Design doc:
https://docs.google.com/document/d/1xkXj94iExFgLWc_OszTNyNGi523ARaKMWPZTeomhI4U
A lot has changed since the last patchset! I recommend revisiting this
design doc and reading the parts in green. I explain the roadmap for
what changes to expect from ETW instrumentation as well as the
instrumentation of this particular CL.
I'll do my best to answer any further questions anyone has about my
particular instrumentation or ETW in general :)
---
This is the first of a series of changelists to round out ETW
instrumentation for V8.
This changelist represents the most minimal change needed to instrument
ETW in V8. In particular, it:
- defines and registers the ETW provider,
- interacts minimally with the rest of V8, by hooking into the
existing TracingController::AddTraceEvent function,
- is designed with a platform-agnostic layer, so that event tracers
for other platforms can be instrumented in teh future.
Some notes on instrumentation (aka I copied stuff from the design doc):
We make heavy use of the TraceLogging API to log events. It differs from
previous methods of emitting ETW events in that it doesn<E2><80><99>t
require the overhead of a separate manifest file to keep track of
metadata; rather, events using this API are self-descriptive.
Here are the five major steps to instrument the TraceLogging API:
- Forward declare the provider (from provider-win.h)
- Define the provider in a .cc file (from provider-win.cc)
- Register the provider (called from v8.cc).
- Write events (called from libplatform/tracing-controller.cc)
- Unregister the provider (called from v8.cc)
At the base, we have an abstract provider class that encapsulates the
functionality of an event provider. These are things like registering
and unregistering the provider, and the actual event-logging.
The provider class is split into provider-win and provider-mac
(currently not instantiated) classes, with OS-dependent implementations
of the above functions.
In particular, the TraceLogging API is used only in provider-win. It is
here that we forward declare and define the provider, as well as write
ETW events.
Finally, there is a v8-provider class that serves as a top-level API and
is exposed to the rest of V8. It acts as a wrapper for the
platform-specific providers.
The .wprp file is needed so that Windows Performance Recorder knows how
to capture our events.
Some considerations:
- Is TracingController::AddTraceEvent the best place from which to
write my events?
- Is src/libplatform/tracing the best place to put my instrumentation?
- Right now, I fail the preupload because of this, which tells me my
files are probably not in the best location:
You added one or more #includes that violate checkdeps rules.
src\init\v8.cc Illegal include: "src/libplatform/tracing/v8-provider.h"
Because of "-src/libplatform" from src's include_rules.
Change-Id: Id53e4a034c9e526524a17000da0a647a95d93edf
Bug: v8:11043
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2233407
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Peter Marshall <petermarshall@chromium.org>
Commit-Queue: Sara Tang <sartang@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#71918}
The category name changed in https://crrev.com/c/2159690 (in
depot_tools), hence presubmit tests start failing whenever someone
touches a file which still refers to the old category.
R=zhin@chromium.orgTBR=machenbach@chromium.org
No-Try: true
No-Tree-Checks: true
Change-Id: I62813a42d63e512d421c4fe94229d04a56056978
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2165760
Reviewed-by: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#67403}
This patch replaces V8's tracing implementation (i.e., the TRACE_EVENT
macros) with the track event base implementation from Perfetto. The
advantages of doing this are:
1) This allows us to remove most tracing-related backend code from V8.
2) V8 can start writing strongly typed trace event arguments, which
are more compact, easier to process and more extensible than legacy
JSON-based trace arguments.
For the time being, we still support the old trace macros when V8 is
embedded into Chrome and other embedders.
Design doc: https://docs.google.com/document/d/1f7tt4cb-JcA5bQFR1oXk60ncJPpkL02_Hi_Bc6MfTQk/edit#heading=h.398p6b4eaen2
Bug: chromium:1006766
Change-Id: Ie71474fbe065821772b13d851487ebbca680c4ae
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1947688
Commit-Queue: Sami Kyöstilä <skyostil@chromium.org>
Auto-Submit: Sami Kyöstilä <skyostil@chromium.org>
Reviewed-by: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#67217}
All the data types defined under ArgValue are 8 bytes
expect "bool as_bool". When casting to <uint64_t> under
"tracing/trace-event.h", boolean gets placed on the lower
byte of the memory on LE, and on the higher byte on BE machines.
When using a "Union" to read back the value as a boolean, only
the lower byte of the memory location is read which makes it
fine on LE machines, however the value will not be present on BE
machines.
