1767882 - Parent process hang on macOS

Status: RESOLVED DUPLICATE of bug 1767918

(Core :: Networking, defect)

Description

[Emilio Cobos Álvarez (:emilio)]

Attachment: fx-nightly-hang.txt

After updating to the latest macOS nightly the parent process is hung up and doesn't respond at all.

Filing under networking due to the stack from activity monitor for now.

Comment 1

[Markus Stange [:mstange]]

mozilla::AutoProfilerThreadSleep::~AutoProfilerThreadSleep() at /builds/worker/workspace/obj-build/dist/include/mozilla/ProfilerThreadSleep.h:52
 (inlined by) mozilla::AutoProfilerThreadSleep::~AutoProfilerThreadSleep() at /builds/worker/workspace/obj-build/dist/include/mozilla/ProfilerThreadSleep.h:52
 (inlined by) mozilla::OffTheBooksCondVar::Wait() at /builds/worker/workspace/obj-build/dist/include/mozilla/CondVar.h:59
 (inlined by) mozilla::Monitor::Wait() at /builds/worker/workspace/obj-build/dist/include/mozilla/Monitor.h:35
 (inlined by) mozilla::detail::BaseMonitorAutoLock<mozilla::Monitor>::Wait() at /builds/worker/workspace/obj-build/dist/include/mozilla/Monitor.h:134
 (inlined by) (anonymous namespace)::BufferWriter::WriteAsync() at /builds/worker/checkouts/gecko/netwerk/base/nsNetUtil.cpp:1533
 (inlined by) (anonymous namespace)::BufferWriter::Write() at /builds/worker/checkouts/gecko/netwerk/base/nsNetUtil.cpp:1412
 (inlined by) NS_ReadInputStreamToBuffer(nsIInputStream*, void**, long long, unsigned long long*) at /builds/worker/checkouts/gecko/netwerk/base/nsNetUtil.cpp:1650

Comment 2

[Markus Stange [:mstange]]

Nevermind, wrong address. The actual inline stack is:

mozilla::OffTheBooksCondVar::Wait() at /builds/worker/workspace/obj-build/dist/include/mozilla/CondVar.h:58
 (inlined by) mozilla::Monitor::Wait() at /builds/worker/workspace/obj-build/dist/include/mozilla/Monitor.h:35
 (inlined by) mozilla::detail::BaseMonitorAutoLock<mozilla::Monitor>::Wait() at /builds/worker/workspace/obj-build/dist/include/mozilla/Monitor.h:134
 (inlined by) (anonymous namespace)::BufferWriter::WriteAsync() at /builds/worker/checkouts/gecko/netwerk/base/nsNetUtil.cpp:1533
 (inlined by) (anonymous namespace)::BufferWriter::Write() at /builds/worker/checkouts/gecko/netwerk/base/nsNetUtil.cpp:1412
 (inlined by) NS_ReadInputStreamToBuffer(nsIInputStream*, void**, long long, unsigned long long*) at /builds/worker/checkouts/gecko/netwerk/base/nsNetUtil.cpp:1650

Comment 3

[Emilio Cobos Álvarez (:emilio)]

https://crash-stats.mozilla.org/report/index/39e9bf0c-6a75-47a0-801d-9a8f60220504

Comment 4

[Emilio Cobos Álvarez (:emilio)]

(That crash was forced via kill -SIGABRT <pid>)

After restarting, I got another extremely similar crash (not hang!): https://crash-stats.mozilla.org/report/index/7d812641-c946-476e-9111-5e8050220504

And now nightly seems to work fine again, so far

Comment 5

[Emilio Cobos Álvarez (:emilio)]

So that's https://searchfox.org/mozilla-central/rev/d989c65584ded72c2de85cb40bede7ac2f176387/netwerk/base/nsNetUtil.cpp#1533, but that code hasn't changed in a while...

Comment 6

[Jeff Muizelaar [:jrmuizel]]

I've seen this kind of thing as well

Comment 7

[Markus Stange [:mstange]]

It's being called from DataPipeReceiver::AsyncWait, and that code is fairly new (from bug 1346446), so maybe Nika has an idea what might be going on?