1 | //===-- ObjectFileELF.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_SOURCE_PLUGINS_OBJECTFILE_ELF_OBJECTFILEELF_H |
10 | #define LLDB_SOURCE_PLUGINS_OBJECTFILE_ELF_OBJECTFILEELF_H |
11 | |
12 | #include <cstdint> |
13 | |
14 | #include <optional> |
15 | #include <vector> |
16 | |
17 | #include "lldb/Symbol/ObjectFile.h" |
18 | #include "lldb/Utility/ArchSpec.h" |
19 | #include "lldb/Utility/FileSpec.h" |
20 | #include "lldb/Utility/UUID.h" |
21 | #include "lldb/lldb-private.h" |
22 | |
23 | #include "ELFHeader.h" |
24 | |
25 | struct ELFNote { |
26 | elf::elf_word n_namesz = 0; |
27 | elf::elf_word n_descsz = 0; |
28 | elf::elf_word n_type = 0; |
29 | |
30 | std::string n_name; |
31 | |
32 | ELFNote() = default; |
33 | |
34 | /// Parse an ELFNote entry from the given DataExtractor starting at position |
35 | /// \p offset. |
36 | /// |
37 | /// \param[in] data |
38 | /// The DataExtractor to read from. |
39 | /// |
40 | /// \param[in,out] offset |
41 | /// Pointer to an offset in the data. On return the offset will be |
42 | /// advanced by the number of bytes read. |
43 | /// |
44 | /// \return |
45 | /// True if the ELFRel entry was successfully read and false otherwise. |
46 | bool (const lldb_private::DataExtractor &data, lldb::offset_t *offset); |
47 | |
48 | size_t GetByteSize() const { |
49 | return 12 + llvm::alignTo(Value: n_namesz, Align: 4) + llvm::alignTo(Value: n_descsz, Align: 4); |
50 | } |
51 | }; |
52 | |
53 | /// \class ObjectFileELF |
54 | /// Generic ELF object file reader. |
55 | /// |
56 | /// This class provides a generic ELF (32/64 bit) reader plugin implementing |
57 | /// the ObjectFile protocol. |
58 | class ObjectFileELF : public lldb_private::ObjectFile { |
59 | public: |
60 | // Static Functions |
61 | static void Initialize(); |
62 | |
63 | static void Terminate(); |
64 | |
65 | static llvm::StringRef GetPluginNameStatic() { return "elf" ; } |
66 | |
67 | static llvm::StringRef GetPluginDescriptionStatic() { |
68 | return "ELF object file reader." ; |
69 | } |
70 | |
71 | static lldb_private::ObjectFile * |
72 | CreateInstance(const lldb::ModuleSP &module_sp, lldb::DataBufferSP data_sp, |
73 | lldb::offset_t data_offset, const lldb_private::FileSpec *file, |
74 | lldb::offset_t file_offset, lldb::offset_t length); |
75 | |
76 | static lldb_private::ObjectFile *CreateMemoryInstance( |
77 | const lldb::ModuleSP &module_sp, lldb::WritableDataBufferSP data_sp, |
78 | const lldb::ProcessSP &process_sp, lldb::addr_t ); |
79 | |
80 | static size_t GetModuleSpecifications(const lldb_private::FileSpec &file, |
81 | lldb::DataBufferSP &data_sp, |
82 | lldb::offset_t data_offset, |
83 | lldb::offset_t file_offset, |
84 | lldb::offset_t length, |
85 | lldb_private::ModuleSpecList &specs); |
86 | |
87 | static bool MagicBytesMatch(lldb::DataBufferSP &data_sp, lldb::addr_t offset, |
88 | lldb::addr_t length); |
89 | |
90 | // PluginInterface protocol |
91 | llvm::StringRef GetPluginName() override { return GetPluginNameStatic(); } |
92 | |
93 | // LLVM RTTI support |
94 | static char ID; |
95 | bool isA(const void *ClassID) const override { |
96 | return ClassID == &ID || ObjectFile::isA(ClassID); |
97 | } |
98 | static bool classof(const ObjectFile *obj) { return obj->isA(ClassID: &ID); } |
99 | |
100 | // ObjectFile Protocol. |
101 | bool () override; |
102 | |
103 | bool SetLoadAddress(lldb_private::Target &target, lldb::addr_t value, |
104 | bool value_is_offset) override; |
105 | |
106 | lldb::ByteOrder GetByteOrder() const override; |
107 | |
108 | bool IsExecutable() const override; |
109 | |
110 | uint32_t GetAddressByteSize() const override; |
111 | |
112 | lldb_private::AddressClass GetAddressClass(lldb::addr_t file_addr) override; |
113 | |
114 | void ParseSymtab(lldb_private::Symtab &symtab) override; |
