v8/test/cctest/libplatform/test-tracing.cc

512 lines
19 KiB
C++
Raw Normal View History

// Copyright 2016 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include <limits>
#include "include/libplatform/v8-tracing.h"
[tracing] Fix races in TracingController implementation The default TracingController (used by d8 and Node) has some concurrency issues. The new test flushes these out, when a second thread logs trace events while the main thread calls StopTracing(). - Use an acquire load in UpdateCategoryGroupEnabledFlags() because this was racing with GetCategoryGroupEnabled() where a new category is added in the slow path. g_category_groups is append-only, but reads/writes to g_category_index need to be correctly ordered so that new categories are added and only then is the change to the index visible. The relaxed load ignored this and caused unsynchronized read/write. - Use a relaxed load in ~ScopedTracer() to access category_group_enabled as this previously used a non-atomic operation which caused a race with UpdateCategoryGroupEnabledFlag() which does a relaxed store. - Replace TracingController::mode_ with an atomic bool as read/writes to mode_ were not synchronized and caused TSAN errors. It only has two states and it doesn't seem like we will extend this so just convert it to bool. - Take the lock around calling trace_object->Initialize in AddTraceEvent(), and around trace_buffer_->Flush() in StopTracing(). These two raced previously as the underlying TraceBufferRingBuffer passes out pointers to TraceObjects in a synchronized way, but the caller (AddTraceEvent) then writes into the object without synchronization. This leads to races when Flush() is called, at which time TraceBufferRingBuffer assumes that all the pointers it handed out are to valid, initialized TraceObjects - which is not true because AddTraceEvent may still be calling Initialize on them. This could be the cause of issues in Node.js where the last line of tracing/logging sometimes gets cut off. This is kind of a band-aid solution - access to the TraceObjects handed out by the ring buffer really needs proper synchronization which at this point would require redesign. It's quite likely we will replace this with Perfetto in the near future so not much point investing in this code right now. - Enable TracingCpuProfiler test which was flaky due to these bugs. Bug: v8:8821 Change-Id: I141296800c6906ac0e7f3f21dd16d861b07dae62 Reviewed-on: https://chromium-review.googlesource.com/c/1477283 Commit-Queue: Peter Marshall <petermarshall@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Ali Ijaz Sheikh <ofrobots@google.com> Cr-Commit-Position: refs/heads/master@{#59752}
2019-02-20 09:02:06 +00:00
#include "src/base/platform/platform.h"
#include "src/libplatform/default-platform.h"
#include "src/tracing/trace-event.h"
#include "test/cctest/cctest.h"
namespace v8 {
namespace platform {
namespace tracing {
TEST(TestTraceConfig) {
LocalContext env;
TraceConfig* trace_config = new TraceConfig();
trace_config->AddIncludedCategory("v8");
trace_config->AddIncludedCategory(TRACE_DISABLED_BY_DEFAULT("v8.runtime"));
CHECK_EQ(trace_config->IsSystraceEnabled(), false);
CHECK_EQ(trace_config->IsArgumentFilterEnabled(), false);
CHECK_EQ(trace_config->IsCategoryGroupEnabled("v8"), true);
CHECK_EQ(trace_config->IsCategoryGroupEnabled("v8.cpu_profile"), false);
CHECK_EQ(trace_config->IsCategoryGroupEnabled("v8.cpu_profile.hires"), false);
CHECK_EQ(trace_config->IsCategoryGroupEnabled(
TRACE_DISABLED_BY_DEFAULT("v8.runtime")),
true);
CHECK_EQ(trace_config->IsCategoryGroupEnabled("v8,v8.cpu_profile"), true);
CHECK_EQ(
trace_config->IsCategoryGroupEnabled("v8,disabled-by-default-v8.runtime"),
true);
CHECK_EQ(trace_config->IsCategoryGroupEnabled(
"v8_cpu_profile,v8.cpu_profile.hires"),
false);
delete trace_config;
}
TEST(TestTraceObject) {
TraceObject trace_object;
uint8_t category_enabled_flag = 41;
trace_object.Initialize('X', &category_enabled_flag, "Test.Trace",
"Test.Scope", 42, 123, 0, nullptr, nullptr, nullptr,
nullptr, 0, 1729, 4104);
CHECK_EQ('X', trace_object.phase());
CHECK_EQ(category_enabled_flag, *trace_object.category_enabled_flag());
CHECK_EQ(std::string("Test.Trace"), std::string(trace_object.name()));
CHECK_EQ(std::string("Test.Scope"), std::string(trace_object.scope()));
CHECK_EQ(0u, trace_object.duration());
CHECK_EQ(0u, trace_object.cpu_duration());
}
class ConvertableToTraceFormatMock : public v8::ConvertableToTraceFormat {
public:
explicit ConvertableToTraceFormatMock(int value) : value_(value) {}
void AppendAsTraceFormat(std::string* out) const override {
*out += "[" + std::to_string(value_) + "," + std::to_string(value_) + "]";
}
private:
int value_;
DISALLOW_COPY_AND_ASSIGN(ConvertableToTraceFormatMock);
};
class MockTraceWriter : public TraceWriter {
public:
void AppendTraceEvent(TraceObject* trace_event) override {
// TraceObject might not have been initialized.
