GitLab

IPCs to WorkerThreadDispatcher arrive initially on main thread via
WorkerThreadDispatcher::OnControlMessageReceived() [1]. This IPC is
forwarded to worker thread, via
WorkerThreadDispatcher::OnMessageReceivedOnWorkerThread() [2].

If the worker destruction
(Dispatcher::WillDestroyServiceWorkerContextOnWorkerThread) happens
on worker thread between [1] and [2], IPC forwarding succeeds
(b/c the worker is alive at [1]), but IPC handling doesn't (as
worker data is already cleared while we run [2]).

Fix this by checking worker data's existence while handling the
IPC on worker thread and bailing out.

Bug: 1008143
Test: Locally running ServiceWorkerTest.Update*Extension thousand times.
Change-Id: Id5dd1b4c3d24995228aeb0f1fdd72edc1b77f747
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1825169Reviewed-by: David Bertoni <dbertoni@chromium.org>
Commit-Queue: Istiaque Ahmed <lazyboy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#700087}