1 //===- DWARFDebugLine.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 LLVM_DEBUGINFO_DWARFDEBUGLINE_H
10 #define LLVM_DEBUGINFO_DWARFDEBUGLINE_H
11 
12 #include "llvm/ADT/Optional.h"
13 #include "llvm/ADT/StringRef.h"
14 #include "llvm/DebugInfo/DIContext.h"
15 #include "llvm/DebugInfo/DWARF/DWARFCompileUnit.h"
16 #include "llvm/DebugInfo/DWARF/DWARFDataExtractor.h"
17 #include "llvm/DebugInfo/DWARF/DWARFFormValue.h"
18 #include "llvm/DebugInfo/DWARF/DWARFRelocMap.h"
19 #include "llvm/DebugInfo/DWARF/DWARFTypeUnit.h"
20 #include "llvm/Support/MD5.h"
21 #include "llvm/Support/Path.h"
22 #include <cstdint>
23 #include <map>
24 #include <string>
25 #include <vector>
26 
27 namespace llvm {
28 
29 class DWARFUnit;
30 class raw_ostream;
31 
32 class DWARFDebugLine {
33 public:
34   struct FileNameEntry {
35     FileNameEntry() = default;
36 
37     DWARFFormValue Name;
38     uint64_t DirIdx = 0;
39     uint64_t ModTime = 0;
40     uint64_t Length = 0;
41     MD5::MD5Result Checksum;
42     DWARFFormValue Source;
43   };
44 
45   /// Tracks which optional content types are present in a DWARF file name
46   /// entry format.
47   struct ContentTypeTracker {
48     ContentTypeTracker() = default;
49 
50     /// Whether filename entries provide a modification timestamp.
51     bool HasModTime = false;
52     /// Whether filename entries provide a file size.
53     bool HasLength = false;
54     /// For v5, whether filename entries provide an MD5 checksum.
55     bool HasMD5 = false;
56     /// For v5, whether filename entries provide source text.
57     bool HasSource = false;
58 
59     /// Update tracked content types with \p ContentType.
60     void trackContentType(dwarf::LineNumberEntryFormat ContentType);
61   };
62 
63   struct Prologue {
64     Prologue();
65 
66     /// The size in bytes of the statement information for this compilation unit
67     /// (not including the total_length field itself).
68     uint64_t TotalLength;
69     /// Version, address size (starting in v5), and DWARF32/64 format; these
70     /// parameters affect interpretation of forms (used in the directory and
71     /// file tables starting with v5).
72     dwarf::FormParams FormParams;
73     /// The number of bytes following the prologue_length field to the beginning
74     /// of the first byte of the statement program itself.
75     uint64_t PrologueLength;
76     /// In v5, size in bytes of a segment selector.
77     uint8_t SegSelectorSize;
78     /// The size in bytes of the smallest target machine instruction. Statement
79     /// program opcodes that alter the address register first multiply their
80     /// operands by this value.
81     uint8_t MinInstLength;
82     /// The maximum number of individual operations that may be encoded in an
83     /// instruction.
84     uint8_t MaxOpsPerInst;
85     /// The initial value of theis_stmtregister.
86     uint8_t DefaultIsStmt;
87     /// This parameter affects the meaning of the special opcodes. See below.
88     int8_t LineBase;
89     /// This parameter affects the meaning of the special opcodes. See below.
90     uint8_t LineRange;
91     /// The number assigned to the first special opcode.
92     uint8_t OpcodeBase;
93     /// This tracks which optional file format content types are present.
94     ContentTypeTracker ContentTypes;
95     std::vector<uint8_t> StandardOpcodeLengths;
96     std::vector<DWARFFormValue> IncludeDirectories;
97     std::vector<FileNameEntry> FileNames;
98 
99     const dwarf::FormParams getFormParams() const { return FormParams; }
100     uint16_t getVersion() const { return FormParams.Version; }
101     uint8_t getAddressSize() const { return FormParams.AddrSize; }
102     bool isDWARF64() const { return FormParams.Format == dwarf::DWARF64; }
103 
104     uint32_t sizeofTotalLength() const { return isDWARF64() ? 12 : 4; }
105 
106     uint32_t sizeofPrologueLength() const { return isDWARF64() ? 8 : 4; }
107 
108     bool totalLengthIsValid() const;
109 
110     /// Length of the prologue in bytes.