const char* name = trace_event->name() ? trace_event->name() : "";
events_.push_back(name);
}
void Flush() override {}
std::vector<std::string> events() { return events_; }
private:
std::vector<std::string> events_;
};
TEST(TestTraceBufferRingBuffer) {
// We should be able to add kChunkSize * 2 + 1 trace events.
const int HANDLES_COUNT = TraceBufferChunk::kChunkSize * 2 + 1;
MockTraceWriter* writer = new MockTraceWriter();
TraceBuffer* ring_buffer =
TraceBuffer::CreateTraceBufferRingBuffer(2, writer);
std::string names[HANDLES_COUNT];
for (int i = 0; i < HANDLES_COUNT; ++i) {
names[i] = "Test.EventNo" + std::to_string(i);
}
std::vector<uint64_t> handles(HANDLES_COUNT);
uint8_t category_enabled_flag = 41;
for (size_t i = 0; i < handles.size(); ++i) {
TraceObject* trace_object = ring_buffer->AddTraceEvent(&handles[i]);
CHECK_NOT_NULL(trace_object);
trace_object->Initialize('X', &category_enabled_flag, names[i].c_str(),
"Test.Scope", 42, 123, 0, nullptr, nullptr,
nullptr, nullptr, 0, 1729, 4104);
trace_object = ring_buffer->GetEventByHandle(handles[i]);
CHECK_NOT_NULL(trace_object);
CHECK_EQ('X', trace_object->phase());
CHECK_EQ(names[i], std::string(trace_object->name()));
CHECK_EQ(category_enabled_flag, *trace_object->category_enabled_flag());
}
// We should only be able to retrieve the last kChunkSize + 1.
for (size_t i = 0; i < TraceBufferChunk::kChunkSize; ++i) {
CHECK_NULL(ring_buffer->GetEventByHandle(handles[i]));
}
for (size_t i = TraceBufferChunk::kChunkSize; i < handles.size(); ++i) {
TraceObject* trace_object = ring_buffer->GetEventByHandle(handles[i]);
CHECK_NOT_NULL(trace_object);
// The object properties should be correct.
CHECK_EQ('X', trace_object->phase());
CHECK_EQ(names[i], std::string(trace_object->name()));
CHECK_EQ(category_enabled_flag, *trace_object->category_enabled_flag());
}
// Check Flush(), that the writer wrote the last kChunkSize 1 event names.
ring_buffer->Flush();
auto events = writer->events();
CHECK_EQ(TraceBufferChunk::kChunkSize + 1, events.size());
for (size_t i = TraceBufferChunk::kChunkSize; i < handles.size(); ++i) {
CHECK_EQ(names[i], events[i - TraceBufferChunk::kChunkSize]);
}
delete ring_buffer;
}
void PopulateJSONWriter(TraceWriter* writer) {
v8::Platform* old_platform = i::V8::GetCurrentPlatform();
std::unique_ptr<v8::Platform> default_platform(
v8::platform::NewDefaultPlatform());
i::V8::SetPlatformForTesting(default_platform.get());
auto tracing = base::make_unique<v8::platform::tracing::TracingController>();
v8::platform::tracing::TracingController* tracing_controller = tracing.get();
static_cast<v8::platform::DefaultPlatform*>(default_platform.get())
->SetTracingController(std::move(tracing));
TraceBuffer* ring_buffer =
TraceBuffer::CreateTraceBufferRingBuffer(1, writer);
tracing_controller->Initialize(ring_buffer);
TraceConfig* trace_config = new TraceConfig();
trace_config->AddIncludedCategory("v8-cat");
tracing_controller->StartTracing(trace_config);
TraceObject trace_object;
trace_object.InitializeForTesting(
'X', tracing_controller->GetCategoryGroupEnabled("v8-cat"), "Test0",
v8::internal::tracing::kGlobalScope, 42, 0x1234, 0, nullptr, nullptr,
nullptr, nullptr, TRACE_EVENT_FLAG_HAS_ID, 11, 22, 100, 50, 33, 44);
writer->AppendTraceEvent(&trace_object);
trace_object.InitializeForTesting(
'Y', tracing_controller->GetCategoryGroupEnabled("v8-cat"), "Test1",
v8::internal::tracing::kGlobalScope, 43, 0x5678, 0, nullptr, nullptr,
nullptr, nullptr, TRACE_EVENT_FLAG_FLOW_IN | TRACE_EVENT_FLAG_FLOW_OUT,
55, 66, 110, 55, 77, 88);
writer->AppendTraceEvent(&trace_object);
tracing_controller->StopTracing();
i::V8::SetPlatformForTesting(old_platform);
}
TEST(TestJSONTraceWriter) {
std::ostringstream stream;
TraceWriter* writer = TraceWriter::CreateJSONTraceWriter(stream);
PopulateJSONWriter(writer);
std::string trace_str = stream.str();
std::string expected_trace_str =
"{\"traceEvents\":[{\"pid\":11,\"tid\":22,\"ts\":100,\"tts\":50,"
"\"ph\":\"X\",\"cat\":\"v8-cat\",\"name\":\"Test0\",\"dur\":33,"
"\"tdur\":44,\"id\":\"0x2a\",\"args\":{}},{\"pid\":55,\"tid\":66,"
"\"ts\":110,\"tts\":55,\"ph\":\"Y\",\"cat\":\"v8-cat\",\"name\":"
"\"Test1\",\"dur\":77,\"tdur\":88,\"bind_id\":\"0x5678\","
"\"flow_in\":true,\"flow_out\":true,\"args\":{}}]}";
CHECK_EQ(expected_trace_str, trace_str);
}
TEST(TestJSONTraceWriterWithCustomtag) {
std::ostringstream stream;
