1 //===-- Trace.h -------------------------------------------------*- C++ -*-===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
9 #ifndef LLDB_TARGET_TRACE_H
10 #define LLDB_TARGET_TRACE_H
11 
12 #include <optional>
13 #include <unordered_map>
14 
15 #include "llvm/Support/JSON.h"
16 
17 #include "lldb/Core/PluginInterface.h"
18 #include "lldb/Target/Thread.h"
19 #include "lldb/Target/TraceCursor.h"
20 #include "lldb/Utility/ArchSpec.h"
21 #include "lldb/Utility/TraceGDBRemotePackets.h"
22 #include "lldb/Utility/UnimplementedError.h"
23 #include "lldb/lldb-private.h"
24 #include "lldb/lldb-types.h"
25 
26 namespace lldb_private {
27 
28 /// \class Trace Trace.h "lldb/Target/Trace.h"
29 /// A plug-in interface definition class for trace information.
30 ///
31 /// Trace plug-ins allow processor trace information to be loaded into LLDB so
32 /// that the data can be dumped, used for reverse and forward stepping to allow
33 /// introspection into the reason your process crashed or found its way to its
34 /// current state.
35 ///
36 /// Trace information can be loaded into a target without a process to allow
37 /// introspection of the trace information during post mortem analysis, such as
38 /// when loading core files.
39 ///
40 /// Processor trace information can also be fetched through the process
41 /// interfaces during a live debug session if your process supports gathering
42 /// this information.
43 ///
44 /// In order to support live tracing, the name of the plug-in should match the
45 /// name of the tracing type returned by the gdb-remote packet
46 /// \a jLLDBTraceSupported.
47 class Trace : public PluginInterface,
48               public std::enable_shared_from_this<Trace> {
49 public:
50   /// Dump the trace data that this plug-in has access to.
51   ///
52   /// This function will dump all of the trace data for all threads in a user
53   /// readable format. Options for dumping can be added as this API is iterated
54   /// on.
55   ///
56   /// \param[in] s
57   ///     A stream object to dump the information to.
58   virtual void Dump(Stream *s) const = 0;
59 
60   /// Save the trace to the specified directory, which will be created if
61   /// needed. This will also create a file \a <directory>/trace.json with the
62   /// main properties of the trace session, along with others files which
63   /// contain the actual trace data. The trace.json file can be used later as
64   /// input for the "trace load" command to load the trace in LLDB.
65   ///
66   /// \param[in] directory
67   ///   The directory where the trace files will be saved.
68   ///
69   /// \param[in] compact
70   ///   Try not to save to disk information irrelevant to the traced processes.
71   ///   Each trace plug-in implements this in a different fashion.
72   ///
73   /// \return
74   ///   A \a FileSpec pointing to the bundle description file, or an \a
75   ///   llvm::Error otherwise.
76   virtual llvm::Expected<FileSpec> SaveToDisk(FileSpec directory,
77                                               bool compact) = 0;
78 
79   /// Find a trace plug-in using JSON data.
80   ///
81   /// When loading trace data from disk, the information for the trace data
82   /// can be contained in multiple files and require plug-in specific
83   /// information about the CPU. Using data like JSON provides an
84   /// easy way to specify all of the settings and information that we will need
85   /// to load trace data into LLDB. This structured data can include:
86   ///   - The plug-in name (this allows a specific plug-in to be selected)
87   ///   - Architecture or target triple
88   ///   - one or more paths to the trace data file on disk
89   ///     - cpu trace data
90   ///     - thread events or related information
91   ///   - shared library load information to use for this trace data that
92   ///     allows a target to be created so the trace information can be
93   ///     symbolicated so that the trace information can be displayed to the
94   ///     user
95   ///     - shared library path
96   ///     - load address
97   ///     - information on how to fetch the shared library
98   ///       - path to locally cached file on disk
99   ///       - URL to download the file
100   ///   - Any information needed to load the trace file
101   ///     - CPU information
102   ///     - Custom plug-in information needed to decode the trace information
103   ///       correctly.
