1 //===-- ObjectFileELF.cpp -------------------------------------------------===//
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 #include "ObjectFileELF.h"
10 
11 #include <algorithm>
12 #include <cassert>
13 #include <unordered_map>
14 
15 #include "lldb/Core/FileSpecList.h"
16 #include "lldb/Core/Module.h"
17 #include "lldb/Core/ModuleSpec.h"
18 #include "lldb/Core/PluginManager.h"
19 #include "lldb/Core/Section.h"
20 #include "lldb/Host/FileSystem.h"
21 #include "lldb/Host/LZMA.h"
22 #include "lldb/Symbol/DWARFCallFrameInfo.h"
23 #include "lldb/Symbol/SymbolContext.h"
24 #include "lldb/Target/SectionLoadList.h"
25 #include "lldb/Target/Target.h"
26 #include "lldb/Utility/ArchSpec.h"
27 #include "lldb/Utility/DataBufferHeap.h"
28 #include "lldb/Utility/Log.h"
29 #include "lldb/Utility/RangeMap.h"
30 #include "lldb/Utility/Status.h"
31 #include "lldb/Utility/Stream.h"
32 #include "lldb/Utility/Timer.h"
33 #include "llvm/ADT/IntervalMap.h"
34 #include "llvm/ADT/PointerUnion.h"
35 #include "llvm/ADT/StringRef.h"
36 #include "llvm/BinaryFormat/ELF.h"
37 #include "llvm/Object/Decompressor.h"
38 #include "llvm/Support/ARMBuildAttributes.h"
39 #include "llvm/Support/CRC.h"
40 #include "llvm/Support/MathExtras.h"
41 #include "llvm/Support/MemoryBuffer.h"
42 #include "llvm/Support/MipsABIFlags.h"
43 
44 #define CASE_AND_STREAM(s, def, width)                                         \
45   case def:                                                                    \
46     s->Printf("%-*s", width, #def);                                            \
47     break;
48 
49 using namespace lldb;
50 using namespace lldb_private;
51 using namespace elf;
52 using namespace llvm::ELF;
53 
54 LLDB_PLUGIN_DEFINE(ObjectFileELF)
55 
56 namespace {
57 
58 // ELF note owner definitions
59 const char *const LLDB_NT_OWNER_FREEBSD = "FreeBSD";
60 const char *const LLDB_NT_OWNER_GNU = "GNU";
61 const char *const LLDB_NT_OWNER_NETBSD = "NetBSD";
62 const char *const LLDB_NT_OWNER_NETBSDCORE = "NetBSD-CORE";
63 const char *const LLDB_NT_OWNER_OPENBSD = "OpenBSD";
64 const char *const LLDB_NT_OWNER_ANDROID = "Android";
65 const char *const LLDB_NT_OWNER_CORE = "CORE";
66 const char *const LLDB_NT_OWNER_LINUX = "LINUX";
67 
68 // ELF note type definitions
69 const elf_word LLDB_NT_FREEBSD_ABI_TAG = 0x01;
70 const elf_word LLDB_NT_FREEBSD_ABI_SIZE = 4;
71 
72 const elf_word LLDB_NT_GNU_ABI_TAG = 0x01;
73 const elf_word LLDB_NT_GNU_ABI_SIZE = 16;
74 
75 const elf_word LLDB_NT_GNU_BUILD_ID_TAG = 0x03;
76 
77 const elf_word LLDB_NT_NETBSD_IDENT_TAG = 1;
78 const elf_word LLDB_NT_NETBSD_IDENT_DESCSZ = 4;
79 const elf_word LLDB_NT_NETBSD_IDENT_NAMESZ = 7;
80 const elf_word LLDB_NT_NETBSD_PROCINFO = 1;
81 
82 // GNU ABI note OS constants
83 const elf_word LLDB_NT_GNU_ABI_OS_LINUX = 0x00;
84 const elf_word LLDB_NT_GNU_ABI_OS_HURD = 0x01;
85 const elf_word LLDB_NT_GNU_ABI_OS_SOLARIS = 0x02;
86 
87 //===----------------------------------------------------------------------===//
88 /// \class ELFRelocation
89 /// Generic wrapper for ELFRel and ELFRela.
90 ///
91 /// This helper class allows us to parse both ELFRel and ELFRela relocation
92 /// entries in a generic manner.
93 class ELFRelocation {
94 public:
95   /// Constructs an ELFRelocation entry with a personality as given by @p
96   /// type.
97   ///
98   /// \param type Either DT_REL or DT_RELA.  Any other value is invalid.