TraceWriter* writer = TraceWriter::CreateJSONTraceWriter(stream, "customTag");
PopulateJSONWriter(writer);
std::string trace_str = stream.str();
std::string expected_trace_str =
"{\"customTag\":[{\"pid\":11,\"tid\":22,\"ts\":100,\"tts\":50,"
"\"ph\":\"X\",\"cat\":\"v8-cat\",\"name\":\"Test0\",\"dur\":33,"
"\"tdur\":44,\"id\":\"0x2a\",\"args\":{}},{\"pid\":55,\"tid\":66,"
"\"ts\":110,\"tts\":55,\"ph\":\"Y\",\"cat\":\"v8-cat\",\"name\":"
"\"Test1\",\"dur\":77,\"tdur\":88,\"bind_id\":\"0x5678\","
"\"flow_in\":true,\"flow_out\":true,\"args\":{}}]}";
CHECK_EQ(expected_trace_str, trace_str);
}
TEST(TestTracingController) {
v8::Platform* old_platform = i::V8::GetCurrentPlatform();
std::unique_ptr<v8::Platform> default_platform(
v8::platform::NewDefaultPlatform());
i::V8::SetPlatformForTesting(default_platform.get());
auto tracing = base::make_unique<v8::platform::tracing::TracingController>();
v8::platform::tracing::TracingController* tracing_controller = tracing.get();
static_cast<v8::platform::DefaultPlatform*>(default_platform.get())
->SetTracingController(std::move(tracing));
MockTraceWriter* writer = new MockTraceWriter();
TraceBuffer* ring_buffer =
TraceBuffer::CreateTraceBufferRingBuffer(1, writer);
tracing_controller->Initialize(ring_buffer);
TraceConfig* trace_config = new TraceConfig();
trace_config->AddIncludedCategory("v8");
tracing_controller->StartTracing(trace_config);
TRACE_EVENT0("v8", "v8.Test");
// cat category is not included in default config
TRACE_EVENT0("cat", "v8.Test2");
TRACE_EVENT0("v8", "v8.Test3");
tracing_controller->StopTracing();
CHECK_EQ(2u, writer->events().size());
CHECK_EQ(std::string("v8.Test"), writer->events()[0]);
CHECK_EQ(std::string("v8.Test3"), writer->events()[1]);
i::V8::SetPlatformForTesting(old_platform);
}
void GetJSONStrings(std::vector<std::string>& ret, std::string str,
std::string param, std::string start_delim,
std::string end_delim) {
size_t pos = str.find(param);
while (pos != std::string::npos) {
size_t start_pos = str.find(start_delim, pos + param.length());
size_t end_pos = str.find(end_delim, start_pos + 1);
CHECK_NE(start_pos, std::string::npos);
CHECK_NE(end_pos, std::string::npos);
ret.push_back(str.substr(start_pos + 1, end_pos - start_pos - 1));
pos = str.find(param, pos + 1);
}
}
TEST(TestTracingControllerMultipleArgsAndCopy) {
Revert "[tracing] Add a configurable output stream for perfetto tracing" This reverts commit a03ed626797be8e9424bef58a7dde4415eb9181f. Reason for revert: Fails on win32: https://ci.chromium.org/p/v8/builders/ci/V8%20Win32/21252 Original change's description: > [tracing] Add a configurable output stream for perfetto tracing > > Add the ability to provide perfetto with an output stream for the JSON > consumer rather than hardcode it. D8 will use this interface exclusively > once the old trace controller is removed. > > Also add a test for scope-managed trace events and their duration - this > was leftover from a previous CL. > > Cq-Include-Trybots: luci.v8.try:v8_linux64_perfetto_dbg_ng > Bug: v8:8339 > Change-Id: I1c45e17e528b549a4cfdaecabd33c7ac4ab4af77 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1611801 > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Commit-Queue: Peter Marshall <petermarshall@chromium.org> > Cr-Commit-Position: refs/heads/master@{#61753} TBR=ulan@chromium.org,yangguo@chromium.org,jgruber@chromium.org,petermarshall@chromium.org Change-Id: Ie7dbe4fc5f9e496fafc8e3ad2b05011b88c9e2c3 No-Presubmit: true No-Tree-Checks: true No-Try: true Bug: v8:8339 Cq-Include-Trybots: luci.v8.try:v8_linux64_perfetto_dbg_ng Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1625465 Reviewed-by: Clemens Hammacher <clemensh@chromium.org> Commit-Queue: Clemens Hammacher <clemensh@chromium.org> Cr-Commit-Position: refs/heads/master@{#61754}
2019-05-22 15:57:58 +00:00
std::ostringstream stream;
uint64_t aa = 11;
unsigned int bb = 22;
uint16_t cc = 33;
unsigned char dd = 44;
int64_t ee = -55;
int ff = -66;
int16_t gg = -77;
signed char hh = -88;
bool ii1 = true;
bool ii2 = false;
double jj1 = 99.0;
double jj2 = 1e100;
double jj3 = std::numeric_limits<double>::quiet_NaN();
double jj4 = std::numeric_limits<double>::infinity();
double jj5 = -std::numeric_limits<double>::infinity();
void* kk = &aa;
const char* ll = "100";
std::string mm = "INIT";
std::string mmm = "\"INIT\"";
// Create a scope for the tracing controller to terminate the trace writer.