104   ///
105   /// \param[in] debugger
106   ///     The debugger instance where new Targets will be created as part of the
107   ///     JSON data parsing.
108   ///
109   /// \param[in] bundle_description
110   ///     The trace bundle description object describing the trace session.
111   ///
112   /// \param[in] bundle_dir
113   ///     The path to the directory that contains the trace bundle.
114   static llvm::Expected<lldb::TraceSP>
115   FindPluginForPostMortemProcess(Debugger &debugger,
116                                  const llvm::json::Value &bundle_description,
117                                  llvm::StringRef session_file_dir);
118 
119   /// Find a trace plug-in to trace a live process.
120   ///
121   /// \param[in] plugin_name
122   ///     Plug-in name to search.
123   ///
124   /// \param[in] process
125   ///     Live process to trace.
126   ///
127   /// \return
128   ///     A \a TraceSP instance, or an \a llvm::Error if the plug-in name
129   ///     doesn't match any registered plug-ins or tracing couldn't be
130   ///     started.
131   static llvm::Expected<lldb::TraceSP>
132   FindPluginForLiveProcess(llvm::StringRef plugin_name, Process &process);
133 
134   /// Get the schema of a Trace plug-in given its name.
135   ///
136   /// \param[in] plugin_name
137   ///     Name of the trace plugin.
138   static llvm::Expected<llvm::StringRef>
139   FindPluginSchema(llvm::StringRef plugin_name);
140 
141   /// Load a trace from a trace description file and create Targets,
142   /// Processes and Threads based on the contents of such file.
143   ///
144   /// \param[in] debugger
145   ///     The debugger instance where new Targets will be created as part of the
146   ///     JSON data parsing.
147   ///
148   /// \param[in] trace_description_file
149   ///   The file containing the necessary information to load the trace.
150   ///
151   /// \return
152   ///     A \a TraceSP instance, or an \a llvm::Error if loading the trace
153   ///     fails.
154   static llvm::Expected<lldb::TraceSP>
155   LoadPostMortemTraceFromFile(Debugger &debugger,
156                               const FileSpec &trace_description_file);
157 
158   /// Get the command handle for the "process trace start" command.
159   virtual lldb::CommandObjectSP
160   GetProcessTraceStartCommand(CommandInterpreter &interpreter) = 0;
161 
162   /// Get the command handle for the "thread trace start" command.
163   virtual lldb::CommandObjectSP
164   GetThreadTraceStartCommand(CommandInterpreter &interpreter) = 0;
165 
166   /// \return
167   ///     The JSON schema of this Trace plug-in.
168   virtual llvm::StringRef GetSchema() = 0;
169 
170   /// Get a \a TraceCursor for the given thread's trace.
171   ///
172   /// \return
173   ///     A \a TraceCursorSP. If the thread is not traced or its trace
174   ///     information failed to load, an \a llvm::Error is returned.
175   virtual llvm::Expected<lldb::TraceCursorSP>
176   CreateNewCursor(Thread &thread) = 0;
177 
178   /// Dump general info about a given thread's trace. Each Trace plug-in
179   /// decides which data to show.
180   ///
181   /// \param[in] thread
182   ///     The thread that owns the trace in question.
183   ///
184   /// \param[in] s
185   ///     The stream object where the info will be printed printed.
186   ///
187   /// \param[in] verbose
188   ///     If \b true, print detailed info
189   ///     If \b false, print compact info
190   virtual void DumpTraceInfo(Thread &thread, Stream &s, bool verbose,
191                              bool json) = 0;
192 
193   /// Check if a thread is currently traced by this object.
194   ///
195   /// \param[in] tid
196   ///     The id of the thread in question.
197   ///
198   /// \return
199   ///     \b true if the thread is traced by this instance, \b false otherwise.
200   virtual bool IsTraced(lldb::tid_t tid) = 0;
201 
202   /// \return
203   ///     A description of the parameters to use for the \a Trace::Start method.
204   virtual const char *GetStartConfigurationHelp() = 0;
205 
206   /// Start tracing a live process.
