Quelle nsMemoryReporterManager.h Sprache: C

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* 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/. */

#ifndef nsMemoryReporterManager_h__
#define nsMemoryReporterManager_h__

#include "mozilla/Mutex.h"
#include "nsTHashMap.h"
#include "nsHashKeys.h"
#include "nsIMemoryReporter.h"
#include "nsISupports.h"
#include "nsServiceManagerUtils.h"

#ifdef XP_WIN
#  include <windows.h>
#endif  // XP_WIN

#if defined(MOZ_MEMORY)
#  define HAVE_JEMALLOC_STATS 1
#  include "mozmemory.h"
#endif  // MOZ_MEMORY

namespace mozilla {
class MemoryReportingProcess;
namespace dom {
class MemoryReport;
}  // namespace dom
}  // namespace mozilla

class mozIDOMWindowProxy;
class nsIEventTarget;
class nsIRunnable;
class nsITimer;

class nsMemoryReporterManager final : public nsIMemoryReporterManager,
                                      public nsIMemoryReporter {
  virtual ~nsMemoryReporterManager();

public:
  NS_DECL_THREADSAFE_ISUPPORTS
  NS_DECL_NSIMEMORYREPORTERMANAGER
  NS_DECL_NSIMEMORYREPORTER

  MOZ_DEFINE_MALLOC_SIZE_OF(MallocSizeOf)

  nsMemoryReporterManager();

  // Gets the memory reporter manager service.
  static already_AddRefed<nsMemoryReporterManager> GetOrCreate() {
    nsCOMPtr<nsIMemoryReporterManager> imgr =
        do_GetService("@mozilla.org/memory-reporter-manager;1");
    return imgr.forget().downcast<nsMemoryReporterManager>();
  }

  typedef AutoTArray<nsCOMPtr<nsIMemoryReporter>, 32> StrongReportersArray;

  typedef nsTHashMap<nsRefPtrHashKey<nsIMemoryReporter>, bool>
      StrongReportersTable;
  typedef nsTHashMap<nsPtrHashKey<nsIMemoryReporter>, bool> WeakReportersTable;

  // Inter-process memory reporting proceeds as follows.
  //
  // - GetReports() (declared within NS_DECL_NSIMEMORYREPORTERMANAGER)
  //   synchronously gets memory reports for the current process, sets up some
  //   state (mPendingProcessesState) for when child processes report back --
  //   including a timer -- and starts telling child processes to get memory
  //   reports.  Control then returns to the main event loop.
  //
  //   The number of concurrent child process reports is limited by the pref
  //   "memory.report_concurrency" in order to prevent the memory overhead of
  //   memory reporting from causing problems, especially on B2G when swapping
  //   to compressed RAM; see bug 1154053.
  //
  // - HandleChildReport() is called (asynchronously) once per child process
  //   reporter callback.
  //
  // - EndProcessReport() is called (asynchronously) once per process that
  //   finishes reporting back, including the parent.  If all processes do so
  //   before time-out, the timer is cancelled.  If there are child processes
  //   whose requests have not yet been sent, they will be started until the
  //   concurrency limit is (again) reached.
  //
  // - TimeoutCallback() is called (asynchronously) if all the child processes
  //   don't respond within the time threshold.
  //
  // - FinishReporting() finishes things off.  It is *always* called -- either
  //   from EndChildReport() (if all child processes have reported back) or
  //   from TimeoutCallback() (if time-out occurs).
  //
  // All operations occur on the main thread.
  //
  // The above sequence of steps is a "request".  A partially-completed request
  // is described as "in flight".
  //
  // Each request has a "generation", a unique number that identifies it.  This
  // is used to ensure that each reports from a child process corresponds to
  // the appropriate request from the parent process.  (It's easier to
  // implement a generation system than to implement a child report request
  // cancellation mechanism.)
  //
  // Failures are mostly ignored, because it's (a) typically the most sensible
  // thing to do, and (b) often hard to do anything else.  The following are
  // the failure cases of note.
  //
  // - If a request is made while the previous request is in flight, the new
  //   request is ignored, as per getReports()'s specification.  No error is
  //   reported, because the previous request will complete soon enough.
  //
  // - If one or more child processes fail to respond within the time limit,
  //   things will proceed as if they don't exist.  No error is reported,
  //   because partial information is better than nothing.
  //
  // - If a child process reports after the time-out occurs, it is ignored.
  //   (Generation checking will ensure it is ignored even if a subsequent
  //   request is in flight;  this is the main use of generations.)  No error
  //   is reported, because there's nothing sensible to be done about it at
  //   this late stage.
  //
  // - If the time-out occurs after a child process has sent some reports but
  //   before it has signaled completion (see bug 1151597), then what it
  //   successfully sent will be included, with no explicit indication that it
  //   is incomplete.
  //
  // Now, what what happens if a child process is created/destroyed in the
  // middle of a request?  Well, PendingProcessesState is initialized with an
  // array of child process actors as of when the report started.  So...
  //
  // - If a process is created after reporting starts, it won't be sent a
  //   request for reports.  So the reported data will reflect how things were
  //   when the request began.
  //
  // - If a process is destroyed before it starts reporting back, the reported
  //   data will reflect how things are when the request ends.
  //
  // - If a process is destroyed after it starts reporting back but before it
  //   finishes, the reported data will contain a partial report for it.
  //
  // - If a process is destroyed after reporting back, but before all other
  //   child processes have reported back, it will be included in the reported
  //   data.  So the reported data will reflect how things were when the
  //   request began.
  //
  // The inconsistencies between these cases are unfortunate but difficult to
  // avoid.  It's enough of an edge case to not be worth doing more.
  //
  void HandleChildReport(uint32_t aGeneration,
                         const mozilla::dom::MemoryReport& aChildReport);
  void EndProcessReport(uint32_t aGeneration, bool aSuccess);