{
v8::Platform* old_platform = i::V8::GetCurrentPlatform();
std::unique_ptr<v8::Platform> default_platform(
v8::platform::NewDefaultPlatform());
i::V8::SetPlatformForTesting(default_platform.get());
auto tracing =
base::make_unique<v8::platform::tracing::TracingController>();
v8::platform::tracing::TracingController* tracing_controller =
tracing.get();
static_cast<v8::platform::DefaultPlatform*>(default_platform.get())
->SetTracingController(std::move(tracing));
TraceWriter* writer = TraceWriter::CreateJSONTraceWriter(stream);
TraceBuffer* ring_buffer =
TraceBuffer::CreateTraceBufferRingBuffer(1, writer);
tracing_controller->Initialize(ring_buffer);
TraceConfig* trace_config = new TraceConfig();
trace_config->AddIncludedCategory("v8");
tracing_controller->StartTracing(trace_config);
{
TRACE_EVENT1("v8", "v8.Test.aa", "aa", aa);
TRACE_EVENT1("v8", "v8.Test.bb", "bb", bb);
TRACE_EVENT1("v8", "v8.Test.cc", "cc", cc);
TRACE_EVENT1("v8", "v8.Test.dd", "dd", dd);
TRACE_EVENT1("v8", "v8.Test.ee", "ee", ee);
TRACE_EVENT1("v8", "v8.Test.ff", "ff", ff);
TRACE_EVENT1("v8", "v8.Test.gg", "gg", gg);
TRACE_EVENT1("v8", "v8.Test.hh", "hh", hh);
TRACE_EVENT1("v8", "v8.Test.ii", "ii1", ii1);
TRACE_EVENT1("v8", "v8.Test.ii", "ii2", ii2);
TRACE_EVENT1("v8", "v8.Test.jj1", "jj1", jj1);
TRACE_EVENT1("v8", "v8.Test.jj2", "jj2", jj2);
TRACE_EVENT1("v8", "v8.Test.jj3", "jj3", jj3);
TRACE_EVENT1("v8", "v8.Test.jj4", "jj4", jj4);
TRACE_EVENT1("v8", "v8.Test.jj5", "jj5", jj5);
TRACE_EVENT1("v8", "v8.Test.kk", "kk", kk);
TRACE_EVENT1("v8", "v8.Test.ll", "ll", ll);
TRACE_EVENT1("v8", "v8.Test.mm", "mm", TRACE_STR_COPY(mmm.c_str()));
TRACE_EVENT2("v8", "v8.Test2.1", "aa", aa, "ll", ll);
TRACE_EVENT2("v8", "v8.Test2.2", "mm1", TRACE_STR_COPY(mm.c_str()), "mm2",
TRACE_STR_COPY(mmm.c_str()));
// Check copies are correct.
TRACE_EVENT_COPY_INSTANT0("v8", mm.c_str(), TRACE_EVENT_SCOPE_THREAD);
TRACE_EVENT_COPY_INSTANT2("v8", mm.c_str(), TRACE_EVENT_SCOPE_THREAD,
"mm1", mm.c_str(), "mm2", mmm.c_str());
mm = "CHANGED";
mmm = "CHANGED";
TRACE_EVENT_INSTANT1("v8", "v8.Test", TRACE_EVENT_SCOPE_THREAD, "a1",
new ConvertableToTraceFormatMock(42));
std::unique_ptr<ConvertableToTraceFormatMock> trace_event_arg(
new ConvertableToTraceFormatMock(42));
TRACE_EVENT_INSTANT2("v8", "v8.Test", TRACE_EVENT_SCOPE_THREAD, "a1",
std::move(trace_event_arg), "a2",
new ConvertableToTraceFormatMock(123));
}
tracing_controller->StopTracing();
i::V8::SetPlatformForTesting(old_platform);
}
std::string trace_str = stream.str();
std::vector<std::string> all_args, all_names, all_cats;
GetJSONStrings(all_args, trace_str, "\"args\"", "{", "}");
GetJSONStrings(all_names, trace_str, "\"name\"", "\"", "\"");
GetJSONStrings(all_cats, trace_str, "\"cat\"", "\"", "\"");
CHECK_EQ(all_args.size(), 24u);
CHECK_EQ(all_args[0], "\"aa\":11");
CHECK_EQ(all_args[1], "\"bb\":22");
CHECK_EQ(all_args[2], "\"cc\":33");
CHECK_EQ(all_args[3], "\"dd\":44");
CHECK_EQ(all_args[4], "\"ee\":-55");
CHECK_EQ(all_args[5], "\"ff\":-66");
CHECK_EQ(all_args[6], "\"gg\":-77");
CHECK_EQ(all_args[7], "\"hh\":-88");
CHECK_EQ(all_args[8], "\"ii1\":true");
CHECK_EQ(all_args[9], "\"ii2\":false");
CHECK_EQ(all_args[10], "\"jj1\":99.0");
CHECK_EQ(all_args[11], "\"jj2\":1e+100");
CHECK_EQ(all_args[12], "\"jj3\":\"NaN\"");
CHECK_EQ(all_args[13], "\"jj4\":\"Infinity\"");
CHECK_EQ(all_args[14], "\"jj5\":\"-Infinity\"");
std::ostringstream pointer_stream;
pointer_stream << "\"kk\":\"" << &aa << "\"";
CHECK_EQ(all_args[15], pointer_stream.str());
CHECK_EQ(all_args[16], "\"ll\":\"100\"");
CHECK_EQ(all_args[17], "\"mm\":\"\\\"INIT\\\"\"");
CHECK_EQ(all_names[18], "v8.Test2.1");
CHECK_EQ(all_args[18], "\"aa\":11,\"ll\":\"100\"");
CHECK_EQ(all_args[19], "\"mm1\":\"INIT\",\"mm2\":\"\\\"INIT\\\"\"");
CHECK_EQ(all_names[20], "INIT");
CHECK_EQ(all_names[21], "INIT");
CHECK_EQ(all_args[21], "\"mm1\":\"INIT\",\"mm2\":\"\\\"INIT\\\"\"");
CHECK_EQ(all_args[22], "\"a1\":[42,42]");
CHECK_EQ(all_args[23], "\"a1\":[42,42],\"a2\":[123,123]");
}
namespace {
class TraceStateObserverImpl : public TracingController::TraceStateObserver {
public:
void OnTraceEnabled() override { ++enabled_count; }
void OnTraceDisabled() override { ++disabled_count; }
int enabled_count = 0;
int disabled_count = 0;
};
} // namespace
TEST(TracingObservers) {
v8::Platform* old_platform = i::V8::GetCurrentPlatform();
std::unique_ptr<v8::Platform> default_platform(
v8::platform::NewDefaultPlatform());
i::V8::SetPlatformForTesting(default_platform.get());
auto tracing = base::make_unique<v8::platform::tracing::TracingController>();
v8::platform::tracing::TracingController* tracing_controller = tracing.get();
static_cast<v8::platform::DefaultPlatform*>(default_platform.get())
->SetTracingController(std::move(tracing));
MockTraceWriter* writer = new MockTraceWriter();
v8::platform::tracing::TraceBuffer* ring_buffer =
v8::platform::tracing::TraceBuffer::CreateTraceBufferRingBuffer(1,
writer);
tracing_controller->Initialize(ring_buffer);
v8::platform::tracing::TraceConfig* trace_config =
new v8::platform::tracing::TraceConfig();
trace_config->AddIncludedCategory("v8");
TraceStateObserverImpl observer;
