netwerk/base/nsSocketTransportService2.cpp
author Daniel Varga <dvarga@mozilla.com>
Fri, 20 Sep 2019 17:06:40 +0000
changeset 494466 07eb2cc7e1c31f2bc6ae24ade3847f39e1403acf
parent 478702 e3bd88cde7938f9a4b4348bf0174dfa537a21563
permissions -rw-r--r--
Bug 1544147 - disable test_process_error.xul for frequent failures on linux64-qr. r=gbrown Differential Revision: https://phabricator.services.mozilla.com/D46599

// vim:set sw=2 sts=2 et cin:
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

#include "nsSocketTransportService2.h"
#include "nsSocketTransport2.h"
#include "IOActivityMonitor.h"
#include "mozilla/IntegerPrintfMacros.h"
#include "mozilla/Preferences.h"
#include "nsIOService.h"
#include "nsASocketHandler.h"
#include "nsError.h"
#include "prnetdb.h"
#include "prerror.h"
#include "nsIPrefService.h"
#include "nsIPrefBranch.h"
#include "nsServiceManagerUtils.h"
#include "nsIObserverService.h"
#include "mozilla/Atomics.h"
#include "mozilla/Services.h"
#include "mozilla/Likely.h"
#include "mozilla/PublicSSL.h"
#include "mozilla/ChaosMode.h"
#include "mozilla/PodOperations.h"
#include "mozilla/Telemetry.h"
#include "nsThreadUtils.h"
#include "nsIFile.h"
#include "nsIWidget.h"

#ifdef MOZ_TASK_TRACER
#  include "GeckoTaskTracer.h"
#endif

namespace mozilla {
namespace net {

LazyLogModule gSocketTransportLog("nsSocketTransport");
LazyLogModule gUDPSocketLog("UDPSocket");
LazyLogModule gTCPSocketLog("TCPSocket");

nsSocketTransportService* gSocketTransportService = nullptr;
static Atomic<PRThread*, Relaxed> gSocketThread;

#define SEND_BUFFER_PREF "network.tcp.sendbuffer"
#define KEEPALIVE_ENABLED_PREF "network.tcp.keepalive.enabled"
#define KEEPALIVE_IDLE_TIME_PREF "network.tcp.keepalive.idle_time"
#define KEEPALIVE_RETRY_INTERVAL_PREF "network.tcp.keepalive.retry_interval"
#define KEEPALIVE_PROBE_COUNT_PREF "network.tcp.keepalive.probe_count"
#define SOCKET_LIMIT_TARGET 1000U
#define SOCKET_LIMIT_MIN 50U
#define MAX_TIME_BETWEEN_TWO_POLLS \
  "network.sts.max_time_for_events_between_two_polls"
#define POLL_BUSY_WAIT_PERIOD "network.sts.poll_busy_wait_period"
#define POLL_BUSY_WAIT_PERIOD_TIMEOUT \
  "network.sts.poll_busy_wait_period_timeout"
#define MAX_TIME_FOR_PR_CLOSE_DURING_SHUTDOWN \
  "network.sts.max_time_for_pr_close_during_shutdown"
#define POLLABLE_EVENT_TIMEOUT "network.sts.pollable_event_timeout"
#define ESNI_ENABLED "network.security.esni.enabled"
#define ESNI_DISABLED_MITM "security.pki.mitm_detected"

#define REPAIR_POLLABLE_EVENT_TIME 10

uint32_t nsSocketTransportService::gMaxCount;
PRCallOnceType nsSocketTransportService::gMaxCountInitOnce;

// Utility functions
bool OnSocketThread() { return PR_GetCurrentThread() == gSocketThread; }

//-----------------------------------------------------------------------------

bool nsSocketTransportService::SocketContext::IsTimedOut(
    PRIntervalTime now) const {
  return TimeoutIn(now) == 0;
}

void nsSocketTransportService::SocketContext::EnsureTimeout(
    PRIntervalTime now) {
  SOCKET_LOG(("SocketContext::EnsureTimeout socket=%p", mHandler));
  if (!mPollStartEpoch) {
    SOCKET_LOG(("  engaging"));
    mPollStartEpoch = now;
  }
}

void nsSocketTransportService::SocketContext::DisengageTimeout() {
  SOCKET_LOG(("SocketContext::DisengageTimeout socket=%p", mHandler));
  mPollStartEpoch = 0;
}

PRIntervalTime nsSocketTransportService::SocketContext::TimeoutIn(
    PRIntervalTime now) const {
  SOCKET_LOG(("SocketContext::TimeoutIn socket=%p, timeout=%us", mHandler,
              mHandler->mPollTimeout));

  if (mHandler->mPollTimeout == UINT16_MAX || !mPollStartEpoch) {
    SOCKET_LOG(("  not engaged"));
    return NS_SOCKET_POLL_TIMEOUT;
  }

  PRIntervalTime elapsed = (now - mPollStartEpoch);
  PRIntervalTime timeout = PR_SecondsToInterval(mHandler->mPollTimeout);

  if (elapsed >= timeout) {
    SOCKET_LOG(("  timed out!"));
    return 0;
  }
  SOCKET_LOG(("  remains %us", PR_IntervalToSeconds(timeout - elapsed)));
  return timeout - elapsed;
}

void nsSocketTransportService::SocketContext::MaybeResetEpoch() {
  if (mPollStartEpoch && mHandler->mPollTimeout == UINT16_MAX) {
    mPollStartEpoch = 0;
  }
}

//-----------------------------------------------------------------------------
// ctor/dtor (called on the main/UI thread by the service manager)

nsSocketTransportService::nsSocketTransportService()
    : mThread(nullptr),
      mLock("nsSocketTransportService::mLock"),
      mInitialized(false),
      mShuttingDown(false),
      mOffline(false),
      mGoingOffline(false),
      mRawThread(nullptr),
      mActiveListSize(SOCKET_LIMIT_MIN),
      mIdleListSize(SOCKET_LIMIT_MIN),
      mActiveCount(0),
      mIdleCount(0),
      mSentBytesCount(0),
      mReceivedBytesCount(0),
      mSendBufferSize(0),
      mKeepaliveIdleTimeS(600),
      mKeepaliveRetryIntervalS(1),
      mKeepaliveProbeCount(kDefaultTCPKeepCount),
      mKeepaliveEnabledPref(false),
      mPollableEventTimeout(TimeDuration::FromSeconds(6)),
      mServingPendingQueue(false),
      mMaxTimePerPollIter(100),
      mMaxTimeForPrClosePref(PR_SecondsToInterval(5)),
      mLastNetworkLinkChangeTime(0),
      mNetworkLinkChangeBusyWaitPeriod(PR_SecondsToInterval(50)),
      mNetworkLinkChangeBusyWaitTimeout(PR_SecondsToInterval(7)),
      mSleepPhase(false),
      mProbedMaxCount(false)
#if defined(XP_WIN)
      ,
      mPolling(false)
#endif
      ,
      mEsniEnabled(false),
      mTrustedMitmDetected(false),
      mNotTrustedMitmDetected(false) {
  NS_ASSERTION(NS_IsMainThread(), "wrong thread");

  PR_CallOnce(&gMaxCountInitOnce, DiscoverMaxCount);
  mActiveList =
      (SocketContext*)moz_xmalloc(sizeof(SocketContext) * mActiveListSize);
  mIdleList =
      (SocketContext*)moz_xmalloc(sizeof(SocketContext) * mIdleListSize);
  mPollList =
      (PRPollDesc*)moz_xmalloc(sizeof(PRPollDesc) * (mActiveListSize + 1));

  NS_ASSERTION(!gSocketTransportService, "must not instantiate twice");
  gSocketTransportService = this;
}

nsSocketTransportService::~nsSocketTransportService() {
  NS_ASSERTION(NS_IsMainThread(), "wrong thread");
  NS_ASSERTION(!mInitialized, "not shutdown properly");

  free(mActiveList);
  free(mIdleList);
  free(mPollList);
  gSocketTransportService = nullptr;
}

//-----------------------------------------------------------------------------
// event queue (any thread)

already_AddRefed<nsIThread> nsSocketTransportService::GetThreadSafely() {
  MutexAutoLock lock(mLock);
  nsCOMPtr<nsIThread> result = mThread;
  return result.forget();
}

