply to Sotaro Ikeda [:sotaro] from comment #7)
> > Greg, can you re-test again with IPC log enabled?
> 
> Greg, is it possible?
> Without the IPC log enabled log, it is not possible to investigate about the
> problem.

Greg, Tapas, is it possible re-test by enabling gecko IPC log?

Comment 18

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

(In reply to Sotaro Ikeda [:sotaro] from comment #17)
> (In reply to Sotaro Ikeda [:sotaro] from comment #10)
> > (In reply to Sotaro Ikeda [:sotaro] from comment #7)
> > > Greg, can you re-test again with IPC log enabled?
> > 
> > Greg, is it possible?
> > Without the IPC log enabled log, it is not possible to investigate about the
> > problem.
> 
> Greg, Tapas, is it possible re-test by enabling gecko IPC log?

We asked our test folks to reproduce it with IPC logs. we will get back to you asap

Comment 19

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

Is it OK if enable only minimal gecko IPC traces from bug 1046084 comment 35. Do we still need FULL IPC log here. ? I am asking again because our test team will try to reproduce both bug 1056893 and bug 1046084 on same build.

Comment 20

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

I don't think so.  But we might be able to do targeted logging here too.  Sotaro what information do you actually need here?

Comment 21

[Sotaro Ikeda [:sotaro]]

(In reply to Kyle Huey [:khuey] ([e-mail]) from comment #20)
> I don't think so.  But we might be able to do targeted logging here too. 
> Sotaro what information do you actually need here?

I have no idea what triggers IPC error, therefore if would be nice if we could get full gecko IPC log. But bug 1046084 comment 35 seems OK for me now.

Comment 22

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

(In reply to Sotaro Ikeda [:sotaro] from comment #21)
> (In reply to Kyle Huey [:khuey] ([e-mail]) from comment #20)
> > I don't think so.  But we might be able to do targeted logging here too. 
> > Sotaro what information do you actually need here?
> 
> I have no idea what triggers IPC error, therefore if would be nice if we
> could get full gecko IPC log. But bug 1046084 comment 35 seems OK for me now.

That code only logs errors that go through MessageChannel::MaybeHandleError.  It doesn't look like this does.

Comment 23

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

bent, can we end up in OnChannelError for something other than our pipe being screwed up?

Comment 24

[Sotaro Ikeda [:sotaro]]

(In reply to Kyle Huey [:khuey] ([e-mail]) from comment #22)
> 
> That code only logs errors that go through MessageChannel::MaybeHandleError.
> It doesn't look like this does.

Hmm, if so, we need full gecko IPC log.

Comment 25

[bhavana bajaj [:bajaj]]

(In reply to Sotaro Ikeda [:sotaro] from comment #15)
> (In reply to cafbot (PoC: ggrisco) from comment #14)
> > Observed on: 
> > 
> > Device: msm8226
> > Gonk Version: AU_LINUX_GECKO_B2G_KK_3.6.01.04.00.000.060
> > Moz BuildID: 20140810000201
> > B2G Version: 2.0
> > Gecko Version: 32.0
> > Gaia: 
> > http://git.mozilla.org/?p=releases/gaia.git;a=commit;
> > h=de28796a8956a48bb98ca67df6a33e0622d642d1
> > Gecko:
> > http://git.mozilla.org/?p=releases/gecko.git;a=commit;
> > h=2b27becae85092d46bfadcd4fb5605e82e1e1093
> 
> This gecko does not include Bug 1053204 fix.

Just wanted to be rather explicit that this crash report that CAF filed does not include  gecko fix include from Bug 1053204.

(If we think this is an independent issue and has no connection to 1053204, we shall continue investigating... They have picked this patch from 1053204 in their AU63 which is currently undergoing testing)

Comment 26

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

If we're randomly stomping on fds because bug 1053204 is not fixed yet that could explain this bug ...

Comment 27

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

My recommendation would be to take the logging for bug 1046084 and not turn on full IPC logging.  If this is not fixed by bug 1053204 we at least know what subsystem it is in, whereas we don't even know where to start looking for bug 1046084.

Comment 28

[Tapas[:tkundu on #b2g/gaia/memshrink/gfx] (always NI me)]

(In reply to Kyle Huey [:khuey] ([e-mail]) from comment #27)
> My recommendation would be to take the logging for bug 1046084 and not turn
> on full IPC logging.  If this is not fixed by bug 1053204 we at least know
> what subsystem it is in, whereas we don't even know where to start looking
> for bug 1046084.

Done. We waiting logs from test team. We will update asap

Comment 29

[Sotaro Ikeda [:sotaro]]

Bug 1054929 did invalid fd close. It could be one possible cause.

Comment 30

[Greg Grisco]

(In reply to Sotaro Ikeda [:sotaro] from comment #29)
> Bug 1054929 did invalid fd close. It could be one possible cause.

We applied patches from bug 1057220, maybe that will help catch invalid fd close if that's the cause.

Comment 31

[Greg Grisco]

We're still not seeing this reproduce, closing.