gecko-dev/toolkit/components/telemetry/Telemetry.cpp

3029 строки
88 KiB
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/. */
#include <algorithm>
#include <fstream>
#include <prio.h>
#include "mozilla/Attributes.h"
#include "mozilla/DebugOnly.h"
#include "mozilla/Likely.h"
#include "mozilla/MathAlgorithms.h"
#include "base/histogram.h"
#include "base/pickle.h"
#include "nsIComponentManager.h"
#include "nsIServiceManager.h"
#include "nsThreadManager.h"
#include "nsCOMArray.h"
#include "nsCOMPtr.h"
#include "nsXPCOMPrivate.h"
#include "nsIXULAppInfo.h"
#include "nsVersionComparator.h"
#include "mozilla/MemoryReporting.h"
#include "mozilla/ModuleUtils.h"
#include "nsIXPConnect.h"
#include "mozilla/Services.h"
#include "jsapi.h"
#include "jsfriendapi.h"
#include "js/GCAPI.h"
#include "nsString.h"
#include "nsITelemetry.h"
#include "nsIFile.h"
#include "nsIFileStreams.h"
#include "nsIMemoryReporter.h"
#include "nsISeekableStream.h"
#include "Telemetry.h"
#include "nsTHashtable.h"
#include "nsHashKeys.h"
#include "nsBaseHashtable.h"
#include "nsXULAppAPI.h"
#include "nsReadableUtils.h"
#include "nsThreadUtils.h"
#if defined(XP_WIN)
#include "nsUnicharUtils.h"
#endif
#include "nsNetCID.h"
#include "nsNetUtil.h"
#include "plstr.h"
#include "nsAppDirectoryServiceDefs.h"
#include "mozilla/BackgroundHangMonitor.h"
#include "mozilla/ThreadHangStats.h"
#include "mozilla/ProcessedStack.h"
#include "mozilla/Mutex.h"
#include "mozilla/FileUtils.h"
#include "mozilla/Preferences.h"
#include "mozilla/StaticPtr.h"
#include "mozilla/IOInterposer.h"
#include "mozilla/PoisonIOInterposer.h"
#include "mozilla/StartupTimeline.h"
#if defined(MOZ_ENABLE_PROFILER_SPS)
#include "shared-libraries.h"
#endif
#define EXPIRED_ID "__expired__"
namespace {
using namespace base;
using namespace mozilla;
template<class EntryType>
class AutoHashtable : public nsTHashtable<EntryType>
{
public:
AutoHashtable(uint32_t initSize = PL_DHASH_MIN_SIZE);
typedef bool (*ReflectEntryFunc)(EntryType *entry, JSContext *cx, JS::Handle<JSObject*> obj);
bool ReflectIntoJS(ReflectEntryFunc entryFunc, JSContext *cx, JS::Handle<JSObject*> obj);
private:
struct EnumeratorArgs {
JSContext *cx;
JS::Handle<JSObject*> obj;
ReflectEntryFunc entryFunc;
};
static PLDHashOperator ReflectEntryStub(EntryType *entry, void *arg);
};
template<class EntryType>
AutoHashtable<EntryType>::AutoHashtable(uint32_t initSize)
: nsTHashtable<EntryType>(initSize)
{
}
template<typename EntryType>
PLDHashOperator
AutoHashtable<EntryType>::ReflectEntryStub(EntryType *entry, void *arg)
{
EnumeratorArgs *args = static_cast<EnumeratorArgs *>(arg);
if (!args->entryFunc(entry, args->cx, args->obj)) {
return PL_DHASH_STOP;
}
return PL_DHASH_NEXT;
}
/**
* Reflect the individual entries of table into JS, usually by defining
* some property and value of obj. entryFunc is called for each entry.
*/
template<typename EntryType>
bool
AutoHashtable<EntryType>::ReflectIntoJS(ReflectEntryFunc entryFunc,
JSContext *cx, JS::Handle<JSObject*> obj)
{
EnumeratorArgs args = { cx, obj, entryFunc };
uint32_t num = this->EnumerateEntries(ReflectEntryStub, static_cast<void*>(&args));
return num == this->Count();
}
// This class is conceptually a list of ProcessedStack objects, but it represents them
// more efficiently by keeping a single global list of modules.
class CombinedStacks {
public:
typedef std::vector<Telemetry::ProcessedStack::Frame> Stack;
const Telemetry::ProcessedStack::Module& GetModule(unsigned aIndex) const;
size_t GetModuleCount() const;
const Stack& GetStack(unsigned aIndex) const;
void AddStack(const Telemetry::ProcessedStack& aStack);
size_t GetStackCount() const;
size_t SizeOfExcludingThis() const;
private:
std::vector<Telemetry::ProcessedStack::Module> mModules;
std::vector<Stack> mStacks;
};
static JSObject *
CreateJSStackObject(JSContext *cx, const CombinedStacks &stacks);
size_t
CombinedStacks::GetModuleCount() const {
return mModules.size();
}
const Telemetry::ProcessedStack::Module&
CombinedStacks::GetModule(unsigned aIndex) const {
return mModules[aIndex];
}
void
CombinedStacks::AddStack(const Telemetry::ProcessedStack& aStack) {
mStacks.resize(mStacks.size() + 1);
CombinedStacks::Stack& adjustedStack = mStacks.back();
size_t stackSize = aStack.GetStackSize();
for (size_t i = 0; i < stackSize; ++i) {
const Telemetry::ProcessedStack::Frame& frame = aStack.GetFrame(i);
uint16_t modIndex;
if (frame.mModIndex == std::numeric_limits<uint16_t>::max()) {
modIndex = frame.mModIndex;
} else {
const Telemetry::ProcessedStack::Module& module =
aStack.GetModule(frame.mModIndex);
std::vector<Telemetry::ProcessedStack::Module>::iterator modIterator =
std::find(mModules.begin(), mModules.end(), module);
if (modIterator == mModules.end()) {
mModules.push_back(module);
modIndex = mModules.size() - 1;
} else {
modIndex = modIterator - mModules.begin();
}
}
Telemetry::ProcessedStack::Frame adjustedFrame = { frame.mOffset, modIndex };
adjustedStack.push_back(adjustedFrame);
}
}
const CombinedStacks::Stack&
CombinedStacks::GetStack(unsigned aIndex) const {
return mStacks[aIndex];
}
size_t
CombinedStacks::GetStackCount() const {
return mStacks.size();
}
size_t
CombinedStacks::SizeOfExcludingThis() const {
// This is a crude approximation. We would like to do something like
// aMallocSizeOf(&mModules[0]), but on linux aMallocSizeOf will call
// malloc_usable_size which is only safe on the pointers returned by malloc.
// While it works on current libstdc++, it is better to be safe and not assume
// that &vec[0] points to one. We could use a custom allocator, but
// it doesn't seem worth it.
size_t n = 0;
n += mModules.capacity() * sizeof(Telemetry::ProcessedStack::Module);
n += mStacks.capacity() * sizeof(Stack);
for (std::vector<Stack>::const_iterator i = mStacks.begin(),
e = mStacks.end(); i != e; ++i) {
const Stack& s = *i;
n += s.capacity() * sizeof(Telemetry::ProcessedStack::Frame);
}
return n;
}
class HangReports {
public:
size_t SizeOfExcludingThis() const;
void AddHang(const Telemetry::ProcessedStack& aStack, uint32_t aDuration,
int32_t aSystemUptime, int32_t aFirefoxUptime);
uint32_t GetDuration(unsigned aIndex) const;
int32_t GetSystemUptime(unsigned aIndex) const;
int32_t GetFirefoxUptime(unsigned aIndex) const;
const CombinedStacks& GetStacks() const;
private:
struct HangInfo {
// Hang duration (in seconds)
uint32_t mDuration;
// System uptime (in minutes) at the time of the hang
int32_t mSystemUptime;
// Firefox uptime (in minutes) at the time of the hang
int32_t mFirefoxUptime;
};
std::vector<HangInfo> mHangInfo;
CombinedStacks mStacks;
};
void
HangReports::AddHang(const Telemetry::ProcessedStack& aStack,
uint32_t aDuration,
int32_t aSystemUptime,
int32_t aFirefoxUptime) {
HangInfo info = { aDuration, aSystemUptime, aFirefoxUptime };
mHangInfo.push_back(info);
mStacks.AddStack(aStack);
}
size_t
HangReports::SizeOfExcludingThis() const {
size_t n = 0;
n += mStacks.SizeOfExcludingThis();
// This is a crude approximation. See comment on
// CombinedStacks::SizeOfExcludingThis.
n += mHangInfo.capacity() * sizeof(HangInfo);
return n;
}
const CombinedStacks&
HangReports::GetStacks() const {
return mStacks;
}
uint32_t
HangReports::GetDuration(unsigned aIndex) const {
return mHangInfo[aIndex].mDuration;
}
int32_t
HangReports::GetSystemUptime(unsigned aIndex) const {
return mHangInfo[aIndex].mSystemUptime;
}
int32_t
HangReports::GetFirefoxUptime(unsigned aIndex) const {
return mHangInfo[aIndex].mFirefoxUptime;
}
/**
* IOInterposeObserver recording statistics of main-thread I/O during execution,
* aimed at consumption by TelemetryImpl
*/
class TelemetryIOInterposeObserver : public IOInterposeObserver
{
/** File-level statistics structure */
struct FileStats {
FileStats()
: creates(0)
, reads(0)
, writes(0)
, fsyncs(0)
, stats(0)
, totalTime(0)
{}
uint32_t creates; /** Number of create/open operations */
uint32_t reads; /** Number of read operations */
uint32_t writes; /** Number of write operations */
uint32_t fsyncs; /** Number of fsync operations */
uint32_t stats; /** Number of stat operations */
double totalTime; /** Accumulated duration of all operations */
};
typedef nsBaseHashtableET<nsStringHashKey, FileStats> FileIOEntryType;
struct SafeDir {
SafeDir(const nsAString& aPath, const nsAString& aSubstName)
: mPath(aPath)
, mSubstName(aSubstName)
{}
size_t SizeOfExcludingThis(mozilla::MallocSizeOf aMallocSizeOf) const {
return mPath.SizeOfExcludingThisIfUnshared(aMallocSizeOf) +
mSubstName.SizeOfExcludingThisIfUnshared(aMallocSizeOf);
}
nsString mPath; /** Path to the directory */
nsString mSubstName; /** Name to substitute with */
};
public:
TelemetryIOInterposeObserver(nsIFile* aXreDir);
/**
* An implementation of Observe that records statistics of all
* file IO operations.
*/
void Observe(Observation& aOb);
/**
* Reflect recorded file IO statistics into Javascript
*/
bool ReflectIntoJS(JSContext *cx, JS::Handle<JSObject*> rootObj);
/**
* Adds a path for inclusion in main thread I/O report.