99   ELFRelocation(unsigned type);
100 
101   ~ELFRelocation();
102 
103   bool Parse(const lldb_private::DataExtractor &data, lldb::offset_t *offset);
104 
105   static unsigned RelocType32(const ELFRelocation &rel);
106 
107   static unsigned RelocType64(const ELFRelocation &rel);
108 
109   static unsigned RelocSymbol32(const ELFRelocation &rel);
110 
111   static unsigned RelocSymbol64(const ELFRelocation &rel);
112 
113   static unsigned RelocOffset32(const ELFRelocation &rel);
114 
115   static unsigned RelocOffset64(const ELFRelocation &rel);
116 
117   static unsigned RelocAddend32(const ELFRelocation &rel);
118 
119   static unsigned RelocAddend64(const ELFRelocation &rel);
120 
121 private:
122   typedef llvm::PointerUnion<ELFRel *, ELFRela *> RelocUnion;
123 
124   RelocUnion reloc;
125 };
126 
ELFRelocation(unsigned type)127 ELFRelocation::ELFRelocation(unsigned type) {
128   if (type == DT_REL || type == SHT_REL)
129     reloc = new ELFRel();
130   else if (type == DT_RELA || type == SHT_RELA)
131     reloc = new ELFRela();
132   else {
133     assert(false && "unexpected relocation type");
134     reloc = static_cast<ELFRel *>(nullptr);
135   }
136 }
137 
~ELFRelocation()138 ELFRelocation::~ELFRelocation() {
139   if (reloc.is<ELFRel *>())
140     delete reloc.get<ELFRel *>();
141   else
142     delete reloc.get<ELFRela *>();
143 }
144 
Parse(const lldb_private::DataExtractor & data,lldb::offset_t * offset)145 bool ELFRelocation::Parse(const lldb_private::DataExtractor &data,
146                           lldb::offset_t *offset) {
147   if (reloc.is<ELFRel *>())
148     return reloc.get<ELFRel *>()->Parse(data, offset);
149   else
150     return reloc.get<ELFRela *>()->Parse(data, offset);
151 }
152 
RelocType32(const ELFRelocation & rel)153 unsigned ELFRelocation::RelocType32(const ELFRelocation &rel) {
154   if (rel.reloc.is<ELFRel *>())
155     return ELFRel::RelocType32(*rel.reloc.get<ELFRel *>());
156   else
157     return ELFRela::RelocType32(*rel.reloc.get<ELFRela *>());
158 }
159 
RelocType64(const ELFRelocation & rel)160 unsigned ELFRelocation::RelocType64(const ELFRelocation &rel) {
161   if (rel.reloc.is<ELFRel *>())
162     return ELFRel::RelocType64(*rel.reloc.get<ELFRel *>());
163   else
164     return ELFRela::RelocType64(*rel.reloc.get<ELFRela *>());
165 }
166 
RelocSymbol32(const ELFRelocation & rel)167 unsigned ELFRelocation::RelocSymbol32(const ELFRelocation &rel) {
168   if (rel.reloc.is<ELFRel *>())
169     return ELFRel::RelocSymbol32(*rel.reloc.get<ELFRel *>());
170   else
171     return ELFRela::RelocSymbol32(*rel.reloc.get<ELFRela *>());
172 }
173 
RelocSymbol64(const ELFRelocation & rel)174 unsigned ELFRelocation::RelocSymbol64(const ELFRelocation &rel) {
175   if (rel.reloc.is<ELFRel *>())
176     return ELFRel::RelocSymbol64(*rel.reloc.get<ELFRel *>());
177   else
178     return ELFRela::RelocSymbol64(*rel.reloc.get<ELFRela *>());
179 }
180 
RelocOffset32(const ELFRelocation & rel)181 unsigned ELFRelocation::RelocOffset32(const ELFRelocation &rel) {
182   if (rel.reloc.is<ELFRel *>())
183     return rel.reloc.get<ELFRel *>()->r_offset;
184   else
185     return rel.reloc.get<ELFRela *>()->r_offset;
186 }
187 
RelocOffset64(const ELFRelocation & rel)188 unsigned ELFRelocation::RelocOffset64(const ELFRelocation &rel) {
189   if (rel.reloc.is<ELFRel *>())
190     return rel.reloc.get<ELFRel *>()->r_offset;
191   else
192     return rel.reloc.get<ELFRela *>()->r_offset;
193 }
194 
RelocAddend32(const ELFRelocation & rel)195 unsigned ELFRelocation::RelocAddend32(const ELFRelocation &rel) {
196   if (rel.reloc.is<ELFRel *>())
197     return 0;
198   else
199     return rel.reloc.get<ELFRela *>()->r_addend;
200 }
201 
RelocAddend64(const ELFRelocation & rel)202 unsigned ELFRelocation::RelocAddend64(const ELFRelocation &rel) {
203   if (rel.reloc.is<ELFRel *>())
204     return 0;
205   else
206     return rel.reloc.get<ELFRela *>()->r_addend;
207 }
208 
209 } // end anonymous namespace
210 
SegmentID(size_t PHdrIndex)211 static user_id_t SegmentID(size_t PHdrIndex) {
212   return ~user_id_t(PHdrIndex);
213 }
214 
Parse(const DataExtractor & data,lldb::offset_t * offset)215 bool ELFNote::Parse(const DataExtractor &data, lldb::offset_t *offset) {
216   // Read all fields.
217   if (data.GetU32(offset, &n_namesz, 3) == nullptr)
218     return false;
219 
220   // The name field is required to be nul-terminated, and n_namesz includes the
221   // terminating nul in observed implementations (contrary to the ELF-64 spec).
222   // A special case is needed for cores generated by some older Linux versions,
223   // which write a note named "CORE" without a nul terminator and n_namesz = 4.