207   ///
208   /// \param[in] configuration
209   ///     See \a SBTrace::Start(const lldb::SBStructuredData &) for more
210   ///     information.
211   ///
212   /// \return
213   ///     \a llvm::Error::success if the operation was successful, or
214   ///     \a llvm::Error otherwise.
215   virtual llvm::Error Start(
216       StructuredData::ObjectSP configuration = StructuredData::ObjectSP()) = 0;
217 
218   /// Start tracing live threads.
219   ///
220   /// \param[in] tids
221   ///     Threads to trace. This method tries to trace as many threads as
222   ///     possible.
223   ///
224   /// \param[in] configuration
225   ///     See \a SBTrace::Start(const lldb::SBThread &, const
226   ///     lldb::SBStructuredData &) for more information.
227   ///
228   /// \return
229   ///     \a llvm::Error::success if the operation was successful, or
230   ///     \a llvm::Error otherwise.
231   virtual llvm::Error Start(
232       llvm::ArrayRef<lldb::tid_t> tids,
233       StructuredData::ObjectSP configuration = StructuredData::ObjectSP()) = 0;
234 
235   /// Stop tracing live threads.
236   ///
237   /// \param[in] tids
238   ///     The threads to stop tracing on.
239   ///
240   /// \return
241   ///     \a llvm::Error::success if the operation was successful, or
242   ///     \a llvm::Error otherwise.
243   llvm::Error Stop(llvm::ArrayRef<lldb::tid_t> tids);
244 
245   /// Stop tracing all current and future threads of a live process.
246   ///
247   /// \param[in] request
248   ///     The information determining which threads or process to stop tracing.
249   ///
250   /// \return
251   ///     \a llvm::Error::success if the operation was successful, or
252   ///     \a llvm::Error otherwise.
253   llvm::Error Stop();
254 
255   /// \return
256   ///     The stop ID of the live process being traced, or an invalid stop ID
257   ///     if the trace is in an error or invalid state.
258   uint32_t GetStopID();
259 
260   using OnBinaryDataReadCallback =
261       std::function<llvm::Error(llvm::ArrayRef<uint8_t> data)>;
262   using OnCpusBinaryDataReadCallback = std::function<llvm::Error(
263       const llvm::DenseMap<lldb::cpu_id_t, llvm::ArrayRef<uint8_t>>
264           &cpu_to_data)>;
265 
266   /// Fetch binary data associated with a thread, either live or postmortem, and
267   /// pass it to the given callback. The reason of having a callback is to free
268   /// the caller from having to manage the life cycle of the data and to hide
269   /// the different data fetching procedures that exist for live and post mortem
270   /// threads.
271   ///
272   /// The fetched data is not persisted after the callback is invoked.
273   ///
274   /// \param[in] tid
275   ///     The tid who owns the data.
276   ///
277   /// \param[in] kind
278   ///     The kind of data to read.
279   ///
280   /// \param[in] callback
281   ///     The callback to be invoked once the data was successfully read. Its
282   ///     return value, which is an \a llvm::Error, is returned by this
283   ///     function.
284   ///
285   /// \return
286   ///     An \a llvm::Error if the data couldn't be fetched, or the return value
287   ///     of the callback, otherwise.
288   llvm::Error OnThreadBinaryDataRead(lldb::tid_t tid, llvm::StringRef kind,
289                                      OnBinaryDataReadCallback callback);
290 
291   /// Fetch binary data associated with a cpu, either live or postmortem, and
292   /// pass it to the given callback. The reason of having a callback is to free
293   /// the caller from having to manage the life cycle of the data and to hide
294   /// the different data fetching procedures that exist for live and post mortem
295   /// cpus.
296   ///
297   /// The fetched data is not persisted after the callback is invoked.
298   ///
299   /// \param[in] cpu_id
300   ///     The cpu who owns the data.
301   ///
302   /// \param[in] kind
303   ///     The kind of data to read.
304   ///
305   /// \param[in] callback
306   ///     The callback to be invoked once the data was successfully read. Its
307   ///     return value, which is an \a llvm::Error, is returned by this
308   ///     function.