111     uint32_t getLength() const {
112       return PrologueLength + sizeofTotalLength() + sizeof(getVersion()) +
113              sizeofPrologueLength();
114     }
115 
116     /// Length of the line table data in bytes (not including the prologue).
117     uint32_t getStatementTableLength() const {
118       return TotalLength + sizeofTotalLength() - getLength();
119     }
120 
121     int32_t getMaxLineIncrementForSpecialOpcode() const {
122       return LineBase + (int8_t)LineRange - 1;
123     }
124 
125     /// Get DWARF-version aware access to the file name entry at the provided
126     /// index.
127     const llvm::DWARFDebugLine::FileNameEntry &
128     getFileNameEntry(uint64_t Index) const;
129 
130     bool hasFileAtIndex(uint64_t FileIndex) const;
131 
132     bool
133     getFileNameByIndex(uint64_t FileIndex, StringRef CompDir,
134                        DILineInfoSpecifier::FileLineInfoKind Kind,
135                        std::string &Result,
136                        sys::path::Style Style = sys::path::Style::native) const;
137 
138     void clear();
139     void dump(raw_ostream &OS, DIDumpOptions DumpOptions) const;
140     Error parse(const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
141                 const DWARFContext &Ctx, const DWARFUnit *U = nullptr);
142   };
143 
144   /// Standard .debug_line state machine structure.
145   struct Row {
146     explicit Row(bool DefaultIsStmt = false);
147 
148     /// Called after a row is appended to the matrix.
149     void postAppend();
150     void reset(bool DefaultIsStmt);
151     void dump(raw_ostream &OS) const;
152 
153     static void dumpTableHeader(raw_ostream &OS);
154 
155     static bool orderByAddress(const Row &LHS, const Row &RHS) {
156       return std::tie(LHS.Address.SectionIndex, LHS.Address.Address) <
157              std::tie(RHS.Address.SectionIndex, RHS.Address.Address);
158     }
159 
160     /// The program-counter value corresponding to a machine instruction
161     /// generated by the compiler and section index pointing to the section
162     /// containg this PC. If relocation information is present then section
163     /// index is the index of the section which contains above address.
164     /// Otherwise this is object::SectionedAddress::Undef value.
165     object::SectionedAddress Address;
166     /// An unsigned integer indicating a source line number. Lines are numbered
167     /// beginning at 1. The compiler may emit the value 0 in cases where an
168     /// instruction cannot be attributed to any source line.
169     uint32_t Line;
170     /// An unsigned integer indicating a column number within a source line.
171     /// Columns are numbered beginning at 1. The value 0 is reserved to indicate
172     /// that a statement begins at the 'left edge' of the line.
173     uint16_t Column;
174     /// An unsigned integer indicating the identity of the source file
175     /// corresponding to a machine instruction.
176     uint16_t File;
177     /// An unsigned integer representing the DWARF path discriminator value
178     /// for this location.
179     uint32_t Discriminator;
180     /// An unsigned integer whose value encodes the applicable instruction set
181     /// architecture for the current instruction.
182     uint8_t Isa;
183     /// A boolean indicating that the current instruction is the beginning of a
184     /// statement.
185     uint8_t IsStmt : 1,
186         /// A boolean indicating that the current instruction is the
187         /// beginning of a basic block.
188         BasicBlock : 1,
189         /// A boolean indicating that the current address is that of the
190         /// first byte after the end of a sequence of target machine
191         /// instructions.
192         EndSequence : 1,
193         /// A boolean indicating that the current address is one (of possibly
194         /// many) where execution should be suspended for an entry breakpoint
195         /// of a function.