224   if (n_namesz == 4) {
225     char buf[4];
226     if (data.ExtractBytes(*offset, 4, data.GetByteOrder(), buf) != 4)
227       return false;
228     if (strncmp(buf, "CORE", 4) == 0) {
229       n_name = "CORE";
230       *offset += 4;
231       return true;
232     }
233   }
234 
235   const char *cstr = data.GetCStr(offset, llvm::alignTo(n_namesz, 4));
236   if (cstr == nullptr) {
237     Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS));
238     LLDB_LOGF(log, "Failed to parse note name lacking nul terminator");
239 
240     return false;
241   }
242   n_name = cstr;
243   return true;
244 }
245 
mipsVariantFromElfFlags(const elf::ELFHeader & header)246 static uint32_t mipsVariantFromElfFlags (const elf::ELFHeader &header) {
247   const uint32_t mips_arch = header.e_flags & llvm::ELF::EF_MIPS_ARCH;
248   uint32_t endian = header.e_ident[EI_DATA];
249   uint32_t arch_variant = ArchSpec::eMIPSSubType_unknown;
250   uint32_t fileclass = header.e_ident[EI_CLASS];
251 
252   // If there aren't any elf flags available (e.g core elf file) then return
253   // default
254   // 32 or 64 bit arch (without any architecture revision) based on object file's class.
255   if (header.e_type == ET_CORE) {
256     switch (fileclass) {
257     case llvm::ELF::ELFCLASS32:
258       return (endian == ELFDATA2LSB) ? ArchSpec::eMIPSSubType_mips32el
259                                      : ArchSpec::eMIPSSubType_mips32;
260     case llvm::ELF::ELFCLASS64:
261       return (endian == ELFDATA2LSB) ? ArchSpec::eMIPSSubType_mips64el
262                                      : ArchSpec::eMIPSSubType_mips64;
263     default:
264       return arch_variant;
265     }
266   }
267 
268   switch (mips_arch) {
269   case llvm::ELF::EF_MIPS_ARCH_1:
270   case llvm::ELF::EF_MIPS_ARCH_2:
271   case llvm::ELF::EF_MIPS_ARCH_32:
272     return (endian == ELFDATA2LSB) ? ArchSpec::eMIPSSubType_mips32el
273                                    : ArchSpec::eMIPSSubType_mips32;
274   case llvm::ELF::EF_MIPS_ARCH_32R2:
275     return (endian == ELFDATA2LSB) ? ArchSpec::eMIPSSubType_mips32r2el
276                                    : ArchSpec::eMIPSSubType_mips32r2;
277   case llvm::ELF::EF_MIPS_ARCH_32R6:
278     return (endian == ELFDATA2LSB) ? ArchSpec::eMIPSSubType_mips32r6el
279                                    : ArchSpec::eMIPSSubType_mips32r6;
280   case llvm::ELF::EF_MIPS_ARCH_3:
281   case llvm::ELF::EF_MIPS_ARCH_4:
282   case llvm::ELF::EF_MIPS_ARCH_5:
283   case llvm::ELF::EF_MIPS_ARCH_64:
284     return (endian == ELFDATA2LSB) ? ArchSpec::eMIPSSubType_mips64el
285                                    : ArchSpec::eMIPSSubType_mips64;
286   case llvm::ELF::EF_MIPS_ARCH_64R2:
287     return (endian == ELFDATA2LSB) ? ArchSpec::eMIPSSubType_mips64r2el
288                                    : ArchSpec::eMIPSSubType_mips64r2;
289   case llvm::ELF::EF_MIPS_ARCH_64R6:
290     return (endian == ELFDATA2LSB) ? ArchSpec::eMIPSSubType_mips64r6el
291                                    : ArchSpec::eMIPSSubType_mips64r6;
292   default:
293     break;
294   }
295 
296   return arch_variant;
297 }
298 
subTypeFromElfHeader(const elf::ELFHeader & header)299 static uint32_t subTypeFromElfHeader(const elf::ELFHeader &header) {
300   if (header.e_machine == llvm::ELF::EM_MIPS)
301     return mipsVariantFromElfFlags(header);
302 
303   return LLDB_INVALID_CPUTYPE;
304 }
305 
306 char ObjectFileELF::ID;
307 
308 // Arbitrary constant used as UUID prefix for core files.
309 const uint32_t ObjectFileELF::g_core_uuid_magic(0xE210C);
310 
311 // Static methods.