Using an 8 byte data type as boolean will assure the entire filed
is read instead of only the lower byte.
Change-Id: I0740b9c019588c963a4c7878af60c6df04827141
TBR: petermarshall@chromium.org
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1896835
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Junliang Yan <jyan@ca.ibm.com>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64859}
Update the TracingCpuProfiler test to work properly with perfetto.
Roll perfetto to get fixes for bugs encountered with starting/stopping
tracing rapidly, which happens in the test for the tracing profiler.
Add a check that the DataSource::Register call was successful to flush
out any errors there (although they are fixed by the perfetto roll).
Emit a fake trace event when stopping tracing in order to avoid
losing the final trace event that the user provided. Remove the ad-hoc
fake final trace events that the cctests for perfetto added.
Add a test StartAndStopRepeated which flushes out the issues fixed by
the perfetto roll.
TBR=yangguo@chromium.org
Cq-Include-Trybots: luci.v8.try:v8_linux64_perfetto_dbg_ng
Bug: v8:8339
Change-Id: I042d7385486bf42c86f1631406974693868a477f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1731006
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#63083}
The client API provides a much simpler interface so that we don't have
to deal with producers, consumers etc. directly. This CL removes all the
code that dealt with the more complex API used previously.
The architecture used here requires that the embedder call into
Tracing::Initialize() to set up the tracing backend. The tracing
controller then connects to this backend when calling
DataSource::Register() and Tracing::NewTrace(). This will ultimately
avoid the need for a virtual call (or two) for every trace event that
need to be dispatched over the API - chrome can provide a backend
and V8 will connect to it opaquely with the same code when tracing is
enabled.
Cq-Include-Trybots: luci.v8.try:v8_linux64_perfetto_dbg_ng
Bug: v8:8339
Change-Id: I6b74fbb49ffcc89638caeb59ed3d5cc81238f3e8
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1634916
Reviewed-by: Yang Guo <yangguo@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#62568}
Add a new abstract class TraceEventListener which is just an interface
for consuming trace events. This separates the V8-specific stuff that
an actual perfetto consumer needs to do e.g. handling the has_more flag
and signalling back to the controller with a semaphore.
This is a change from the previous plan of making the PerfettoConsumer
class sub-classable to implement custom consumption of trace events.
This will be difficult when the consumer is created outside of the
PerfettoTracingController as we can't hook up the
consumer_finished_semaphore_ that belongs to the controller.
Now the PerfettoTracingController is responsible for the Consumer life-
cycle and hides it entirely from callers. We add the
AddTraceEventListener() method to allow callers to register a listener
either for testing or a JSON listener for real tracing.
This lets us write tests that can store all the trace events in memory
without first converting them to JSON, letting us write test more
easily. There's an example test add to test-tracing - more tests using
this style will follow.
Cq-Include-Trybots: luci.v8.try:v8_linux64_perfetto_dbg_ng
Bug: v8:8339
Change-Id: I2d2b0f408b1c7bed954144163e1968f40d772c1b
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628789
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Auto-Submit: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#61854}
This is a reland of a03ed62679
Removed the added test which was failing on win32. The test was unrelated
to the CL; we can add it later.
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=jgruber@chromium.org, ulan@chromium.org
Bug: v8:8339
Change-Id: I3442a4d111e12947c107e7d0c226ae934acd06e4
Cq-Include-Trybots: luci.v8.try:v8_linux64_perfetto_dbg_ng
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1627334
Reviewed-by: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#61780}
This reverts commit a03ed62679.
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}
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}
Based on Primiano's prototype:
https://chromium-review.googlesource.com/c/v8/v8/+/1290549
This is still behind a build flag. I'll add functionality incrementally
rather than land everything in one giant CL.
This CL sets up the basic classes that will be used for the Perfetto
implementation, e.g. the producer, consumer, controller and task runner.
This implementation produces a binary proto file in the current
directory named v8_trace.proto. It doesn't yet produce JSON output,
that is coming in a following CL.
Currently the old tracing and perfetto tracing are both run alongside
each other if the build flag is enabled.