tracing_controller->AddTraceStateObserver(&observer);
CHECK_EQ(0, observer.enabled_count);
CHECK_EQ(0, observer.disabled_count);
tracing_controller->StartTracing(trace_config);
CHECK_EQ(1, observer.enabled_count);
CHECK_EQ(0, observer.disabled_count);
TraceStateObserverImpl observer2;
tracing_controller->AddTraceStateObserver(&observer2);
CHECK_EQ(1, observer2.enabled_count);
CHECK_EQ(0, observer2.disabled_count);
tracing_controller->RemoveTraceStateObserver(&observer2);
CHECK_EQ(1, observer2.enabled_count);
CHECK_EQ(0, observer2.disabled_count);
tracing_controller->StopTracing();
CHECK_EQ(1, observer.enabled_count);
CHECK_EQ(1, observer.disabled_count);
CHECK_EQ(1, observer2.enabled_count);
CHECK_EQ(0, observer2.disabled_count);
tracing_controller->RemoveTraceStateObserver(&observer);
CHECK_EQ(1, observer.enabled_count);
CHECK_EQ(1, observer.disabled_count);
trace_config = new v8::platform::tracing::TraceConfig();
tracing_controller->StartTracing(trace_config);
tracing_controller->StopTracing();
CHECK_EQ(1, observer.enabled_count);
CHECK_EQ(1, observer.disabled_count);
i::V8::SetPlatformForTesting(old_platform);
}
[tracing] Fix races in TracingController implementation The default TracingController (used by d8 and Node) has some concurrency issues. The new test flushes these out, when a second thread logs trace events while the main thread calls StopTracing(). - Use an acquire load in UpdateCategoryGroupEnabledFlags() because this was racing with GetCategoryGroupEnabled() where a new category is added in the slow path. g_category_groups is append-only, but reads/writes to g_category_index need to be correctly ordered so that new categories are added and only then is the change to the index visible. The relaxed load ignored this and caused unsynchronized read/write. - Use a relaxed load in ~ScopedTracer() to access category_group_enabled as this previously used a non-atomic operation which caused a race with UpdateCategoryGroupEnabledFlag() which does a relaxed store. - Replace TracingController::mode_ with an atomic bool as read/writes to mode_ were not synchronized and caused TSAN errors. It only has two states and it doesn't seem like we will extend this so just convert it to bool. - Take the lock around calling trace_object->Initialize in AddTraceEvent(), and around trace_buffer_->Flush() in StopTracing(). These two raced previously as the underlying TraceBufferRingBuffer passes out pointers to TraceObjects in a synchronized way, but the caller (AddTraceEvent) then writes into the object without synchronization. This leads to races when Flush() is called, at which time TraceBufferRingBuffer assumes that all the pointers it handed out are to valid, initialized TraceObjects - which is not true because AddTraceEvent may still be calling Initialize on them. This could be the cause of issues in Node.js where the last line of tracing/logging sometimes gets cut off. This is kind of a band-aid solution - access to the TraceObjects handed out by the ring buffer really needs proper synchronization which at this point would require redesign. It's quite likely we will replace this with Perfetto in the near future so not much point investing in this code right now. - Enable TracingCpuProfiler test which was flaky due to these bugs. Bug: v8:8821 Change-Id: I141296800c6906ac0e7f3f21dd16d861b07dae62 Reviewed-on: https://chromium-review.googlesource.com/c/1477283 Commit-Queue: Peter Marshall <petermarshall@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Ali Ijaz Sheikh <ofrobots@google.com> Cr-Commit-Position: refs/heads/master@{#59752}
2019-02-20 09:02:06 +00:00
class TraceWritingThread : public base::Thread {
public:
TraceWritingThread(
v8::platform::tracing::TracingController* tracing_controller)
: base::Thread(base::Thread::Options("TraceWritingThread")),
tracing_controller_(tracing_controller) {}
void Run() override {
running_.store(true);
while (running_.load()) {
[tracing] Fix races in TracingController implementation The default TracingController (used by d8 and Node) has some concurrency issues. The new test flushes these out, when a second thread logs trace events while the main thread calls StopTracing(). - Use an acquire load in UpdateCategoryGroupEnabledFlags() because this was racing with GetCategoryGroupEnabled() where a new category is added in the slow path. g_category_groups is append-only, but reads/writes to g_category_index need to be correctly ordered so that new categories are added and only then is the change to the index visible. The relaxed load ignored this and caused unsynchronized read/write. - Use a relaxed load in ~ScopedTracer() to access category_group_enabled as this previously used a non-atomic operation which caused a race with UpdateCategoryGroupEnabledFlag() which does a relaxed store. - Replace TracingController::mode_ with an atomic bool as read/writes to mode_ were not synchronized and caused TSAN errors. It only has two states and it doesn't seem like we will extend this so just convert it to bool. - Take the lock around calling trace_object->Initialize in AddTraceEvent(), and