309   ///
310   /// \return
311   ///     An \a llvm::Error if the data couldn't be fetched, or the return value
312   ///     of the callback, otherwise.
313   llvm::Error OnCpuBinaryDataRead(lldb::cpu_id_t cpu_id, llvm::StringRef kind,
314                                   OnBinaryDataReadCallback callback);
315 
316   /// Similar to \a OnCpuBinaryDataRead but this is able to fetch the same data
317   /// from all cpus at once.
318   llvm::Error OnAllCpusBinaryDataRead(llvm::StringRef kind,
319                                       OnCpusBinaryDataReadCallback callback);
320 
321   /// \return
322   ///     All the currently traced processes.
323   std::vector<Process *> GetAllProcesses();
324 
325   /// \return
326   ///     The list of cpus being traced. Might be empty depending on the
327   ///     plugin.
328   llvm::ArrayRef<lldb::cpu_id_t> GetTracedCpus();
329 
330   /// Helper method for reading a data file and passing its data to the given
331   /// callback.
332   static llvm::Error OnDataFileRead(FileSpec file,
333                                     OnBinaryDataReadCallback callback);
334 
335 protected:
336   /// Get the currently traced live process.
337   ///
338   /// \return
339   ///     If it's not a live process, return \a nullptr.
340   Process *GetLiveProcess();
341 
342   /// Get the currently traced postmortem processes.
343   ///
344   /// \return
345   ///     If it's not a live process session, return an empty list.
346   llvm::ArrayRef<Process *> GetPostMortemProcesses();
347 
348   /// Dispatcher for live trace data requests with some additional error
349   /// checking.
350   llvm::Expected<std::vector<uint8_t>>
351   GetLiveTraceBinaryData(const TraceGetBinaryDataRequest &request,
352                          uint64_t expected_size);
353 
354   /// Implementation of \a OnThreadBinaryDataRead() for live threads.
355   llvm::Error OnLiveThreadBinaryDataRead(lldb::tid_t tid, llvm::StringRef kind,
356                                          OnBinaryDataReadCallback callback);
357 
358   /// Implementation of \a OnLiveBinaryDataRead() for live cpus.
359   llvm::Error OnLiveCpuBinaryDataRead(lldb::cpu_id_t cpu, llvm::StringRef kind,
360                                       OnBinaryDataReadCallback callback);
361 
362   /// Implementation of \a OnThreadBinaryDataRead() for post mortem threads.
363   llvm::Error
364   OnPostMortemThreadBinaryDataRead(lldb::tid_t tid, llvm::StringRef kind,
365                                    OnBinaryDataReadCallback callback);
366 
367   /// Implementation of \a OnCpuBinaryDataRead() for post mortem cpus.
368   llvm::Error OnPostMortemCpuBinaryDataRead(lldb::cpu_id_t cpu_id,
369                                             llvm::StringRef kind,
370                                             OnBinaryDataReadCallback callback);
371 
372   /// Get the file path containing data of a postmortem thread given a data
373   /// identifier.
374   ///
375   /// \param[in] tid
376   ///     The thread whose data is requested.
377   ///
378   /// \param[in] kind
379   ///     The kind of data requested.
380   ///
381   /// \return
382   ///     The file spec containing the requested data, or an \a llvm::Error in
383   ///     case of failures.
384   llvm::Expected<FileSpec> GetPostMortemThreadDataFile(lldb::tid_t tid,
385                                                        llvm::StringRef kind);
386 
387   /// Get the file path containing data of a postmortem cpu given a data
388   /// identifier.
389   ///
390   /// \param[in] cpu_id
391   ///     The cpu whose data is requested.
392   ///
393   /// \param[in] kind
394   ///     The kind of data requested.
395   ///
396   /// \return
397   ///     The file spec containing the requested data, or an \a llvm::Error in
398   ///     case of failures.
399   llvm::Expected<FileSpec> GetPostMortemCpuDataFile(lldb::cpu_id_t cpu_id,
400                                                     llvm::StringRef kind);
401 
402   /// Associate a given thread with a data file using a data identifier.