* @param aPath Directory path
* @param aSubstName Name to substitute for aPath for privacy reasons
*/
void AddPath(const nsAString& aPath, const nsAString& aSubstName);
/**
* Get size of hash table with file stats
*/
size_t SizeOfIncludingThis(mozilla::MallocSizeOf aMallocSizeOf) const {
return aMallocSizeOf(this) + SizeOfExcludingThis(aMallocSizeOf);
}
size_t SizeOfExcludingThis(mozilla::MallocSizeOf aMallocSizeOf) const {
size_t size;
size = mFileStats.SizeOfExcludingThis(SizeOfFileIOEntryTypeExcludingThis,
aMallocSizeOf) +
mSafeDirs.SizeOfExcludingThis(aMallocSizeOf);
uint32_t safeDirsLen = mSafeDirs.Length();
for (uint32_t i = 0; i < safeDirsLen; ++i) {
size += mSafeDirs[i].SizeOfExcludingThis(aMallocSizeOf);
}
return size;
}
private:
// Statistics for each filename
AutoHashtable<FileIOEntryType> mFileStats;
// Container for whitelisted directories
nsTArray<SafeDir> mSafeDirs;
/**
* Reflect a FileIOEntryType object to a Javascript property on obj with
* filename as key containing array:
* [totalTime, creates, reads, writes, fsyncs, stats]
*/
static bool ReflectFileStats(FileIOEntryType* entry, JSContext *cx,
JS::Handle<JSObject*> obj);
static size_t SizeOfFileIOEntryTypeExcludingThis(FileIOEntryType* aEntry,
mozilla::MallocSizeOf mallocSizeOf,
void*)
{
return aEntry->GetKey().SizeOfExcludingThisIfUnshared(mallocSizeOf);
}
};
TelemetryIOInterposeObserver::TelemetryIOInterposeObserver(nsIFile* aXreDir)
{
nsAutoString xreDirPath;
nsresult rv = aXreDir->GetPath(xreDirPath);
if (NS_SUCCEEDED(rv)) {
AddPath(xreDirPath, NS_LITERAL_STRING("{xre}"));
}
}
void TelemetryIOInterposeObserver::AddPath(const nsAString& aPath,
const nsAString& aSubstName)
{
mSafeDirs.AppendElement(SafeDir(aPath, aSubstName));
}
void TelemetryIOInterposeObserver::Observe(Observation& aOb)
{
// We only report main-thread I/O
if (!NS_IsMainThread()) {
return;
}
// Get the filename
const char16_t* filename = aOb.Filename();
// Discard observations without filename
if (!filename) {
return;
}
#if defined(XP_WIN)
nsCaseInsensitiveStringComparator comparator;
#else
nsDefaultStringComparator comparator;
#endif
nsAutoString processedName;
nsDependentString filenameStr(filename);
uint32_t safeDirsLen = mSafeDirs.Length();
for (uint32_t i = 0; i < safeDirsLen; ++i) {
if (StringBeginsWith(filenameStr, mSafeDirs[i].mPath, comparator)) {
processedName = mSafeDirs[i].mSubstName;
processedName += Substring(filenameStr, mSafeDirs[i].mPath.Length());
break;
}
}
if (processedName.IsEmpty()) {
return;
}
// Create a new entry or retrieve the existing one
FileIOEntryType* entry = mFileStats.PutEntry(processedName);
if (entry) {
// Update the statistics
entry->mData.totalTime += (double) aOb.Duration().ToMilliseconds();
switch (aOb.ObservedOperation()) {
case OpCreateOrOpen:
entry->mData.creates += 1;
break;
case OpRead:
entry->mData.reads += 1;
break;
case OpWrite:
entry->mData.writes += 1;
break;
case OpFSync:
entry->mData.fsyncs += 1;
break;
case OpStat:
entry->mData.stats += 1;
break;
default:
break;
}
}
}
bool TelemetryIOInterposeObserver::ReflectFileStats(FileIOEntryType* entry,
JSContext *cx,
JS::Handle<JSObject*> obj)
{
// Let's not report arrays containing [0, 0, 0, 0, 0, 0]
if (entry->mData.totalTime == 0 && entry->mData.creates == 0 &&
entry->mData.reads == 0 && entry->mData.writes == 0 &&
entry->mData.fsyncs == 0 && entry->mData.stats == 0) {
return true;
}
// Array we want to report
JS::AutoValueArray<6> stats(cx);
stats[0].setNumber(entry->mData.totalTime);
stats[1].setNumber(entry->mData.creates);
stats[2].setNumber(entry->mData.reads);
stats[3].setNumber(entry->mData.writes);
stats[4].setNumber(entry->mData.fsyncs);
stats[5].setNumber(entry->mData.stats);
// Create jsEntry as array of elements above
JS::RootedObject jsEntry(cx, JS_NewArrayObject(cx, stats));
if (!jsEntry) {
return false;
}
// Add jsEntry to top-level dictionary
const nsAString& key = entry->GetKey();
return JS_DefineUCProperty(cx, obj, key.Data(), key.Length(),
OBJECT_TO_JSVAL(jsEntry), NULL, NULL,
JSPROP_ENUMERATE);
}
bool TelemetryIOInterposeObserver::ReflectIntoJS(JSContext *cx,
JS::Handle<JSObject*> rootObj)
{
return mFileStats.ReflectIntoJS(ReflectFileStats, cx, rootObj);
}
// This is not a member of TelemetryImpl because we want to record I/O during
// startup.
StaticAutoPtr<TelemetryIOInterposeObserver> sTelemetryIOObserver;
void
ClearIOReporting()
{
if (!sTelemetryIOObserver) {
return;
}
IOInterposer::Unregister(IOInterposeObserver::OpAll, sTelemetryIOObserver);
sTelemetryIOObserver = nullptr;
}
class TelemetryImpl MOZ_FINAL
: public nsITelemetry
, public nsIMemoryReporter
{
NS_DECL_THREADSAFE_ISUPPORTS
NS_DECL_NSITELEMETRY
NS_DECL_NSIMEMORYREPORTER
public:
~TelemetryImpl();
void InitMemoryReporter();
static bool CanRecord();
static already_AddRefed<nsITelemetry> CreateTelemetryInstance();
static void ShutdownTelemetry();
static void RecordSlowStatement(const nsACString &sql, const nsACString &dbName,
uint32_t delay);
#if defined(MOZ_ENABLE_PROFILER_SPS)
static void RecordChromeHang(uint32_t aDuration,
Telemetry::ProcessedStack &aStack,
int32_t aSystemUptime,
int32_t aFirefoxUptime);
#endif
static void RecordThreadHangStats(Telemetry::ThreadHangStats& aStats);
static nsresult GetHistogramEnumId(const char *name, Telemetry::ID *id);
size_t SizeOfIncludingThis(mozilla::MallocSizeOf aMallocSizeOf);
struct Stat {
uint32_t hitCount;
uint32_t totalTime;
};
struct StmtStats {
struct Stat mainThread;
struct Stat otherThreads;
};
typedef nsBaseHashtableET<nsCStringHashKey, StmtStats> SlowSQLEntryType;
private:
TelemetryImpl();
static nsCString SanitizeSQL(const nsACString& sql);
enum SanitizedState { Sanitized, Unsanitized };
static void StoreSlowSQL(const nsACString &offender, uint32_t delay,
SanitizedState state);
static bool ReflectMainThreadSQL(SlowSQLEntryType *entry, JSContext *cx,
JS::Handle<JSObject*> obj);
static bool ReflectOtherThreadsSQL(SlowSQLEntryType *entry, JSContext *cx,
JS::Handle<JSObject*> obj);
static bool ReflectSQL(const SlowSQLEntryType *entry, const Stat *stat,
JSContext *cx, JS::Handle<JSObject*> obj);
bool AddSQLInfo(JSContext *cx, JS::Handle<JSObject*> rootObj, bool mainThread,
bool privateSQL);
bool GetSQLStats(JSContext *cx, JS::MutableHandle<JS::Value> ret,
bool includePrivateSql);
// Like GetHistogramById, but returns the underlying C++ object, not the JS one.
nsresult GetHistogramByName(const nsACString &name, Histogram **ret);
bool ShouldReflectHistogram(Histogram *h);
void IdentifyCorruptHistograms(StatisticsRecorder::Histograms &hs);
typedef StatisticsRecorder::Histograms::iterator HistogramIterator;
struct AddonHistogramInfo {
uint32_t min;
uint32_t max;
uint32_t bucketCount;
uint32_t histogramType;
Histogram *h;
};
typedef nsBaseHashtableET<nsCStringHashKey, AddonHistogramInfo> AddonHistogramEntryType;
typedef AutoHashtable<AddonHistogramEntryType> AddonHistogramMapType;
typedef nsBaseHashtableET<nsCStringHashKey, AddonHistogramMapType *> AddonEntryType;
typedef AutoHashtable<AddonEntryType> AddonMapType;
static bool AddonHistogramReflector(AddonHistogramEntryType *entry,
JSContext *cx, JS::Handle<JSObject*> obj);
static bool AddonReflector(AddonEntryType *entry, JSContext *cx, JS::Handle<JSObject*> obj);
static bool CreateHistogramForAddon(const nsACString &name,
AddonHistogramInfo &info);
void ReadLateWritesStacks(nsIFile* aProfileDir);
AddonMapType mAddonMap;
// This is used for speedy string->Telemetry::ID conversions
typedef nsBaseHashtableET<nsCharPtrHashKey, Telemetry::ID> CharPtrEntryType;
typedef AutoHashtable<CharPtrEntryType> HistogramMapType;
HistogramMapType mHistogramMap;
bool mCanRecord;
static TelemetryImpl *sTelemetry;
AutoHashtable<SlowSQLEntryType> mPrivateSQL;
AutoHashtable<SlowSQLEntryType> mSanitizedSQL;
// This gets marked immutable in debug builds, so we can't use
// AutoHashtable here.
nsTHashtable<nsCStringHashKey> mTrackedDBs;
Mutex mHashMutex;
HangReports mHangReports;
Mutex mHangReportsMutex;
// mThreadHangStats stores recorded, inactive thread hang stats
Vector<Telemetry::ThreadHangStats> mThreadHangStats;
Mutex mThreadHangStatsMutex;
CombinedStacks mLateWritesStacks; // This is collected out of the main thread.
bool mCachedTelemetryData;
uint32_t mLastShutdownTime;
uint32_t mFailedLockCount;
nsCOMArray<nsIFetchTelemetryDataCallback> mCallbacks;
friend class nsFetchTelemetryData;
};
TelemetryImpl* TelemetryImpl::sTelemetry = nullptr;
MOZ_DEFINE_MALLOC_SIZE_OF(TelemetryMallocSizeOf)
NS_IMETHODIMP
TelemetryImpl::CollectReports(nsIHandleReportCallback* aHandleReport,
nsISupports* aData)
{
return MOZ_COLLECT_REPORT(
"explicit/telemetry", KIND_HEAP, UNITS_BYTES,
SizeOfIncludingThis(TelemetryMallocSizeOf),
"Memory used by the telemetry system.");
}
// A initializer to initialize histogram collection
StatisticsRecorder gStatisticsRecorder;
// Hardcoded probes
struct TelemetryHistogram {
uint32_t min;
uint32_t max;
uint32_t bucketCount;
uint32_t histogramType;
uint32_t id_offset;
uint32_t expiration_offset;
bool extendedStatisticsOK;
const char *id() const;
const char *expiration() const;
};
#include "TelemetryHistogramData.inc"
bool gCorruptHistograms[Telemetry::HistogramCount];
const char *
TelemetryHistogram::id() const
{
return &gHistogramStringTable[this->id_offset];
}
const char *
TelemetryHistogram::expiration() const
{
return &gHistogramStringTable[this->expiration_offset];
}
bool
IsExpired(const char *expiration){
static Version current_version = Version(MOZ_APP_VERSION);
MOZ_ASSERT(expiration);
return strcmp(expiration, "never") && (mozilla::Version(expiration) <= current_version);
}
bool
IsExpired(const Histogram *histogram){
return histogram->histogram_name() == EXPIRED_ID;
}
nsresult
HistogramGet(const char *name, const char *expiration, uint32_t min, uint32_t max,
uint32_t bucketCount, uint32_t histogramType, Histogram **result)
{
if (histogramType != nsITelemetry::HISTOGRAM_BOOLEAN
&& histogramType != nsITelemetry::HISTOGRAM_FLAG) {
// Sanity checks for histogram parameters.
if (min >= max)
return NS_ERROR_ILLEGAL_VALUE;
if (bucketCount <= 2)
return NS_ERROR_ILLEGAL_VALUE;
if (min < 1)
return NS_ERROR_ILLEGAL_VALUE;
}
if (IsExpired(expiration)) {
name = EXPIRED_ID;
min = 1;
max = 2;
bucketCount = 3;
histogramType = nsITelemetry::HISTOGRAM_LINEAR;
}
switch (histogramType) {
case nsITelemetry::HISTOGRAM_EXPONENTIAL:
*result = Histogram::FactoryGet(name, min, max, bucketCount, Histogram::kUmaTargetedHistogramFlag);
break;
case nsITelemetry::HISTOGRAM_LINEAR:
*result = LinearHistogram::FactoryGet(name, min, max, bucketCount, Histogram::kUmaTargetedHistogramFlag);
break;
case nsITelemetry::HISTOGRAM_BOOLEAN:
*result = BooleanHistogram::FactoryGet(name, Histogram::kUmaTargetedHistogramFlag);
break;
case nsITelemetry::HISTOGRAM_FLAG:
*result = FlagHistogram::FactoryGet(name, Histogram::kUmaTargetedHistogramFlag);
break;
default:
return NS_ERROR_INVALID_ARG;
}
return NS_OK;
}
// O(1) histogram lookup by numeric id
nsresult
GetHistogramByEnumId(Telemetry::ID id, Histogram **ret)
{
static Histogram* knownHistograms[Telemetry::HistogramCount] = {0};
Histogram *h = knownHistograms[id];
if (h) {
*ret = h;
return NS_OK;
}
const TelemetryHistogram &p = gHistograms[id];
nsresult rv = HistogramGet(p.id(), p.expiration(), p.min, p.max, p.bucketCount, p.histogramType, &h);
if (NS_FAILED(rv))
return rv;
#ifdef DEBUG
// Check that the C++ Histogram code computes the same ranges as the
// Python histogram code.