115 | |
116 | bool IsStripped() override; |
117 | |
118 | void CreateSections(lldb_private::SectionList &unified_section_list) override; |
119 | |
120 | void Dump(lldb_private::Stream *s) override; |
121 | |
122 | lldb_private::ArchSpec GetArchitecture() override; |
123 | |
124 | lldb_private::UUID GetUUID() override; |
125 | |
126 | /// Return the contents of the .gnu_debuglink section, if the object file |
127 | /// contains it. |
128 | std::optional<lldb_private::FileSpec> GetDebugLink(); |
129 | |
130 | uint32_t GetDependentModules(lldb_private::FileSpecList &files) override; |
131 | |
132 | lldb_private::Address |
133 | GetImageInfoAddress(lldb_private::Target *target) override; |
134 | |
135 | lldb_private::Address GetEntryPointAddress() override; |
136 | |
137 | lldb_private::Address GetBaseAddress() override; |
138 | |
139 | ObjectFile::Type CalculateType() override; |
140 | |
141 | ObjectFile::Strata CalculateStrata() override; |
142 | |
143 | size_t ReadSectionData(lldb_private::Section *section, |
144 | lldb::offset_t section_offset, void *dst, |
145 | size_t dst_len) override; |
146 | |
147 | size_t (lldb_private::Section *section, |
148 | lldb_private::DataExtractor §ion_data) override; |
149 | |
150 | llvm::ArrayRef<elf::ELFProgramHeader> (); |
151 | lldb_private::DataExtractor (const elf::ELFProgramHeader &H); |
152 | |
153 | llvm::StringRef |
154 | StripLinkerSymbolAnnotations(llvm::StringRef symbol_name) const override; |
155 | |
156 | void RelocateSection(lldb_private::Section *section) override; |
157 | |
158 | protected: |
159 | |
160 | std::vector<LoadableData> |
161 | GetLoadableData(lldb_private::Target &target) override; |
162 | |
163 | static lldb::WritableDataBufferSP |
164 | MapFileDataWritable(const lldb_private::FileSpec &file, uint64_t Size, |
165 | uint64_t Offset); |
166 | |
167 | private: |
168 | ObjectFileELF(const lldb::ModuleSP &module_sp, lldb::DataBufferSP data_sp, |
169 | lldb::offset_t data_offset, const lldb_private::FileSpec *file, |
170 | lldb::offset_t offset, lldb::offset_t length); |
171 | |
172 | ObjectFileELF(const lldb::ModuleSP &module_sp, |
173 | lldb::DataBufferSP , |
174 | const lldb::ProcessSP &process_sp, lldb::addr_t ); |
175 | |
176 | typedef std::vector<elf::ELFProgramHeader> ; |
177 | |
178 | struct : public elf::ELFSectionHeader { |
179 | lldb_private::ConstString ; |
180 | }; |
181 | |
182 | typedef std::vector<ELFSectionHeaderInfo> ; |
183 | typedef SectionHeaderColl::iterator ; |
184 | typedef SectionHeaderColl::const_iterator ; |
185 | |
186 | struct ELFDynamicWithName { |
187 | elf::ELFDynamic symbol; |
188 | std::string name; |
189 | }; |
190 | typedef std::vector<ELFDynamicWithName> DynamicSymbolColl; |
191 | typedef DynamicSymbolColl::iterator DynamicSymbolCollIter; |
192 | typedef DynamicSymbolColl::const_iterator DynamicSymbolCollConstIter; |
193 | |
194 | /// An ordered map of file address to address class. Used on architectures |
195 | /// like Arm where there is an alternative ISA mode like Thumb. The container |
196 | /// is ordered so that it can be binary searched. |
197 | typedef std::map<lldb::addr_t, lldb_private::AddressClass> |
198 | FileAddressToAddressClassMap; |
199 | |
200 | /// Version of this reader common to all plugins based on this class. |
201 | static const uint32_t m_plugin_version = 1; |
202 | static const uint32_t g_core_uuid_magic; |
203 | |
204 | /// ELF file header. |
205 | elf::ELFHeader ; |
206 | |
207 | /// ELF build ID. |
208 | lldb_private::UUID m_uuid; |
209 | |
210 | /// ELF .gnu_debuglink file and crc data if available. |
211 | std::string m_gnu_debuglink_file; |
212 | uint32_t m_gnu_debuglink_crc = 0; |
213 | |
214 | /// Collection of program headers. |
215 | ProgramHeaderColl ; |
216 | |
217 | /// Collection of section headers. |