Initialize()312 void ObjectFileELF::Initialize() {
313   PluginManager::RegisterPlugin(GetPluginNameStatic(),
314                                 GetPluginDescriptionStatic(), CreateInstance,
315                                 CreateMemoryInstance, GetModuleSpecifications);
316 }
317 
Terminate()318 void ObjectFileELF::Terminate() {
319   PluginManager::UnregisterPlugin(CreateInstance);
320 }
321 
GetPluginNameStatic()322 lldb_private::ConstString ObjectFileELF::GetPluginNameStatic() {
323   static ConstString g_name("elf");
324   return g_name;
325 }
326 
GetPluginDescriptionStatic()327 const char *ObjectFileELF::GetPluginDescriptionStatic() {
328   return "ELF object file reader.";
329 }
330 
CreateInstance(const lldb::ModuleSP & module_sp,DataBufferSP & data_sp,lldb::offset_t data_offset,const lldb_private::FileSpec * file,lldb::offset_t file_offset,lldb::offset_t length)331 ObjectFile *ObjectFileELF::CreateInstance(const lldb::ModuleSP &module_sp,
332                                           DataBufferSP &data_sp,
333                                           lldb::offset_t data_offset,
334                                           const lldb_private::FileSpec *file,
335                                           lldb::offset_t file_offset,
336                                           lldb::offset_t length) {
337   if (!data_sp) {
338     data_sp = MapFileData(*file, length, file_offset);
339     if (!data_sp)
340       return nullptr;
341     data_offset = 0;
342   }
343 
344   assert(data_sp);
345 
346   if (data_sp->GetByteSize() <= (llvm::ELF::EI_NIDENT + data_offset))
347     return nullptr;
348 
349   const uint8_t *magic = data_sp->GetBytes() + data_offset;
350   if (!ELFHeader::MagicBytesMatch(magic))
351     return nullptr;
352 
353   // Update the data to contain the entire file if it doesn't already
354   if (data_sp->GetByteSize() < length) {
355     data_sp = MapFileData(*file, length, file_offset);
356     if (!data_sp)
357       return nullptr;
358     data_offset = 0;
359     magic = data_sp->GetBytes();
360   }
361 
362   unsigned address_size = ELFHeader::AddressSizeInBytes(magic);
363   if (address_size == 4 || address_size == 8) {
364     std::unique_ptr<ObjectFileELF> objfile_up(new ObjectFileELF(
365         module_sp, data_sp, data_offset, file, file_offset, length));
366     ArchSpec spec = objfile_up->GetArchitecture();
367     if (spec && objfile_up->SetModulesArchitecture(spec))
368       return objfile_up.release();
369   }
370 
371   return nullptr;
372 }
373 
CreateMemoryInstance(const lldb::ModuleSP & module_sp,DataBufferSP & data_sp,const lldb::ProcessSP & process_sp,lldb::addr_t header_addr)374 ObjectFile *ObjectFileELF::CreateMemoryInstance(
375     const lldb::ModuleSP &module_sp, DataBufferSP &data_sp,
376     const lldb::ProcessSP &process_sp, lldb::addr_t header_addr) {
377   if (data_sp && data_sp->GetByteSize() > (llvm::ELF::EI_NIDENT)) {
378     const uint8_t *magic = data_sp->GetBytes();
379     if (ELFHeader::MagicBytesMatch(magic)) {
380       unsigned address_size = ELFHeader::AddressSizeInBytes(magic);
381       if (address_size == 4 || address_size == 8) {
382         std::unique_ptr<ObjectFileELF> objfile_up(
383             new ObjectFileELF(module_sp, data_sp, process_sp, header_addr));
384         ArchSpec spec = objfile_up->GetArchitecture();
385         if (spec && objfile_up->SetModulesArchitecture(spec))
386           return objfile_up.release();
387       }
388     }
389   }
390   return nullptr;
391 }
392 
MagicBytesMatch(DataBufferSP & data_sp,lldb::addr_t data_offset,lldb::addr_t data_length)393 bool ObjectFileELF::MagicBytesMatch(DataBufferSP &data_sp,
394                                     lldb::addr_t data_offset,
395                                     lldb::addr_t data_length) {
396   if (data_sp &&
397       data_sp->GetByteSize() > (llvm::ELF::EI_NIDENT + data_offset)) {
398     const uint8_t *magic = data_sp->GetBytes() + data_offset;
399     return ELFHeader::MagicBytesMatch(magic);
400   }
401   return false;
402 }
403 
calc_crc32(uint32_t init,const DataExtractor & data)404 static uint32_t calc_crc32(uint32_t init, const DataExtractor &data) {
405   return llvm::crc32(
406       init, llvm::makeArrayRef(data.GetDataStart(), data.GetByteSize()));
407 }
408 
CalculateELFNotesSegmentsCRC32(const ProgramHeaderColl & program_headers,DataExtractor & object_data)409 uint32_t ObjectFileELF::CalculateELFNotesSegmentsCRC32(
410     const ProgramHeaderColl &program_headers, DataExtractor &object_data) {
411 
412   uint32_t core_notes_crc = 0;
413 
414   for (const ELFProgramHeader &H : program_headers) {
415     if (H.p_type == llvm::ELF::PT_NOTE) {
416       const elf_off ph_offset = H.p_offset;
417       const size_t ph_size = H.p_filesz;
418 
419       DataExtractor segment_data;
420       if (segment_data.SetData(object_data, ph_offset, ph_size) != ph_size) {
421         // The ELF program header contained incorrect data, probably corefile
422         // is incomplete or corrupted.