if (!IsExpired(p.expiration())) {
const struct bounds &b = gBucketLowerBoundIndex[id];
if (b.length != 0) {
MOZ_ASSERT(size_t(b.length) == h->bucket_count(),
"C++/Python bucket # mismatch");
for (int i = 0; i < b.length; ++i) {
MOZ_ASSERT(gBucketLowerBounds[b.offset + i] == h->ranges(i),
"C++/Python bucket mismatch");
}
}
}
#endif
if (p.extendedStatisticsOK) {
h->SetFlags(Histogram::kExtendedStatisticsFlag);
}
*ret = knownHistograms[id] = h;
return NS_OK;
}
bool
FillRanges(JSContext *cx, JS::Handle<JSObject*> array, Histogram *h)
{
JS::Rooted<JS::Value> range(cx);
for (size_t i = 0; i < h->bucket_count(); i++) {
range = INT_TO_JSVAL(h->ranges(i));
if (!JS_DefineElement(cx, array, i, range, nullptr, nullptr, JSPROP_ENUMERATE))
return false;
}
return true;
}
enum reflectStatus {
REFLECT_OK,
REFLECT_CORRUPT,
REFLECT_FAILURE
};
enum reflectStatus
ReflectHistogramAndSamples(JSContext *cx, JS::Handle<JSObject*> obj, Histogram *h,
const Histogram::SampleSet &ss)
{
// We don't want to reflect corrupt histograms.
if (h->FindCorruption(ss) != Histogram::NO_INCONSISTENCIES) {
return REFLECT_CORRUPT;
}
if (!(JS_DefineProperty(cx, obj, "min", INT_TO_JSVAL(h->declared_min()), nullptr, nullptr, JSPROP_ENUMERATE)
&& JS_DefineProperty(cx, obj, "max", INT_TO_JSVAL(h->declared_max()), nullptr, nullptr, JSPROP_ENUMERATE)
&& JS_DefineProperty(cx, obj, "histogram_type", INT_TO_JSVAL(h->histogram_type()), nullptr, nullptr, JSPROP_ENUMERATE)
&& JS_DefineProperty(cx, obj, "sum", DOUBLE_TO_JSVAL(ss.sum()), nullptr, nullptr, JSPROP_ENUMERATE))) {
return REFLECT_FAILURE;
}
if (h->histogram_type() == Histogram::HISTOGRAM) {
if (!(JS_DefineProperty(cx, obj, "log_sum", DOUBLE_TO_JSVAL(ss.log_sum()), nullptr, nullptr, JSPROP_ENUMERATE)
&& JS_DefineProperty(cx, obj, "log_sum_squares", DOUBLE_TO_JSVAL(ss.log_sum_squares()), nullptr, nullptr, JSPROP_ENUMERATE))) {
return REFLECT_FAILURE;
}
} else {
// Export |sum_squares| as two separate 32-bit properties so that we
// can accurately reconstruct it on the analysis side.
uint64_t sum_squares = ss.sum_squares();
// Cast to avoid implicit truncation warnings.
uint32_t lo = static_cast<uint32_t>(sum_squares);
uint32_t hi = static_cast<uint32_t>(sum_squares >> 32);
if (!(JS_DefineProperty(cx, obj, "sum_squares_lo", INT_TO_JSVAL(lo), nullptr, nullptr, JSPROP_ENUMERATE)
&& JS_DefineProperty(cx, obj, "sum_squares_hi", INT_TO_JSVAL(hi), nullptr, nullptr, JSPROP_ENUMERATE))) {
return REFLECT_FAILURE;
}
}
const size_t count = h->bucket_count();
JS::Rooted<JSObject*> rarray(cx, JS_NewArrayObject(cx, count));
if (!rarray) {
return REFLECT_FAILURE;
}
if (!(FillRanges(cx, rarray, h)
&& JS_DefineProperty(cx, obj, "ranges", OBJECT_TO_JSVAL(rarray),
nullptr, nullptr, JSPROP_ENUMERATE))) {
return REFLECT_FAILURE;
}
JS::Rooted<JSObject*> counts_array(cx, JS_NewArrayObject(cx, count));
if (!counts_array) {
return REFLECT_FAILURE;
}
if (!JS_DefineProperty(cx, obj, "counts", OBJECT_TO_JSVAL(counts_array),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return REFLECT_FAILURE;
}
for (size_t i = 0; i < count; i++) {
if (!JS_DefineElement(cx, counts_array, i, INT_TO_JSVAL(ss.counts(i)),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return REFLECT_FAILURE;
}
}
return REFLECT_OK;
}
enum reflectStatus
ReflectHistogramSnapshot(JSContext *cx, JS::Handle<JSObject*> obj, Histogram *h)
{
Histogram::SampleSet ss;
h->SnapshotSample(&ss);
return ReflectHistogramAndSamples(cx, obj, h, ss);
}
bool
IsEmpty(const Histogram *h)
{
Histogram::SampleSet ss;
h->SnapshotSample(&ss);
return ss.counts(0) == 0 && ss.sum() == 0;
}
bool
JSHistogram_Add(JSContext *cx, unsigned argc, JS::Value *vp)
{
JS::CallArgs args = CallArgsFromVp(argc, vp);
if (!args.length()) {
JS_ReportError(cx, "Expected one argument");
return false;
}
if (!(args[0].isNumber() || args[0].isBoolean())) {
JS_ReportError(cx, "Not a number");
return false;
}
int32_t value;
if (!JS::ToInt32(cx, args[0], &value)) {
return false;
}
if (TelemetryImpl::CanRecord()) {
JSObject *obj = JS_THIS_OBJECT(cx, vp);
if (!obj) {
return false;
}
Histogram *h = static_cast<Histogram*>(JS_GetPrivate(obj));
h->Add(value);
}
return true;
}
bool
JSHistogram_Snapshot(JSContext *cx, unsigned argc, JS::Value *vp)
{
JSObject *obj = JS_THIS_OBJECT(cx, vp);
if (!obj) {
return false;
}
Histogram *h = static_cast<Histogram*>(JS_GetPrivate(obj));
JS::Rooted<JSObject*> snapshot(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!snapshot)
return false;
switch (ReflectHistogramSnapshot(cx, snapshot, h)) {
case REFLECT_FAILURE:
return false;
case REFLECT_CORRUPT:
JS_ReportError(cx, "Histogram is corrupt");
return false;
case REFLECT_OK:
JS_SET_RVAL(cx, vp, OBJECT_TO_JSVAL(snapshot));
return true;
default:
MOZ_CRASH("unhandled reflection status");
}
}
bool
JSHistogram_Clear(JSContext *cx, unsigned argc, JS::Value *vp)
{
JSObject *obj = JS_THIS_OBJECT(cx, vp);
if (!obj) {
return false;
}
Histogram *h = static_cast<Histogram*>(JS_GetPrivate(obj));
h->Clear();
return true;
}
nsresult
WrapAndReturnHistogram(Histogram *h, JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
static const JSClass JSHistogram_class = {
"JSHistogram", /* name */
JSCLASS_HAS_PRIVATE, /* flags */
JS_PropertyStub, JS_DeletePropertyStub, JS_PropertyStub, JS_StrictPropertyStub,
JS_EnumerateStub, JS_ResolveStub, JS_ConvertStub
};
JS::Rooted<JSObject*> obj(cx, JS_NewObject(cx, &JSHistogram_class, JS::NullPtr(), JS::NullPtr()));
if (!obj)
return NS_ERROR_FAILURE;
if (!(JS_DefineFunction(cx, obj, "add", JSHistogram_Add, 1, 0)
&& JS_DefineFunction(cx, obj, "snapshot", JSHistogram_Snapshot, 0, 0)
&& JS_DefineFunction(cx, obj, "clear", JSHistogram_Clear, 0, 0))) {
return NS_ERROR_FAILURE;
}
JS_SetPrivate(obj, h);
ret.setObject(*obj);
return NS_OK;
}
static uint32_t
ReadLastShutdownDuration(const char *filename) {
FILE *f = fopen(filename, "r");
if (!f) {
return 0;
}
int shutdownTime;
int r = fscanf(f, "%d\n", &shutdownTime);
fclose(f);
if (r != 1) {
return 0;
}
return shutdownTime;
}
const int32_t kMaxFailedProfileLockFileSize = 10;
bool
GetFailedLockCount(nsIInputStream* inStream, uint32_t aCount,
unsigned int& result)
{
nsAutoCString bufStr;
nsresult rv;
rv = NS_ReadInputStreamToString(inStream, bufStr, aCount);
NS_ENSURE_SUCCESS(rv, false);
result = bufStr.ToInteger(&rv);
return NS_SUCCEEDED(rv) && result > 0;
}
nsresult
GetFailedProfileLockFile(nsIFile* *aFile, nsIFile* aProfileDir)
{
NS_ENSURE_ARG_POINTER(aProfileDir);
nsresult rv = aProfileDir->Clone(aFile);
NS_ENSURE_SUCCESS(rv, rv);
(*aFile)->AppendNative(NS_LITERAL_CSTRING("Telemetry.FailedProfileLocks.txt"));
return NS_OK;
}
class nsFetchTelemetryData : public nsRunnable
{
public:
nsFetchTelemetryData(const char* aShutdownTimeFilename,
nsIFile* aFailedProfileLockFile,
nsIFile* aProfileDir)
: mShutdownTimeFilename(aShutdownTimeFilename),
mFailedProfileLockFile(aFailedProfileLockFile),
mTelemetry(TelemetryImpl::sTelemetry),
mProfileDir(aProfileDir)
{
}
private:
const char* mShutdownTimeFilename;
nsCOMPtr<nsIFile> mFailedProfileLockFile;
nsRefPtr<TelemetryImpl> mTelemetry;
nsCOMPtr<nsIFile> mProfileDir;
public:
void MainThread() {
mTelemetry->mCachedTelemetryData = true;
for (unsigned int i = 0, n = mTelemetry->mCallbacks.Count(); i < n; ++i) {
mTelemetry->mCallbacks[i]->Complete();
}
mTelemetry->mCallbacks.Clear();
}
NS_IMETHOD Run() {
LoadFailedLockCount(mTelemetry->mFailedLockCount);
mTelemetry->mLastShutdownTime =
ReadLastShutdownDuration(mShutdownTimeFilename);
mTelemetry->ReadLateWritesStacks(mProfileDir);
nsCOMPtr<nsIRunnable> e =
NS_NewRunnableMethod(this, &nsFetchTelemetryData::MainThread);
NS_ENSURE_STATE(e);
NS_DispatchToMainThread(e, NS_DISPATCH_NORMAL);
return NS_OK;
}
private:
nsresult
LoadFailedLockCount(uint32_t& failedLockCount)
{
failedLockCount = 0;
int64_t fileSize = 0;
nsresult rv = mFailedProfileLockFile->GetFileSize(&fileSize);
if (NS_FAILED(rv)) {
return rv;
}
NS_ENSURE_TRUE(fileSize <= kMaxFailedProfileLockFileSize,
NS_ERROR_UNEXPECTED);
nsCOMPtr<nsIInputStream> inStream;
rv = NS_NewLocalFileInputStream(getter_AddRefs(inStream),
mFailedProfileLockFile,
PR_RDONLY);
NS_ENSURE_SUCCESS(rv, rv);
NS_ENSURE_TRUE(GetFailedLockCount(inStream, fileSize, failedLockCount),
NS_ERROR_UNEXPECTED);
inStream->Close();
mFailedProfileLockFile->Remove(false);
return NS_OK;
}
};
static TimeStamp gRecordedShutdownStartTime;
static bool gAlreadyFreedShutdownTimeFileName = false;
static char *gRecordedShutdownTimeFileName = nullptr;
static char *
GetShutdownTimeFileName()
{
if (gAlreadyFreedShutdownTimeFileName) {
return nullptr;
}
if (!gRecordedShutdownTimeFileName) {
nsCOMPtr<nsIFile> mozFile;
NS_GetSpecialDirectory(NS_APP_USER_PROFILE_50_DIR, getter_AddRefs(mozFile));
if (!mozFile)
return nullptr;
mozFile->AppendNative(NS_LITERAL_CSTRING("Telemetry.ShutdownTime.txt"));
nsAutoCString nativePath;
nsresult rv = mozFile->GetNativePath(nativePath);
if (!NS_SUCCEEDED(rv))
return nullptr;
gRecordedShutdownTimeFileName = PL_strdup(nativePath.get());
}
return gRecordedShutdownTimeFileName;
}
NS_IMETHODIMP
TelemetryImpl::GetLastShutdownDuration(uint32_t *aResult)
{
// The user must call AsyncFetchTelemetryData first. We return zero instead of
// reporting a failure so that the rest of telemetry can uniformly handle
// the read not being available yet.