  // Functions that (a) implement distinguished amounts, and (b) are outside of
  // this module.
  struct AmountFns {
    mozilla::InfallibleAmountFn mJSMainRuntimeGCHeap = nullptr;
    mozilla::InfallibleAmountFn mJSMainRuntimeTemporaryPeak = nullptr;
    mozilla::InfallibleAmountFn mJSMainRuntimeCompartmentsSystem = nullptr;
    mozilla::InfallibleAmountFn mJSMainRuntimeCompartmentsUser = nullptr;
    mozilla::InfallibleAmountFn mJSMainRuntimeRealmsSystem = nullptr;
    mozilla::InfallibleAmountFn mJSMainRuntimeRealmsUser = nullptr;

    mozilla::InfallibleAmountFn mImagesContentUsedUncompressed = nullptr;

    mozilla::InfallibleAmountFn mStorageSQLite = nullptr;

    mozilla::InfallibleAmountFn mLowMemoryEventsPhysical = nullptr;

    mozilla::InfallibleAmountFn mGhostWindows = nullptr;
  };
  AmountFns mAmountFns;

  // Convenience function to get RSS easily from other code.  This is useful
  // when debugging transient memory spikes with printf instrumentation.
  static int64_t ResidentFast();

  // Convenience function to get peak RSS easily from other code.
  static int64_t ResidentPeak();

  // Convenience function to get USS easily from other code.  This is useful
  // when debugging unshared memory pages for forked processes.
  //
  // Returns 0 if, for some reason, the resident unique memory cannot be
  // determined - typically if there is a race between us and someone else
  // closing the process and we lost that race.
#ifdef XP_WIN
  static int64_t ResidentUnique(HANDLE aProcess = nullptr);
#elif XP_MACOSX
  // On MacOS this can sometimes be significantly slow. It should not be used
  // except in debugging or at the request of a user (eg about:memory).
  static int64_t ResidentUnique(mach_port_t aPort = 0);
#else
  static int64_t ResidentUnique(pid_t aPid = 0);
#endif  // XP_{WIN, MACOSX, LINUX, *}

#ifdef XP_MACOSX
  // Retrive the "phys_footprint" memory statistic on MacOS.
  static int64_t PhysicalFootprint(mach_port_t aPort = 0);
#endif

  // Functions that measure per-tab memory consumption.
  struct SizeOfTabFns {
    mozilla::JSSizeOfTabFn mJS = nullptr;
    mozilla::NonJSSizeOfTabFn mNonJS = nullptr;
  };
  SizeOfTabFns mSizeOfTabFns;

#ifdef HAVE_JEMALLOC_STATS
  // These C++ only versions of HeapAllocated and HeapOverheadFraction avoid
  // extra calls to jemalloc_stats;
  static size_t HeapAllocated(const jemalloc_stats_t& stats);
  static int64_t HeapOverheadFraction(const jemalloc_stats_t& stats);
#endif

private:
  bool IsRegistrationBlocked() MOZ_EXCLUDES(mMutex) {
    mozilla::MutexAutoLock lock(mMutex);
    return mIsRegistrationBlocked;
  }

  [[nodiscard]] nsresult RegisterReporterHelper(nsIMemoryReporter* aReporter,
                                                bool aForce, bool aStrongRef,
                                                bool aIsAsync);

  [[nodiscard]] nsresult StartGettingReports();
  // No [[nodiscard]] here because ignoring the result is common and reasonable.
  nsresult FinishReporting();

  void DispatchReporter(nsIMemoryReporter* aReporter, bool aIsAsync,
                        nsIHandleReportCallback* aHandleReport,
                        nsISupports* aHandleReportData, bool aAnonymize);

  static void TimeoutCallback(nsITimer* aTimer, void* aData);
  // Note: this timeout needs to be long enough to allow for the
  // possibility of DMD reports and/or running on a low-end phone.
  static const uint32_t kTimeoutLengthMS = 180000;

  mozilla::Mutex mMutex;
  bool mIsRegistrationBlocked MOZ_GUARDED_BY(mMutex);

