The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/79f4898b566b0e580f7ebdf368ec52af5d3b9fa2

commit 79f4898b566b0e580f7ebdf368ec52af5d3b9fa2
Author: Ken Rockot <rockot@chromium.org>
Date: Fri Jun 01 20:24:38 2018

Mojo: don't inherit Source for nested RequestContext

Mojo trap events distinguish between local API call triggers and
triggers from external events (e.g. actual IPC). This signal is
leveraged to allow synchronous event dispatch without fear of
re-entrancy into Mojo implementation.

Trap events are aggregated in a thread-local RequestContext during the
extent of any interesting event (API call or incoming IPC) and then
dispatched during stack unwind to avoid re-entrancy.

The executed event handlers during this time may trigger further events,
which are thus accumulated by a new RequestContext, ad infinitum. This
means that chains of event handlers can cause arbitrarily deep
~RequestContext recursion.

Normally this is OK because higher-level Mojo library code does not
abuse the core APIs in a way that could ever actually cause such
recursion, with the exception of one case: external events which happen
to trigger an event handler on the same task sequence that processes
Mojo IPC. In this case, mojo::SimpleWatcher synchronously dispatches out
to user code which may do arbitrary stuff like bind or close new message
pipes. In such cases it's fairly trivial to construct some code which
triggers a stack overflow under certain (e.g. shutdown) conditions where
bindings are doomed to fail immediately.

This CL forces nested RequestContexts to dispatch their events as if
they came from local API calls, regardless of whether the context was
nested within an external event RequestContext, thus significantly
reducing the potential footgun surface at higher layers. This has the
side effect of forcing SimpleWatcher to schedule async dispatch tasks for
all but the outermost external event which triggers it.

Bug:  848809 
Change-Id: If4e4491ad887da73f3c23c8e9639478847c7f0c2
Reviewed-on: https://chromium-review.googlesource.com/1082921
Reviewed-by: Jay Civelli <jcivelli@chromium.org>
Commit-Queue: Ken Rockot <rockot@chromium.org>
Cr-Commit-Position: refs/heads/master@{#563774}
[modify] https://crrev.com/79f4898b566b0e580f7ebdf368ec52af5d3b9fa2/mojo/edk/system/request_context.cc