if (!mCachedTelemetryData) {
*aResult = 0;
return NS_OK;
}
*aResult = mLastShutdownTime;
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::GetFailedProfileLockCount(uint32_t* aResult)
{
// The user must call AsyncFetchTelemetryData first. We return zero instead of
// reporting a failure so that the rest of telemetry can uniformly handle
// the read not being available yet.
if (!mCachedTelemetryData) {
*aResult = 0;
return NS_OK;
}
*aResult = mFailedLockCount;
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::AsyncFetchTelemetryData(nsIFetchTelemetryDataCallback *aCallback)
{
// We have finished reading the data already, just call the callback.
if (mCachedTelemetryData) {
aCallback->Complete();
return NS_OK;
}
// We already have a read request running, just remember the callback.
if (mCallbacks.Count() != 0) {
mCallbacks.AppendObject(aCallback);
return NS_OK;
}
// We make this check so that GetShutdownTimeFileName() doesn't get
// called; calling that function without telemetry enabled violates
// assumptions that the write-the-shutdown-timestamp machinery makes.
if (!Telemetry::CanRecord()) {
mCachedTelemetryData = true;
aCallback->Complete();
return NS_OK;
}
// Send the read to a background thread provided by the stream transport
// service to avoid a read in the main thread.
nsCOMPtr<nsIEventTarget> targetThread =
do_GetService(NS_STREAMTRANSPORTSERVICE_CONTRACTID);
if (!targetThread) {
mCachedTelemetryData = true;
aCallback->Complete();
return NS_OK;
}
// We have to get the filename from the main thread.
const char *shutdownTimeFilename = GetShutdownTimeFileName();
if (!shutdownTimeFilename) {
mCachedTelemetryData = true;
aCallback->Complete();
return NS_OK;
}
nsCOMPtr<nsIFile> profileDir;
nsresult rv = NS_GetSpecialDirectory(NS_APP_USER_PROFILE_50_DIR,
getter_AddRefs(profileDir));
if (NS_FAILED(rv)) {
mCachedTelemetryData = true;
aCallback->Complete();
return NS_OK;
}
nsCOMPtr<nsIFile> failedProfileLockFile;
rv = GetFailedProfileLockFile(getter_AddRefs(failedProfileLockFile),
profileDir);
if (NS_FAILED(rv)) {
mCachedTelemetryData = true;
aCallback->Complete();
return NS_OK;
}
mCallbacks.AppendObject(aCallback);
nsCOMPtr<nsIRunnable> event = new nsFetchTelemetryData(shutdownTimeFilename,
failedProfileLockFile,
profileDir);
targetThread->Dispatch(event, NS_DISPATCH_NORMAL);
return NS_OK;
}
TelemetryImpl::TelemetryImpl():
mHistogramMap(Telemetry::HistogramCount),
mCanRecord(XRE_GetProcessType() == GeckoProcessType_Default),
mHashMutex("Telemetry::mHashMutex"),
mHangReportsMutex("Telemetry::mHangReportsMutex"),
mThreadHangStatsMutex("Telemetry::mThreadHangStatsMutex"),
mCachedTelemetryData(false),
mLastShutdownTime(0),
mFailedLockCount(0)
{
// A whitelist to prevent Telemetry reporting on Addon & Thunderbird DBs
const char *trackedDBs[] = {
"addons.sqlite", "content-prefs.sqlite", "cookies.sqlite",
"downloads.sqlite", "extensions.sqlite", "formhistory.sqlite",
"index.sqlite", "healthreport.sqlite", "netpredictions.sqlite",
"permissions.sqlite", "places.sqlite", "search.sqlite", "signons.sqlite",
"urlclassifier3.sqlite", "webappsstore.sqlite"
};
for (size_t i = 0; i < ArrayLength(trackedDBs); i++)
mTrackedDBs.PutEntry(nsDependentCString(trackedDBs[i]));
#ifdef DEBUG
// Mark immutable to prevent asserts on simultaneous access from multiple threads
mTrackedDBs.MarkImmutable();
#endif
}
TelemetryImpl::~TelemetryImpl() {
UnregisterWeakMemoryReporter(this);
}
void
TelemetryImpl::InitMemoryReporter() {
RegisterWeakMemoryReporter(this);
}
NS_IMETHODIMP
TelemetryImpl::NewHistogram(const nsACString &name, const nsACString &expiration, uint32_t min,
uint32_t max, uint32_t bucketCount, uint32_t histogramType,
JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
Histogram *h;
nsresult rv = HistogramGet(PromiseFlatCString(name).get(), PromiseFlatCString(expiration).get(),
min, max, bucketCount, histogramType, &h);
if (NS_FAILED(rv))
return rv;
h->ClearFlags(Histogram::kUmaTargetedHistogramFlag);
h->SetFlags(Histogram::kExtendedStatisticsFlag);
return WrapAndReturnHistogram(h, cx, ret);
}
bool
TelemetryImpl::ReflectSQL(const SlowSQLEntryType *entry,
const Stat *stat,
JSContext *cx,
JS::Handle<JSObject*> obj)
{
if (stat->hitCount == 0)
return true;
const nsACString &sql = entry->GetKey();
JS::Rooted<JSObject*> arrayObj(cx, JS_NewArrayObject(cx, 0));
if (!arrayObj) {
return false;
}
return (JS_SetElement(cx, arrayObj, 0, stat->hitCount)
&& JS_SetElement(cx, arrayObj, 1, stat->totalTime)
&& JS_DefineProperty(cx, obj,
sql.BeginReading(),
OBJECT_TO_JSVAL(arrayObj),
nullptr, nullptr, JSPROP_ENUMERATE));
}
bool
TelemetryImpl::ReflectMainThreadSQL(SlowSQLEntryType *entry, JSContext *cx,
JS::Handle<JSObject*> obj)
{
return ReflectSQL(entry, &entry->mData.mainThread, cx, obj);
}
bool
TelemetryImpl::ReflectOtherThreadsSQL(SlowSQLEntryType *entry, JSContext *cx,
JS::Handle<JSObject*> obj)
{
return ReflectSQL(entry, &entry->mData.otherThreads, cx, obj);
}
bool
TelemetryImpl::AddSQLInfo(JSContext *cx, JS::Handle<JSObject*> rootObj, bool mainThread,
bool privateSQL)
{
JS::Rooted<JSObject*> statsObj(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!statsObj)
return false;
AutoHashtable<SlowSQLEntryType> &sqlMap =
(privateSQL ? mPrivateSQL : mSanitizedSQL);
AutoHashtable<SlowSQLEntryType>::ReflectEntryFunc reflectFunction =
(mainThread ? ReflectMainThreadSQL : ReflectOtherThreadsSQL);
if (!sqlMap.ReflectIntoJS(reflectFunction, cx, statsObj)) {
return false;
}
return JS_DefineProperty(cx, rootObj,
mainThread ? "mainThread" : "otherThreads",
OBJECT_TO_JSVAL(statsObj),
nullptr, nullptr, JSPROP_ENUMERATE);
}
nsresult
TelemetryImpl::GetHistogramEnumId(const char *name, Telemetry::ID *id)
{
if (!sTelemetry) {
return NS_ERROR_FAILURE;
}
// Cache names
// Note the histogram names are statically allocated
TelemetryImpl::HistogramMapType *map = &sTelemetry->mHistogramMap;
if (!map->Count()) {
for (uint32_t i = 0; i < Telemetry::HistogramCount; i++) {
CharPtrEntryType *entry = map->PutEntry(gHistograms[i].id());
if (MOZ_UNLIKELY(!entry)) {
map->Clear();
return NS_ERROR_OUT_OF_MEMORY;
}
entry->mData = (Telemetry::ID) i;
}
}
CharPtrEntryType *entry = map->GetEntry(name);
if (!entry) {
return NS_ERROR_INVALID_ARG;
}
*id = entry->mData;
return NS_OK;
}
nsresult
TelemetryImpl::GetHistogramByName(const nsACString &name, Histogram **ret)
{
Telemetry::ID id;
nsresult rv = GetHistogramEnumId(PromiseFlatCString(name).get(), &id);
if (NS_FAILED(rv)) {
return rv;
}
rv = GetHistogramByEnumId(id, ret);
if (NS_FAILED(rv))
return rv;
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::HistogramFrom(const nsACString &name, const nsACString &existing_name,
JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
Telemetry::ID id;
nsresult rv = GetHistogramEnumId(PromiseFlatCString(existing_name).get(), &id);
if (NS_FAILED(rv)) {
return rv;
}
const TelemetryHistogram &p = gHistograms[id];
Histogram *existing;
rv = GetHistogramByEnumId(id, &existing);
if (NS_FAILED(rv)) {
return rv;
}
Histogram *clone;
rv = HistogramGet(PromiseFlatCString(name).get(), p.expiration(),
existing->declared_min(), existing->declared_max(),
existing->bucket_count(), p.histogramType, &clone);
if (NS_FAILED(rv))
return rv;
Histogram::SampleSet ss;
existing->SnapshotSample(&ss);
clone->AddSampleSet(ss);
return WrapAndReturnHistogram(clone, cx, ret);
}
void
TelemetryImpl::IdentifyCorruptHistograms(StatisticsRecorder::Histograms &hs)
{
for (HistogramIterator it = hs.begin(); it != hs.end(); ++it) {
Histogram *h = *it;
Telemetry::ID id;
nsresult rv = GetHistogramEnumId(h->histogram_name().c_str(), &id);
// This histogram isn't a static histogram, just ignore it.
if (NS_FAILED(rv)) {
continue;
}
if (gCorruptHistograms[id]) {
continue;
}
Histogram::SampleSet ss;
h->SnapshotSample(&ss);
Histogram::Inconsistencies check = h->FindCorruption(ss);
bool corrupt = (check != Histogram::NO_INCONSISTENCIES);
if (corrupt) {
Telemetry::ID corruptID = Telemetry::HistogramCount;
if (check & Histogram::RANGE_CHECKSUM_ERROR) {
corruptID = Telemetry::RANGE_CHECKSUM_ERRORS;
} else if (check & Histogram::BUCKET_ORDER_ERROR) {
corruptID = Telemetry::BUCKET_ORDER_ERRORS;
} else if (check & Histogram::COUNT_HIGH_ERROR) {
corruptID = Telemetry::TOTAL_COUNT_HIGH_ERRORS;
} else if (check & Histogram::COUNT_LOW_ERROR) {
corruptID = Telemetry::TOTAL_COUNT_LOW_ERRORS;
}
Telemetry::Accumulate(corruptID, 1);
}
gCorruptHistograms[id] = corrupt;
}
}
bool
TelemetryImpl::ShouldReflectHistogram(Histogram *h)
{
const char *name = h->histogram_name().c_str();
Telemetry::ID id;
nsresult rv = GetHistogramEnumId(name, &id);
if (NS_FAILED(rv)) {
// GetHistogramEnumId generally should not fail. But a lookup
// failure shouldn't prevent us from reflecting histograms into JS.