196         PrologueEnd : 1,
197         /// A boolean indicating that the current address is one (of possibly
198         /// many) where execution should be suspended for an exit breakpoint
199         /// of a function.
200         EpilogueBegin : 1;
201   };
202 
203   /// Represents a series of contiguous machine instructions. Line table for
204   /// each compilation unit may consist of multiple sequences, which are not
205   /// guaranteed to be in the order of ascending instruction address.
206   struct Sequence {
207     Sequence();
208 
209     /// Sequence describes instructions at address range [LowPC, HighPC)
210     /// and is described by line table rows [FirstRowIndex, LastRowIndex).
211     uint64_t LowPC;
212     uint64_t HighPC;
213     /// If relocation information is present then this is the index of the
214     /// section which contains above addresses. Otherwise this is
215     /// object::SectionedAddress::Undef value.
216     uint64_t SectionIndex;
217     unsigned FirstRowIndex;
218     unsigned LastRowIndex;
219     bool Empty;
220 
221     void reset();
222 
223     static bool orderByHighPC(const Sequence &LHS, const Sequence &RHS) {
224       return std::tie(LHS.SectionIndex, LHS.HighPC) <
225              std::tie(RHS.SectionIndex, RHS.HighPC);
226     }
227 
228     bool isValid() const {
229       return !Empty && (LowPC < HighPC) && (FirstRowIndex < LastRowIndex);
230     }
231 
232     bool containsPC(object::SectionedAddress PC) const {
233       return SectionIndex == PC.SectionIndex &&
234              (LowPC <= PC.Address && PC.Address < HighPC);
235     }
236   };
237 
238   struct LineTable {
239     LineTable();
240 
241     /// Represents an invalid row
242     const uint32_t UnknownRowIndex = UINT32_MAX;
243 
244     void appendRow(const DWARFDebugLine::Row &R) { Rows.push_back(R); }
245 
246     void appendSequence(const DWARFDebugLine::Sequence &S) {
247       Sequences.push_back(S);
248     }
249 
250     /// Returns the index of the row with file/line info for a given address,
251     /// or UnknownRowIndex if there is no such row.
252     uint32_t lookupAddress(object::SectionedAddress Address) const;
253 
254     bool lookupAddressRange(object::SectionedAddress Address, uint64_t Size,
255                             std::vector<uint32_t> &Result) const;
256 
257     bool hasFileAtIndex(uint64_t FileIndex) const {
258       return Prologue.hasFileAtIndex(FileIndex);
259     }
260 
261     /// Extracts filename by its index in filename table in prologue.
262     /// In Dwarf 4, the files are 1-indexed and the current compilation file
263     /// name is not represented in the list. In DWARF v5, the files are
264     /// 0-indexed and the primary source file has the index 0.
265     /// Returns true on success.
266     bool getFileNameByIndex(uint64_t FileIndex, StringRef CompDir,
267                             DILineInfoSpecifier::FileLineInfoKind Kind,
268                             std::string &Result) const {
269       return Prologue.getFileNameByIndex(FileIndex, CompDir, Kind, Result);
270     }
271 
272     /// Fills the Result argument with the file and line information
273     /// corresponding to Address. Returns true on success.
274     bool getFileLineInfoForAddress(object::SectionedAddress Address,
275                                    const char *CompDir,
276                                    DILineInfoSpecifier::FileLineInfoKind Kind,
277                                    DILineInfo &Result) const;
278 
279     void dump(raw_ostream &OS, DIDumpOptions DumpOptions) const;
280     void clear();
281 
282     /// Parse prologue and all rows.