NS_IMETHODIMP
nsSocketTransportService::DispatchFromScript(nsIRunnable* event,
                                             uint32_t flags) {
  nsCOMPtr<nsIRunnable> event_ref(event);
  return Dispatch(event_ref.forget(), flags);
}

NS_IMETHODIMP
nsSocketTransportService::Dispatch(already_AddRefed<nsIRunnable> event,
                                   uint32_t flags) {
  nsCOMPtr<nsIRunnable> event_ref(event);
  SOCKET_LOG(("STS dispatch [%p]\n", event_ref.get()));

  nsCOMPtr<nsIThread> thread = GetThreadSafely();
  nsresult rv;
  rv = thread ? thread->Dispatch(event_ref.forget(), flags)
              : NS_ERROR_NOT_INITIALIZED;
  if (rv == NS_ERROR_UNEXPECTED) {
    // Thread is no longer accepting events. We must have just shut it
    // down on the main thread. Pretend we never saw it.
    rv = NS_ERROR_NOT_INITIALIZED;
  }
  return rv;
}

NS_IMETHODIMP
nsSocketTransportService::DelayedDispatch(already_AddRefed<nsIRunnable>,
                                          uint32_t) {
  return NS_ERROR_NOT_IMPLEMENTED;
}

NS_IMETHODIMP
nsSocketTransportService::IsOnCurrentThread(bool* result) {
  nsCOMPtr<nsIThread> thread = GetThreadSafely();
  NS_ENSURE_TRUE(thread, NS_ERROR_NOT_INITIALIZED);
  return thread->IsOnCurrentThread(result);
}

NS_IMETHODIMP_(bool)
nsSocketTransportService::IsOnCurrentThreadInfallible() {
  nsCOMPtr<nsIThread> thread = GetThreadSafely();
  NS_ENSURE_TRUE(thread, false);
  return thread->IsOnCurrentThread();
}

//-----------------------------------------------------------------------------
// socket api (socket thread only)

NS_IMETHODIMP
nsSocketTransportService::NotifyWhenCanAttachSocket(nsIRunnable* event) {
  SOCKET_LOG(("nsSocketTransportService::NotifyWhenCanAttachSocket\n"));

  MOZ_ASSERT(OnSocketThread(), "not on socket thread");

  if (CanAttachSocket()) {
    return Dispatch(event, NS_DISPATCH_NORMAL);
  }

  auto* runnable = new LinkedRunnableEvent(event);
  mPendingSocketQueue.insertBack(runnable);
  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::AttachSocket(PRFileDesc* fd,
                                       nsASocketHandler* handler) {
  SOCKET_LOG(
      ("nsSocketTransportService::AttachSocket [handler=%p]\n", handler));

  MOZ_ASSERT(OnSocketThread(), "not on socket thread");

  if (!CanAttachSocket()) {
    return NS_ERROR_NOT_AVAILABLE;
  }

  SocketContext sock;
  sock.mFD = fd;
  sock.mHandler = handler;
  sock.mPollStartEpoch = 0;

  nsresult rv = AddToIdleList(&sock);
  if (NS_SUCCEEDED(rv)) NS_ADDREF(handler);
  return rv;
}

// the number of sockets that can be attached at any given time is
// limited.  this is done because some operating systems (e.g., Win9x)
// limit the number of sockets that can be created by an application.
// AttachSocket will fail if the limit is exceeded.  consumers should
// call CanAttachSocket and check the result before creating a socket.

bool nsSocketTransportService::CanAttachSocket() {
  static bool reported900FDLimit = false;

  uint32_t total = mActiveCount + mIdleCount;
  bool rv = total < gMaxCount;

  if (Telemetry::CanRecordPrereleaseData() &&
      (((total >= 900) || !rv) && !reported900FDLimit)) {
    reported900FDLimit = true;
    Telemetry::Accumulate(Telemetry::NETWORK_SESSION_AT_900FD, true);
  }

  return rv;
}

nsresult nsSocketTransportService::DetachSocket(SocketContext* listHead,
                                                SocketContext* sock) {
  SOCKET_LOG(("nsSocketTransportService::DetachSocket [handler=%p]\n",
              sock->mHandler));
  MOZ_ASSERT((listHead == mActiveList) || (listHead == mIdleList),
             "DetachSocket invalid head");

  {
#ifdef MOZ_TASK_TRACER
    tasktracer::AutoSourceEvent taskTracerEvent(
        tasktracer::SourceEventType::SocketIO);
#endif
    // inform the handler that this socket is going away
    sock->mHandler->OnSocketDetached(sock->mFD);
  }
  mSentBytesCount += sock->mHandler->ByteCountSent();
  mReceivedBytesCount += sock->mHandler->ByteCountReceived();

  // cleanup
  sock->mFD = nullptr;
  NS_RELEASE(sock->mHandler);

  if (listHead == mActiveList)
    RemoveFromPollList(sock);
  else
    RemoveFromIdleList(sock);

  // NOTE: sock is now an invalid pointer

  //
  // notify the first element on the pending socket queue...
  //
  nsCOMPtr<nsIRunnable> event;
  LinkedRunnableEvent* runnable = mPendingSocketQueue.getFirst();
  if (runnable) {
    event = runnable->TakeEvent();
    runnable->remove();
    delete runnable;
  }
  if (event) {
    // move event from pending queue to dispatch queue
    return Dispatch(event, NS_DISPATCH_NORMAL);
  }
  return NS_OK;
}

nsresult nsSocketTransportService::AddToPollList(SocketContext* sock) {
  MOZ_ASSERT(!(static_cast<uint32_t>(sock - mActiveList) < mActiveListSize),
             "AddToPollList Socket Already Active");

  SOCKET_LOG(("nsSocketTransportService::AddToPollList [handler=%p]\n",
              sock->mHandler));
  if (mActiveCount == mActiveListSize) {
    SOCKET_LOG(("  Active List size of %d met\n", mActiveCount));
    if (!GrowActiveList()) {
      NS_ERROR("too many active sockets");
      return NS_ERROR_OUT_OF_MEMORY;
    }
  }

  uint32_t newSocketIndex = mActiveCount;
  if (ChaosMode::isActive(ChaosFeature::NetworkScheduling)) {
    newSocketIndex = ChaosMode::randomUint32LessThan(mActiveCount + 1);
    PodMove(mActiveList + newSocketIndex + 1, mActiveList + newSocketIndex,
            mActiveCount - newSocketIndex);
    PodMove(mPollList + newSocketIndex + 2, mPollList + newSocketIndex + 1,
            mActiveCount - newSocketIndex);
  }

  sock->EnsureTimeout(PR_IntervalNow());
  mActiveList[newSocketIndex] = *sock;
  mActiveCount++;

  mPollList[newSocketIndex + 1].fd = sock->mFD;
  mPollList[newSocketIndex + 1].in_flags = sock->mHandler->mPollFlags;
  mPollList[newSocketIndex + 1].out_flags = 0;

  SOCKET_LOG(("  active=%u idle=%u\n", mActiveCount, mIdleCount));
  return NS_OK;
}

void nsSocketTransportService::RemoveFromPollList(SocketContext* sock) {
  SOCKET_LOG(("nsSocketTransportService::RemoveFromPollList [handler=%p]\n",
              sock->mHandler));

  uint32_t index = sock - mActiveList;
  MOZ_ASSERT(index < mActiveListSize, "invalid index");

  SOCKET_LOG(("  index=%u mActiveCount=%u\n", index, mActiveCount));

  if (index != mActiveCount - 1) {
    mActiveList[index] = mActiveList[mActiveCount - 1];
    mPollList[index + 1] = mPollList[mActiveCount];
  }
  mActiveCount--;

  SOCKET_LOG(("  active=%u idle=%u\n", mActiveCount, mIdleCount));
}

nsresult nsSocketTransportService::AddToIdleList(SocketContext* sock) {
  MOZ_ASSERT(!(static_cast<uint32_t>(sock - mIdleList) < mIdleListSize),
             "AddToIdlelList Socket Already Idle");

  SOCKET_LOG(("nsSocketTransportService::AddToIdleList [handler=%p]\n",
              sock->mHandler));
  if (mIdleCount == mIdleListSize) {
    SOCKET_LOG(("  Idle List size of %d met\n", mIdleCount));
    if (!GrowIdleList()) {
      NS_ERROR("too many idle sockets");
      return NS_ERROR_OUT_OF_MEMORY;
    }
  }

  mIdleList[mIdleCount] = *sock;
  mIdleCount++;