423         break;
424       }
425 
426       core_notes_crc = calc_crc32(core_notes_crc, segment_data);
427     }
428   }
429 
430   return core_notes_crc;
431 }
432 
OSABIAsCString(unsigned char osabi_byte)433 static const char *OSABIAsCString(unsigned char osabi_byte) {
434 #define _MAKE_OSABI_CASE(x)                                                    \
435   case x:                                                                      \
436     return #x
437   switch (osabi_byte) {
438     _MAKE_OSABI_CASE(ELFOSABI_NONE);
439     _MAKE_OSABI_CASE(ELFOSABI_HPUX);
440     _MAKE_OSABI_CASE(ELFOSABI_NETBSD);
441     _MAKE_OSABI_CASE(ELFOSABI_GNU);
442     _MAKE_OSABI_CASE(ELFOSABI_HURD);
443     _MAKE_OSABI_CASE(ELFOSABI_SOLARIS);
444     _MAKE_OSABI_CASE(ELFOSABI_AIX);
445     _MAKE_OSABI_CASE(ELFOSABI_IRIX);
446     _MAKE_OSABI_CASE(ELFOSABI_FREEBSD);
447     _MAKE_OSABI_CASE(ELFOSABI_TRU64);
448     _MAKE_OSABI_CASE(ELFOSABI_MODESTO);
449     _MAKE_OSABI_CASE(ELFOSABI_OPENBSD);
450     _MAKE_OSABI_CASE(ELFOSABI_OPENVMS);
451     _MAKE_OSABI_CASE(ELFOSABI_NSK);
452     _MAKE_OSABI_CASE(ELFOSABI_AROS);
453     _MAKE_OSABI_CASE(ELFOSABI_FENIXOS);
454     _MAKE_OSABI_CASE(ELFOSABI_C6000_ELFABI);
455     _MAKE_OSABI_CASE(ELFOSABI_C6000_LINUX);
456     _MAKE_OSABI_CASE(ELFOSABI_ARM);
457     _MAKE_OSABI_CASE(ELFOSABI_STANDALONE);
458   default:
459     return "<unknown-osabi>";
460   }
461 #undef _MAKE_OSABI_CASE
462 }
463 
464 //
465 // WARNING : This function is being deprecated
466 // It's functionality has moved to ArchSpec::SetArchitecture This function is
467 // only being kept to validate the move.
468 //
469 // TODO : Remove this function
GetOsFromOSABI(unsigned char osabi_byte,llvm::Triple::OSType & ostype)470 static bool GetOsFromOSABI(unsigned char osabi_byte,
471                            llvm::Triple::OSType &ostype) {
472   switch (osabi_byte) {
473   case ELFOSABI_AIX:
474     ostype = llvm::Triple::OSType::AIX;
475     break;
476   case ELFOSABI_FREEBSD:
477     ostype = llvm::Triple::OSType::FreeBSD;
478     break;
479   case ELFOSABI_GNU:
480     ostype = llvm::Triple::OSType::Linux;
481     break;
482   case ELFOSABI_NETBSD:
483     ostype = llvm::Triple::OSType::NetBSD;
484     break;
485   case ELFOSABI_OPENBSD:
486     ostype = llvm::Triple::OSType::OpenBSD;
487     break;
488   case ELFOSABI_SOLARIS:
489     ostype = llvm::Triple::OSType::Solaris;
490     break;
491   default:
492     ostype = llvm::Triple::OSType::UnknownOS;
493   }
494   return ostype != llvm::Triple::OSType::UnknownOS;
495 }
496 
GetModuleSpecifications(const lldb_private::FileSpec & file,lldb::DataBufferSP & data_sp,lldb::offset_t data_offset,lldb::offset_t file_offset,lldb::offset_t length,lldb_private::ModuleSpecList & specs)497 size_t ObjectFileELF::GetModuleSpecifications(
498     const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp,
499     lldb::offset_t data_offset, lldb::offset_t file_offset,
500     lldb::offset_t length, lldb_private::ModuleSpecList &specs) {
501   Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_MODULES));
502 
503   const size_t initial_count = specs.GetSize();
504 
505   if (ObjectFileELF::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) {
506     DataExtractor data;
507     data.SetData(data_sp);
508     elf::ELFHeader header;
509     lldb::offset_t header_offset = data_offset;
510     if (header.Parse(data, &header_offset)) {
511       if (data_sp) {
512         ModuleSpec spec(file);
513 
514         const uint32_t sub_type = subTypeFromElfHeader(header);
515         spec.GetArchitecture().SetArchitecture(
516             eArchTypeELF, header.e_machine, sub_type, header.e_ident[EI_OSABI]);
517 
518         if (spec.GetArchitecture().IsValid()) {
519           llvm::Triple::OSType ostype;
520           llvm::Triple::VendorType vendor;
521           llvm::Triple::OSType spec_ostype =
522               spec.GetArchitecture().GetTriple().getOS();
523 
524           LLDB_LOGF(log, "ObjectFileELF::%s file '%s' module OSABI: %s",
525                     __FUNCTION__, file.GetPath().c_str(),
526                     OSABIAsCString(header.e_ident[EI_OSABI]));
527 
528           // SetArchitecture should have set the vendor to unknown
529           vendor = spec.GetArchitecture().GetTriple().getVendor();
530           assert(vendor == llvm::Triple::UnknownVendor);
531           UNUSED_IF_ASSERT_DISABLED(vendor);
532 
533           //
534           // Validate it is ok to remove GetOsFromOSABI
535           GetOsFromOSABI(header.e_ident[EI_OSABI], ostype);
536           assert(spec_ostype == ostype);
537           if (spec_ostype != llvm::Triple::OSType::UnknownOS) {
538             LLDB_LOGF(log,
539                       "ObjectFileELF::%s file '%s' set ELF module OS type "
540                       "from ELF header OSABI.",
541                       __FUNCTION__, file.GetPath().c_str());
542           }
543 
544           if (data_sp->GetByteSize() < length)
545             data_sp = MapFileData(file, -1, file_offset);
546           if (data_sp)
547             data.SetData(data_sp);
548           // In case there is header extension in the section #0, the header we
549           // parsed above could have sentinel values for e_phnum, e_shnum, and
550           // e_shstrndx.  In this case we need to reparse the header with a
551           // bigger data source to get the actual values.
