• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //===- FunctionInfo.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_GSYM_FUNCTIONINFO_H
10 #define LLVM_DEBUGINFO_GSYM_FUNCTIONINFO_H
11 
12 #include "llvm/ADT/SmallString.h"
13 #include "llvm/DebugInfo/GSYM/ExtractRanges.h"
14 #include "llvm/DebugInfo/GSYM/InlineInfo.h"
15 #include "llvm/DebugInfo/GSYM/LineTable.h"
16 #include "llvm/DebugInfo/GSYM/LookupResult.h"
17 #include "llvm/DebugInfo/GSYM/MergedFunctionsInfo.h"
18 #include "llvm/DebugInfo/GSYM/StringTable.h"
19 #include <cstdint>
20 
21 namespace llvm {
22 class raw_ostream;
23 
24 namespace gsym {
25 
26 class GsymReader;
27 /// Function information in GSYM files encodes information for one contiguous
28 /// address range. If a function has discontiguous address ranges, they will
29 /// need to be encoded using multiple FunctionInfo objects.
30 ///
31 /// ENCODING
32 ///
33 /// The function information gets the function start address as an argument
34 /// to the FunctionInfo::decode(...) function. This information is calculated
35 /// from the GSYM header and an address offset from the GSYM address offsets
36 /// table. The encoded FunctionInfo information must be aligned to a 4 byte
37 /// boundary.
38 ///
39 /// The encoded data for a FunctionInfo starts with fixed data that all
40 /// function info objects have:
41 ///
42 /// ENCODING  NAME        DESCRIPTION
43 /// ========= =========== ====================================================
44 /// uint32_t  Size        The size in bytes of this function.
45 /// uint32_t  Name        The string table offset of the function name.
46 ///
47 /// The optional data in a FunctionInfo object follows this fixed information
48 /// and consists of a stream of tuples that consist of:
49 ///
50 /// ENCODING  NAME        DESCRIPTION
51 /// ========= =========== ====================================================
52 /// uint32_t  InfoType    An "InfoType" enumeration that describes the type
53 ///                       of optional data that is encoded.
54 /// uint32_t  InfoLength  The size in bytes of the encoded data that
55 ///                       immediately follows this length if this value is
56 ///                       greater than zero.
57 /// uint8_t[] InfoData    Encoded bytes that represent the data for the
58 ///                       "InfoType". These bytes are only present if
59 ///                       "InfoLength" is greater than zero.
60 ///
61 /// The "InfoType" is an enumeration:
62 ///
63 ///   enum InfoType {
64 ///     EndOfList = 0u,
65 ///     LineTableInfo = 1u,
66 ///     InlineInfo = 2u
67 ///   };
68 ///
69 /// This stream of tuples is terminated by a "InfoType" whose value is
70 /// InfoType::EndOfList and a zero for "InfoLength". This signifies the end of
71 /// the optional information list. This format allows us to add new optional
72 /// information data to a FunctionInfo object over time and allows older
73 /// clients to still parse the format and skip over any data that they don't
74 /// understand or want to parse.
75 ///
76 /// So the function information encoding essientially looks like:
77 ///
78 /// struct {
79 ///   uint32_t Size;
80 ///   uint32_t Name;
81 ///   struct {
82 ///     uint32_t InfoType;
83 ///     uint32_t InfoLength;
84 ///     uint8_t InfoData[InfoLength];
85 ///   }[N];
86 /// }
87 ///
88 /// Where "N" is the number of tuples.
89 struct FunctionInfo {
90   AddressRange Range;
91   uint32_t Name; ///< String table offset in the string table.
92   std::optional<LineTable> OptLineTable;
93   std::optional<InlineInfo> Inline;
94   std::optional<MergedFunctionsInfo> MergedFunctions;
95   /// If we encode a FunctionInfo during segmenting so we know its size, we can
96   /// cache that encoding here so we don't need to re-encode it when saving the
97   /// GSYM file.
98   SmallString<32> EncodingCache;
99 
100   FunctionInfo(uint64_t Addr = 0, uint64_t Size = 0, uint32_t N = 0)
101       : Range(Addr, Addr + Size), Name(N) {}
102 
103   /// Query if a FunctionInfo has rich debug info.
104   ///
105   /// \returns A bool that indicates if this object has something else than
106   /// range and name. When converting information from a symbol table and from
107   /// debug info, we might end up with multiple FunctionInfo objects for the
108   /// same range and we need to be able to tell which one is the better object
109   /// to use.
hasRichInfoFunctionInfo110   bool hasRichInfo() const { return OptLineTable || Inline; }
111 
112   /// Query if a FunctionInfo object is valid.
113   ///
114   /// Address and size can be zero and there can be no line entries for a
115   /// symbol so the only indication this entry is valid is if the name is
116   /// not zero. This can happen when extracting information from symbol
117   /// tables that do not encode symbol sizes. In that case only the
118   /// address and name will be filled in.
119   ///
120   /// \returns A boolean indicating if this FunctionInfo is valid.
isValidFunctionInfo121   bool isValid() const {
122     return Name != 0;
123   }
124 
125   /// Decode an object from a binary data stream.
126   ///
127   /// \param Data The binary stream to read the data from. This object must
128   /// have the data for the object starting at offset zero. The data
129   /// can contain more data than needed.