//
// However, these two histograms are created by Histogram itself for
// tracking corruption. We have our own histograms for that, so
// ignore these two.
if (strcmp(name, "Histogram.InconsistentCountHigh") == 0
|| strcmp(name, "Histogram.InconsistentCountLow") == 0) {
return false;
}
return true;
} else {
return !gCorruptHistograms[id];
}
}
// Compute the name to pass into Histogram for the addon histogram
// 'name' from the addon 'id'. We can't use 'name' directly because it
// might conflict with other histograms in other addons or even with our
// own.
void
AddonHistogramName(const nsACString &id, const nsACString &name,
nsACString &ret)
{
ret.Append(id);
ret.Append(NS_LITERAL_CSTRING(":"));
ret.Append(name);
}
NS_IMETHODIMP
TelemetryImpl::RegisterAddonHistogram(const nsACString &id,
const nsACString &name,
uint32_t min, uint32_t max,
uint32_t bucketCount,
uint32_t histogramType)
{
AddonEntryType *addonEntry = mAddonMap.GetEntry(id);
if (!addonEntry) {
addonEntry = mAddonMap.PutEntry(id);
if (MOZ_UNLIKELY(!addonEntry)) {
return NS_ERROR_OUT_OF_MEMORY;
}
addonEntry->mData = new AddonHistogramMapType();
}
AddonHistogramMapType *histogramMap = addonEntry->mData;
AddonHistogramEntryType *histogramEntry = histogramMap->GetEntry(name);
// Can't re-register the same histogram.
if (histogramEntry) {
return NS_ERROR_FAILURE;
}
histogramEntry = histogramMap->PutEntry(name);
if (MOZ_UNLIKELY(!histogramEntry)) {
return NS_ERROR_OUT_OF_MEMORY;
}
AddonHistogramInfo &info = histogramEntry->mData;
info.min = min;
info.max = max;
info.bucketCount = bucketCount;
info.histogramType = histogramType;
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::GetAddonHistogram(const nsACString &id, const nsACString &name,
JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
AddonEntryType *addonEntry = mAddonMap.GetEntry(id);
// The given id has not been registered.
if (!addonEntry) {
return NS_ERROR_INVALID_ARG;
}
AddonHistogramMapType *histogramMap = addonEntry->mData;
AddonHistogramEntryType *histogramEntry = histogramMap->GetEntry(name);
// The given histogram name has not been registered.
if (!histogramEntry) {
return NS_ERROR_INVALID_ARG;
}
AddonHistogramInfo &info = histogramEntry->mData;
if (!info.h) {
nsAutoCString actualName;
AddonHistogramName(id, name, actualName);
if (!CreateHistogramForAddon(actualName, info)) {
return NS_ERROR_FAILURE;
}
}
return WrapAndReturnHistogram(info.h, cx, ret);
}
NS_IMETHODIMP
TelemetryImpl::UnregisterAddonHistograms(const nsACString &id)
{
AddonEntryType *addonEntry = mAddonMap.GetEntry(id);
if (addonEntry) {
// Histogram's destructor is private, so this is the best we can do.
// The histograms the addon created *will* stick around, but they
// will be deleted if and when the addon registers histograms with
// the same names.
delete addonEntry->mData;
mAddonMap.RemoveEntry(id);
}
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::GetHistogramSnapshots(JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
JS::Rooted<JSObject*> root_obj(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!root_obj)
return NS_ERROR_FAILURE;
ret.setObject(*root_obj);
// Ensure that all the HISTOGRAM_FLAG histograms have been created, so
// that their values are snapshotted.
for (size_t i = 0; i < Telemetry::HistogramCount; ++i) {
if (gHistograms[i].histogramType == nsITelemetry::HISTOGRAM_FLAG) {
Histogram *h;
DebugOnly<nsresult> rv = GetHistogramByEnumId(Telemetry::ID(i), &h);
MOZ_ASSERT(NS_SUCCEEDED(rv));
}
};
StatisticsRecorder::Histograms hs;
StatisticsRecorder::GetHistograms(&hs);
// We identify corrupt histograms first, rather than interspersing it
// in the loop below, to ensure that our corruption statistics don't
// depend on histogram enumeration order.
//
// Of course, we hope that all of these corruption-statistics
// histograms are not themselves corrupt...
IdentifyCorruptHistograms(hs);
// OK, now we can actually reflect things.
JS::Rooted<JSObject*> hobj(cx);
for (HistogramIterator it = hs.begin(); it != hs.end(); ++it) {
Histogram *h = *it;
if (!ShouldReflectHistogram(h) || IsEmpty(h) || IsExpired(h)) {
continue;
}
hobj = JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr());
if (!hobj) {
return NS_ERROR_FAILURE;
}
switch (ReflectHistogramSnapshot(cx, hobj, h)) {
case REFLECT_CORRUPT:
// We can still hit this case even if ShouldReflectHistograms
// returns true. The histogram lies outside of our control
// somehow; just skip it.
continue;
case REFLECT_FAILURE:
return NS_ERROR_FAILURE;
case REFLECT_OK:
if (!JS_DefineProperty(cx, root_obj, h->histogram_name().c_str(),
OBJECT_TO_JSVAL(hobj), nullptr, nullptr, JSPROP_ENUMERATE)) {
return NS_ERROR_FAILURE;
}
}
}
return NS_OK;
}
bool
TelemetryImpl::CreateHistogramForAddon(const nsACString &name,
AddonHistogramInfo &info)
{
Histogram *h;
nsresult rv = HistogramGet(PromiseFlatCString(name).get(), "never",
info.min, info.max, info.bucketCount,
info.histogramType, &h);
if (NS_FAILED(rv)) {
return false;
}
// Don't let this histogram be reported via the normal means
// (e.g. Telemetry.registeredHistograms); we'll make it available in
// other ways.
h->ClearFlags(Histogram::kUmaTargetedHistogramFlag);
info.h = h;
return true;
}
bool
TelemetryImpl::AddonHistogramReflector(AddonHistogramEntryType *entry,
JSContext *cx, JS::Handle<JSObject*> obj)
{
AddonHistogramInfo &info = entry->mData;
// Never even accessed the histogram.
if (!info.h) {
// Have to force creation of HISTOGRAM_FLAG histograms.
if (info.histogramType != nsITelemetry::HISTOGRAM_FLAG)
return true;
if (!CreateHistogramForAddon(entry->GetKey(), info)) {
return false;
}
}
if (IsEmpty(info.h)) {
return true;
}
JS::Rooted<JSObject*> snapshot(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!snapshot) {
// Just consider this to be skippable.
return true;
}
switch (ReflectHistogramSnapshot(cx, snapshot, info.h)) {
case REFLECT_FAILURE:
case REFLECT_CORRUPT:
return false;
case REFLECT_OK:
const nsACString &histogramName = entry->GetKey();
if (!JS_DefineProperty(cx, obj,
PromiseFlatCString(histogramName).get(),
OBJECT_TO_JSVAL(snapshot), nullptr, nullptr,
JSPROP_ENUMERATE)) {
return false;
}
break;
}
return true;
}
bool
TelemetryImpl::AddonReflector(AddonEntryType *entry,
JSContext *cx, JS::Handle<JSObject*> obj)
{
const nsACString &addonId = entry->GetKey();
JS::Rooted<JSObject*> subobj(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!subobj) {
return false;
}
AddonHistogramMapType *map = entry->mData;
if (!(map->ReflectIntoJS(AddonHistogramReflector, cx, subobj)
&& JS_DefineProperty(cx, obj,
PromiseFlatCString(addonId).get(),
OBJECT_TO_JSVAL(subobj), nullptr, nullptr,
JSPROP_ENUMERATE))) {
return false;
}
return true;
}
NS_IMETHODIMP
TelemetryImpl::GetAddonHistogramSnapshots(JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
JS::Rooted<JSObject*> obj(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!obj) {
return NS_ERROR_FAILURE;
}
if (!mAddonMap.ReflectIntoJS(AddonReflector, cx, obj)) {
return NS_ERROR_FAILURE;
}
ret.setObject(*obj);
return NS_OK;
}
bool
TelemetryImpl::GetSQLStats(JSContext *cx, JS::MutableHandle<JS::Value> ret, bool includePrivateSql)
{
JS::Rooted<JSObject*> root_obj(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!root_obj)
return false;
ret.setObject(*root_obj);
MutexAutoLock hashMutex(mHashMutex);
// Add info about slow SQL queries on the main thread
if (!AddSQLInfo(cx, root_obj, true, includePrivateSql))
return false;
// Add info about slow SQL queries on other threads
if (!AddSQLInfo(cx, root_obj, false, includePrivateSql))
return false;
return true;
}
NS_IMETHODIMP
TelemetryImpl::GetSlowSQL(JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
if (GetSQLStats(cx, ret, false))
return NS_OK;
return NS_ERROR_FAILURE;
}
NS_IMETHODIMP
TelemetryImpl::GetDebugSlowSQL(JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
bool revealPrivateSql =
Preferences::GetBool("toolkit.telemetry.debugSlowSql", false);
if (GetSQLStats(cx, ret, revealPrivateSql))
return NS_OK;
return NS_ERROR_FAILURE;
}
NS_IMETHODIMP
TelemetryImpl::GetMaximalNumberOfConcurrentThreads(uint32_t *ret)
{
*ret = nsThreadManager::get()->GetHighestNumberOfThreads();
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::GetChromeHangs(JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
MutexAutoLock hangReportMutex(mHangReportsMutex);
const CombinedStacks& stacks = mHangReports.GetStacks();
JS::Rooted<JSObject*> fullReportObj(cx, CreateJSStackObject(cx, stacks));
if (!fullReportObj) {
return NS_ERROR_FAILURE;
}
ret.setObject(*fullReportObj);
JS::Rooted<JSObject*> durationArray(cx, JS_NewArrayObject(cx, 0));
JS::Rooted<JSObject*> systemUptimeArray(cx, JS_NewArrayObject(cx, 0));
JS::Rooted<JSObject*> firefoxUptimeArray(cx, JS_NewArrayObject(cx, 0));
if (!durationArray || !systemUptimeArray || !firefoxUptimeArray) {
return NS_ERROR_FAILURE;
}
bool ok = JS_DefineProperty(cx, fullReportObj, "durations",
OBJECT_TO_JSVAL(durationArray),
nullptr, nullptr, JSPROP_ENUMERATE);
if (!ok) {
return NS_ERROR_FAILURE;
}
ok = JS_DefineProperty(cx, fullReportObj, "systemUptime",
OBJECT_TO_JSVAL(systemUptimeArray),
nullptr, nullptr, JSPROP_ENUMERATE);
if (!ok) {
return NS_ERROR_FAILURE;
}
ok = JS_DefineProperty(cx, fullReportObj, "firefoxUptime",
OBJECT_TO_JSVAL(firefoxUptimeArray),
nullptr, nullptr, JSPROP_ENUMERATE);
if (!ok) {
return NS_ERROR_FAILURE;
}
const size_t length = stacks.GetStackCount();
for (size_t i = 0; i < length; ++i) {
if (!JS_SetElement(cx, durationArray, i, mHangReports.GetDuration(i))) {
return NS_ERROR_FAILURE;
}
if (!JS_SetElement(cx, systemUptimeArray, i, mHangReports.GetSystemUptime(i))) {
return NS_ERROR_FAILURE;
}
if (!JS_SetElement(cx, firefoxUptimeArray, i, mHangReports.GetFirefoxUptime(i))) {
return NS_ERROR_FAILURE;
}
}
return NS_OK;
}
static JSObject *
CreateJSStackObject(JSContext *cx, const CombinedStacks &stacks) {
JS::Rooted<JSObject*> ret(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!ret) {
return nullptr;
}
JS::Rooted<JSObject*> moduleArray(cx, JS_NewArrayObject(cx, 0));
if (!moduleArray) {
return nullptr;
}
bool ok = JS_DefineProperty(cx, ret, "memoryMap",
OBJECT_TO_JSVAL(moduleArray),
nullptr, nullptr, JSPROP_ENUMERATE);
if (!ok) {
return nullptr;
}
const size_t moduleCount = stacks.GetModuleCount();
for (size_t moduleIndex = 0; moduleIndex < moduleCount; ++moduleIndex) {
// Current module
const Telemetry::ProcessedStack::Module& module =
stacks.GetModule(moduleIndex);
JS::Rooted<JSObject*> moduleInfoArray(cx, JS_NewArrayObject(cx, 0));
if (!moduleInfoArray) {
return nullptr;
}
if (!JS_SetElement(cx, moduleArray, moduleIndex, moduleInfoArray)) {
return nullptr;
}
unsigned index = 0;
// Module name
JS::Rooted<JSString*> str(cx, JS_NewStringCopyZ(cx, module.mName.c_str()));
if (!str) {
return nullptr;
}
if (!JS_SetElement(cx, moduleInfoArray, index++, str)) {
return nullptr;
}
// Module breakpad identifier
JS::Rooted<JSString*> id(cx, JS_NewStringCopyZ(cx, module.mBreakpadId.c_str()));
if (!id) {
return nullptr;
}
if (!JS_SetElement(cx, moduleInfoArray, index++, id)) {
return nullptr;
}
}
JS::Rooted<JSObject*> reportArray(cx, JS_NewArrayObject(cx, 0));
if (!reportArray) {
return nullptr;
}
ok = JS_DefineProperty(cx, ret, "stacks",
OBJECT_TO_JSVAL(reportArray),
nullptr, nullptr, JSPROP_ENUMERATE);
if (!ok) {
return nullptr;
}
const size_t length = stacks.GetStackCount();
for (size_t i = 0; i < length; ++i) {
// Represent call stack PCs as (module index, offset) pairs.
