Karl Tomlinson (:karlt) Assignee
	Description • 6 years ago

+++ This bug was initially created as a clone of Bug #1471953 +++

On ThreadedDriver thread:

1. MediaStreamGraphInitThreadRunnable::Run()
   dispatches AsyncCubebOperation::SHUTDOWN to an AsyncCubebTask thread.

https://searchfox.org/mozilla-central/rev/97d488a17a848ce3bebbfc83dc916cf20b88451c/dom/media/GraphDriver.cpp#194

2. ThreadedDriver::RunThread() can hand the graph back to the main thread.

On the main thread:

3. MediaStreamGraphShutDownRunnable::Run() waits for shutdown of the
   ThreadedDriver thread and then deletes the graph.

https://searchfox.org/mozilla-central/rev/97d488a17a848ce3bebbfc83dc916cf20b88451c/dom/media/MediaStreamGraph.cpp#1466

On the AsyncCubebTask thread:

4. CompleteAudioContextOperations tries to use the graph and calls a virtual
   function on a stale reference.

https://searchfox.org/mozilla-central/rev/97d488a17a848ce3bebbfc83dc916cf20b88451c/dom/media/GraphDriver.cpp#510
https://searchfox.org/mozilla-central/rev/97d488a17a848ce3bebbfc83dc916cf20b88451c/dom/media/MediaStreamGraph.cpp#4125

This is a very unlikely sequence of events, but I don't see anything to
guarantee it won't happen.

Detected by resetting GraphDriver::mGraphImpl when no longer safe to use
https://treeherder.mozilla.org/#/jobs?repo=try&revision=ebb91a780f7eb6be87903ae68659d7f482682fbf&selectedJob=186125558

	Karl Tomlinson (:karlt) Assignee
	Comment 1 • 6 years ago

This should work provided nothing tries to wake up the MSG again.  I think
we're safe from that because mPendingUpdateRunnables will not be run.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=bc2db90dbbd2b8c20cef220d7bf3fad5fd17c172&selectedJob=186138137

	Karl Tomlinson (:karlt) Assignee
	Comment 2 • 6 years ago

Comment on attachment 8989339 [details] [diff] [review]
keep a strong reference to MediaStreamGraph from GraphDriver

[Security approval request comment]
How easily could an exploit be constructed based on the patch?
Not easily.  We don't have any evidence that this crash can happen in
practice, but the existence of bug 1471953 is evidence that similar races can
have unexpected winners.

Do comments in the patch, the check-in comment, or tests included in the patch paint a bulls-eye on the security problem?
The patch identifies the security problem.  It does not identify exactly where
the stale pointer is referenced, but
https://bugzilla.mozilla.org/attachment.cgi?id=8989335
would identify one case if landed at the same time.

Which older supported branches are affected by this flaw?
All branches since 40 may be affected.

If not all supported branches, which bug introduced the flaw?
Possibly bug 1094764.

Do you have backports for the affected branches? If not, how different, hard to create, and risky will they be?
The patch would need to be modified for other branches, but this would be
easy.
There is a low risk of introducing a leak, but automated tests should catch if
there is such a problem.

How likely is this patch to cause regressions; how much testing does it need?
Automated testing is sufficient.

	Al Billings [:abillings - ex-MoCo]
	Comment 3 • 6 years ago

sec-approval+ for trunk.
Can I get Beta and ESR60 patches made and nominated as well, please?

	Karl Tomlinson (:karlt) Assignee
	Comment 4 • 6 years ago

https://hg.mozilla.org/integration/mozilla-inbound/rev/5a0ddea9ad624286eb863e8419e144d5cbae4d58

	Karl Tomlinson (:karlt) Assignee
	Comment 5 • 6 years ago

Approval Request Comment
[Feature/Bug causing the regression]:
Possibly bug 1094764.
[User impact if declined]:
Potential sec-high.
[Is this code covered by automated tests?]:
The code is exercised by automated tests,
but we do not have a way to cause the race to be run in a way that
reproduces the bug.
[Has the fix been verified in Nightly?]:
No.
[Needs manual test from QE? If yes, steps to reproduce]:
No.
[List of other uplifts needed for the feature/fix]:
None.
[Is the change risky?]:
There is a low risk of introducing a leak, but automated tests should catch if
there is such a problem.
[Why is the change risky/not risky?]:
It changes object lifetimes.
[String changes made/needed]:
None.

	Karl Tomlinson (:karlt) Assignee
	Comment 6 • 6 years ago

[Approval Request Comment]
If this is not a sec:{high,crit} bug, please state case for ESR consideration:
sec-high.
User impact if declined:
Potential sec-high.
Fix Landed on Version:
63
Risk to taking this patch (and alternatives if risky):
There is a low risk of introducing a leak, but automated tests should catch if
there is such a problem.
String or UUID changes made by this patch:
None

	Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)
	Comment 7 • 6 years ago

https://hg.mozilla.org/mozilla-central/rev/5a0ddea9ad62

	Liz Henry (:lizzard) (relman/hg->git project)
	Comment 8 • 6 years ago

From talking with Karl on irc, sounds like this fixes both this and bug 1471953.  Let's uplift this for beta and esr60 since it already landed on trunk.

	Liz Henry (:lizzard) (relman/hg->git project)
	Comment 9 • 6 years ago

Comment on attachment 8990798 [details] [diff] [review]
beta patch

Fix for UAF, let's uplift for beta 8.

	Ryan VanderMeulen [:RyanVM]
	Comment 10 • 6 years ago
uplift

https://hg.mozilla.org/releases/mozilla-beta/rev/47b9b8ecc3f1

	Ryan VanderMeulen [:RyanVM]
	Comment 11 • 6 years ago

Comment on attachment 8990799 [details] [diff] [review]
esr60 patch

Approved for ESR 60.2 also.

	Ryan VanderMeulen [:RyanVM]
	Comment 12 • 6 years ago
uplift

https://hg.mozilla.org/releases/mozilla-esr60/rev/511fca97fb60