552           if (header.HasHeaderExtension()) {
553             lldb::offset_t header_offset = data_offset;
554             header.Parse(data, &header_offset);
555           }
556 
557           uint32_t gnu_debuglink_crc = 0;
558           std::string gnu_debuglink_file;
559           SectionHeaderColl section_headers;
560           lldb_private::UUID &uuid = spec.GetUUID();
561 
562           GetSectionHeaderInfo(section_headers, data, header, uuid,
563                                gnu_debuglink_file, gnu_debuglink_crc,
564                                spec.GetArchitecture());
565 
566           llvm::Triple &spec_triple = spec.GetArchitecture().GetTriple();
567 
568           LLDB_LOGF(log,
569                     "ObjectFileELF::%s file '%s' module set to triple: %s "
570                     "(architecture %s)",
571                     __FUNCTION__, file.GetPath().c_str(),
572                     spec_triple.getTriple().c_str(),
573                     spec.GetArchitecture().GetArchitectureName());
574 
575           if (!uuid.IsValid()) {
576             uint32_t core_notes_crc = 0;
577 
578             if (!gnu_debuglink_crc) {
579               static Timer::Category func_cat(LLVM_PRETTY_FUNCTION);
580               lldb_private::Timer scoped_timer(
581                   func_cat,
582                   "Calculating module crc32 %s with size %" PRIu64 " KiB",
583                   file.GetLastPathComponent().AsCString(),
584                   (length - file_offset) / 1024);
585 
586               // For core files - which usually don't happen to have a
587               // gnu_debuglink, and are pretty bulky - calculating whole
588               // contents crc32 would be too much of luxury.  Thus we will need
589               // to fallback to something simpler.
590               if (header.e_type == llvm::ELF::ET_CORE) {
591                 ProgramHeaderColl program_headers;
592                 GetProgramHeaderInfo(program_headers, data, header);
593 
594                 core_notes_crc =
595                     CalculateELFNotesSegmentsCRC32(program_headers, data);
596               } else {
597                 gnu_debuglink_crc = calc_crc32(0, data);
598               }
599             }
600             using u32le = llvm::support::ulittle32_t;
601             if (gnu_debuglink_crc) {
602               // Use 4 bytes of crc from the .gnu_debuglink section.
603               u32le data(gnu_debuglink_crc);
604               uuid = UUID::fromData(&data, sizeof(data));
605             } else if (core_notes_crc) {
606               // Use 8 bytes - first 4 bytes for *magic* prefix, mainly to make
607               // it look different form .gnu_debuglink crc followed by 4 bytes
608               // of note segments crc.
609               u32le data[] = {u32le(g_core_uuid_magic), u32le(core_notes_crc)};
610               uuid = UUID::fromData(data, sizeof(data));
611             }
612           }
613 
614           specs.Append(spec);
615         }
616       }
617     }
618   }
619 
620   return specs.GetSize() - initial_count;
621 }
622 
623 // PluginInterface protocol
GetPluginName()624 lldb_private::ConstString ObjectFileELF::GetPluginName() {
625   return GetPluginNameStatic();
626 }
627 
GetPluginVersion()628 uint32_t ObjectFileELF::GetPluginVersion() { return m_plugin_version; }
629 // ObjectFile protocol
630 
ObjectFileELF(const lldb::ModuleSP & module_sp,DataBufferSP & data_sp,lldb::offset_t data_offset,const FileSpec * file,lldb::offset_t file_offset,lldb::offset_t length)631 ObjectFileELF::ObjectFileELF(const lldb::ModuleSP &module_sp,
632                              DataBufferSP &data_sp, lldb::offset_t data_offset,
633                              const FileSpec *file, lldb::offset_t file_offset,
634                              lldb::offset_t length)
635     : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset) {
636   if (file)
637     m_file = *file;
638 }
639 
ObjectFileELF(const lldb::ModuleSP & module_sp,DataBufferSP & header_data_sp,const lldb::ProcessSP & process_sp,addr_t header_addr)640 ObjectFileELF::ObjectFileELF(const lldb::ModuleSP &module_sp,
641                              DataBufferSP &header_data_sp,
642                              const lldb::ProcessSP &process_sp,
643                              addr_t header_addr)
644     : ObjectFile(module_sp, process_sp, header_addr, header_data_sp) {}
645 
IsExecutable() const646 bool ObjectFileELF::IsExecutable() const {
647   return ((m_header.e_type & ET_EXEC) != 0) || (m_header.e_entry != 0);
648 }
649 
SetLoadAddress(Target & target,lldb::addr_t value,bool value_is_offset)650 bool ObjectFileELF::SetLoadAddress(Target &target, lldb::addr_t value,
651                                    bool value_is_offset) {
652   ModuleSP module_sp = GetModule();
653   if (module_sp) {
654     size_t num_loaded_sections = 0;
655     SectionList *section_list = GetSectionList();
656     if (section_list) {
657       if (!value_is_offset) {
658         addr_t base = GetBaseAddress().GetFileAddress();
659         if (base == LLDB_INVALID_ADDRESS)
660           return false;
661         value -= base;
662       }
663 
664       const size_t num_sections = section_list->GetSize();
665       size_t sect_idx = 0;
666 
667       for (sect_idx = 0; sect_idx < num_sections; ++sect_idx) {
668         // Iterate through the object file sections to find all of the sections
669         // that have SHF_ALLOC in their flag bits.