403   ///
404   /// \param[in] tid
405   ///     The thread associated with the data file.
406   ///
407   /// \param[in] kind
408   ///     The kind of data being registered.
409   ///
410   /// \param[in] file_spec
411   ///     The path of the data file.
412   void SetPostMortemThreadDataFile(lldb::tid_t tid, llvm::StringRef kind,
413                                    FileSpec file_spec);
414 
415   /// Associate a given cpu with a data file using a data identifier.
416   ///
417   /// \param[in] cpu_id
418   ///     The cpu associated with the data file.
419   ///
420   /// \param[in] kind
421   ///     The kind of data being registered.
422   ///
423   /// \param[in] file_spec
424   ///     The path of the data file.
425   void SetPostMortemCpuDataFile(lldb::cpu_id_t cpu_id, llvm::StringRef kind,
426                                 FileSpec file_spec);
427 
428   /// Get binary data of a live thread given a data identifier.
429   ///
430   /// \param[in] tid
431   ///     The thread whose data is requested.
432   ///
433   /// \param[in] kind
434   ///     The kind of data requested.
435   ///
436   /// \return
437   ///     A vector of bytes with the requested data, or an \a llvm::Error in
438   ///     case of failures.
439   llvm::Expected<std::vector<uint8_t>>
440   GetLiveThreadBinaryData(lldb::tid_t tid, llvm::StringRef kind);
441 
442   /// Get binary data of a live cpu given a data identifier.
443   ///
444   /// \param[in] cpu_id
445   ///     The cpu whose data is requested.
446   ///
447   /// \param[in] kind
448   ///     The kind of data requested.
449   ///
450   /// \return
451   ///     A vector of bytes with the requested data, or an \a llvm::Error in
452   ///     case of failures.
453   llvm::Expected<std::vector<uint8_t>>
454   GetLiveCpuBinaryData(lldb::cpu_id_t cpu_id, llvm::StringRef kind);
455 
456   /// Get binary data of the current process given a data identifier.
457   ///
458   /// \param[in] kind
459   ///     The kind of data requested.
460   ///
461   /// \return
462   ///     A vector of bytes with the requested data, or an \a llvm::Error in
463   ///     case of failures.
464   llvm::Expected<std::vector<uint8_t>>
465   GetLiveProcessBinaryData(llvm::StringRef kind);
466 
467   /// Get the size of the data returned by \a GetLiveThreadBinaryData
468   std::optional<uint64_t> GetLiveThreadBinaryDataSize(lldb::tid_t tid,
469                                                       llvm::StringRef kind);
470 
471   /// Get the size of the data returned by \a GetLiveCpuBinaryData
472   std::optional<uint64_t> GetLiveCpuBinaryDataSize(lldb::cpu_id_t cpu_id,
473                                                    llvm::StringRef kind);
474 
475   /// Get the size of the data returned by \a GetLiveProcessBinaryData
476   std::optional<uint64_t> GetLiveProcessBinaryDataSize(llvm::StringRef kind);
477 
478   /// Constructor for post mortem processes
479   Trace(llvm::ArrayRef<lldb::ProcessSP> postmortem_processes,
480         std::optional<std::vector<lldb::cpu_id_t>> postmortem_cpus);
481 
482   /// Constructor for a live process
Trace(Process & live_process)483   Trace(Process &live_process) : m_live_process(&live_process) {}
484 
485   /// Start tracing a live process or its threads.
486   ///
487   /// \param[in] request
488   ///     JSON object with the information necessary to start tracing. In the
489   ///     case of gdb-remote processes, this JSON object should conform to the
490   ///     jLLDBTraceStart packet.
491   ///
492   /// \return
493   ///     \a llvm::Error::success if the operation was successful, or
494   ///     \a llvm::Error otherwise.
495   llvm::Error Start(const llvm::json::Value &request);
496 
497   /// Get the current tracing state of a live process and its threads.