  SOCKET_LOG(("  active=%u idle=%u\n", mActiveCount, mIdleCount));
  return NS_OK;
}

void nsSocketTransportService::RemoveFromIdleList(SocketContext* sock) {
  SOCKET_LOG(("nsSocketTransportService::RemoveFromIdleList [handler=%p]\n",
              sock->mHandler));

  uint32_t index = sock - mIdleList;
  NS_ASSERTION(index < mIdleListSize, "invalid index in idle list");

  if (index != mIdleCount - 1) mIdleList[index] = mIdleList[mIdleCount - 1];
  mIdleCount--;

  SOCKET_LOG(("  active=%u idle=%u\n", mActiveCount, mIdleCount));
}

void nsSocketTransportService::MoveToIdleList(SocketContext* sock) {
  nsresult rv = AddToIdleList(sock);
  if (NS_FAILED(rv))
    DetachSocket(mActiveList, sock);
  else
    RemoveFromPollList(sock);
}

void nsSocketTransportService::MoveToPollList(SocketContext* sock) {
  nsresult rv = AddToPollList(sock);
  if (NS_FAILED(rv))
    DetachSocket(mIdleList, sock);
  else
    RemoveFromIdleList(sock);
}

bool nsSocketTransportService::GrowActiveList() {
  int32_t toAdd = gMaxCount - mActiveListSize;
  if (toAdd > 100) {
    toAdd = 100;
  } else if (toAdd < 1) {
    MOZ_ASSERT(false, "CanAttachSocket() should prevent this");
    return false;
  }

  mActiveListSize += toAdd;
  mActiveList = (SocketContext*)moz_xrealloc(
      mActiveList, sizeof(SocketContext) * mActiveListSize);
  mPollList = (PRPollDesc*)moz_xrealloc(
      mPollList, sizeof(PRPollDesc) * (mActiveListSize + 1));
  return true;
}

bool nsSocketTransportService::GrowIdleList() {
  int32_t toAdd = gMaxCount - mIdleListSize;
  if (toAdd > 100) {
    toAdd = 100;
  } else if (toAdd < 1) {
    MOZ_ASSERT(false, "CanAttachSocket() should prevent this");
    return false;
  }

  mIdleListSize += toAdd;
  mIdleList = (SocketContext*)moz_xrealloc(
      mIdleList, sizeof(SocketContext) * mIdleListSize);
  return true;
}

PRIntervalTime nsSocketTransportService::PollTimeout(PRIntervalTime now) {
  if (mActiveCount == 0) {
    return NS_SOCKET_POLL_TIMEOUT;
  }

  // compute minimum time before any socket timeout expires.
  PRIntervalTime minR = NS_SOCKET_POLL_TIMEOUT;
  for (uint32_t i = 0; i < mActiveCount; ++i) {
    const SocketContext& s = mActiveList[i];
    PRIntervalTime r = s.TimeoutIn(now);
    if (r < minR) {
      minR = r;
    }
  }
  if (minR == NS_SOCKET_POLL_TIMEOUT) {
    SOCKET_LOG(("poll timeout: none\n"));
    return NS_SOCKET_POLL_TIMEOUT;
  }
  SOCKET_LOG(("poll timeout: %" PRIu32 "\n", PR_IntervalToSeconds(minR)));
  return minR;
}

int32_t nsSocketTransportService::Poll(TimeDuration* pollDuration,
                                       PRIntervalTime ts) {
  PRPollDesc* pollList;
  uint32_t pollCount;
  PRIntervalTime pollTimeout;
  *pollDuration = nullptr;

  // If there are pending events for this thread then
  // DoPollIteration() should service the network without blocking.
  bool pendingEvents = false;
  mRawThread->HasPendingEvents(&pendingEvents);

  if (mPollList[0].fd) {
    mPollList[0].out_flags = 0;
    pollList = mPollList;
    pollCount = mActiveCount + 1;
    pollTimeout = pendingEvents ? PR_INTERVAL_NO_WAIT : PollTimeout(ts);
  } else {
    // no pollable event, so busy wait...
    pollCount = mActiveCount;
    if (pollCount)
      pollList = &mPollList[1];
    else
      pollList = nullptr;
    pollTimeout =
        pendingEvents ? PR_INTERVAL_NO_WAIT : PR_MillisecondsToInterval(25);
  }

  if ((ts - mLastNetworkLinkChangeTime) < mNetworkLinkChangeBusyWaitPeriod) {
    // Being here means we are few seconds after a network change has
    // been detected.
    PRIntervalTime to = mNetworkLinkChangeBusyWaitTimeout;
    if (to) {
      pollTimeout = std::min(to, pollTimeout);
      SOCKET_LOG(("  timeout shorthened after network change event"));
    }
  }

  TimeStamp pollStart;
  if (Telemetry::CanRecordPrereleaseData()) {
    pollStart = TimeStamp::NowLoRes();
  }

  SOCKET_LOG(("    timeout = %i milliseconds\n",
              PR_IntervalToMilliseconds(pollTimeout)));

  int32_t rv = [&]() {
    if (pollTimeout != PR_INTERVAL_NO_WAIT) {
      // There will be an actual non-zero wait, let the profiler record
      // idle time and mark thread as sleeping around the polling call.
      AUTO_PROFILER_LABEL("nsSocketTransportService::Poll", IDLE);
      AUTO_PROFILER_THREAD_SLEEP;
      return PR_Poll(pollList, pollCount, pollTimeout);
    }
    return PR_Poll(pollList, pollCount, pollTimeout);
  }();

  if (Telemetry::CanRecordPrereleaseData() && !pollStart.IsNull()) {
    *pollDuration = TimeStamp::NowLoRes() - pollStart;
  }

  SOCKET_LOG(("    ...returned after %i milliseconds\n",
              PR_IntervalToMilliseconds(PR_IntervalNow() - ts)));

  return rv;
}

//-----------------------------------------------------------------------------
// xpcom api

NS_IMPL_ISUPPORTS(nsSocketTransportService, nsISocketTransportService,
                  nsIRoutedSocketTransportService, nsIEventTarget,
                  nsISerialEventTarget, nsIThreadObserver, nsIRunnable,
                  nsPISocketTransportService, nsIObserver)

static const char* gCallbackPrefs[] = {
    SEND_BUFFER_PREF,
    KEEPALIVE_ENABLED_PREF,
    KEEPALIVE_IDLE_TIME_PREF,
    KEEPALIVE_RETRY_INTERVAL_PREF,
    KEEPALIVE_PROBE_COUNT_PREF,
    MAX_TIME_BETWEEN_TWO_POLLS,
    MAX_TIME_FOR_PR_CLOSE_DURING_SHUTDOWN,
    POLLABLE_EVENT_TIMEOUT,
    ESNI_ENABLED,
    ESNI_DISABLED_MITM,
    nullptr,
};

/* static */
void nsSocketTransportService::PrefCallback(const char* aPref,
                                            nsSocketTransportService* aSelf) {
  aSelf->UpdatePrefs();
}

// called from main thread only
NS_IMETHODIMP
nsSocketTransportService::Init() {
  if (!NS_IsMainThread()) {
    NS_ERROR("wrong thread");
    return NS_ERROR_UNEXPECTED;
  }

  if (mInitialized) return NS_OK;

  if (mShuttingDown) return NS_ERROR_UNEXPECTED;

  nsCOMPtr<nsIThread> thread;
  nsresult rv =
      NS_NewNamedThread("Socket Thread", getter_AddRefs(thread), this);
  if (NS_FAILED(rv)) return rv;

  {
    MutexAutoLock lock(mLock);
    // Install our mThread, protecting against concurrent readers
    thread.swap(mThread);
  }