around trace_buffer_->Flush() in StopTracing(). These two raced previously as the underlying TraceBufferRingBuffer passes out pointers to TraceObjects in a synchronized way, but the caller (AddTraceEvent) then writes into the object without synchronization. This leads to races when Flush() is called, at which time TraceBufferRingBuffer assumes that all the pointers it handed out are to valid, initialized TraceObjects - which is not true because AddTraceEvent may still be calling Initialize on them. This could be the cause of issues in Node.js where the last line of tracing/logging sometimes gets cut off. This is kind of a band-aid solution - access to the TraceObjects handed out by the ring buffer really needs proper synchronization which at this point would require redesign. It's quite likely we will replace this with Perfetto in the near future so not much point investing in this code right now. - Enable TracingCpuProfiler test which was flaky due to these bugs. Bug: v8:8821 Change-Id: I141296800c6906ac0e7f3f21dd16d861b07dae62 Reviewed-on: https://chromium-review.googlesource.com/c/1477283 Commit-Queue: Peter Marshall <petermarshall@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Ali Ijaz Sheikh <ofrobots@google.com> Cr-Commit-Position: refs/heads/master@{#59752}
2019-02-20 09:02:06 +00:00
TRACE_EVENT0("v8", "v8.Test");
tracing_controller_->AddTraceEvent('A', nullptr, "v8", "", 1, 1, 0,
nullptr, nullptr, nullptr, nullptr, 0);
tracing_controller_->AddTraceEventWithTimestamp('A', nullptr, "v8", "", 1,
1, 0, nullptr, nullptr,
nullptr, nullptr, 0, 0);
}
}
void Stop() { running_.store(false); }
[tracing] Fix races in TracingController implementation The default TracingController (used by d8 and Node) has some concurrency issues. The new test flushes these out, when a second thread logs trace events while the main thread calls StopTracing(). - Use an acquire load in UpdateCategoryGroupEnabledFlags() because this was racing with GetCategoryGroupEnabled() where a new category is added in the slow path. g_category_groups is append-only, but reads/writes to g_category_index need to be correctly ordered so that new categories are added and only then is the change to the index visible. The relaxed load ignored this and caused unsynchronized read/write. - Use a relaxed load in ~ScopedTracer() to access category_group_enabled as this previously used a non-atomic operation which caused a race with UpdateCategoryGroupEnabledFlag() which does a relaxed store. - Replace TracingController::mode_ with an atomic bool as read/writes to mode_ were not synchronized and caused TSAN errors. It only has two states and it doesn't seem like we will extend this so just convert it to bool. - Take the lock around calling trace_object->Initialize in AddTraceEvent(), and around trace_buffer_->Flush() in StopTracing(). These two raced previously as the underlying TraceBufferRingBuffer passes out pointers to TraceObjects in a synchronized way, but the caller (AddTraceEvent) then writes into the object without synchronization. This leads to races when Flush() is called, at which time TraceBufferRingBuffer assumes that all the pointers it handed out are to valid, initialized TraceObjects - which is not true because AddTraceEvent may still be calling Initialize on them. This could be the cause of issues in Node.js where the last line of tracing/logging sometimes gets cut off. This is kind of a band-aid solution - access to the TraceObjects handed out by the ring buffer really needs proper synchronization which at this point would require redesign. It's quite likely we will replace this with Perfetto in the near future so not much point investing in this code right now. - Enable TracingCpuProfiler test which was flaky due to these bugs. Bug: v8:8821 Change-Id: I141296800c6906ac0e7f3f21dd16d861b07dae62 Reviewed-on: https://chromium-review.googlesource.com/c/1477283 Commit-Queue: Peter Marshall <petermarshall@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Ali Ijaz Sheikh <ofrobots@google.com> Cr-Commit-Position: refs/heads/master@{#59752}
2019-02-20 09:02:06 +00:00
private:
std::atomic_bool running_{false};
[tracing] Fix races in TracingController implementation The default TracingController (used by d8 and Node) has some concurrency issues. The new test flushes these out, when a second thread logs trace events while the main thread calls StopTracing(). - Use an acquire load in UpdateCategoryGroupEnabledFlags() because this was racing with GetCategoryGroupEnabled() where a new category is added in the slow path. g_category_groups is append-only, but reads/writes to g_category_index need to be correctly ordered so that new categories are added and only then is the change to the index visible. The relaxed load ignored this and caused unsynchronized read/write. - Use a relaxed load in ~ScopedTracer() to access category_group_enabled as this previously used a non-atomic operation which caused a race with UpdateCategoryGroupEnabledFlag() which does a relaxed store. - Replace TracingController::mode_ with an atomic bool as read/writes to mode_ were not