  // This array is used for strong reporters that are not supposed to ever be
  // unregistered before the manager goes away. Currently this is only used
  // for reporters the manager defines itself. These reporters must be sync!
  mozilla::UniquePtr<StrongReportersArray> mStrongEternalReporters
      MOZ_GUARDED_BY(mMutex);

  // These hash tables are used for all additional reporters registered via
  // our nsIMemoryReporterManager interface.
  mozilla::UniquePtr<StrongReportersTable> mStrongReporters
      MOZ_GUARDED_BY(mMutex);
  mozilla::UniquePtr<WeakReportersTable> mWeakReporters MOZ_GUARDED_BY(mMutex);

  // These three are only used for testing purposes.
  mozilla::UniquePtr<StrongReportersArray> mSavedStrongEternalReporters
      MOZ_GUARDED_BY(mMutex);
  mozilla::UniquePtr<StrongReportersTable> mSavedStrongReporters
      MOZ_GUARDED_BY(mMutex);
  mozilla::UniquePtr<WeakReportersTable> mSavedWeakReporters
      MOZ_GUARDED_BY(mMutex);

  uint32_t mNextGeneration;  // MainThread only

  // Used to keep track of state of which processes are currently running and
  // waiting to run memory reports. Holds references to parameters needed when
  // requesting a memory report and finishing reporting.
  struct PendingProcessesState {
    uint32_t mGeneration;
    bool mAnonymize;
    bool mMinimize;
    nsCOMPtr<nsITimer> mTimer;
    nsTArray<RefPtr<mozilla::MemoryReportingProcess>> mChildrenPending;
    uint32_t mNumProcessesRunning;
    uint32_t mNumProcessesCompleted;
    uint32_t mConcurrencyLimit;
    nsCOMPtr<nsIHandleReportCallback> mHandleReport;
    nsCOMPtr<nsISupports> mHandleReportData;
    nsCOMPtr<nsIFinishReportingCallback> mFinishReporting;
    nsCOMPtr<nsISupports> mFinishReportingData;
    nsString mDMDDumpIdent;

    PendingProcessesState(uint32_t aGeneration, bool aAnonymize, bool aMinimize,
                          uint32_t aConcurrencyLimit,
                          nsIHandleReportCallback* aHandleReport,
                          nsISupports* aHandleReportData,
                          nsIFinishReportingCallback* aFinishReporting,
                          nsISupports* aFinishReportingData,
                          const nsAString& aDMDDumpIdent);
  };

  // Used to keep track of the state of the asynchronously run memory
  // reporters. The callback and file handle used when all memory reporters
  // have finished are also stored here.
  struct PendingReportersState {
    // Number of memory reporters currently running.
    uint32_t mReportsPending;

    // Callback for when all memory reporters have completed.
    nsCOMPtr<nsIFinishReportingCallback> mFinishReporting;
    nsCOMPtr<nsISupports> mFinishReportingData;

    // File handle to write a DMD report to if requested.
    FILE* mDMDFile;

    PendingReportersState(nsIFinishReportingCallback* aFinishReporting,
                          nsISupports* aFinishReportingData, FILE* aDMDFile)
        : mReportsPending(0),
          mFinishReporting(aFinishReporting),
          mFinishReportingData(aFinishReportingData),
          mDMDFile(aDMDFile) {}
  };

  // When this is non-null, a request is in flight.  Note: We use manual
  // new/delete for this because its lifetime doesn't match block scope or
  // anything like that.
  PendingProcessesState* mPendingProcessesState;  // MainThread only

  // This is reinitialized each time a call to GetReports is initiated.
  PendingReportersState* mPendingReportersState;  // MainThread only

  // Used in GetHeapAllocatedAsync() to run jemalloc_stats async.
  nsCOMPtr<nsIEventTarget> mThreadPool MOZ_GUARDED_BY(mMutex);

  PendingProcessesState* GetStateForGeneration(uint32_t aGeneration);
  [[nodiscard]] static bool StartChildReport(
      mozilla::MemoryReportingProcess* aChild,
      const PendingProcessesState* aState);
};

#define NS_MEMORY_REPORTER_MANAGER_CID              \
  {                                                 \
    0xfb97e4f5, 0x32dd, 0x497a, {                   \
      0xba, 0xa2, 0x7d, 0x1e, 0x55, 0x7, 0x99, 0x10 \
    }                                               \
  }

#endif  // nsMemoryReporterManager_h__

quality100%

¤ Dauer der Verarbeitung: 0.2 Sekunden (vorverarbeitet) ¤

Wurzel

Suchen

Beweissystem der NASA

Beweissystem Isabelle

NIST Cobol Testsuite

Cephes Mathematical Library

Wiener Entwicklungsmethode

Haftungshinweis

Die Informationen auf dieser Webseite wurden nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit, noch Qualität der bereit gestellten Informationen zugesichert.

Bemerkung:

Die farbliche Syntaxdarstellung ist noch experimentell.