218 | SectionHeaderColl ; |
219 | |
220 | /// The file address of the .dynamic section. This can be found in the p_vaddr |
221 | /// of the PT_DYNAMIC program header. |
222 | lldb::addr_t m_dynamic_base_addr = LLDB_INVALID_ADDRESS; |
223 | |
224 | /// Collection of symbols from the dynamic table. |
225 | DynamicSymbolColl m_dynamic_symbols; |
226 | |
227 | /// Object file parsed from .gnu_debugdata section (\sa |
228 | /// GetGnuDebugDataObjectFile()) |
229 | std::shared_ptr<ObjectFileELF> m_gnu_debug_data_object_file; |
230 | |
231 | /// List of file specifications corresponding to the modules (shared |
232 | /// libraries) on which this object file depends. |
233 | mutable std::unique_ptr<lldb_private::FileSpecList> m_filespec_up; |
234 | |
235 | /// Cached value of the entry point for this module. |
236 | lldb_private::Address m_entry_point_address; |
237 | |
238 | /// The architecture detected from parsing elf file contents. |
239 | lldb_private::ArchSpec m_arch_spec; |
240 | |
241 | /// The address class for each symbol in the elf file |
242 | FileAddressToAddressClassMap m_address_class_map; |
243 | |
244 | /// Returns the index of the given section header. |
245 | size_t (const SectionHeaderCollIter &I); |
246 | |
247 | /// Returns the index of the given section header. |
248 | size_t (const SectionHeaderCollConstIter &I) const; |
249 | |
250 | // Parses the ELF program headers. |
251 | static size_t (ProgramHeaderColl &, |
252 | lldb_private::DataExtractor &object_data, |
253 | const elf::ELFHeader &); |
254 | |
255 | // Finds PT_NOTE segments and calculates their crc sum. |
256 | static uint32_t |
257 | (const ProgramHeaderColl &, |
258 | lldb_private::DataExtractor &data); |
259 | |
260 | /// Parses all section headers present in this object file and populates |
261 | /// m_program_headers. This method will compute the header list only once. |
262 | /// Returns true iff the headers have been successfully parsed. |
263 | bool (); |
264 | |
265 | /// Parses all section headers present in this object file and populates |
266 | /// m_section_headers. This method will compute the header list only once. |
267 | /// Returns the number of headers parsed. |
268 | size_t (); |
269 | |
270 | lldb::SectionType (const ELFSectionHeaderInfo &H) const; |
271 | |
272 | static void (lldb_private::DataExtractor &data, |
273 | uint64_t length, |
274 | lldb_private::ArchSpec &arch_spec); |
275 | |
276 | /// Parses the elf section headers and returns the uuid, debug link name, |
277 | /// crc, archspec. |
278 | static size_t (SectionHeaderColl &, |
279 | lldb_private::DataExtractor &object_data, |
280 | const elf::ELFHeader &, |
281 | lldb_private::UUID &uuid, |
282 | std::string &gnu_debuglink_file, |
283 | uint32_t &gnu_debuglink_crc, |
284 | lldb_private::ArchSpec &arch_spec); |
285 | |
286 | /// Scans the dynamic section and locates all dependent modules (shared |
287 | /// libraries) populating m_filespec_up. This method will compute the |
288 | /// dependent module list only once. Returns the number of dependent |
289 | /// modules parsed. |
290 | size_t ParseDependentModules(); |
291 | |
292 | /// Parses the dynamic symbol table and populates m_dynamic_symbols. The |
293 | /// vector retains the order as found in the object file. Returns the |
294 | /// number of dynamic symbols parsed. |
295 | size_t ParseDynamicSymbols(); |
296 | |
297 | /// Populates the symbol table with all non-dynamic linker symbols. This |
298 | /// method will parse the symbols only once. Returns the number of symbols |
299 | /// parsed and a map of address types (used by targets like Arm that have |
300 | /// an alternative ISA mode like Thumb). |
301 | std::pair<unsigned, FileAddressToAddressClassMap> |