synchronized and caused TSAN errors. It only has two states and it doesn't seem like we will extend this so just convert it to bool. - Take the lock around calling trace_object->Initialize in AddTraceEvent(), and around trace_buffer_->Flush() in StopTracing(). These two raced previously as the underlying TraceBufferRingBuffer passes out pointers to TraceObjects in a synchronized way, but the caller (AddTraceEvent) then writes into the object without synchronization. This leads to races when Flush() is called, at which time TraceBufferRingBuffer assumes that all the pointers it handed out are to valid, initialized TraceObjects - which is not true because AddTraceEvent may still be calling Initialize on them. This could be the cause of issues in Node.js where the last line of tracing/logging sometimes gets cut off. This is kind of a band-aid solution - access to the TraceObjects handed out by the ring buffer really needs proper synchronization which at this point would require redesign. It's quite likely we will replace this with Perfetto in the near future so not much point investing in this code right now. - Enable TracingCpuProfiler test which was flaky due to these bugs. Bug: v8:8821 Change-Id: I141296800c6906ac0e7f3f21dd16d861b07dae62 Reviewed-on: https://chromium-review.googlesource.com/c/1477283 Commit-Queue: Peter Marshall <petermarshall@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Ali Ijaz Sheikh <ofrobots@google.com> Cr-Commit-Position: refs/heads/master@{#59752}
2019-02-20 09:02:06 +00:00
v8::platform::tracing::TracingController* tracing_controller_;
};
TEST(AddTraceEventMultiThreaded) {
v8::Platform* old_platform = i::V8::GetCurrentPlatform();
std::unique_ptr<v8::Platform> default_platform(
v8::platform::NewDefaultPlatform());
i::V8::SetPlatformForTesting(default_platform.get());
auto tracing = base::make_unique<v8::platform::tracing::TracingController>();
v8::platform::tracing::TracingController* tracing_controller = tracing.get();
static_cast<v8::platform::DefaultPlatform*>(default_platform.get())
->SetTracingController(std::move(tracing));
MockTraceWriter* writer = new MockTraceWriter();
TraceBuffer* ring_buffer =
TraceBuffer::CreateTraceBufferRingBuffer(1, writer);
tracing_controller->Initialize(ring_buffer);
TraceConfig* trace_config = new TraceConfig();
trace_config->AddIncludedCategory("v8");
tracing_controller->StartTracing(trace_config);
TraceWritingThread thread(tracing_controller);
thread.StartSynchronously();
TRACE_EVENT0("v8", "v8.Test2");
TRACE_EVENT0("v8", "v8.Test2");
[tracing] Fix races in TracingController implementation The default TracingController (used by d8 and Node) has some concurrency issues. The new test flushes these out, when a second thread logs trace events while the main thread calls StopTracing(). - Use an acquire load in UpdateCategoryGroupEnabledFlags() because this was racing with GetCategoryGroupEnabled() where a new category is added in the slow path. g_category_groups is append-only, but reads/writes to g_category_index need to be correctly ordered so that new categories are added and only then is the change to the index visible. The relaxed load ignored this and caused unsynchronized read/write. - Use a relaxed load in ~ScopedTracer() to access category_group_enabled as this previously used a non-atomic operation which caused a race with UpdateCategoryGroupEnabledFlag() which does a relaxed store. - Replace TracingController::mode_ with an atomic bool as read/writes to mode_ were not synchronized and caused TSAN errors. It only has two states and it doesn't seem like we will extend this so just convert it to bool. - Take the lock around calling trace_object->Initialize in AddTraceEvent(), and around trace_buffer_->Flush() in StopTracing(). These two raced previously as the underlying TraceBufferRingBuffer passes out pointers to TraceObjects in a synchronized way, but the caller (AddTraceEvent) then writes into the object without synchronization. This leads to races when Flush() is called, at which time TraceBufferRingBuffer assumes that all the pointers it handed out are to valid, initialized TraceObjects - which is not true because AddTraceEvent may still be calling Initialize on them. This could be the cause of issues in Node.js where the last line of tracing/logging sometimes gets cut off. This is kind of a band-aid solution - access to the TraceObjects handed out by the ring buffer really needs proper synchronization which at this point would require redesign. It's quite likely we will replace this with Perfetto in the near future so not much point investing in this code right now. - Enable TracingCpuProfiler test which was flaky due to these bugs. Bug: v8:8821 Change-Id: I141296800c6906ac0e7f3f21dd16d861b07dae62 Reviewed-on: https://chromium-review.googlesource.com/c/1477283 Commit-Queue: Peter Marshall <petermarshall@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Ali Ijaz Sheikh <ofrobots@google.com> Cr-Commit-Position: refs/heads/master@{#59752}
2019-02-20 09:02:06 +00:00
base::OS::Sleep(base::TimeDelta::FromMilliseconds(10));
