• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1.. _module-pw_trace_tokenized:
2
3==================
4pw_trace_tokenized
5==================
6Pigweed's tracing module provides facilities for applications to trace
7information about the execution of their application. The module is split into
8two components:
9
101. The facade, provided elsewhere, which is only a macro interface layer
112. The backend (this module), is one implemention of the low level tracing.
12
13------
14Status
15------
16This module is currently in development, and is therefore still undergoing
17significant changes.
18
19Future work will:
20
211. Add a more complete API for how to retrieve data from ring_buffer.
222. Add a Python library to decode the trace data.
233. Add examples with sample output (especially for filtering and triggering).
244. Add tools to retrieve trace data.
255. Add more sinks, such as RTT.
266. Add support to more platforms.
277. Improve the locking behaviour and provide default trace locking
28   implementions.
29
30--------
31Overview
32--------
33The tokenized trace backend aims to be a reasonable tradeoff of trace features
34and event size for most applications. It works by encoding all compile time data
35for a trace event into a tokenized number. This provides a good amount of
36compression, while maintaining the full trace feature set.
37
38In addition the tokenized trace backend adds flexibility through callbacks,
39which allows the application to do things such as filtering trace_events and
40triggering tracing to turn on and off. This flexibility can help maximize the
41effectiveness of a limited trace buffer as well as be a valuable tool while
42debugging.
43
44
45Compatibility
46-------------
47Most of this module is compatible with C and C++, the only exception to this is
48the ``RegisterCallbackWhenCreated`` helper class.
49
50Dependencies
51------------
52``pw_assert``
53``pw_log``
54``pw_preprocessor``
55``pw_status``
56``pw_tokenizer``
57``pw_trace:facade``
58``pw_varint``
59
60---------
61Macro API
62---------
63All code should use the trace API facade directly. This backend fully
64implements all features of the tracing facade.
65
66
67Event Callbacks & Data Sinks
68----------------------------
69The tokenized trace module adds both event callbacks and data sinks which
70provide hooks into tracing.
71
72The *event callbacks* are called when trace events occur, with the trace event
73data, before the event is encoded or sent to the sinks. The callbacks may
74modify the run-time fields of the trace event, i.e. ``trace_id``,
75``data_buffer`` and ``data_size``. Using the return flags, these callbacks can
76be used to adjust the trace behaviour at runtime in response to specific events.
77
78If requested (using ``called_on_every_event``) the callback will be called on
79every trace event regardless if tracing is currently enabled or not. Using this,
80the application can trigger tracing on or off when specific traces or patterns
81of traces are observed, or can selectively filter traces to preserve the trace
82buffer.
83
84The event callback is called in the context of the traced task. It must be
85ISR-safe to support tracing within ISRs. It must be lightweight to prevent
86performance issues in the trace tasks.
87
88The return flags ``pw_trace_TraceEventReturnFlags`` support the following
89behaviors:
90
91* ``PW_TRACE_EVENT_RETURN_FLAGS_SKIP_EVENT`` can be set true to skip this
92  sample.
93* ``PW_TRACE_EVENT_RETURN_FLAGS_DISABLE_AFTER_PROCESSING`` can be set true to
94  disable tracing after this sample.
95
96.. cpp:function:: pw_trace_TraceEventReturnFlags pw_trace_EventCallback( \
97    void* user_data,  \
98    pw_trace_tokenized_TraceEvent* event)
99.. cpp:function:: pw_Status pw_trace_RegisterEventCallback( \
100    pw_trace_EventCallback callback, \
101    pw_trace_EventCallbackFlags flags, \
102    void* user_data, \
103    pw_trace_EventCallbackHandle* handle)
104.. cpp:function:: pw_Status pw_trace_UnregisterEventCallback( \
105    pw_trace_EventCallbackHandle handle)
106
107
108The *data sinks* are called only for trace events which get processed (tracing
109is enabled, and the sample not skipped). The sink callback is called with the
110encoded bytes of the trace event, which can be used by the application to
111connect different data sinks. The callback is broken into three callbacks
112``pw_trace_SinkStartBlock``, ``pw_trace_SinkAddBytes``, and
113``pw_trace_SinkEndBlock``. ``Start`` is called with the size of the block,
114before any bytes are emitted and can be used if needed to allocate space.
115``AddBytes`` is then called multiple times with chunks of bytes. Finally ``End``
116is called to allow any cleanup to be done by the sink if neccessary. Not all
117callbacks are required, it is acceptible to provide nullptr for any callbacks
118which you don't require.
119
120.. cpp:function:: void pw_trace_SinkStartBlock(void* user_data, size_t size)
121.. cpp:function:: void pw_trace_SinkAddBytes( \
122    void* user_data, \
123    const void* bytes, \
124    size_t size)
125.. cpp:function:: void pw_trace_SinkEndBlock(void* user_data)
126.. cpp:function:: pw_Status pw_trace_RegisterSink( \
127    pw_trace_SinkStartBlock start, \
128    pw_trace_SinkAddBytes add_bytes, \
129    pw_trace_SinkEndBlock end_block, \
130    void* user_data, \
131    pw_trace_SinkHandle* handle)
132.. cpp:function:: pw_Status pw_trace_UnregisterSink(pw_trace_SinkHandle handle)
133
134Trace Reference
135---------------
136Some use-cases might involve referencing a specific trace event, for example
137to use it as a trigger or filtering. Since the trace events are tokenized, a
138macro is provided to generate the token to use as a reference. All the fields
139must match exactly to generate the correct trace reference. If the trace does
140not have a group, use ``PW_TRACE_GROUP_LABEL_DEFAULT``.
