1 //===-- sanitizer_symbolizer_internal.h -------------------------*- C++ -*-===//
2 //
3 // This file is distributed under the University of Illinois Open Source
4 // License. See LICENSE.TXT for details.
5 //
6 //===----------------------------------------------------------------------===//
7 //
8 // Header for internal classes and functions to be used by implementations of
9 // symbolizers.
10 //
11 //===----------------------------------------------------------------------===//
12 #ifndef SANITIZER_SYMBOLIZER_INTERNAL_H
13 #define SANITIZER_SYMBOLIZER_INTERNAL_H
14 
15 #include "sanitizer_symbolizer.h"
16 #include "sanitizer_file.h"
17 
18 namespace __sanitizer {
19 
20 // Parsing helpers, 'str' is searched for delimiter(s) and a string or uptr
21 // is extracted. When extracting a string, a newly allocated (using
22 // InternalAlloc) and null-terminataed buffer is returned. They return a pointer
23 // to the next characted after the found delimiter.
24 const char *ExtractToken(const char *str, const char *delims, char **result);
25 const char *ExtractInt(const char *str, const char *delims, int *result);
26 const char *ExtractUptr(const char *str, const char *delims, uptr *result);
27 const char *ExtractTokenUpToDelimiter(const char *str, const char *delimiter,
28                                       char **result);
29 
30 const char *DemangleSwiftAndCXX(const char *name);
31 
32 // SymbolizerTool is an interface that is implemented by individual "tools"
33 // that can perform symbolication (external llvm-symbolizer, libbacktrace,
34 // Windows DbgHelp symbolizer, etc.).
35 class SymbolizerTool {
36  public:
37   // The main |Symbolizer| class implements a "fallback chain" of symbolizer
38   // tools. In a request to symbolize an address, if one tool returns false,
39   // the next tool in the chain will be tried.
40   SymbolizerTool *next;
41 
SymbolizerTool()42   SymbolizerTool() : next(nullptr) { }
43 
44   // Can't declare pure virtual functions in sanitizer runtimes:
45   // __cxa_pure_virtual might be unavailable.
46 
47   // The |stack| parameter is inout. It is pre-filled with the address,
48   // module base and module offset values and is to be used to construct
49   // other stack frames.
SymbolizePC(uptr addr,SymbolizedStack * stack)50   virtual bool SymbolizePC(uptr addr, SymbolizedStack *stack) {
51     UNIMPLEMENTED();
52   }
53 
54   // The |info| parameter is inout. It is pre-filled with the module base
55   // and module offset values.
SymbolizeData(uptr addr,DataInfo * info)56   virtual bool SymbolizeData(uptr addr, DataInfo *info) {
57     UNIMPLEMENTED();
58   }
59 
Flush()60   virtual void Flush() {}
61 
62   // Return nullptr to fallback to the default platform-specific demangler.
Demangle(const char * name)63   virtual const char *Demangle(const char *name) {
64     return nullptr;
65   }
66 };
67 
68 // SymbolizerProcess encapsulates communication between the tool and
69 // external symbolizer program, running in a different subprocess.
70 // SymbolizerProcess may not be used from two threads simultaneously.
71 class SymbolizerProcess {
72  public:
73   explicit SymbolizerProcess(const char *path, bool use_forkpty = false);
74   const char *SendCommand(const char *command);
75 
76  protected:
ReachedEndOfOutput(const char * buffer,uptr length)77   virtual bool ReachedEndOfOutput(const char *buffer, uptr length) const {
78     UNIMPLEMENTED();
79   }
80 
81   /// The maximum number of arguments required to invoke a tool process.
82   enum { kArgVMax = 6 };
83 
84   /// Fill in an argv array to invoke the child process.
GetArgV(const char * path_to_binary,const char * (& argv)[kArgVMax])85   virtual void GetArgV(const char *path_to_binary,
86                        const char *(&argv)[kArgVMax]) const {
87     UNIMPLEMENTED();
88   }
89 
90   virtual bool ReadFromSymbolizer(char *buffer, uptr max_length);
91 
92  private:
93   bool Restart();
94   const char *SendCommandImpl(const char *command);
95   bool WriteToSymbolizer(const char *buffer, uptr length);
96   bool StartSymbolizerSubprocess();
97 
98   const char *path_;
99   fd_t input_fd_;
100   fd_t output_fd_;
101 
102   static const uptr kBufferSize = 16 * 1024;
103   char buffer_[kBufferSize];
104 
105   static const uptr kMaxTimesRestarted = 5;
106   static const int kSymbolizerStartupTimeMillis = 10;
107   uptr times_restarted_;
108   bool failed_to_start_;
109   bool reported_invalid_path_;
110   bool use_forkpty_;
111 };
112 
113 class LLVMSymbolizerProcess;
114 
115 // This tool invokes llvm-symbolizer in a subprocess. It should be as portable
116 // as the llvm-symbolizer tool is.
117 class LLVMSymbolizer : public SymbolizerTool {
118  public:
119   explicit LLVMSymbolizer(const char *path, LowLevelAllocator *allocator);
120 
121   bool SymbolizePC(uptr addr, SymbolizedStack *stack) override;
122 
123   bool SymbolizeData(uptr addr, DataInfo *info) override;
124 
125  private:
126   const char *FormatAndSendCommand(bool is_data, const char *module_name,
127                                    uptr module_offset, ModuleArch arch);
128 
129   LLVMSymbolizerProcess *symbolizer_process_;
130   static const uptr kBufferSize = 16 * 1024;
131   char buffer_[kBufferSize];
132 };
133 
134 // Parses one or more two-line strings in the following format:
135 //   <function_name>
136 //   <file_name>:<line_number>[:<column_number>]
137 // Used by LLVMSymbolizer, Addr2LinePool and InternalSymbolizer, since all of
138 // them use the same output format.  Returns true if any useful debug
139 // information was found.
140 void ParseSymbolizePCOutput(const char *str, SymbolizedStack *res);
141 
142 // Parses a two-line string in the following format:
143 //   <symbol_name>
144 //   <start_address> <size>
145 // Used by LLVMSymbolizer and InternalSymbolizer.
146 void ParseSymbolizeDataOutput(const char *str, DataInfo *info);
147 
148 }  // namespace __sanitizer
149 
150 #endif  // SANITIZER_SYMBOLIZER_INTERNAL_H
151