JS::Rooted<JSObject*> pcArray(cx, JS_NewArrayObject(cx, 0));
if (!pcArray) {
return nullptr;
}
if (!JS_SetElement(cx, reportArray, i, pcArray)) {
return nullptr;
}
const CombinedStacks::Stack& stack = stacks.GetStack(i);
const uint32_t pcCount = stack.size();
for (size_t pcIndex = 0; pcIndex < pcCount; ++pcIndex) {
const Telemetry::ProcessedStack::Frame& frame = stack[pcIndex];
JS::Rooted<JSObject*> framePair(cx, JS_NewArrayObject(cx, 0));
if (!framePair) {
return nullptr;
}
int modIndex = (std::numeric_limits<uint16_t>::max() == frame.mModIndex) ?
-1 : frame.mModIndex;
if (!JS_SetElement(cx, framePair, 0, modIndex)) {
return nullptr;
}
if (!JS_SetElement(cx, framePair, 1, static_cast<double>(frame.mOffset))) {
return nullptr;
}
if (!JS_SetElement(cx, pcArray, pcIndex, framePair)) {
return nullptr;
}
}
}
return ret;
}
static bool
IsValidBreakpadId(const std::string &breakpadId) {
if (breakpadId.size() < 33) {
return false;
}
for (unsigned i = 0, n = breakpadId.size(); i < n; ++i) {
char c = breakpadId[i];
if ((c < '0' || c > '9') && (c < 'A' || c > 'F')) {
return false;
}
}
return true;
}
// Read a stack from the given file name. In case of any error, aStack is
// unchanged.
static void
ReadStack(const char *aFileName, Telemetry::ProcessedStack &aStack)
{
std::ifstream file(aFileName);
size_t numModules;
file >> numModules;
if (file.fail()) {
return;
}
char newline = file.get();
if (file.fail() || newline != '\n') {
return;
}
Telemetry::ProcessedStack stack;
for (size_t i = 0; i < numModules; ++i) {
std::string breakpadId;
file >> breakpadId;
if (file.fail() || !IsValidBreakpadId(breakpadId)) {
return;
}
char space = file.get();
if (file.fail() || space != ' ') {
return;
}
std::string moduleName;
getline(file, moduleName);
if (file.fail() || moduleName[0] == ' ') {
return;
}
Telemetry::ProcessedStack::Module module = {
moduleName,
breakpadId
};
stack.AddModule(module);
}
size_t numFrames;
file >> numFrames;
if (file.fail()) {
return;
}
newline = file.get();
if (file.fail() || newline != '\n') {
return;
}
for (size_t i = 0; i < numFrames; ++i) {
uint16_t index;
file >> index;
uintptr_t offset;
file >> std::hex >> offset >> std::dec;
if (file.fail()) {
return;
}
Telemetry::ProcessedStack::Frame frame = {
offset,
index
};
stack.AddFrame(frame);
}
aStack = stack;
}
static JSObject*
CreateJSTimeHistogram(JSContext* cx, const Telemetry::TimeHistogram& time)
{
/* Create JS representation of TimeHistogram,
in the format of Chromium-style histograms. */
JS::RootedObject ret(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!ret) {
return nullptr;
}
if (!JS_DefineProperty(cx, ret, "min",
UINT_TO_JSVAL(time.GetBucketMin(0)),
nullptr, nullptr, JSPROP_ENUMERATE) ||
!JS_DefineProperty(cx, ret, "max",
UINT_TO_JSVAL(time.GetBucketMax(
ArrayLength(time) - 1)),
nullptr, nullptr, JSPROP_ENUMERATE) ||
!JS_DefineProperty(cx, ret, "histogram_type",
INT_TO_JSVAL(nsITelemetry::HISTOGRAM_EXPONENTIAL),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return nullptr;
}
// TODO: calculate "sum", "log_sum", and "log_sum_squares"
if (!JS_DefineProperty(cx, ret, "sum", INT_TO_JSVAL(0),
nullptr, nullptr, JSPROP_ENUMERATE) ||
!JS_DefineProperty(cx, ret, "log_sum", DOUBLE_TO_JSVAL(0.0),
nullptr, nullptr, JSPROP_ENUMERATE) ||
!JS_DefineProperty(cx, ret, "log_sum_squares", DOUBLE_TO_JSVAL(0.0),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return nullptr;
}
JS::RootedObject ranges(
cx, JS_NewArrayObject(cx, ArrayLength(time) + 1));
JS::RootedObject counts(
cx, JS_NewArrayObject(cx, ArrayLength(time) + 1));
if (!ranges || !counts) {
return nullptr;
}
/* In a Chromium-style histogram, the first bucket is an "under" bucket
that represents all values below the histogram's range. */
if (!JS_SetElement(cx, ranges, 0, time.GetBucketMin(0)) ||
!JS_SetElement(cx, counts, 0, 0)) {
return nullptr;
}
for (size_t i = 0; i < ArrayLength(time); i++) {
if (!JS_SetElement(cx, ranges, i + 1, time.GetBucketMax(i)) ||
!JS_SetElement(cx, counts, i + 1, time[i])) {
return nullptr;
}
}
if (!JS_DefineProperty(cx, ret, "ranges", OBJECT_TO_JSVAL(ranges),
nullptr, nullptr, JSPROP_ENUMERATE) ||
!JS_DefineProperty(cx, ret, "counts", OBJECT_TO_JSVAL(counts),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return nullptr;
}
return ret;
}
static JSObject*
CreateJSHangHistogram(JSContext* cx, const Telemetry::HangHistogram& hang)
{
JS::RootedObject ret(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!ret) {
return nullptr;
}
const Telemetry::HangHistogram::Stack& hangStack = hang.GetStack();
JS::RootedObject stack(cx,
JS_NewArrayObject(cx, hangStack.length()));
if (!ret) {
return nullptr;
}
for (size_t i = 0; i < hangStack.length(); i++) {
JS::RootedString string(cx, JS_NewStringCopyZ(cx, hangStack[i]));
if (!JS_SetElement(cx, stack, i, string)) {
return nullptr;
}
}
JS::RootedObject time(cx, CreateJSTimeHistogram(cx, hang));
if (!time ||
!JS_DefineProperty(cx, ret, "stack", OBJECT_TO_JSVAL(stack),
nullptr, nullptr, JSPROP_ENUMERATE) ||
!JS_DefineProperty(cx, ret, "histogram", OBJECT_TO_JSVAL(time),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return nullptr;
}
return ret;
}
static JSObject*
CreateJSThreadHangStats(JSContext* cx, const Telemetry::ThreadHangStats& thread)
{
JS::RootedObject ret(cx, JS_NewObject(cx, nullptr, JS::NullPtr(), JS::NullPtr()));
if (!ret) {
return nullptr;
}
JS::RootedString name(cx, JS_NewStringCopyZ(cx, thread.GetName()));
if (!name ||
!JS_DefineProperty(cx, ret, "name", STRING_TO_JSVAL(name),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return nullptr;
}
JS::RootedObject activity(cx, CreateJSTimeHistogram(cx, thread.mActivity));
if (!activity ||
!JS_DefineProperty(cx, ret, "activity", OBJECT_TO_JSVAL(activity),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return nullptr;
}
JS::RootedObject hangs(cx, JS_NewArrayObject(cx, 0));
if (!hangs) {
return nullptr;
}
for (size_t i = 0; i < thread.mHangs.length(); i++) {
JS::RootedObject obj(cx, CreateJSHangHistogram(cx, thread.mHangs[i]));
if (!JS_SetElement(cx, hangs, i, obj)) {
return nullptr;
}
}
if (!JS_DefineProperty(cx, ret, "hangs", OBJECT_TO_JSVAL(hangs),
nullptr, nullptr, JSPROP_ENUMERATE)) {
return nullptr;
}
return ret;
}
NS_IMETHODIMP
TelemetryImpl::GetThreadHangStats(JSContext* cx, JS::MutableHandle<JS::Value> ret)
{
JS::RootedObject retObj(cx, JS_NewArrayObject(cx, 0));
if (!retObj) {
return NS_ERROR_FAILURE;
}
size_t threadIndex = 0;
#ifdef MOZ_ENABLE_BACKGROUND_HANG_MONITOR
/* First add active threads; we need to hold |iter| (and its lock)
throughout this method to avoid a race condition where a thread can
be recorded twice if the thread is destroyed while this method is
running */
BackgroundHangMonitor::ThreadHangStatsIterator iter;
for (Telemetry::ThreadHangStats* histogram = iter.GetNext();
histogram; histogram = iter.GetNext()) {
JS::RootedObject obj(cx,
CreateJSThreadHangStats(cx, *histogram));
if (!JS_SetElement(cx, retObj, threadIndex++, obj)) {
return NS_ERROR_FAILURE;
}
}
#endif
// Add saved threads next
MutexAutoLock autoLock(mThreadHangStatsMutex);
for (size_t i = 0; i < mThreadHangStats.length(); i++) {
JS::RootedObject obj(cx,
CreateJSThreadHangStats(cx, mThreadHangStats[i]));
if (!JS_SetElement(cx, retObj, threadIndex++, obj)) {
return NS_ERROR_FAILURE;
}
}
ret.setObject(*retObj);
return NS_OK;
}
void
TelemetryImpl::ReadLateWritesStacks(nsIFile* aProfileDir)
{
nsAutoCString nativePath;
nsresult rv = aProfileDir->GetNativePath(nativePath);
if (NS_FAILED(rv)) {
return;
}
const char *name = nativePath.get();
PRDir *dir = PR_OpenDir(name);
if (!dir) {
return;
}
PRDirEntry *ent;
const char *prefix = "Telemetry.LateWriteFinal-";
unsigned int prefixLen = strlen(prefix);
while ((ent = PR_ReadDir(dir, PR_SKIP_NONE))) {
if (strncmp(prefix, ent->name, prefixLen) != 0) {
continue;
}
nsAutoCString stackNativePath = nativePath;
stackNativePath += XPCOM_FILE_PATH_SEPARATOR;
stackNativePath += nsDependentCString(ent->name);
Telemetry::ProcessedStack stack;
ReadStack(stackNativePath.get(), stack);
if (stack.GetStackSize() != 0) {
mLateWritesStacks.AddStack(stack);
}
// Delete the file so that we don't report it again on the next run.
PR_Delete(stackNativePath.get());
}
PR_CloseDir(dir);
}
NS_IMETHODIMP
TelemetryImpl::GetLateWrites(JSContext *cx, JS::MutableHandle<JS::Value> ret)
{
// The user must call AsyncReadTelemetryData first. We return an empty list
// instead of reporting a failure so that the rest of telemetry can uniformly
// handle the read not being available yet.
// FIXME: we allocate the js object again and again in the getter. We should
// figure out a way to cache it. In order to do that we have to call
// JS_AddNamedObjectRoot. A natural place to do so is in the TelemetryImpl
// constructor, but it is not clear how to get a JSContext in there.
// Another option would be to call it in here when we first call
// CreateJSStackObject, but we would still need to figure out where to call
// JS_RemoveObjectRoot. Would it be ok to never call JS_RemoveObjectRoot
// and just set the pointer to nullptr is the telemetry destructor?