302 | ParseSymbolTable(lldb_private::Symtab *symbol_table, lldb::user_id_t start_id, |
303 | lldb_private::Section *symtab); |
304 | |
305 | /// Helper routine for ParseSymbolTable(). |
306 | std::pair<unsigned, FileAddressToAddressClassMap> |
307 | (lldb_private::Symtab *symbol_table, lldb::user_id_t start_id, |
308 | lldb_private::SectionList *section_list, |
309 | const size_t num_symbols, |
310 | const lldb_private::DataExtractor &symtab_data, |
311 | const lldb_private::DataExtractor &strtab_data); |
312 | |
313 | /// Scans the relocation entries and adds a set of artificial symbols to the |
314 | /// given symbol table for each PLT slot. Returns the number of symbols |
315 | /// added. |
316 | unsigned (lldb_private::Symtab *symbol_table, |
317 | lldb::user_id_t start_id, |
318 | const ELFSectionHeaderInfo *rela_hdr, |
319 | lldb::user_id_t section_id); |
320 | |
321 | void ParseUnwindSymbols(lldb_private::Symtab *symbol_table, |
322 | lldb_private::DWARFCallFrameInfo *eh_frame); |
323 | |
324 | /// Relocates debug sections |
325 | unsigned (const elf::ELFSectionHeader *rel_hdr, |
326 | lldb::user_id_t rel_id, |
327 | lldb_private::Symtab *thetab); |
328 | |
329 | unsigned (lldb_private::Symtab *symtab, |
330 | const elf::ELFHeader *hdr, |
331 | const elf::ELFSectionHeader *rel_hdr, |
332 | const elf::ELFSectionHeader *symtab_hdr, |
333 | const elf::ELFSectionHeader *debug_hdr, |
334 | lldb_private::DataExtractor &rel_data, |
335 | lldb_private::DataExtractor &symtab_data, |
336 | lldb_private::DataExtractor &debug_data, |
337 | lldb_private::Section *rel_section); |
338 | |
339 | /// Loads the section name string table into m_shstr_data. Returns the |
340 | /// number of bytes constituting the table. |
341 | size_t (); |
342 | |
343 | /// Utility method for looking up a section given its name. Returns the |
344 | /// index of the corresponding section or zero if no section with the given |
345 | /// name can be found (note that section indices are always 1 based, and so |
346 | /// section index 0 is never valid). |
347 | lldb::user_id_t GetSectionIndexByName(const char *name); |
348 | |
349 | /// Returns the section header with the given id or NULL. |
350 | const ELFSectionHeaderInfo *(lldb::user_id_t id); |
351 | |
352 | /// \name ELF header dump routines |
353 | //@{ |
354 | static void (lldb_private::Stream *s, |
355 | const elf::ELFHeader &); |
356 | |
357 | static void (lldb_private::Stream *s, |
358 | unsigned char ei_data); |
359 | |
360 | static void (lldb_private::Stream *s, |
361 | elf::elf_half e_type); |
362 | //@} |
363 | |
364 | /// \name ELF program header dump routines |
365 | //@{ |
366 | void (lldb_private::Stream *s); |
367 | |
368 | static void (lldb_private::Stream *s, |
369 | const elf::ELFProgramHeader &ph); |
370 | |
371 | static void (lldb_private::Stream *s, |
372 | elf::elf_word p_type); |
373 | |
374 | static void (lldb_private::Stream *s, |
375 | elf::elf_word p_flags); |
376 | //@} |
377 | |
378 | /// \name ELF section header dump routines |
379 | //@{ |
380 | void (lldb_private::Stream *s); |
381 | |
382 | static void (lldb_private::Stream *s, |
383 | const ELFSectionHeaderInfo &sh); |
384 | |
385 | static void (lldb_private::Stream *s, |
386 | elf::elf_word sh_type); |
387 | |
388 | static void (lldb_private::Stream *s, |
389 | elf::elf_xword sh_flags); |
390 | //@} |
391 | |
392 | /// ELF dependent module dump routine. |
393 | void DumpDependentModules(lldb_private::Stream *s); |
394 | |
395 | /// ELF dump the .dynamic section |
396 | void DumpELFDynamic(lldb_private::Stream *s); |
397 | |
398 | const elf::ELFDynamic *FindDynamicSymbol(unsigned tag); |
399 | |
400 | unsigned PLTRelocationType(); |
401 | |
402 | static lldb_private::Status |
403 | (lldb_private::DataExtractor &data, |
404 | lldb_private::ArchSpec &arch_spec, |
