Pieter (CC'ed) was trying to use the DevTools profiling functionality on Chrome Canary 62 but starting the profiler kept hanging the render processes. He then switched to Chrome Stable 60.0.3112.113 but its devtools profiler would also hang the render process.

Then a short time later, the Chrome Stable browser completely froze, and Pieter captured the attached hang stacks.


Looking at the stacks, it looks like page JavaScript was setting a cookie, which triggered a Sync event:

blink::V8Document::cookieAttributeGetterCallback [ExceptionState.h:120]
blink::Document::cookie [Document.cpp:0]
blink::Cookies [CookieJar.cpp:52]
content::RendererWebCookieJarImpl::Cookies [renderer_webcookiejar_impl.cc:31]
content::mojom::RenderFrameMessageFilterProxy::GetCookies [memory:2580]
mojo::InterfaceEndpointClient::AcceptWithResponder [weak_ptr.h:225]
mojo::SyncEventWatcher::SyncWatch [sync_event_watcher.cc:42]
mojo::SyncHandleRegistry::Wait [sync_handle_registry.cc:96]
mojo::WaitSet::State::Wait [wait_set.cc:204]
base::WaitableEvent::WaitMany [waitable_event_posix.cc:129]
_pthread_cond_wait
__psynch_cvwait


But the Browser process never returned from sending the sync event to the network:

base::debug::TaskAnnotator::RunTask [callback_forward.h:29]
syncer::SyncSchedulerImpl::TrySyncCycleJobImpl [sync_scheduler_impl.cc:717]
syncer::SyncSchedulerImpl::DoNudgeSyncCycleJob [sync_scheduler_impl.cc:441]
syncer::Syncer::NormalSyncShare [memory:2556]
syncer::Syncer::BuildAndPostCommits [syncer.cc:171]
syncer::Commit::PostAndProcessResponse [commit.cc:141]
syncer::SyncerProtoUtil::PostClientToServerMessage [syncer_proto_util.cc:381]
syncer::SyncerProtoUtil::PostAndProcessHeaders [syncer_proto_util.cc:337]
syncer::ServerConnectionManager::PostBufferWithCachedAuth [server_connection_manager.cc:252]
syncer::ServerConnectionManager::PostBufferToPath [server_connection_manager.cc:284]
syncer::SyncBridgedConnection::Init [sync_server_connection_manager.cc:54]
syncer::HttpBridge::MakeSynchronousPost [lock.h:26]
base::WaitableEvent::Wait [waitable_event_posix.cc:159]
base::WaitableEvent::TimedWaitUntil [waitable_event_posix.cc:225]
_pthread_cond_wait
__psynch_cvwait


It looks like a tracing event from ServiceWorker might have blocked the I/O thread?

content::BrowserThreadImpl::IOThreadRun [browser_thread_impl.cc:279]
base::RunLoop::Run [run_loop.cc:112]
base::MessagePumpLibevent::Run [message_pump_libevent.cc:224]
event_base_loop   [event.c:382]
base::MessagePumpLibevent::OnLibeventNotification [message_pump_libevent.cc:97]
mojo::edk:: [channel_posix.cc:317]
mojo::edk::Channel::OnReadComplete [memory:2580]
mojo::edk::NodeChannel::OnChannelMessage [node_channel.cc:0]
mojo::edk::RequestContext::~RequestContext [request_context.cc:31]
mojo::edk::Watch::InvokeCallback [lock.h:27]
mojo::SimpleWatcher::Context::CallNotify [simple_watcher.cc:59]
mojo::SimpleWatcher::OnHandleReady [weak_ptr.h:225]
mojo::Connector::ReadAllAvailableMessages [connector.cc:419]
mojo::Connector::ReadSingleMessage [connector.cc:390]
mojo::internal::MultiplexRouter::Accept [multiplex_router.cc:610]
mojo::internal::MultiplexRouter::ProcessIncomingMessage [multiplex_router.cc:0]
content::mojom::ServiceWorkerFetchResponseCallbackStubDispatch::Accept [service_worker_event_dispatcher.mojom.cc:330]
content::ServiceWorkerFetchDispatcher::ResponseCallback::OnFallback [memory:2580]
content::ServiceWorkerFetchDispatcher::ResponseCallback::HandleResponse [memory:2580]
content::ServiceWorkerFetchDispatcher::Complete [memory:2580]
base::internal::Invoker<base::internal::BindState<void  [memory:2580]
content::ServiceWorkerURLRequestJob::DidDispatchFetchEvent [service_worker_url_request_job.cc:812]
net::URLRequest::Restart [url_request.cc:689]
net::URLRequest::PrepareToRestart [memory:2580]
<name omitted>   [net_log_with_source.cc:86]
net::NetLog::AddEntry [net_log.cc:103]
net::TraceNetLogObserver::OnAddEntry [trace_event.h:729]
base::trace_event::TraceLog::AddTraceEventWithThreadIdAndTimestamp [trace_log.cc:1288]
base::trace_event::TraceLog::ThreadLocalEventBuffer::AddTraceEvent [lock.h:27]
base::trace_event::TraceLog::SetDisabledWhileLocked [iterator:1354]
net::NetLog::DeprecatedRemoveObserver [vector:1465]
base::internal::LockImpl::Lock [activity_tracker.h:1284]
_pthread_mutex_lock_wait   [0x7fffe07e2dfa]
__psynch_mutexwait   [0x7fffe06f7c22]


Possibly related: the browser process's main thread was waiting on the compositor during the hang:

base::MessageLoop::RunTask [vector:644]
base::debug::TaskAnnotator::RunTask [callback_forward.h:29]
ui:: [callback_forward.h:29]
ui::Compositor::RequestNewCompositorFrameSink [weak_ptr.h:246]
content::GpuProcessTransportFactory::CreateCompositorFrameSink [callback_internal.h:150]
content::BrowserGpuChannelHostFactory::EstablishGpuChannel [ref_counted.h:472]
base::internal::Invoker<base::internal::BindState<void  [bind_internal.h:343]
void base::internal::FunctorTraits<void  [ref_counted.h:472]
content::GpuProcessTransportFactory::EstablishedGpuChannel [gpu_process_transport_factory.cc:468]
ui::ContextProviderCommandBuffer::BindToCurrentThread [context_provider_command_buffer.cc:258]
gpu::CommandBufferProxyImpl::Create [command_buffer_proxy_impl.cc:91]
gpu::CommandBufferProxyImpl::Initialize [command_buffer_proxy_impl.cc:229]
IPC::SyncChannel::Send [atomic:922]
IPC::SyncChannel::WaitForReply [ref_counted.h:476]
mojo::SyncHandleRegistry::Wait [sync_handle_registry.cc:96]
mojo::WaitSet::State::Wait [wait_set.cc:204]
base::WaitableEvent::WaitMany [waitable_event_posix.cc:129]
_pthread_cond_wait
__psynch_cvwait
 

Deleted: GoogleChromeHelperSymbolicated.txt 82.9 KB

GoogleChromeHelperSymbolicated.txt 82.9 KB View Download

Deleted: GoogleChromeBrowserSymbolicated.txt 128 KB

GoogleChromeBrowserSymbolicated.txt 128 KB View Download