JSObject *report;
if (!mCachedTelemetryData) {
CombinedStacks empty;
report = CreateJSStackObject(cx, empty);
} else {
report = CreateJSStackObject(cx, mLateWritesStacks);
}
if (report == nullptr) {
return NS_ERROR_FAILURE;
}
ret.setObject(*report);
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::RegisteredHistograms(uint32_t *aCount, char*** aHistograms)
{
size_t count = ArrayLength(gHistograms);
size_t offset = 0;
char** histograms = static_cast<char**>(nsMemory::Alloc(count * sizeof(char*)));
for (size_t i = 0; i < count; ++i) {
if (IsExpired(gHistograms[i].expiration())) {
offset++;
continue;
}
const char* h = gHistograms[i].id();
size_t len = strlen(h);
histograms[i - offset] = static_cast<char*>(nsMemory::Clone(h, len+1));
}
*aCount = count - offset;
*aHistograms = histograms;
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::GetHistogramById(const nsACString &name, JSContext *cx,
JS::MutableHandle<JS::Value> ret)
{
Histogram *h;
nsresult rv = GetHistogramByName(name, &h);
if (NS_FAILED(rv))
return rv;
return WrapAndReturnHistogram(h, cx, ret);
}
NS_IMETHODIMP
TelemetryImpl::GetCanRecord(bool *ret) {
*ret = mCanRecord;
return NS_OK;
}
NS_IMETHODIMP
TelemetryImpl::SetCanRecord(bool canRecord) {
mCanRecord = !!canRecord;
return NS_OK;
}
bool
TelemetryImpl::CanRecord() {
return !sTelemetry || sTelemetry->mCanRecord;
}
NS_IMETHODIMP
TelemetryImpl::GetCanSend(bool *ret) {
#if defined(MOZILLA_OFFICIAL) && defined(MOZ_TELEMETRY_REPORTING)
*ret = true;
#else
*ret = false;
#endif
return NS_OK;
}
already_AddRefed<nsITelemetry>
TelemetryImpl::CreateTelemetryInstance()
{
NS_ABORT_IF_FALSE(sTelemetry == nullptr, "CreateTelemetryInstance may only be called once, via GetService()");
sTelemetry = new TelemetryImpl();
// AddRef for the local reference
NS_ADDREF(sTelemetry);
// AddRef for the caller
nsCOMPtr<nsITelemetry> ret = sTelemetry;
sTelemetry->InitMemoryReporter();
return ret.forget();
}
void
TelemetryImpl::ShutdownTelemetry()
{
// No point in collecting IO beyond this point
ClearIOReporting();
NS_IF_RELEASE(sTelemetry);
}
void
TelemetryImpl::StoreSlowSQL(const nsACString &sql, uint32_t delay,
SanitizedState state)
{
AutoHashtable<SlowSQLEntryType> *slowSQLMap = nullptr;
if (state == Sanitized)
slowSQLMap = &(sTelemetry->mSanitizedSQL);
else
slowSQLMap = &(sTelemetry->mPrivateSQL);
MutexAutoLock hashMutex(sTelemetry->mHashMutex);
SlowSQLEntryType *entry = slowSQLMap->GetEntry(sql);
if (!entry) {
entry = slowSQLMap->PutEntry(sql);
if (MOZ_UNLIKELY(!entry))
return;
entry->mData.mainThread.hitCount = 0;
entry->mData.mainThread.totalTime = 0;
entry->mData.otherThreads.hitCount = 0;
entry->mData.otherThreads.totalTime = 0;
}
if (NS_IsMainThread()) {
entry->mData.mainThread.hitCount++;
entry->mData.mainThread.totalTime += delay;
} else {
entry->mData.otherThreads.hitCount++;
entry->mData.otherThreads.totalTime += delay;
}
}
/**
* This method replaces string literals in SQL strings with the word :private
*
* States used in this state machine:
*
* NORMAL:
* - This is the active state when not iterating over a string literal or
* comment
*
* SINGLE_QUOTE:
* - Defined here: http://www.sqlite.org/lang_expr.html
* - This state represents iterating over a string literal opened with
* a single quote.
* - A single quote within the string can be encoded by putting 2 single quotes
* in a row, e.g. 'This literal contains an escaped quote '''
* - Any double quotes found within a single-quoted literal are ignored
* - This state covers BLOB literals, e.g. X'ABC123'
* - The string literal and the enclosing quotes will be replaced with
* the text :private
*
* DOUBLE_QUOTE:
* - Same rules as the SINGLE_QUOTE state.
* - According to http://www.sqlite.org/lang_keywords.html,
* SQLite interprets text in double quotes as an identifier unless it's used in
* a context where it cannot be resolved to an identifier and a string literal
* is allowed. This method removes text in double-quotes for safety.
*
* DASH_COMMENT:
* - http://www.sqlite.org/lang_comment.html
* - A dash comment starts with two dashes in a row,
* e.g. DROP TABLE foo -- a comment
* - Any text following two dashes in a row is interpreted as a comment until
* end of input or a newline character
* - Any quotes found within the comment are ignored and no replacements made
*
* C_STYLE_COMMENT:
* - http://www.sqlite.org/lang_comment.html
* - A C-style comment starts with a forward slash and an asterisk, and ends
* with an asterisk and a forward slash
* - Any text following comment start is interpreted as a comment up to end of
* input or comment end
* - Any quotes found within the comment are ignored and no replacements made
*/
nsCString
TelemetryImpl::SanitizeSQL(const nsACString &sql) {
nsCString output;
int length = sql.Length();
typedef enum {
NORMAL,
SINGLE_QUOTE,
DOUBLE_QUOTE,
DASH_COMMENT,
C_STYLE_COMMENT,
} State;
State state = NORMAL;
int fragmentStart = 0;
for (int i = 0; i < length; i++) {
char character = sql[i];
char nextCharacter = (i + 1 < length) ? sql[i + 1] : '\0';
switch (character) {
case '\'':
case '"':
if (state == NORMAL) {
state = (character == '\'') ? SINGLE_QUOTE : DOUBLE_QUOTE;
output += nsDependentCSubstring(sql, fragmentStart, i - fragmentStart);
output += ":private";
fragmentStart = -1;
} else if ((state == SINGLE_QUOTE && character == '\'') ||
(state == DOUBLE_QUOTE && character == '"')) {
if (nextCharacter == character) {
// Two consecutive quotes within a string literal are a single escaped quote
i++;
} else {
state = NORMAL;
fragmentStart = i + 1;
}
}
break;
case '-':
if (state == NORMAL) {
if (nextCharacter == '-') {
state = DASH_COMMENT;
i++;
}
}
break;
case '\n':
if (state == DASH_COMMENT) {
state = NORMAL;
}
break;
case '/':
if (state == NORMAL) {
if (nextCharacter == '*') {
state = C_STYLE_COMMENT;
i++;
}
}
break;
case '*':
if (state == C_STYLE_COMMENT) {
if (nextCharacter == '/') {
state = NORMAL;
}
}
break;
default:
continue;
}
}
if ((fragmentStart >= 0) && fragmentStart < length)
output += nsDependentCSubstring(sql, fragmentStart, length - fragmentStart);
return output;
}
// Slow SQL statements will be automatically
// trimmed to kMaxSlowStatementLength characters.
// This limit doesn't include the ellipsis and DB name,
// that are appended at the end of the stored statement.
const uint32_t kMaxSlowStatementLength = 1000;
void
TelemetryImpl::RecordSlowStatement(const nsACString &sql,
const nsACString &dbName,
uint32_t delay)
{
if (!sTelemetry || !sTelemetry->mCanRecord)
return;
bool isFirefoxDB = sTelemetry->mTrackedDBs.Contains(dbName);
if (isFirefoxDB) {
nsAutoCString sanitizedSQL(SanitizeSQL(sql));
if (sanitizedSQL.Length() > kMaxSlowStatementLength) {
sanitizedSQL.SetLength(kMaxSlowStatementLength);
sanitizedSQL += "...";
}
sanitizedSQL.AppendPrintf(" /* %s */", nsPromiseFlatCString(dbName).get());
StoreSlowSQL(sanitizedSQL, delay, Sanitized);
} else {
// Report aggregate DB-level statistics for addon DBs
nsAutoCString aggregate;
aggregate.AppendPrintf("Untracked SQL for %s",
nsPromiseFlatCString(dbName).get());
StoreSlowSQL(aggregate, delay, Sanitized);
}
nsAutoCString fullSQL;
fullSQL.AppendPrintf("%s /* %s */",
nsPromiseFlatCString(sql).get(),
nsPromiseFlatCString(dbName).get());
StoreSlowSQL(fullSQL, delay, Unsanitized);
}
#if defined(MOZ_ENABLE_PROFILER_SPS)
void
TelemetryImpl::RecordChromeHang(uint32_t aDuration,
Telemetry::ProcessedStack &aStack,
int32_t aSystemUptime,
int32_t aFirefoxUptime)
{
if (!sTelemetry || !sTelemetry->mCanRecord)
return;
MutexAutoLock hangReportMutex(sTelemetry->mHangReportsMutex);
sTelemetry->mHangReports.AddHang(aStack, aDuration,
aSystemUptime, aFirefoxUptime);
}
#endif
void
TelemetryImpl::RecordThreadHangStats(Telemetry::ThreadHangStats& aStats)
{
if (!sTelemetry || !sTelemetry->mCanRecord)
return;
MutexAutoLock autoLock(sTelemetry->mThreadHangStatsMutex);
sTelemetry->mThreadHangStats.append(Move(aStats));
}
NS_IMPL_ISUPPORTS2(TelemetryImpl, nsITelemetry, nsIMemoryReporter)
NS_GENERIC_FACTORY_SINGLETON_CONSTRUCTOR(nsITelemetry, TelemetryImpl::CreateTelemetryInstance)
#define NS_TELEMETRY_CID \
{0xaea477f2, 0xb3a2, 0x469c, {0xaa, 0x29, 0x0a, 0x82, 0xd1, 0x32, 0xb8, 0x29}}
NS_DEFINE_NAMED_CID(NS_TELEMETRY_CID);
const Module::CIDEntry kTelemetryCIDs[] = {
{ &kNS_TELEMETRY_CID, false, nullptr, nsITelemetryConstructor },
{ nullptr }
};
const Module::ContractIDEntry kTelemetryContracts[] = {
{ "@mozilla.org/base/telemetry;1", &kNS_TELEMETRY_CID },
{ nullptr }
};
const Module kTelemetryModule = {
Module::kVersion,
kTelemetryCIDs,
kTelemetryContracts,
nullptr,
nullptr,
nullptr,
TelemetryImpl::ShutdownTelemetry
};
NS_IMETHODIMP
TelemetryImpl::GetFileIOReports(JSContext *cx, JS::MutableHandleValue ret)
{
if (sTelemetryIOObserver) {
JS::Rooted<JSObject*> obj(cx, JS_NewObject(cx, nullptr, JS::NullPtr(),
JS::NullPtr()));
if (!obj) {
return NS_ERROR_FAILURE;
}
if (!sTelemetryIOObserver->ReflectIntoJS(cx, obj)) {
return NS_ERROR_FAILURE;
}
ret.setObject(*obj);
return NS_OK;
}
ret.setNull();
return NS_OK;
}
size_t
TelemetryImpl::SizeOfIncludingThis(mozilla::MallocSizeOf aMallocSizeOf)
{
size_t n = aMallocSizeOf(this);
// Ignore the hashtables in mAddonMap; they are not significant.
n += mAddonMap.SizeOfExcludingThis(nullptr, aMallocSizeOf);
n += mHistogramMap.SizeOfExcludingThis(nullptr, aMallocSizeOf);
n += mPrivateSQL.SizeOfExcludingThis(nullptr, aMallocSizeOf);
n += mSanitizedSQL.SizeOfExcludingThis(nullptr, aMallocSizeOf);
n += mTrackedDBs.SizeOfExcludingThis(nullptr, aMallocSizeOf);
n += mHangReports.SizeOfExcludingThis();
n += mThreadHangStats.sizeOfExcludingThis(aMallocSizeOf);
// It's a bit gross that we measure this other stuff that lives outside of
// TelemetryImpl... oh well.
if (sTelemetryIOObserver) {
n += sTelemetryIOObserver->SizeOfIncludingThis(aMallocSizeOf);
}
StatisticsRecorder::Histograms hs;
StatisticsRecorder::GetHistograms(&hs);
for (HistogramIterator it = hs.begin(); it != hs.end(); ++it) {
Histogram *h = *it;
n += h->SizeOfIncludingThis(aMallocSizeOf);
}
return n;
}
} // anonymous namespace
namespace mozilla {
void
RecordShutdownStartTimeStamp() {
#ifdef DEBUG
// FIXME: this function should only be called once, since it should be called
// at the earliest point we *know* we are shutting down. Unfortunately
// this assert has been firing. Given that if we are called multiple times
// we just keep the last timestamp, the assert is commented for now.
static bool recorded = false;
// MOZ_ASSERT(!recorded);
(void)recorded; // Silence unused-var warnings (remove when assert re-enabled)
recorded = true;
#endif
if (!Telemetry::CanRecord())
return;
gRecordedShutdownStartTime = TimeStamp::Now();
GetShutdownTimeFileName();
}
void
RecordShutdownEndTimeStamp() {
if (!gRecordedShutdownTimeFileName || gAlreadyFreedShutdownTimeFileName)
return;
nsCString name(gRecordedShutdownTimeFileName);
PL_strfree(gRecordedShutdownTimeFileName);
gRecordedShutdownTimeFileName = nullptr;
gAlreadyFreedShutdownTimeFileName = true;
nsCString tmpName = name;
tmpName += ".tmp";
FILE *f = fopen(tmpName.get(), "w");
if (!f)
return;
// On a normal release build this should be called just before
// calling _exit, but on a debug build or when the user forces a full
// shutdown this is called as late as possible, so we have to
// white list this write as write poisoning will be enabled.