  Preferences::RegisterCallbacks(PrefCallback, gCallbackPrefs, this);
  UpdatePrefs();

  nsCOMPtr<nsIObserverService> obsSvc = services::GetObserverService();
  if (obsSvc) {
    obsSvc->AddObserver(this, "profile-initial-state", false);
    obsSvc->AddObserver(this, "last-pb-context-exited", false);
    obsSvc->AddObserver(this, NS_WIDGET_SLEEP_OBSERVER_TOPIC, true);
    obsSvc->AddObserver(this, NS_WIDGET_WAKE_OBSERVER_TOPIC, true);
    obsSvc->AddObserver(this, "xpcom-shutdown-threads", false);
    obsSvc->AddObserver(this, NS_NETWORK_LINK_TOPIC, false);
  }

  mInitialized = true;
  return NS_OK;
}

// called from main thread only
NS_IMETHODIMP
nsSocketTransportService::Shutdown(bool aXpcomShutdown) {
  SOCKET_LOG(("nsSocketTransportService::Shutdown\n"));

  NS_ENSURE_STATE(NS_IsMainThread());

  if (!mInitialized) return NS_OK;

  if (mShuttingDown) return NS_ERROR_UNEXPECTED;

  {
    MutexAutoLock lock(mLock);

    // signal the socket thread to shutdown
    mShuttingDown = true;

    if (mPollableEvent) {
      mPollableEvent->Signal();
    }
  }

  if (!aXpcomShutdown) {
    return ShutdownThread();
  }

  return NS_OK;
}

nsresult nsSocketTransportService::ShutdownThread() {
  SOCKET_LOG(("nsSocketTransportService::ShutdownThread\n"));

  NS_ENSURE_STATE(NS_IsMainThread());

  if (!mInitialized || !mShuttingDown) return NS_OK;

  // join with thread
  mThread->Shutdown();
  {
    MutexAutoLock lock(mLock);
    // Drop our reference to mThread and make sure that any concurrent
    // readers are excluded
    mThread = nullptr;
  }

  Preferences::UnregisterCallbacks(PrefCallback, gCallbackPrefs, this);

  nsCOMPtr<nsIObserverService> obsSvc = services::GetObserverService();
  if (obsSvc) {
    obsSvc->RemoveObserver(this, "profile-initial-state");
    obsSvc->RemoveObserver(this, "last-pb-context-exited");
    obsSvc->RemoveObserver(this, NS_WIDGET_SLEEP_OBSERVER_TOPIC);
    obsSvc->RemoveObserver(this, NS_WIDGET_WAKE_OBSERVER_TOPIC);
    obsSvc->RemoveObserver(this, "xpcom-shutdown-threads");
    obsSvc->RemoveObserver(this, NS_NETWORK_LINK_TOPIC);
  }

  if (mAfterWakeUpTimer) {
    mAfterWakeUpTimer->Cancel();
    mAfterWakeUpTimer = nullptr;
  }

  IOActivityMonitor::Shutdown();

  mInitialized = false;
  mShuttingDown = false;

  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::GetOffline(bool* offline) {
  *offline = mOffline;
  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::SetOffline(bool offline) {
  MutexAutoLock lock(mLock);
  if (!mOffline && offline) {
    // signal the socket thread to go offline, so it will detach sockets
    mGoingOffline = true;
    mOffline = true;
  } else if (mOffline && !offline) {
    mOffline = false;
  }
  if (mPollableEvent) {
    mPollableEvent->Signal();
  }

  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::GetKeepaliveIdleTime(int32_t* aKeepaliveIdleTimeS) {
  MOZ_ASSERT(aKeepaliveIdleTimeS);
  if (NS_WARN_IF(!aKeepaliveIdleTimeS)) {
    return NS_ERROR_NULL_POINTER;
  }
  *aKeepaliveIdleTimeS = mKeepaliveIdleTimeS;
  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::GetKeepaliveRetryInterval(
    int32_t* aKeepaliveRetryIntervalS) {
  MOZ_ASSERT(aKeepaliveRetryIntervalS);
  if (NS_WARN_IF(!aKeepaliveRetryIntervalS)) {
    return NS_ERROR_NULL_POINTER;
  }
  *aKeepaliveRetryIntervalS = mKeepaliveRetryIntervalS;
  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::GetKeepaliveProbeCount(
    int32_t* aKeepaliveProbeCount) {
  MOZ_ASSERT(aKeepaliveProbeCount);
  if (NS_WARN_IF(!aKeepaliveProbeCount)) {
    return NS_ERROR_NULL_POINTER;
  }
  *aKeepaliveProbeCount = mKeepaliveProbeCount;
  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::CreateTransport(const nsTArray<nsCString>& types,
                                          const nsACString& host, int32_t port,
                                          nsIProxyInfo* proxyInfo,
                                          nsISocketTransport** result) {
  return CreateRoutedTransport(types, host, port, NS_LITERAL_CSTRING(""), 0,
                               proxyInfo, result);
}

NS_IMETHODIMP
nsSocketTransportService::CreateRoutedTransport(
    const nsTArray<nsCString>& types, const nsACString& host, int32_t port,
    const nsACString& hostRoute, int32_t portRoute, nsIProxyInfo* proxyInfo,
    nsISocketTransport** result) {
  NS_ENSURE_TRUE(mInitialized, NS_ERROR_NOT_INITIALIZED);
  NS_ENSURE_TRUE(port >= 0 && port <= 0xFFFF, NS_ERROR_ILLEGAL_VALUE);

  RefPtr<nsSocketTransport> trans = new nsSocketTransport();
  nsresult rv = trans->Init(types, host, port, hostRoute, portRoute, proxyInfo);
  if (NS_FAILED(rv)) {
    return rv;
  }

  trans.forget(result);
  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::CreateUnixDomainTransport(
    nsIFile* aPath, nsISocketTransport** result) {
#ifdef XP_UNIX
  nsresult rv;

  NS_ENSURE_TRUE(mInitialized, NS_ERROR_NOT_INITIALIZED);

  nsAutoCString path;
  rv = aPath->GetNativePath(path);
  if (NS_FAILED(rv)) return rv;

  RefPtr<nsSocketTransport> trans = new nsSocketTransport();

  rv = trans->InitWithFilename(path.get());
  if (NS_FAILED(rv)) return rv;

  trans.forget(result);
  return NS_OK;
#else
  return NS_ERROR_SOCKET_ADDRESS_NOT_SUPPORTED;
#endif
}

NS_IMETHODIMP
nsSocketTransportService::CreateUnixDomainAbstractAddressTransport(
    const nsACString& aName, nsISocketTransport** result) {
  // Abstract socket address is supported on Linux only
#ifdef XP_LINUX
  RefPtr<nsSocketTransport> trans = new nsSocketTransport();
  // First character of Abstract socket address is null
  UniquePtr<char[]> name(new char[aName.Length() + 1]);
  *(name.get()) = 0;
  memcpy(name.get() + 1, aName.BeginReading(), aName.Length());
  nsresult rv = trans->InitWithName(name.get(), aName.Length() + 1);
  if (NS_FAILED(rv)) {
    return rv;
  }

  trans.forget(result);
  return NS_OK;
#else
  return NS_ERROR_SOCKET_ADDRESS_NOT_SUPPORTED;
#endif
}

NS_IMETHODIMP
nsSocketTransportService::OnDispatchedEvent() {
#ifndef XP_WIN
  // On windows poll can hang and this became worse when we introduced the
  // patch for bug 698882 (see also bug 1292181), therefore we reverted the
  // behavior on windows to be as before bug 698882, e.g. write to the socket
  // also if an event dispatch is on the socket thread and writing to the
  // socket for each event.
  if (OnSocketThread()) {
    // this check is redundant to one done inside ::Signal(), but
    // we can do it here and skip obtaining the lock - given that
    // this is a relatively common occurance its worth the
    // redundant code
    SOCKET_LOG(("OnDispatchedEvent Same Thread Skip Signal\n"));
    return NS_OK;
  }
#else
  if (gIOService->IsNetTearingDown()) {
    // Poll can hang sometimes. If we are in shutdown, we are going to
    // start a watchdog. If we do not exit poll within
    // REPAIR_POLLABLE_EVENT_TIME signal a pollable event again.
    StartPollWatchdog();
  }
#endif

  MutexAutoLock lock(mLock);
  if (mPollableEvent) {
    mPollableEvent->Signal();
  }
  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::OnProcessNextEvent(nsIThreadInternal* thread,
                                             bool mayWait) {
  return NS_OK;
}