[tracing] Fix races in TracingController implementation The default TracingController (used by d8 and Node) has some concurrency issues. The new test flushes these out, when a second thread logs trace events while the main thread calls StopTracing(). - Use an acquire load in UpdateCategoryGroupEnabledFlags() because this was racing with GetCategoryGroupEnabled() where a new category is added in the slow path. g_category_groups is append-only, but reads/writes to g_category_index need to be correctly ordered so that new categories are added and only then is the change to the index visible. The relaxed load ignored this and caused unsynchronized read/write. - Use a relaxed load in ~ScopedTracer() to access category_group_enabled as this previously used a non-atomic operation which caused a race with UpdateCategoryGroupEnabledFlag() which does a relaxed store. - Replace TracingController::mode_ with an atomic bool as read/writes to mode_ were not synchronized and caused TSAN errors. It only has two states and it doesn't seem like we will extend this so just convert it to bool. - Take the lock around calling trace_object->Initialize in AddTraceEvent(), and around trace_buffer_->Flush() in StopTracing(). These two raced previously as the underlying TraceBufferRingBuffer passes out pointers to TraceObjects in a synchronized way, but the caller (AddTraceEvent) then writes into the object without synchronization. This leads to races when Flush() is called, at which time TraceBufferRingBuffer assumes that all the pointers it handed out are to valid, initialized TraceObjects - which is not true because AddTraceEvent may still be calling Initialize on them. This could be the cause of issues in Node.js where the last line of tracing/logging sometimes gets cut off. This is kind of a band-aid solution - access to the TraceObjects handed out by the ring buffer really needs proper synchronization which at this point would require redesign. It's quite likely we will replace this with Perfetto in the near future so not much point investing in this code right now. - Enable TracingCpuProfiler test which was flaky due to these bugs. Bug: v8:8821 Change-Id: I141296800c6906ac0e7f3f21dd16d861b07dae62 Reviewed-on: https://chromium-review.googlesource.com/c/1477283 Commit-Queue: Peter Marshall <petermarshall@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Ali Ijaz Sheikh <ofrobots@google.com> Cr-Commit-Position: refs/heads/master@{#59752}
2019-02-20 09:02:06 +00:00
tracing_controller->StopTracing();
thread.Stop();
[tracing] Fix races in TracingController implementation The default TracingController (used by d8 and Node) has some concurrency issues. The new test flushes these out, when a second thread logs trace events while the main thread calls StopTracing(). - Use an acquire load in UpdateCategoryGroupEnabledFlags() because this was racing with GetCategoryGroupEnabled() where a new category is added in the slow path. g_category_groups is append-only, but reads/writes to g_category_index need to be correctly ordered so that new categories are added and only then is the change to the index visible. The relaxed load ignored this and caused unsynchronized read/write. - Use a relaxed load in ~ScopedTracer() to access category_group_enabled as this previously used a non-atomic operation which caused a race with UpdateCategoryGroupEnabledFlag() which does a relaxed store. - Replace TracingController::mode_ with an atomic bool as read/writes to mode_ were not synchronized and caused TSAN errors. It only has two states and it doesn't seem like we will extend this so just convert it to bool. - Take the lock around calling trace_object->Initialize in AddTraceEvent(), and around trace_buffer_->Flush() in StopTracing(). These two raced previously as the underlying TraceBufferRingBuffer passes out pointers to TraceObjects in a synchronized way, but the caller (AddTraceEvent) then writes into the object without synchronization. This leads to races when Flush() is called, at which time TraceBufferRingBuffer assumes that all the pointers it handed out are to valid, initialized TraceObjects - which is not true because AddTraceEvent may still be calling Initialize on them. This could be the cause of issues in Node.js where the last line of tracing/logging sometimes gets cut off. This is kind of a band-aid solution - access to the TraceObjects handed out by the ring buffer really needs proper synchronization which at this point would require redesign. It's quite likely we will replace this with Perfetto in the near future so not much point investing in this code right now. - Enable TracingCpuProfiler test which was flaky due to these bugs. Bug: v8:8821 Change-Id: I141296800c6906ac0e7f3f21dd16d861b07dae62 Reviewed-on: https://chromium-review.googlesource.com/c/1477283 Commit-Queue: Peter Marshall <petermarshall@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Ali Ijaz Sheikh <ofrobots@google.com> Cr-Commit-Position: refs/heads/master@{#59752}
2019-02-20 09:02:06 +00:00
thread.Join();
i::V8::SetPlatformForTesting(old_platform);
}
} // namespace tracing
} // namespace platform
} // namespace v8