MozillaRegisterDebugFILE(f);
TimeStamp now = TimeStamp::Now();
MOZ_ASSERT(now >= gRecordedShutdownStartTime);
TimeDuration diff = now - gRecordedShutdownStartTime;
uint32_t diff2 = diff.ToMilliseconds();
int written = fprintf(f, "%d\n", diff2);
MozillaUnRegisterDebugFILE(f);
int rv = fclose(f);
if (written < 0 || rv != 0) {
PR_Delete(tmpName.get());
return;
}
PR_Delete(name.get());
PR_Rename(tmpName.get(), name.get());
}
namespace Telemetry {
void
Accumulate(ID aHistogram, uint32_t aSample)
{
if (!TelemetryImpl::CanRecord()) {
return;
}
Histogram *h;
nsresult rv = GetHistogramByEnumId(aHistogram, &h);
if (NS_SUCCEEDED(rv))
h->Add(aSample);
}
void
Accumulate(const char* name, uint32_t sample)
{
if (!TelemetryImpl::CanRecord()) {
return;
}
ID id;
nsresult rv = TelemetryImpl::GetHistogramEnumId(name, &id);
if (NS_FAILED(rv)) {
return;
}
Histogram *h;
rv = GetHistogramByEnumId(id, &h);
if (NS_SUCCEEDED(rv)) {
h->Add(sample);
}
}
void
AccumulateTimeDelta(ID aHistogram, TimeStamp start, TimeStamp end)
{
Accumulate(aHistogram,
static_cast<uint32_t>((end - start).ToMilliseconds()));
}
bool
CanRecord()
{
return TelemetryImpl::CanRecord();
}
base::Histogram*
GetHistogramById(ID id)
{
Histogram *h = nullptr;
GetHistogramByEnumId(id, &h);
return h;
}
void
RecordSlowSQLStatement(const nsACString &statement,
const nsACString &dbName,
uint32_t delay)
{
TelemetryImpl::RecordSlowStatement(statement, dbName, delay);
}
void Init()
{
// Make the service manager hold a long-lived reference to the service
nsCOMPtr<nsITelemetry> telemetryService =
do_GetService("@mozilla.org/base/telemetry;1");
MOZ_ASSERT(telemetryService);
}
#if defined(MOZ_ENABLE_PROFILER_SPS)
void RecordChromeHang(uint32_t duration,
ProcessedStack &aStack,
int32_t aSystemUptime,
int32_t aFirefoxUptime)
{
TelemetryImpl::RecordChromeHang(duration, aStack,
aSystemUptime, aFirefoxUptime);
}
#endif
void RecordThreadHangStats(ThreadHangStats& aStats)
{
TelemetryImpl::RecordThreadHangStats(aStats);
}
ProcessedStack::ProcessedStack()
{
}
size_t ProcessedStack::GetStackSize() const
{
return mStack.size();
}
const ProcessedStack::Frame &ProcessedStack::GetFrame(unsigned aIndex) const
{
MOZ_ASSERT(aIndex < mStack.size());
return mStack[aIndex];
}
void ProcessedStack::AddFrame(const Frame &aFrame)
{
mStack.push_back(aFrame);
}
size_t ProcessedStack::GetNumModules() const
{
return mModules.size();
}
const ProcessedStack::Module &ProcessedStack::GetModule(unsigned aIndex) const
{
MOZ_ASSERT(aIndex < mModules.size());
return mModules[aIndex];
}
void ProcessedStack::AddModule(const Module &aModule)
{
mModules.push_back(aModule);
}
void ProcessedStack::Clear() {
mModules.clear();
mStack.clear();
}
bool ProcessedStack::Module::operator==(const Module& aOther) const {
return mName == aOther.mName &&
mBreakpadId == aOther.mBreakpadId;
}
struct StackFrame
{
uintptr_t mPC; // The program counter at this position in the call stack.
uint16_t mIndex; // The number of this frame in the call stack.
uint16_t mModIndex; // The index of module that has this program counter.
};
#ifdef MOZ_ENABLE_PROFILER_SPS
static bool CompareByPC(const StackFrame &a, const StackFrame &b)
{
return a.mPC < b.mPC;
}
static bool CompareByIndex(const StackFrame &a, const StackFrame &b)
{
return a.mIndex < b.mIndex;
}
#endif
ProcessedStack
GetStackAndModules(const std::vector<uintptr_t>& aPCs)
{
std::vector<StackFrame> rawStack;
for (std::vector<uintptr_t>::const_iterator i = aPCs.begin(),
e = aPCs.end(); i != e; ++i) {
uintptr_t aPC = *i;
StackFrame Frame = {aPC, static_cast<uint16_t>(rawStack.size()),
std::numeric_limits<uint16_t>::max()};
rawStack.push_back(Frame);
}
#ifdef MOZ_ENABLE_PROFILER_SPS
// Remove all modules not referenced by a PC on the stack
std::sort(rawStack.begin(), rawStack.end(), CompareByPC);
size_t moduleIndex = 0;
size_t stackIndex = 0;
size_t stackSize = rawStack.size();
SharedLibraryInfo rawModules = SharedLibraryInfo::GetInfoForSelf();
rawModules.SortByAddress();
while (moduleIndex < rawModules.GetSize()) {
const SharedLibrary& module = rawModules.GetEntry(moduleIndex);
uintptr_t moduleStart = module.GetStart();
uintptr_t moduleEnd = module.GetEnd() - 1;
// the interval is [moduleStart, moduleEnd)
bool moduleReferenced = false;
for (;stackIndex < stackSize; ++stackIndex) {
uintptr_t pc = rawStack[stackIndex].mPC;
if (pc >= moduleEnd)
break;
if (pc >= moduleStart) {
// If the current PC is within the current module, mark
// module as used
moduleReferenced = true;
rawStack[stackIndex].mPC -= moduleStart;
rawStack[stackIndex].mModIndex = moduleIndex;
} else {
// PC does not belong to any module. It is probably from
// the JIT. Use a fixed mPC so that we don't get different
// stacks on different runs.
rawStack[stackIndex].mPC =
std::numeric_limits<uintptr_t>::max();
}
}
if (moduleReferenced) {
++moduleIndex;
} else {
// Remove module if no PCs within its address range
rawModules.RemoveEntries(moduleIndex, moduleIndex + 1);
}
}
for (;stackIndex < stackSize; ++stackIndex) {
// These PCs are past the last module.
rawStack[stackIndex].mPC = std::numeric_limits<uintptr_t>::max();
}
std::sort(rawStack.begin(), rawStack.end(), CompareByIndex);
#endif
// Copy the information to the return value.
ProcessedStack Ret;
for (std::vector<StackFrame>::iterator i = rawStack.begin(),
e = rawStack.end(); i != e; ++i) {
const StackFrame &rawFrame = *i;
ProcessedStack::Frame frame = { rawFrame.mPC, rawFrame.mModIndex };
Ret.AddFrame(frame);
}
#ifdef MOZ_ENABLE_PROFILER_SPS
for (unsigned i = 0, n = rawModules.GetSize(); i != n; ++i) {
const SharedLibrary &info = rawModules.GetEntry(i);
const std::string &name = info.GetName();
std::string basename = name;
#ifdef XP_MACOSX
// FIXME: We want to use just the basename as the libname, but the
// current profiler addon needs the full path name, so we compute the
// basename in here.
size_t pos = name.rfind('/');
if (pos != std::string::npos) {
basename = name.substr(pos + 1);
}
#endif
ProcessedStack::Module module = {
basename,
info.GetBreakpadId()
};
Ret.AddModule(module);
}
#endif
return Ret;
}
void
WriteFailedProfileLock(nsIFile* aProfileDir)
{
nsCOMPtr<nsIFile> file;
nsresult rv = GetFailedProfileLockFile(getter_AddRefs(file), aProfileDir);
NS_ENSURE_SUCCESS_VOID(rv);
int64_t fileSize = 0;
rv = file->GetFileSize(&fileSize);
// It's expected that the file might not exist yet
if (NS_FAILED(rv) && rv != NS_ERROR_FILE_NOT_FOUND) {
return;
}
nsCOMPtr<nsIFileStream> fileStream;
rv = NS_NewLocalFileStream(getter_AddRefs(fileStream), file,
PR_RDWR | PR_CREATE_FILE, 0640);
NS_ENSURE_SUCCESS_VOID(rv);
NS_ENSURE_TRUE_VOID(fileSize <= kMaxFailedProfileLockFileSize);
unsigned int failedLockCount = 0;
if (fileSize > 0) {
nsCOMPtr<nsIInputStream> inStream = do_QueryInterface(fileStream);
NS_ENSURE_TRUE_VOID(inStream);
if (!GetFailedLockCount(inStream, fileSize, failedLockCount)) {
failedLockCount = 0;
}
}
++failedLockCount;
nsAutoCString bufStr;
bufStr.AppendInt(static_cast<int>(failedLockCount));
nsCOMPtr<nsISeekableStream> seekStream = do_QueryInterface(fileStream);
NS_ENSURE_TRUE_VOID(seekStream);
// If we read in an existing failed lock count, we need to reset the file ptr
if (fileSize > 0) {
rv = seekStream->Seek(nsISeekableStream::NS_SEEK_SET, 0);
NS_ENSURE_SUCCESS_VOID(rv);
}
nsCOMPtr<nsIOutputStream> outStream = do_QueryInterface(fileStream);
uint32_t bytesLeft = bufStr.Length();
const char* bytes = bufStr.get();
do {
uint32_t written = 0;
rv = outStream->Write(bytes, bytesLeft, &written);
if (NS_FAILED(rv)) {
break;
}
bytes += written;
bytesLeft -= written;
} while (bytesLeft > 0);
seekStream->SetEOF();
}
void
InitIOReporting(nsIFile* aXreDir)
{
// Never initialize twice
if (sTelemetryIOObserver) {
return;
}
// Initialize IO interposing
IOInterposer::Init();
InitPoisonIOInterposer();
sTelemetryIOObserver = new TelemetryIOInterposeObserver(aXreDir);
IOInterposer::Register(IOInterposeObserver::OpAll, sTelemetryIOObserver);
}
void
SetProfileDir(nsIFile* aProfD)
{
if (!sTelemetryIOObserver || !aProfD) {
return;
}
nsAutoString profDirPath;
nsresult rv = aProfD->GetPath(profDirPath);
if (NS_FAILED(rv)) {
return;
}
sTelemetryIOObserver->AddPath(profDirPath, NS_LITERAL_STRING("{profile}"));
}
void
TimeHistogram::Add(PRIntervalTime aTime)
{
uint32_t timeMs = PR_IntervalToMilliseconds(aTime);
size_t index = mozilla::FloorLog2(timeMs);
operator[](index)++;
}
uint32_t
HangHistogram::GetHash(const Stack& aStack)
{
uint32_t hash = 0;
for (const char* const* label = aStack.begin();
label != aStack.end(); label++) {
/* We only need to hash the pointer instead of the text content
because we are assuming constant pointers */
hash = AddToHash(hash, *label);
}
return hash;
}
bool
HangHistogram::operator==(const HangHistogram& aOther) const
{
if (mHash != aOther.mHash) {
return false;
}
if (mStack.length() != aOther.mStack.length()) {
return false;
}
return PodEqual(mStack.begin(), aOther.mStack.begin(), mStack.length());
}
} // namespace Telemetry
} // namespace mozilla
NSMODULE_DEFN(nsTelemetryModule) = &kTelemetryModule;
/**
* The XRE_TelemetryAdd function is to be used by embedding applications
* that can't use mozilla::Telemetry::Accumulate() directly.
*/
void
XRE_TelemetryAccumulate(int aID, uint32_t aSample)
{
mozilla::Telemetry::Accumulate((mozilla::Telemetry::ID) aID, aSample);
}