283     Error parse(DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
284                 const DWARFContext &Ctx, const DWARFUnit *U,
285                 function_ref<void(Error)> RecoverableErrorCallback,
286                 raw_ostream *OS = nullptr);
287 
288     using RowVector = std::vector<Row>;
289     using RowIter = RowVector::const_iterator;
290     using SequenceVector = std::vector<Sequence>;
291     using SequenceIter = SequenceVector::const_iterator;
292 
293     struct Prologue Prologue;
294     RowVector Rows;
295     SequenceVector Sequences;
296 
297   private:
298     uint32_t findRowInSeq(const DWARFDebugLine::Sequence &Seq,
299                           object::SectionedAddress Address) const;
300     Optional<StringRef>
301     getSourceByIndex(uint64_t FileIndex,
302                      DILineInfoSpecifier::FileLineInfoKind Kind) const;
303 
304     uint32_t lookupAddressImpl(object::SectionedAddress Address) const;
305 
306     bool lookupAddressRangeImpl(object::SectionedAddress Address, uint64_t Size,
307                                 std::vector<uint32_t> &Result) const;
308   };
309 
310   const LineTable *getLineTable(uint64_t Offset) const;
311   Expected<const LineTable *>
312   getOrParseLineTable(DWARFDataExtractor &DebugLineData, uint64_t Offset,
313                       const DWARFContext &Ctx, const DWARFUnit *U,
314                       function_ref<void(Error)> RecoverableErrorCallback);
315 
316   /// Helper to allow for parsing of an entire .debug_line section in sequence.
317   class SectionParser {
318   public:
319     using cu_range = DWARFUnitVector::iterator_range;
320     using tu_range = DWARFUnitVector::iterator_range;
321     using LineToUnitMap = std::map<uint64_t, DWARFUnit *>;
322 
323     SectionParser(DWARFDataExtractor &Data, const DWARFContext &C, cu_range CUs,
324                   tu_range TUs);
325 
326     /// Get the next line table from the section. Report any issues via the
327     /// callbacks.
328     ///
329     /// \param RecoverableErrorCallback - any issues that don't prevent further
330     /// parsing of the table will be reported through this callback.
331     /// \param UnrecoverableErrorCallback - any issues that prevent further
332     /// parsing of the table will be reported through this callback.
333     /// \param OS - if not null, the parser will print information about the
334     /// table as it parses it.
335     LineTable
336     parseNext(
337         function_ref<void(Error)> RecoverableErrorCallback,
338         function_ref<void(Error)> UnrecoverableErrorCallback,
339         raw_ostream *OS = nullptr);
340 
341     /// Skip the current line table and go to the following line table (if
342     /// present) immediately.
343     ///
344     /// \param ErrorCallback - report any prologue parsing issues via this
345     /// callback.
346     void skip(function_ref<void(Error)> ErrorCallback);
347 
348     /// Indicates if the parser has parsed as much as possible.
349     ///
350     /// \note Certain problems with the line table structure might mean that
351     /// parsing stops before the end of the section is reached.
352     bool done() const { return Done; }
353 
354     /// Get the offset the parser has reached.
355     uint64_t getOffset() const { return Offset; }
356 
357   private:
358     DWARFUnit *prepareToParse(uint64_t Offset);
359     void moveToNextTable(uint64_t OldOffset, const Prologue &P);
360 
361     LineToUnitMap LineToUnit;
362 
363     DWARFDataExtractor &DebugLineData;
364     const DWARFContext &Context;
365     uint64_t Offset = 0;
366     bool Done = false;
367   };
368 
369 private:
370   struct ParsingState {
371     ParsingState(struct LineTable *LT);
372 
373     void resetRowAndSequence();
374     void appendRowToMatrix();
375 
376     /// Line table we're currently parsing.
377     struct LineTable *LineTable;
378     struct Row Row;
379     struct Sequence Sequence;
380   };
381 
382   using LineTableMapTy = std::map<uint64_t, LineTable>;
383   using LineTableIter = LineTableMapTy::iterator;
384   using LineTableConstIter = LineTableMapTy::const_iterator;
385 
386   LineTableMapTy LineTableMap;
387 };
388 
389 } // end namespace llvm
390 
391 #endif // LLVM_DEBUGINFO_DWARFDEBUGLINE_H
392