NS_IMETHODIMP
nsSocketTransportService::AfterProcessNextEvent(nsIThreadInternal* thread,
                                                bool eventWasProcessed) {
  return NS_OK;
}

void nsSocketTransportService::MarkTheLastElementOfPendingQueue() {
  mServingPendingQueue = false;
}

NS_IMETHODIMP
nsSocketTransportService::Run() {
  SOCKET_LOG(("STS thread init %d sockets\n", gMaxCount));

#if defined(XP_WIN)
  // see bug 1361495, gethostname() triggers winsock initialization.
  // so do it here (on parent and child) to protect against it being done first
  // accidentally on the main thread.. especially via PR_GetSystemInfo(). This
  // will also improve latency of first real winsock operation
  // ..
  // If STS-thread is no longer needed this should still be run before exiting

  char ignoredStackBuffer[255];
  Unused << gethostname(ignoredStackBuffer, 255);
#endif

  psm::InitializeSSLServerCertVerificationThreads();

  gSocketThread = PR_GetCurrentThread();

  {
    MutexAutoLock lock(mLock);
    mPollableEvent.reset(new PollableEvent());
    //
    // NOTE: per bug 190000, this failure could be caused by Zone-Alarm
    // or similar software.
    //
    // NOTE: per bug 191739, this failure could also be caused by lack
    // of a loopback device on Windows and OS/2 platforms (it creates
    // a loopback socket pair on these platforms to implement a pollable
    // event object).  if we can't create a pollable event, then we'll
    // have to "busy wait" to implement the socket event queue :-(
    //
    if (!mPollableEvent->Valid()) {
      mPollableEvent = nullptr;
      NS_WARNING("running socket transport thread without a pollable event");
      SOCKET_LOG(("running socket transport thread without a pollable event"));
    }

    mPollList[0].fd = mPollableEvent ? mPollableEvent->PollableFD() : nullptr;
    mPollList[0].in_flags = PR_POLL_READ | PR_POLL_EXCEPT;
    mPollList[0].out_flags = 0;
  }

  mRawThread = NS_GetCurrentThread();

  // hook ourselves up to observe event processing for this thread
  nsCOMPtr<nsIThreadInternal> threadInt = do_QueryInterface(mRawThread);
  threadInt->SetObserver(this);

  // make sure the pseudo random number generator is seeded on this thread
  srand(static_cast<unsigned>(PR_Now()));

  // For the calculation of the duration of the last cycle (i.e. the last
  // for-loop iteration before shutdown).
  TimeStamp startOfCycleForLastCycleCalc;
  int numberOfPendingEventsLastCycle;

  // For measuring of the poll iteration duration without time spent blocked
  // in poll().
  TimeStamp pollCycleStart;
  // Time blocked in poll().
  TimeDuration singlePollDuration;

  // For calculating the time needed for a new element to run.
  TimeStamp startOfIteration;
  TimeStamp startOfNextIteration;
  int numberOfPendingEvents;

  // If there is too many pending events queued, we will run some poll()
  // between them and the following variable is cumulative time spent
  // blocking in poll().
  TimeDuration pollDuration;

  for (;;) {
    bool pendingEvents = false;

    numberOfPendingEvents = 0;
    numberOfPendingEventsLastCycle = 0;
    if (Telemetry::CanRecordPrereleaseData()) {
      startOfCycleForLastCycleCalc = TimeStamp::NowLoRes();
      startOfNextIteration = TimeStamp::NowLoRes();
    }
    pollDuration = nullptr;

    do {
      if (Telemetry::CanRecordPrereleaseData()) {
        pollCycleStart = TimeStamp::NowLoRes();
      }

      DoPollIteration(&singlePollDuration);

      if (Telemetry::CanRecordPrereleaseData() && !pollCycleStart.IsNull()) {
        Telemetry::Accumulate(Telemetry::STS_POLL_BLOCK_TIME,
                              singlePollDuration.ToMilliseconds());
        Telemetry::AccumulateTimeDelta(Telemetry::STS_POLL_CYCLE,
                                       pollCycleStart + singlePollDuration,
                                       TimeStamp::NowLoRes());
        pollDuration += singlePollDuration;
      }

      mRawThread->HasPendingEvents(&pendingEvents);
      if (pendingEvents) {
        if (!mServingPendingQueue) {
          nsresult rv = Dispatch(
              NewRunnableMethod(
                  "net::nsSocketTransportService::"
                  "MarkTheLastElementOfPendingQueue",
                  this,
                  &nsSocketTransportService::MarkTheLastElementOfPendingQueue),
              nsIEventTarget::DISPATCH_NORMAL);
          if (NS_FAILED(rv)) {
            NS_WARNING(
                "Could not dispatch a new event on the "
                "socket thread.");
          } else {
            mServingPendingQueue = true;
          }

          if (Telemetry::CanRecordPrereleaseData()) {
            startOfIteration = startOfNextIteration;
            // Everything that comes after this point will
            // be served in the next iteration. If no even
            // arrives, startOfNextIteration will be reset at the
            // beginning of each for-loop.
            startOfNextIteration = TimeStamp::NowLoRes();
          }
        }
        TimeStamp eventQueueStart = TimeStamp::NowLoRes();
        do {
          NS_ProcessNextEvent(mRawThread);
          numberOfPendingEvents++;
          pendingEvents = false;
          mRawThread->HasPendingEvents(&pendingEvents);
        } while (pendingEvents && mServingPendingQueue &&
                 ((TimeStamp::NowLoRes() - eventQueueStart).ToMilliseconds() <
                  mMaxTimePerPollIter));

        if (Telemetry::CanRecordPrereleaseData() && !mServingPendingQueue &&
            !startOfIteration.IsNull()) {
          Telemetry::AccumulateTimeDelta(Telemetry::STS_POLL_AND_EVENTS_CYCLE,
                                         startOfIteration + pollDuration,
                                         TimeStamp::NowLoRes());

          Telemetry::Accumulate(Telemetry::STS_NUMBER_OF_PENDING_EVENTS,
                                numberOfPendingEvents);

          numberOfPendingEventsLastCycle += numberOfPendingEvents;
          numberOfPendingEvents = 0;
          pollDuration = nullptr;
        }
      }
    } while (pendingEvents);

    bool goingOffline = false;
    // now that our event queue is empty, check to see if we should exit
    {
      MutexAutoLock lock(mLock);
      if (mShuttingDown) {
        if (Telemetry::CanRecordPrereleaseData() &&
            !startOfCycleForLastCycleCalc.IsNull()) {
          Telemetry::Accumulate(
              Telemetry::STS_NUMBER_OF_PENDING_EVENTS_IN_THE_LAST_CYCLE,
              numberOfPendingEventsLastCycle);
          Telemetry::AccumulateTimeDelta(
              Telemetry::STS_POLL_AND_EVENT_THE_LAST_CYCLE,
              startOfCycleForLastCycleCalc, TimeStamp::NowLoRes());
        }
        break;
      }
      if (mGoingOffline) {
        mGoingOffline = false;
        goingOffline = true;
      }
    }
    // Avoid potential deadlock
    if (goingOffline) Reset(true);
  }

  SOCKET_LOG(("STS shutting down thread\n"));

  // detach all sockets, including locals
  Reset(false);

  // We don't clear gSocketThread so that OnSocketThread() won't be a false
  // alarm for events generated by stopping the SLL threads during shutdown.
  psm::StopSSLServerCertVerificationThreads();

  // Final pass over the event queue. This makes sure that events posted by
  // socket detach handlers get processed.
  NS_ProcessPendingEvents(mRawThread);

  SOCKET_LOG(("STS thread exit\n"));

  return NS_OK;
}

void nsSocketTransportService::DetachSocketWithGuard(bool aGuardLocals,
                                                     SocketContext* socketList,
                                                     int32_t index) {
  bool isGuarded = false;
  if (aGuardLocals) {
    socketList[index].mHandler->IsLocal(&isGuarded);
    if (!isGuarded) socketList[index].mHandler->KeepWhenOffline(&isGuarded);
  }
  if (!isGuarded) DetachSocket(socketList, &socketList[index]);
}

void nsSocketTransportService::Reset(bool aGuardLocals) {
  // detach any sockets
  int32_t i;
  for (i = mActiveCount - 1; i >= 0; --i) {
    DetachSocketWithGuard(aGuardLocals, mActiveList, i);
  }
  for (i = mIdleCount - 1; i >= 0; --i) {
    DetachSocketWithGuard(aGuardLocals, mIdleList, i);
  }
}

nsresult nsSocketTransportService::DoPollIteration(TimeDuration* pollDuration) {
  SOCKET_LOG(("STS poll iter\n"));