141
142.. cpp:function:: PW_TRACE_REF(event_type, module, label, flags, group)
143.. cpp:function:: PW_TRACE_REF_DATA( \
144   event_type, module, label, flags, group, type)
145
146
147-----------
148Time source
149-----------
150Tracing requires the platform to provide the time source for tracing, this can
151be done in one of a few ways.
152
1531. Create a file with the default time functions, and provide as build variable
154   ``pw_trace_tokenized_time``, which will get pulled in as a dependency.
1552. Provide time functions elsewhere in project, and ensure they are included.
1563. Redefine the trace time macros to something else, other then the default
157   trace time functions.
158
159.. cpp:function:: PW_TRACE_TIME_TYPE pw_trace_GetTraceTime()
160.. cpp:function:: PW_TRACE_GET_TIME()
161.. cpp:function:: size_t pw_trace_GetTraceTimeTicksPerSecond()
162.. cpp:function:: PW_TRACE_GET_TIME_TICKS_PER_SECOND()
163
164
165------
166Buffer
167------
168The optional trace buffer adds a ring buffer which contains the encoded trace
169data. This is still a work in progress, in particular better methods for
170retrieving the data still need to be added. Currently there is an accessor for
171the underlying ring buffer object, but this is a short term solution.
172
173.. cpp:function:: void ClearBuffer()
174.. cpp:function:: pw::ring_buffer::PrefixedEntryRingBuffer* GetBuffer()
175
176The buffer has two configurable options:
177
1781. PW_TRACE_BUFFER_SIZE_BYTES: The total size of the ring buffer in bytes.
1792. PW_TRACE_BUFFER_MAX_BLOCK_SIZE_BYTES: The maximum single trace object size.
180   Including the token, time, and any attached data. Any trace object larger
181   then this will be dropped.
182
183.. cpp:function:: ConstByteSpan DeringAndViewRawBuffer()
184
185The DeringAndViewRawBuffer function can be used to get bulk access of the full
186deringed prefixed-ring-buffer data. This might be neccessary for large zero-copy
187bulk transfers. It is the caller's responsibility to disable tracing during
188access to the buffer. The data in the block is defined by the
189prefixed-ring-buffer format without any user-preamble.
190
191
192Added dependencies
193------------------
194``pw_ring_buffer``
195``pw_varint``
196
197
198-------
199Logging
200-------
201The optional trace buffer logging adds support to dump trace buffers to the log.
202Buffers are converted to base64-encoding then split across log lines. Trace logs
203are surrounded by 'begin' and 'end' tags.
204
205Ex. Invoking PW_TRACE_INSTANT with 'test1' and 'test2', then calling this
206function would produce this in the output logs:
207
208.. code:: sh
209
210  [TRACE] begin
211  [TRACE] data: BWdDMRoABWj52YMB
212  [TRACE] end
213
214Added dependencies
215------------------
216``pw_base64``
217``pw_log``
218``pw_ring_buffer``
219``pw_string``
220``pw_tokenizer``
221``pw_varint``
222
223--------------
224Python decoder
225--------------
226The python decoder can be used to convert the binary trace data into json data
227which can be viewed in chrome://tracing.
228
229``get_trace.py`` can be used for retrieveing trace data from devices which are
230using the trace_rpc_server.
231
232``trace_tokenized.py`` can be used to decode a binary file of trace data.
233
234--------
235Examples
236--------
237The examples all use `pw_trace` sample app to provide the trace data. Details
238for how to build, run, and decode the traces are included at the top of each
239example. This is early work, and is provided as an example of how different
240tracing concepts can look.
241
242Basic
243-----
244The basic example turns on tracing and dumps all trace output to a file provided
245on the command line.
246
247Trigger
248-------
249The trigger example demonstrates how a trace event can be used as a trigger to
250start and stop capturing a trace. The examples makes use of ``PW_TRACE_REF``
251and ``PW_TRACE_REF_DATA`` to specify a start and stop event for the capture.
252This can be useful if the trace buffer is small and you wish to capture a
253specific series of events.
254
255Filter
256------
257The filter example demonstrates how a callback can be used to filter which trace
258events get processed and saved. In this example all events from the processing
259task which don't have traceId equal to 3 are removed. Both the other task traces
260are not removed. This can be a useful feature while debugging as it limits the
261amount of events which get stored to the buffer, and only saves the events of
262interest.
263
264--------------------
265Snapshot integration
266--------------------
267Tokenized trace buffers can be captured to a ``pw.snapshot.Snapshot`` or
268``pw.trace.SnapshotTraceInfo`` proto in the ``trace_data`` field. The expected
269format is a de-ringed raw tokenized trace buffer, which can be retrieved via
270``pw::trace::DeringAndViewRawBuffer()``.
271
272``pw_trace_tokenized`` does not yet have Python tooling integration for
273interpretation of serialized snapshots with a populated ``trace_data`` field.
274