670         SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx));
671         if (section_sp->Test(SHF_ALLOC) ||
672             section_sp->GetType() == eSectionTypeContainer) {
673           lldb::addr_t load_addr = section_sp->GetFileAddress();
674           // We don't want to update the load address of a section with type
675           // eSectionTypeAbsoluteAddress as they already have the absolute load
676           // address already specified
677           if (section_sp->GetType() != eSectionTypeAbsoluteAddress)
678             load_addr += value;
679 
680           // On 32-bit systems the load address have to fit into 4 bytes. The
681           // rest of the bytes are the overflow from the addition.
682           if (GetAddressByteSize() == 4)
683             load_addr &= 0xFFFFFFFF;
684 
685           if (target.GetSectionLoadList().SetSectionLoadAddress(section_sp,
686                                                                 load_addr))
687             ++num_loaded_sections;
688         }
689       }
690       return num_loaded_sections > 0;
691     }
692   }
693   return false;
694 }
695 
GetByteOrder() const696 ByteOrder ObjectFileELF::GetByteOrder() const {
697   if (m_header.e_ident[EI_DATA] == ELFDATA2MSB)
698     return eByteOrderBig;
699   if (m_header.e_ident[EI_DATA] == ELFDATA2LSB)
700     return eByteOrderLittle;
701   return eByteOrderInvalid;
702 }
703 
GetAddressByteSize() const704 uint32_t ObjectFileELF::GetAddressByteSize() const {
705   return m_data.GetAddressByteSize();
706 }
707 
GetAddressClass(addr_t file_addr)708 AddressClass ObjectFileELF::GetAddressClass(addr_t file_addr) {
709   Symtab *symtab = GetSymtab();
710   if (!symtab)
711     return AddressClass::eUnknown;
712 
713   // The address class is determined based on the symtab. Ask it from the
714   // object file what contains the symtab information.
715   ObjectFile *symtab_objfile = symtab->GetObjectFile();
716   if (symtab_objfile != nullptr && symtab_objfile != this)
717     return symtab_objfile->GetAddressClass(file_addr);
718 
719   auto res = ObjectFile::GetAddressClass(file_addr);
720   if (res != AddressClass::eCode)
721     return res;
722 
723   auto ub = m_address_class_map.upper_bound(file_addr);
724   if (ub == m_address_class_map.begin()) {
725     // No entry in the address class map before the address. Return default
726     // address class for an address in a code section.
727     return AddressClass::eCode;
728   }
729 
730   // Move iterator to the address class entry preceding address
731   --ub;
732 
733   return ub->second;
734 }
735 
SectionIndex(const SectionHeaderCollIter & I)736 size_t ObjectFileELF::SectionIndex(const SectionHeaderCollIter &I) {
737   return std::distance(m_section_headers.begin(), I);
738 }
739 
SectionIndex(const SectionHeaderCollConstIter & I) const740 size_t ObjectFileELF::SectionIndex(const SectionHeaderCollConstIter &I) const {
741   return std::distance(m_section_headers.begin(), I);
742 }
743 
ParseHeader()744 bool ObjectFileELF::ParseHeader() {
745   lldb::offset_t offset = 0;
746   return m_header.Parse(m_data, &offset);
747 }
748 
GetUUID()749 UUID ObjectFileELF::GetUUID() {
750   // Need to parse the section list to get the UUIDs, so make sure that's been
751   // done.
752   if (!ParseSectionHeaders() && GetType() != ObjectFile::eTypeCoreFile)
753     return UUID();
754 
755   if (!m_uuid) {
756     using u32le = llvm::support::ulittle32_t;
757     if (GetType() == ObjectFile::eTypeCoreFile) {
758       uint32_t core_notes_crc = 0;
759 
760       if (!ParseProgramHeaders())
761         return UUID();
762 
763       core_notes_crc =
764           CalculateELFNotesSegmentsCRC32(m_program_headers, m_data);
765 
766       if (core_notes_crc) {
767         // Use 8 bytes - first 4 bytes for *magic* prefix, mainly to make it
768         // look different form .gnu_debuglink crc - followed by 4 bytes of note
769         // segments crc.
770         u32le data[] = {u32le(g_core_uuid_magic), u32le(core_notes_crc)};
771         m_uuid = UUID::fromData(data, sizeof(data));
772       }
773     } else {
774       if (!m_gnu_debuglink_crc)
775         m_gnu_debuglink_crc = calc_crc32(0, m_data);
776       if (m_gnu_debuglink_crc) {
777         // Use 4 bytes of crc from the .gnu_debuglink section.