  PRIntervalTime now = PR_IntervalNow();

  int32_t i, count;
  //
  // poll loop
  //
  // walk active list backwards to see if any sockets should actually be
  // idle, then walk the idle list backwards to see if any idle sockets
  // should become active.  take care to check only idle sockets that
  // were idle to begin with ;-)
  //
  count = mIdleCount;
  for (i = mActiveCount - 1; i >= 0; --i) {
    //---
    SOCKET_LOG(("  active [%u] { handler=%p condition=%" PRIx32
                " pollflags=%hu }\n",
                i, mActiveList[i].mHandler,
                static_cast<uint32_t>(mActiveList[i].mHandler->mCondition),
                mActiveList[i].mHandler->mPollFlags));
    //---
    if (NS_FAILED(mActiveList[i].mHandler->mCondition)) {
      DetachSocket(mActiveList, &mActiveList[i]);
    } else {
      uint16_t in_flags = mActiveList[i].mHandler->mPollFlags;
      if (in_flags == 0) {
        MoveToIdleList(&mActiveList[i]);
      } else {
        // update poll flags
        mPollList[i + 1].in_flags = in_flags;
        mPollList[i + 1].out_flags = 0;
        mActiveList[i].EnsureTimeout(now);
      }
    }
  }
  for (i = count - 1; i >= 0; --i) {
    //---
    SOCKET_LOG(("  idle [%u] { handler=%p condition=%" PRIx32
                " pollflags=%hu }\n",
                i, mIdleList[i].mHandler,
                static_cast<uint32_t>(mIdleList[i].mHandler->mCondition),
                mIdleList[i].mHandler->mPollFlags));
    //---
    if (NS_FAILED(mIdleList[i].mHandler->mCondition))
      DetachSocket(mIdleList, &mIdleList[i]);
    else if (mIdleList[i].mHandler->mPollFlags != 0)
      MoveToPollList(&mIdleList[i]);
  }

  {
    MutexAutoLock lock(mLock);
    if (mPollableEvent) {
      // we want to make sure the timeout is measured from the time
      // we enter poll().  This method resets the timestamp to 'now',
      // if we were first signalled between leaving poll() and here.
      // If we didn't do this and processing events took longer than
      // the allowed signal timeout, we would detect it as a
      // false-positive.  AdjustFirstSignalTimestamp is then a no-op
      // until mPollableEvent->Clear() is called.
      mPollableEvent->AdjustFirstSignalTimestamp();
    }
  }

  SOCKET_LOG(
      ("  calling PR_Poll [active=%u idle=%u]\n", mActiveCount, mIdleCount));

#if defined(XP_WIN)
  // 30 active connections is the historic limit before firefox 7's 256. A few
  //  windows systems have troubles with the higher limit, so actively probe a
  // limit the first time we exceed 30.
  if ((mActiveCount > 30) && !mProbedMaxCount) ProbeMaxCount();
#endif

  // Measures seconds spent while blocked on PR_Poll
  int32_t n = 0;
  *pollDuration = nullptr;

  if (!gIOService->IsNetTearingDown()) {
    // Let's not do polling during shutdown.
#if defined(XP_WIN)
    StartPolling();
#endif
    n = Poll(pollDuration, now);
#if defined(XP_WIN)
    EndPolling();
#endif
  }

  now = PR_IntervalNow();

  if (n < 0) {
    SOCKET_LOG(("  PR_Poll error [%d] os error [%d]\n", PR_GetError(),
                PR_GetOSError()));
  } else {
    //
    // service "active" sockets...
    //
    uint32_t numberOfOnSocketReadyCalls = 0;
    for (i = 0; i < int32_t(mActiveCount); ++i) {
      PRPollDesc& desc = mPollList[i + 1];
      SocketContext& s = mActiveList[i];
      if (n > 0 && desc.out_flags != 0) {
#ifdef MOZ_TASK_TRACER
        tasktracer::AutoSourceEvent taskTracerEvent(
            tasktracer::SourceEventType::SocketIO);
#endif
        s.DisengageTimeout();
        s.mHandler->OnSocketReady(desc.fd, desc.out_flags);
        numberOfOnSocketReadyCalls++;
      } else if (s.IsTimedOut(now)) {
#ifdef MOZ_TASK_TRACER
        tasktracer::AutoSourceEvent taskTracerEvent(
            tasktracer::SourceEventType::SocketIO);
#endif
        SOCKET_LOG(("socket %p timed out", s.mHandler));
        s.DisengageTimeout();
        s.mHandler->OnSocketReady(desc.fd, -1);
        numberOfOnSocketReadyCalls++;
      } else {
        s.MaybeResetEpoch();
      }
    }
    if (Telemetry::CanRecordPrereleaseData()) {
      Telemetry::Accumulate(Telemetry::STS_NUMBER_OF_ONSOCKETREADY_CALLS,
                            numberOfOnSocketReadyCalls);
    }

    //
    // check for "dead" sockets and remove them (need to do this in
    // reverse order obviously).
    //
    for (i = mActiveCount - 1; i >= 0; --i) {
      if (NS_FAILED(mActiveList[i].mHandler->mCondition))
        DetachSocket(mActiveList, &mActiveList[i]);
    }

    {
      MutexAutoLock lock(mLock);
      // acknowledge pollable event (should not block)
      if (n != 0 &&
          (mPollList[0].out_flags & (PR_POLL_READ | PR_POLL_EXCEPT)) &&
          mPollableEvent &&
          ((mPollList[0].out_flags & PR_POLL_EXCEPT) ||
           !mPollableEvent->Clear())) {
        // On Windows, the TCP loopback connection in the
        // pollable event may become broken when a laptop
        // switches between wired and wireless networks or
        // wakes up from hibernation.  We try to create a
        // new pollable event.  If that fails, we fall back
        // on "busy wait".
        TryRepairPollableEvent();
      }

      if (mPollableEvent &&
          !mPollableEvent->IsSignallingAlive(mPollableEventTimeout)) {
        SOCKET_LOG(("Pollable event signalling failed/timed out"));
        TryRepairPollableEvent();
      }
    }
  }

  return NS_OK;
}

void nsSocketTransportService::UpdateSendBufferPref() {
  int32_t bufferSize;

  // If the pref is set, honor it. 0 means use OS defaults.
  nsresult rv = Preferences::GetInt(SEND_BUFFER_PREF, &bufferSize);
  if (NS_SUCCEEDED(rv)) {
    mSendBufferSize = bufferSize;
    return;
  }

#if defined(XP_WIN)
  mSendBufferSize = 131072 * 4;
#endif
}

nsresult nsSocketTransportService::UpdatePrefs() {
  mSendBufferSize = 0;

  UpdateSendBufferPref();

  // Default TCP Keepalive Values.
  int32_t keepaliveIdleTimeS;
  nsresult rv =
      Preferences::GetInt(KEEPALIVE_IDLE_TIME_PREF, &keepaliveIdleTimeS);
  if (NS_SUCCEEDED(rv))
    mKeepaliveIdleTimeS = clamped(keepaliveIdleTimeS, 1, kMaxTCPKeepIdle);

  int32_t keepaliveRetryIntervalS;
  rv = Preferences::GetInt(KEEPALIVE_RETRY_INTERVAL_PREF,
                           &keepaliveRetryIntervalS);
  if (NS_SUCCEEDED(rv))
    mKeepaliveRetryIntervalS =
        clamped(keepaliveRetryIntervalS, 1, kMaxTCPKeepIntvl);

  int32_t keepaliveProbeCount;
  rv = Preferences::GetInt(KEEPALIVE_PROBE_COUNT_PREF, &keepaliveProbeCount);
  if (NS_SUCCEEDED(rv))
    mKeepaliveProbeCount = clamped(keepaliveProbeCount, 1, kMaxTCPKeepCount);
  bool keepaliveEnabled = false;
  rv = Preferences::GetBool(KEEPALIVE_ENABLED_PREF, &keepaliveEnabled);
  if (NS_SUCCEEDED(rv) && keepaliveEnabled != mKeepaliveEnabledPref) {
    mKeepaliveEnabledPref = keepaliveEnabled;
    OnKeepaliveEnabledPrefChange();
  }

