Closed Bug 1630550 Opened 6 years ago Closed 6 years ago

Intermittent SUMMARY: AddressSanitizer: SEGV /builds/worker/checkouts/gecko/ipc/glue/MessageChannel.cpp:769:9 in mozilla::ipc::MessageChannel::Clear()

Categories

(Core :: Audio/Video: GMP, defect)

defect
Not set
normal

Tracking

()

RESOLVED DUPLICATE of bug 1551615

People

(Reporter: intermittent-bug-filer, Unassigned)

Details

(Keywords: intermittent-failure, Whiteboard: [stockwell needswork:owner])

Filed by: sgiesecke [at] mozilla.com
Parsed log: https://treeherder.mozilla.org/logviewer.html#?job_id=297894985&repo=try
Full log: https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/B0XuI9PTSWemSffBjdgCPQ/runs/0/artifacts/public/logs/live_backing.log


This is actually not an access violation but:

MOZ_CRASH("MessageChannel destroyed without being closed.")
Summary: Intermittent PID 10785 | SUMMARY: AddressSanitizer: SEGV /builds/worker/checkouts/gecko/ipc/glue/MessageChannel.cpp:769:9 in mozilla::ipc::MessageChannel::Clear() → Intermittent SUMMARY: AddressSanitizer: SEGV /builds/worker/checkouts/gecko/ipc/glue/MessageChannel.cpp:769:9 in mozilla::ipc::MessageChannel::Clear()

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

For more information, please visit auto_nag documentation.

Flags: needinfo?(jld)

The failures from yesterday are all in opt-web-platform-tests-e10s-15, and the half dozen I looked at were all in the encrypted-media directory.

It looks like the crashes are happening in ~GMPServiceChild which is being called during shutdown.

Component: IPC → Audio/Video: GMP
Flags: needinfo?(jld)

It looks ~GMPServiceChild is destroying this hashtable:
nsRefPtrHashtable<nsUint64HashKey, GMPContentParent> mContentParents;

which causes us to call ~GMPContentParent, which destroys a message channel.

If the line numbers match SearchFox, it is crashing on this line:
MOZ_CRASH("MessageChannel destroyed without being closed.");

GMPContentParent isn't refcounted by IPDL, but it is refcounted, and it looks like the last reference is being destroyed before IPDL thinks the actor should have gone away.

Bryce, could you take a look? Thanks.

Flags: needinfo?(bvandyk)

I think this is a dupe (ish) of bug 1551615, which I'm currently trying to clobber. There's a cluster of shutdown issues there, this presentation isn't too bad, but hiding just in case.

Group: core-security
Status: NEW → RESOLVED
Closed: 6 years ago
Flags: needinfo?(bvandyk)
Resolution: --- → DUPLICATE
Group: core-security → media-core-security
Group: media-core-security
You need to log in before you can comment on or make changes to this bug.