130   ///
131   /// \param BaseAddr The FunctionInfo's start address and will be used as the
132   /// base address when decoding any contained information like the line table
133   /// and the inline info.
134   ///
135   /// \returns An FunctionInfo or an error describing the issue that was
136   /// encountered during decoding.
137   static llvm::Expected<FunctionInfo> decode(DataExtractor &Data,
138                                              uint64_t BaseAddr);
139 
140   /// Encode this object into FileWriter stream.
141   ///
142   /// \param O The binary stream to write the data to at the current file
143   /// position.
144   ///
145   /// \param NoPadding Directly write the FunctionInfo data, without any padding
146   /// By default, FunctionInfo will be 4-byte aligned by padding with
147   /// 0's at the start. This is OK since the function will return the offset of
148   /// actual data in the stream. However when writing FunctionInfo's as a
149   /// stream, the padding will break the decoding of the data - since the offset
150   /// where the FunctionInfo starts is not kept in this scenario.
151   ///
152   /// \returns An error object that indicates failure or the offset of the
153   /// function info that was successfully written into the stream.
154   llvm::Expected<uint64_t> encode(FileWriter &O, bool NoPadding = false) const;
155 
156   /// Encode this function info into the internal byte cache and return the size
157   /// in bytes.
158   ///
159   /// When segmenting GSYM files we need to know how big each FunctionInfo will
160   /// encode into so we can generate segments of the right size. We don't want
161   /// to have to encode a FunctionInfo twice, so we can cache the encoded bytes
162   /// and re-use then when calling FunctionInfo::encode(...).
163   ///
164   /// \returns The size in bytes of the FunctionInfo if it were to be encoded
165   /// into a byte stream.
166   uint64_t cacheEncoding();
167 
168   /// Lookup an address within a FunctionInfo object's data stream.
169   ///
170   /// Instead of decoding an entire FunctionInfo object when doing lookups,
171   /// we can decode only the information we need from the FunctionInfo's data
172   /// for the specific address. The lookup result information is returned as
173   /// a LookupResult.
174   ///
175   /// \param Data The binary stream to read the data from. This object must
176   /// have the data for the object starting at offset zero. The data
177   /// can contain more data than needed.
178   ///
179   /// \param GR The GSYM reader that contains the string and file table that
180   /// will be used to fill in information in the returned result.
181   ///
182   /// \param FuncAddr The function start address decoded from the GsymReader.
183   ///
184   /// \param Addr The address to lookup.
185   ///
186   /// \returns An LookupResult or an error describing the issue that was
187   /// encountered during decoding. An error should only be returned if the
188   /// address is not contained in the FunctionInfo or if the data is corrupted.
189   static llvm::Expected<LookupResult> lookup(DataExtractor &Data,
190                                              const GsymReader &GR,
191                                              uint64_t FuncAddr,
192                                              uint64_t Addr);
193 
startAddressFunctionInfo194   uint64_t startAddress() const { return Range.start(); }
endAddressFunctionInfo195   uint64_t endAddress() const { return Range.end(); }
sizeFunctionInfo196   uint64_t size() const { return Range.size(); }
197 
clearFunctionInfo198   void clear() {
199     Range = {0, 0};
200     Name = 0;
201     OptLineTable = std::nullopt;
202     Inline = std::nullopt;
203   }
204 };
205 
206 inline bool operator==(const FunctionInfo &LHS, const FunctionInfo &RHS) {
207   return LHS.Range == RHS.Range && LHS.Name == RHS.Name &&
208          LHS.OptLineTable == RHS.OptLineTable && LHS.Inline == RHS.Inline;
209 }
210 inline bool operator!=(const FunctionInfo &LHS, const FunctionInfo &RHS) {
211   return !(LHS == RHS);
212 }
213 /// This sorting will order things consistently by address range first, but
214 /// then followed by increasing levels of debug info like inline information
215 /// and line tables. We might end up with a FunctionInfo from debug info that
216 /// will have the same range as one from the symbol table, but we want to
217 /// quickly be able to sort and use the best version when creating the final
218 /// GSYM file. This function compares the inline information as we have seen
219 /// cases where LTO can generate a wide array of differing inline information,
220 /// mostly due to messing up the address ranges for inlined functions, so the
221 /// inline information with the most entries will appeear last. If the inline
222 /// information match, either by both function infos not having any or both
223 /// being exactly the same, we will then compare line tables. Comparing line
224 /// tables allows the entry with the most line entries to appear last. This
225 /// ensures we are able to save the FunctionInfo with the most debug info into
226 /// the GSYM file.
227 inline bool operator<(const FunctionInfo &LHS, const FunctionInfo &RHS) {
228   // First sort by address range
229   if (LHS.Range != RHS.Range)
230     return LHS.Range < RHS.Range;
231   if (LHS.Inline == RHS.Inline)
232     return LHS.OptLineTable < RHS.OptLineTable;
233   return LHS.Inline < RHS.Inline;
234 }
235 
236 raw_ostream &operator<<(raw_ostream &OS, const FunctionInfo &R);
237 
238 } // namespace gsym
239 } // namespace llvm
240 
241 #endif // LLVM_DEBUGINFO_GSYM_FUNCTIONINFO_H
242