  int32_t maxTimePref;
  rv = Preferences::GetInt(MAX_TIME_BETWEEN_TWO_POLLS, &maxTimePref);
  if (NS_SUCCEEDED(rv) && maxTimePref >= 0) {
    mMaxTimePerPollIter = maxTimePref;
  }

  int32_t pollBusyWaitPeriod;
  rv = Preferences::GetInt(POLL_BUSY_WAIT_PERIOD, &pollBusyWaitPeriod);
  if (NS_SUCCEEDED(rv) && pollBusyWaitPeriod > 0) {
    mNetworkLinkChangeBusyWaitPeriod = PR_SecondsToInterval(pollBusyWaitPeriod);
  }

  int32_t pollBusyWaitPeriodTimeout;
  rv = Preferences::GetInt(POLL_BUSY_WAIT_PERIOD_TIMEOUT,
                           &pollBusyWaitPeriodTimeout);
  if (NS_SUCCEEDED(rv) && pollBusyWaitPeriodTimeout > 0) {
    mNetworkLinkChangeBusyWaitTimeout =
        PR_SecondsToInterval(pollBusyWaitPeriodTimeout);
  }

  int32_t maxTimeForPrClosePref;
  rv = Preferences::GetInt(MAX_TIME_FOR_PR_CLOSE_DURING_SHUTDOWN,
                           &maxTimeForPrClosePref);
  if (NS_SUCCEEDED(rv) && maxTimeForPrClosePref >= 0) {
    mMaxTimeForPrClosePref = PR_MillisecondsToInterval(maxTimeForPrClosePref);
  }

  int32_t pollableEventTimeout;
  rv = Preferences::GetInt(POLLABLE_EVENT_TIMEOUT, &pollableEventTimeout);
  if (NS_SUCCEEDED(rv) && pollableEventTimeout >= 0) {
    MutexAutoLock lock(mLock);
    mPollableEventTimeout = TimeDuration::FromSeconds(pollableEventTimeout);
  }

  bool esniPref = false;
  rv = Preferences::GetBool(ESNI_ENABLED, &esniPref);
  if (NS_SUCCEEDED(rv)) {
    mEsniEnabled = esniPref;
  }

  bool esniMitmPref = false;
  rv = Preferences::GetBool(ESNI_DISABLED_MITM, &esniMitmPref);
  if (NS_SUCCEEDED(rv)) {
    mTrustedMitmDetected = esniMitmPref;
  }

  return NS_OK;
}

void nsSocketTransportService::OnKeepaliveEnabledPrefChange() {
  // Dispatch to socket thread if we're not executing there.
  if (!OnSocketThread()) {
    gSocketTransportService->Dispatch(
        NewRunnableMethod(
            "net::nsSocketTransportService::OnKeepaliveEnabledPrefChange", this,
            &nsSocketTransportService::OnKeepaliveEnabledPrefChange),
        NS_DISPATCH_NORMAL);
    return;
  }

  SOCKET_LOG(("nsSocketTransportService::OnKeepaliveEnabledPrefChange %s",
              mKeepaliveEnabledPref ? "enabled" : "disabled"));

  // Notify each socket that keepalive has been en/disabled globally.
  for (int32_t i = mActiveCount - 1; i >= 0; --i) {
    NotifyKeepaliveEnabledPrefChange(&mActiveList[i]);
  }
  for (int32_t i = mIdleCount - 1; i >= 0; --i) {
    NotifyKeepaliveEnabledPrefChange(&mIdleList[i]);
  }
}

void nsSocketTransportService::NotifyKeepaliveEnabledPrefChange(
    SocketContext* sock) {
  MOZ_ASSERT(sock, "SocketContext cannot be null!");
  MOZ_ASSERT(sock->mHandler, "SocketContext does not have a handler!");

  if (!sock || !sock->mHandler) {
    return;
  }

#ifdef MOZ_TASK_TRACER
  tasktracer::AutoSourceEvent taskTracerEvent(
      tasktracer::SourceEventType::SocketIO);
#endif
  sock->mHandler->OnKeepaliveEnabledPrefChange(mKeepaliveEnabledPref);
}

NS_IMETHODIMP
nsSocketTransportService::Observe(nsISupports* subject, const char* topic,
                                  const char16_t* data) {
  SOCKET_LOG(("nsSocketTransportService::Observe topic=%s", topic));

  if (!strcmp(topic, "profile-initial-state")) {
    if (!Preferences::GetBool(IO_ACTIVITY_ENABLED_PREF, false)) {
      return NS_OK;
    }
    return net::IOActivityMonitor::Init();
  }

  if (!strcmp(topic, "last-pb-context-exited")) {
    nsCOMPtr<nsIRunnable> ev = NewRunnableMethod(
        "net::nsSocketTransportService::ClosePrivateConnections", this,
        &nsSocketTransportService::ClosePrivateConnections);
    nsresult rv = Dispatch(ev, nsIEventTarget::DISPATCH_NORMAL);
    NS_ENSURE_SUCCESS(rv, rv);
  }

  if (!strcmp(topic, NS_TIMER_CALLBACK_TOPIC)) {
    nsCOMPtr<nsITimer> timer = do_QueryInterface(subject);
    if (timer == mAfterWakeUpTimer) {
      mAfterWakeUpTimer = nullptr;
      mSleepPhase = false;
    }

#if defined(XP_WIN)
    if (timer == mPollRepairTimer) {
      DoPollRepair();
    }
#endif

  } else if (!strcmp(topic, NS_WIDGET_SLEEP_OBSERVER_TOPIC)) {
    mSleepPhase = true;
    if (mAfterWakeUpTimer) {
      mAfterWakeUpTimer->Cancel();
      mAfterWakeUpTimer = nullptr;
    }
  } else if (!strcmp(topic, NS_WIDGET_WAKE_OBSERVER_TOPIC)) {
    if (mSleepPhase && !mAfterWakeUpTimer) {
      NS_NewTimerWithObserver(getter_AddRefs(mAfterWakeUpTimer), this, 2000,
                              nsITimer::TYPE_ONE_SHOT);
    }
  } else if (!strcmp(topic, "xpcom-shutdown-threads")) {
    ShutdownThread();
  } else if (!strcmp(topic, NS_NETWORK_LINK_TOPIC)) {
    mLastNetworkLinkChangeTime = PR_IntervalNow();
    mNotTrustedMitmDetected = false;
  }

  return NS_OK;
}

void nsSocketTransportService::ClosePrivateConnections() {
  // Must be called on the socket thread.
#ifdef DEBUG
  bool onSTSThread;
  IsOnCurrentThread(&onSTSThread);
  MOZ_ASSERT(onSTSThread);
#endif

  for (int32_t i = mActiveCount - 1; i >= 0; --i) {
    if (mActiveList[i].mHandler->mIsPrivate) {
      DetachSocket(mActiveList, &mActiveList[i]);
    }
  }
  for (int32_t i = mIdleCount - 1; i >= 0; --i) {
    if (mIdleList[i].mHandler->mIsPrivate) {
      DetachSocket(mIdleList, &mIdleList[i]);
    }
  }

  ClearPrivateSSLState();
}

NS_IMETHODIMP
nsSocketTransportService::GetSendBufferSize(int32_t* value) {
  *value = mSendBufferSize;
  return NS_OK;
}

/// ugly OS specific includes are placed at the bottom of the src for clarity

#if defined(XP_WIN)
#  include <windows.h>
#elif defined(XP_UNIX) && !defined(AIX) && !defined(NEXTSTEP) && !defined(QNX)
#  include <sys/resource.h>
#endif

// Right now the only need to do this is on windows.
#if defined(XP_WIN)
void nsSocketTransportService::ProbeMaxCount() {
  MOZ_ASSERT(OnSocketThread(), "not on socket thread");

  if (mProbedMaxCount) return;
  mProbedMaxCount = true;

  // Allocate and test a PR_Poll up to the gMaxCount number of unconnected
  // sockets. See bug 692260 - windows should be able to handle 1000 sockets
  // in select() without a problem, but LSPs have been known to balk at lower
  // numbers. (64 in the bug).