Cq-Include-Trybots: luci.v8.try:v8_linux64_perfetto_dbg_ng
Bug: v8:8339
Change-Id: I0eb9ecefa191ceead60aadd5b591d75c99395a6e
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1408995
Reviewed-by: Yang Guo <yangguo@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#61209}
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}
- Switch #define constant to static const
- Remove unnecessary Internal version of GetCategoryGroupEnabled()
- Fix a typo in a comment
Change-Id: I4af71dc62c7c4742bdfbcaa1ad336298eb325c42
Reviewed-on: https://chromium-review.googlesource.com/c/1477221
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#59708}
This API will be used by Node.js to provide output compatible with
Chrome devtools.
Cq-Include-Trybots: luci.chromium.try:linux_chromium_rel_ng
Change-Id: I265495f8af39bfc78d7fdbe43ac308f0920e817d
Reviewed-on: https://chromium-review.googlesource.com/1044491
Reviewed-by: Aleksey Kozyatinskiy <kozyatinskiy@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Commit-Queue: Eugene Ostroukhov <eostroukhov@chromium.org>
Cr-Commit-Position: refs/heads/master@{#53041}
Make it possible for embedders to provide their own tracing timetamps by
providing an overridable virtual function on V8's tracing controller.
Bug:
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: I727e633cb7f63d4b41c2e427ecca3c9174c90bfe
Reviewed-on: https://chromium-review.googlesource.com/847690
Reviewed-by: Yang Guo <yangguo@chromium.org>
Reviewed-by: Fadi Meawad <fmeawad@chromium.org>
Commit-Queue: Ali Ijaz Sheikh <ofrobots@google.com>
Cr-Commit-Position: refs/heads/master@{#50489}
This will allow for embedders to easily implement their own Platform
without duplicating the tracing controller code.
BUG=v8:6511
R=fmeawad@chromium.org
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: I7c64933d12b2cf53f0636fbc87f6ad5d22019f5c
Reviewed-on: https://chromium-review.googlesource.com/543015
Commit-Queue: Jochen Eisinger <jochen@chromium.org>
Reviewed-by: Fadi Meawad <fmeawad@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46118}
We only need included categories list, excluded categories list will only work
if we use regular expression in categories list, which is not supported in V8.
TBR=jochen@chromium.org
Review-Url: https://codereview.chromium.org/2462143002
Cr-Commit-Position: refs/heads/master@{#40681}
The state sampling was implemented in chrome, we had an interface for it
V8 but it was not implemented yet.
The chrome version version has been removed in
https://codereview.chromium.org/2406703002/
Therefore following up with its removal in V8 as well.
This CL can land independent of the Chromium related CL.
R=primiano@chromium.org,alph@chromium.org
Review-Url: https://codereview.chromium.org/2410523002
Cr-Commit-Position: refs/heads/master@{#40232}
Reason for revert:
Revert this CL due to V8 Arm Builder failure and V8 Mips Builder failure.
https://build.chromium.org/p/client.v8.ports/builders/V8%20Arm%20-%20builder/builds/2456https://build.chromium.org/p/client.v8.ports/builders/V8%20Mips%20-%20builder/builds/2506
Original issue's description:
> [Tracing] V8 Tracing Controller
>
> V8 has had a trace event macro interface for while, but without a tracing
> controller a standalone V8 would be unable to collect traces.
>
> This CL introduces a complete Tracing Controller system for V8.
> It is fully function except that it does not yet store trace event args.
>
> This CL has a few components,
> The tracing controller itself, contributed by the author of this CL
> The Trace config (including the parser), contributed by lpy@
> The Trace Object, Trace Writer, and Trace Buffer are all contributed by rksang@
>
> BUG=v8:4561
> LOG=N
>
> Committed: https://crrev.com/3d598452679ce208ad9b2f48e0fb3fae352ce375
> Cr-Commit-Position: refs/heads/master@{#38073}
TBR=jochen@chromium.org,mattloring@google.com,rskang@google.com,yangguo@chromium.org,fmeawad@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:4561
Review-Url: https://codereview.chromium.org/2183943002
Cr-Commit-Position: refs/heads/master@{#38074}
V8 has had a trace event macro interface for while, but without a tracing
controller a standalone V8 would be unable to collect traces.
This CL introduces a complete Tracing Controller system for V8.
It is fully function except that it does not yet store trace event args.
This CL has a few components,
The tracing controller itself, contributed by the author of this CL
The Trace config (including the parser), contributed by lpy@
The Trace Object, Trace Writer, and Trace Buffer are all contributed by rksang@
BUG=v8:4561
LOG=N
Review-Url: https://codereview.chromium.org/2137013006
Cr-Commit-Position: refs/heads/master@{#38073}