405 | lldb_private::UUID &uuid); |
406 | |
407 | bool AnySegmentHasPhysicalAddress(); |
408 | |
409 | /// Takes the .gnu_debugdata and returns the decompressed object file that is |
410 | /// stored within that section. |
411 | /// |
412 | /// \returns either the decompressed object file stored within the |
413 | /// .gnu_debugdata section or \c nullptr if an error occured or if there's no |
414 | /// section with that name. |
415 | std::shared_ptr<ObjectFileELF> GetGnuDebugDataObjectFile(); |
416 | |
417 | /// Get the bytes that represent the .dynamic section. |
418 | /// |
419 | /// This function will fetch the data for the .dynamic section in an ELF file. |
420 | /// The PT_DYNAMIC program header will be used to extract the data and this |
421 | /// function will fall back to using the section headers if PT_DYNAMIC isn't |
422 | /// found. |
423 | /// |
424 | /// \return The bytes that represent the string table data or \c std::nullopt |
425 | /// if an error occured. |
426 | std::optional<lldb_private::DataExtractor> GetDynamicData(); |
427 | |
428 | /// Get the bytes that represent the dynamic string table data. |
429 | /// |
430 | /// This function will fetch the data for the string table in an ELF file. If |
431 | /// the ELF file is loaded from a file on disk, it will use the section |
432 | /// headers to extract the data and fall back to using the DT_STRTAB and |
433 | /// DT_STRSZ .dynamic entries. |
434 | /// |
435 | /// \return The bytes that represent the string table data or \c std::nullopt |
436 | /// if an error occured. |
437 | std::optional<lldb_private::DataExtractor> GetDynstrData(); |
438 | |
439 | /// Read the bytes pointed to by the \a dyn dynamic entry. |
440 | /// |
441 | /// ELFDynamic::d_ptr values contain file addresses if we load the ELF file |
442 | /// form a file on disk, or they contain load addresses if they were read |
443 | /// from memory. This function will correctly extract the data in both cases |
444 | /// if it is available. |
445 | /// |
446 | /// \param[in] dyn The dynamic entry to use to fetch the data from. |
447 | /// |
448 | /// \param[in] length The number of bytes to read. |
449 | /// |
450 | /// \param[in] offset The number of bytes to skip after the d_ptr value |
451 | /// before reading data. |
452 | /// |
453 | /// \return The bytes that represent the dynanic entries data or |
454 | /// \c std::nullopt if an error occured or the data is not available. |
455 | std::optional<lldb_private::DataExtractor> |
456 | ReadDataFromDynamic(const elf::ELFDynamic *dyn, uint64_t length, |
457 | uint64_t offset = 0); |
458 | |
459 | /// Get the bytes that represent the dynamic symbol table from the .dynamic |
460 | /// section from process memory. |
461 | /// |
462 | /// This functon uses the DT_SYMTAB value from the .dynamic section to read |
463 | /// the symbols table data from process memory. The number of symbols in the |
464 | /// symbol table is calculated by looking at the DT_HASH or DT_GNU_HASH |
465 | /// values as the symbol count isn't stored in the .dynamic section. |
466 | /// |
467 | /// \return The bytes that represent the symbol table data from the .dynamic |
468 | /// section or section headers or \c std::nullopt if an error |
469 | /// occured or if there is no dynamic symbol data available. |
470 | std::optional<lldb_private::DataExtractor> |
471 | GetDynsymDataFromDynamic(uint32_t &num_symbols); |
472 | |
473 | /// Get the number of symbols from the DT_HASH dynamic entry. |
474 | std::optional<uint32_t> GetNumSymbolsFromDynamicHash(); |
475 | |
476 | /// Get the number of symbols from the DT_GNU_HASH dynamic entry. |
477 | std::optional<uint32_t> GetNumSymbolsFromDynamicGnuHash(); |
478 | }; |
479 | |
480 | #endif // LLDB_SOURCE_PLUGINS_OBJECTFILE_ELF_OBJECTFILEELF_H |
481 | |