  // Allocate
  struct PRPollDesc pfd[SOCKET_LIMIT_TARGET];
  uint32_t numAllocated = 0;

  for (uint32_t index = 0; index < gMaxCount; ++index) {
    pfd[index].in_flags = PR_POLL_READ | PR_POLL_WRITE | PR_POLL_EXCEPT;
    pfd[index].out_flags = 0;
    pfd[index].fd = PR_OpenTCPSocket(PR_AF_INET);
    if (!pfd[index].fd) {
      SOCKET_LOG(("Socket Limit Test index %d failed\n", index));
      if (index < SOCKET_LIMIT_MIN)
        gMaxCount = SOCKET_LIMIT_MIN;
      else
        gMaxCount = index;
      break;
    }
    ++numAllocated;
  }

  // Test
  static_assert(SOCKET_LIMIT_MIN >= 32U, "Minimum Socket Limit is >= 32");
  while (gMaxCount <= numAllocated) {
    int32_t rv = PR_Poll(pfd, gMaxCount, PR_MillisecondsToInterval(0));

    SOCKET_LOG(("Socket Limit Test poll() size=%d rv=%d\n", gMaxCount, rv));

    if (rv >= 0) break;

    SOCKET_LOG(("Socket Limit Test poll confirmationSize=%d rv=%d error=%d\n",
                gMaxCount, rv, PR_GetError()));

    gMaxCount -= 32;
    if (gMaxCount <= SOCKET_LIMIT_MIN) {
      gMaxCount = SOCKET_LIMIT_MIN;
      break;
    }
  }

  // Free
  for (uint32_t index = 0; index < numAllocated; ++index)
    if (pfd[index].fd) PR_Close(pfd[index].fd);

  Telemetry::Accumulate(Telemetry::NETWORK_PROBE_MAXCOUNT, gMaxCount);
  SOCKET_LOG(("Socket Limit Test max was confirmed at %d\n", gMaxCount));
}
#endif  // windows

PRStatus nsSocketTransportService::DiscoverMaxCount() {
  gMaxCount = SOCKET_LIMIT_MIN;

#if defined(XP_UNIX) && !defined(AIX) && !defined(NEXTSTEP) && !defined(QNX)
  // On unix and os x network sockets and file
  // descriptors are the same. OS X comes defaulted at 256,
  // most linux at 1000. We can reliably use [sg]rlimit to
  // query that and raise it if needed.

  struct rlimit rlimitData;
  if (getrlimit(RLIMIT_NOFILE, &rlimitData) == -1)  // rlimit broken - use min
    return PR_SUCCESS;

  if (rlimitData.rlim_cur >= SOCKET_LIMIT_TARGET) {  // larger than target!
    gMaxCount = SOCKET_LIMIT_TARGET;
    return PR_SUCCESS;
  }

  int32_t maxallowed = rlimitData.rlim_max;
  if ((uint32_t)maxallowed <= SOCKET_LIMIT_MIN) {
    return PR_SUCCESS;  // so small treat as if rlimit is broken
  }

  if ((maxallowed == -1) ||  // no hard cap - ok to set target
      ((uint32_t)maxallowed >= SOCKET_LIMIT_TARGET)) {
    maxallowed = SOCKET_LIMIT_TARGET;
  }

  rlimitData.rlim_cur = maxallowed;
  setrlimit(RLIMIT_NOFILE, &rlimitData);
  if ((getrlimit(RLIMIT_NOFILE, &rlimitData) != -1) &&
      (rlimitData.rlim_cur > SOCKET_LIMIT_MIN)) {
    gMaxCount = rlimitData.rlim_cur;
  }

#elif defined(XP_WIN) && !defined(WIN_CE)
  // >= XP is confirmed to have at least 1000
  static_assert(SOCKET_LIMIT_TARGET <= 1000,
                "SOCKET_LIMIT_TARGET max value is 1000");
  gMaxCount = SOCKET_LIMIT_TARGET;
#else
  // other platforms are harder to test - so leave at safe legacy value
#endif

  return PR_SUCCESS;
}

// Used to return connection info to Dashboard.cpp
void nsSocketTransportService::AnalyzeConnection(nsTArray<SocketInfo>* data,
                                                 struct SocketContext* context,
                                                 bool aActive) {
  if (context->mHandler->mIsPrivate) return;
  PRFileDesc* aFD = context->mFD;

  PRFileDesc* idLayer = PR_GetIdentitiesLayer(aFD, PR_NSPR_IO_LAYER);

  NS_ENSURE_TRUE_VOID(idLayer);

  bool tcp = PR_GetDescType(idLayer) == PR_DESC_SOCKET_TCP;

  PRNetAddr peer_addr;
  PodZero(&peer_addr);
  PRStatus rv = PR_GetPeerName(aFD, &peer_addr);
  if (rv != PR_SUCCESS) return;

  char host[64] = {0};
  rv = PR_NetAddrToString(&peer_addr, host, sizeof(host));
  if (rv != PR_SUCCESS) return;

  uint16_t port;
  if (peer_addr.raw.family == PR_AF_INET)
    port = peer_addr.inet.port;
  else
    port = peer_addr.ipv6.port;
  port = PR_ntohs(port);
  uint64_t sent = context->mHandler->ByteCountSent();
  uint64_t received = context->mHandler->ByteCountReceived();
  SocketInfo info = {nsCString(host), sent, received, port, aActive, tcp};

  data->AppendElement(info);
}

void nsSocketTransportService::GetSocketConnections(
    nsTArray<SocketInfo>* data) {
  MOZ_ASSERT(OnSocketThread(), "not on socket thread");
  for (uint32_t i = 0; i < mActiveCount; i++)
    AnalyzeConnection(data, &mActiveList[i], true);
  for (uint32_t i = 0; i < mIdleCount; i++)
    AnalyzeConnection(data, &mIdleList[i], false);
}

bool nsSocketTransportService::IsTelemetryEnabledAndNotSleepPhase() {
  return Telemetry::CanRecordPrereleaseData() && !mSleepPhase;
}

#if defined(XP_WIN)
void nsSocketTransportService::StartPollWatchdog() {
  // Start off the timer from a runnable off of the main thread in order to
  // avoid a deadlock, see bug 1370448.
  RefPtr<nsSocketTransportService> self(this);
  NS_DispatchToMainThread(NS_NewRunnableFunction(
      "nsSocketTransportService::StartPollWatchdog", [self] {
        MutexAutoLock lock(self->mLock);

        // Poll can hang sometimes. If we are in shutdown, we are going to start
        // a watchdog. If we do not exit poll within REPAIR_POLLABLE_EVENT_TIME
        // signal a pollable event again.
        MOZ_ASSERT(gIOService->IsNetTearingDown());
        if (self->mPolling && !self->mPollRepairTimer) {
          NS_NewTimerWithObserver(getter_AddRefs(self->mPollRepairTimer), self,
                                  REPAIR_POLLABLE_EVENT_TIME,
                                  nsITimer::TYPE_REPEATING_SLACK);
        }
      }));
}

void nsSocketTransportService::DoPollRepair() {
  MutexAutoLock lock(mLock);
  if (mPolling && mPollableEvent) {
    mPollableEvent->Signal();
  } else if (mPollRepairTimer) {
    mPollRepairTimer->Cancel();
  }
}

void nsSocketTransportService::StartPolling() {
  MutexAutoLock lock(mLock);
  mPolling = true;
}

void nsSocketTransportService::EndPolling() {
  MutexAutoLock lock(mLock);
  mPolling = false;
  if (mPollRepairTimer) {
    mPollRepairTimer->Cancel();
  }
}

#endif

void nsSocketTransportService::TryRepairPollableEvent() {
  mLock.AssertCurrentThreadOwns();

  NS_WARNING("Trying to repair mPollableEvent");
  mPollableEvent.reset(new PollableEvent());
  if (!mPollableEvent->Valid()) {
    mPollableEvent = nullptr;
  }
  SOCKET_LOG(
      ("running socket transport thread without "
       "a pollable event now valid=%d",
       !!mPollableEvent));
  mPollList[0].fd = mPollableEvent ? mPollableEvent->PollableFD() : nullptr;
  mPollList[0].in_flags = PR_POLL_READ | PR_POLL_EXCEPT;
  mPollList[0].out_flags = 0;
}

}  // namespace net
}  // namespace mozilla