498   ///
499   /// \return
500   ///     A JSON object string with custom data depending on the trace
501   ///     technology, or an \a llvm::Error in case of errors.
502   llvm::Expected<std::string> GetLiveProcessState();
503 
504   /// Method to be overriden by the plug-in to refresh its own state.
505   ///
506   /// This is invoked by RefreshLiveProcessState when a new state is found.
507   ///
508   /// \param[in] state
509   ///     The jLLDBTraceGetState response.
510   ///
511   /// \param[in] json_response
512   ///     The original JSON response as a string. It might be useful to redecode
513   ///     it if it contains custom data for a specific trace plug-in.
514   ///
515   /// \return
516   ///     \b Error::success() if this operation succeedes, or an actual error
517   ///     otherwise.
518   virtual llvm::Error
519   DoRefreshLiveProcessState(TraceGetStateResponse state,
520                             llvm::StringRef json_response) = 0;
521 
522   /// Return the list of processes traced by this instance. None of the returned
523   /// pointers are invalid.
524   std::vector<Process *> GetTracedProcesses();
525 
526   /// Method to be invoked by the plug-in to refresh the live process state. It
527   /// will invoked DoRefreshLiveProcessState at some point, which should be
528   /// implemented by the plug-in for custom state handling.
529   ///
530   /// The result is cached through the same process stop. Even in the case of
531   /// errors, it caches the error.
532   ///
533   /// \return
534   ///   An error message if this operation failed, or \b nullptr otherwise.
535   const char *RefreshLiveProcessState();
536 
537 private:
538   uint32_t m_stop_id = LLDB_INVALID_STOP_ID;
539 
540   /// Process traced by this object if doing live tracing. Otherwise it's null.
541   Process *m_live_process = nullptr;
542 
543   /// We package all the data that can change upon process stops to make sure
544   /// this contract is very visible.
545   /// This variable should only be accessed directly by constructores or live
546   /// process data refreshers.
547   struct Storage {
548     /// Portmortem processes traced by this object if doing non-live tracing.
549     /// Otherwise it's empty.
550     std::vector<Process *> postmortem_processes;
551 
552     /// These data kinds are returned by lldb-server when fetching the state of
553     /// the tracing session. The size in bytes can be used later for fetching
554     /// the data in batches.
555     /// \{
556 
557     /// tid -> data kind -> size
558     llvm::DenseMap<lldb::tid_t, llvm::DenseMap<ConstString, uint64_t>>
559         live_thread_data;
560 
561     /// cpu id -> data kind -> size
562     llvm::DenseMap<lldb::cpu_id_t, llvm::DenseMap<ConstString, uint64_t>>
563         live_cpu_data_sizes;
564     /// cpu id -> data kind -> bytes
565     llvm::DenseMap<lldb::cpu_id_t,
566                    llvm::DenseMap<ConstString, std::vector<uint8_t>>>
567         live_cpu_data;
568 
569     /// data kind -> size
570     llvm::DenseMap<ConstString, uint64_t> live_process_data;
571     /// \}
572 
573     /// The list of cpus being traced. Might be \b std::nullopt depending on the
574     /// plug-in.
575     std::optional<std::vector<lldb::cpu_id_t>> cpus;
576 
577     /// Postmortem traces can specific additional data files, which are
578     /// represented in this variable using a data kind identifier for each file.
579     /// \{
580 
581     /// tid -> data kind -> file
582     llvm::DenseMap<lldb::tid_t, llvm::DenseMap<ConstString, FileSpec>>
583         postmortem_thread_data;
584 
585     /// cpu id -> data kind -> file
586     llvm::DenseMap<lldb::cpu_id_t, llvm::DenseMap<ConstString, FileSpec>>
587         postmortem_cpu_data;
588 
589     /// \}
590 
591     std::optional<std::string> live_refresh_error;
592   } m_storage;
593 
594   /// Get the storage after refreshing the data in the case of a live process.
595   Storage &GetUpdatedStorage();
596 };
597 
598 } // namespace lldb_private
599 
600 #endif // LLDB_TARGET_TRACE_H
601