778         u32le data(m_gnu_debuglink_crc);
779         m_uuid = UUID::fromData(&data, sizeof(data));
780       }
781     }
782   }
783 
784   return m_uuid;
785 }
786 
GetDebugLink()787 llvm::Optional<FileSpec> ObjectFileELF::GetDebugLink() {
788   if (m_gnu_debuglink_file.empty())
789     return llvm::None;
790   return FileSpec(m_gnu_debuglink_file);
791 }
792 
GetDependentModules(FileSpecList & files)793 uint32_t ObjectFileELF::GetDependentModules(FileSpecList &files) {
794   size_t num_modules = ParseDependentModules();
795   uint32_t num_specs = 0;
796 
797   for (unsigned i = 0; i < num_modules; ++i) {
798     if (files.AppendIfUnique(m_filespec_up->GetFileSpecAtIndex(i)))
799       num_specs++;
800   }
801 
802   return num_specs;
803 }
804 
GetImageInfoAddress(Target * target)805 Address ObjectFileELF::GetImageInfoAddress(Target *target) {
806   if (!ParseDynamicSymbols())
807     return Address();
808 
809   SectionList *section_list = GetSectionList();
810   if (!section_list)
811     return Address();
812 
813   // Find the SHT_DYNAMIC (.dynamic) section.
814   SectionSP dynsym_section_sp(
815       section_list->FindSectionByType(eSectionTypeELFDynamicLinkInfo, true));
816   if (!dynsym_section_sp)
817     return Address();
818   assert(dynsym_section_sp->GetObjectFile() == this);
819 
820   user_id_t dynsym_id = dynsym_section_sp->GetID();
821   const ELFSectionHeaderInfo *dynsym_hdr = GetSectionHeaderByIndex(dynsym_id);
822   if (!dynsym_hdr)
823     return Address();
824 
825   for (size_t i = 0; i < m_dynamic_symbols.size(); ++i) {
826     ELFDynamic &symbol = m_dynamic_symbols[i];
827 
828     if (symbol.d_tag == DT_DEBUG) {
829       // Compute the offset as the number of previous entries plus the size of
830       // d_tag.
831       addr_t offset = i * dynsym_hdr->sh_entsize + GetAddressByteSize();
832       return Address(dynsym_section_sp, offset);
833     }
834     // MIPS executables uses DT_MIPS_RLD_MAP_REL to support PIE. DT_MIPS_RLD_MAP
835     // exists in non-PIE.
836     else if ((symbol.d_tag == DT_MIPS_RLD_MAP ||
837               symbol.d_tag == DT_MIPS_RLD_MAP_REL) &&
838              target) {
839       addr_t offset = i * dynsym_hdr->sh_entsize + GetAddressByteSize();
840       addr_t dyn_base = dynsym_section_sp->GetLoadBaseAddress(target);
841       if (dyn_base == LLDB_INVALID_ADDRESS)
842         return Address();
843 
844       Status error;
845       if (symbol.d_tag == DT_MIPS_RLD_MAP) {
846         // DT_MIPS_RLD_MAP tag stores an absolute address of the debug pointer.
847         Address addr;
848         if (target->ReadPointerFromMemory(dyn_base + offset, false, error,
849                                           addr))
850           return addr;
851       }
852       if (symbol.d_tag == DT_MIPS_RLD_MAP_REL) {
853         // DT_MIPS_RLD_MAP_REL tag stores the offset to the debug pointer,
854         // relative to the address of the tag.
855         uint64_t rel_offset;
856         rel_offset = target->ReadUnsignedIntegerFromMemory(
857             dyn_base + offset, false, GetAddressByteSize(), UINT64_MAX, error);
858         if (error.Success() && rel_offset != UINT64_MAX) {
859           Address addr;
860           addr_t debug_ptr_address =
861               dyn_base + (offset - GetAddressByteSize()) + rel_offset;
862           addr.SetOffset(debug_ptr_address);
863           return addr;
864         }
865       }
866     }
867   }
868 
869   return Address();
870 }
871 
GetEntryPointAddress()872 lldb_private::Address ObjectFileELF::GetEntryPointAddress() {
873   if (m_entry_point_address.IsValid())
874     return m_entry_point_address;
875 
876   if (!ParseHeader() || !IsExecutable())
877     return m_entry_point_address;
878 
879   SectionList *section_list = GetSectionList();
880   addr_t offset = m_header.e_entry;
881 
882   if (!section_list)
883     m_entry_point_address.SetOffset(offset);
884   else
885     m_entry_point_address.ResolveAddressUsingFileSections(offset, section_list);
886   return m_entry_point_address;
887 }
888 
GetBaseAddress()889 Address ObjectFileELF::GetBaseAddress() {
890   for (const auto &EnumPHdr : llvm::enumerate(ProgramHeaders())) {
891     const ELFProgramHeader &H = EnumPHdr.value();
892     if (H.p_type != PT_LOAD)
893       continue;
894 
895     return Address(
896         GetSectionList()->FindSectionByID(SegmentID(EnumPHdr.index())), 0);
897   }
898   return LLDB_INVALID_ADDRESS;
899 }
900 
901 // ParseDependentModules
ParseDependentModules()902 size_t ObjectFileELF::ParseDependentModules() {
903   if (m_filespec_up)
904     return m_filespec_up->GetSize();
905 
906   m_filespec_up = std::make_unique<FileSpecList>();
907 
908   if (!ParseSectionHeaders())
909     return 0;
910 
911   SectionList *section_list = GetSectionList();
912   if (!section_list)
913     return 0;
914 
915   // Find the SHT_DYNAMIC section.
916   Section *dynsym =
917       section_list->FindSectionByType(eSectionTypeELFDynamicLinkInfo, true)
918           .get();
919   if (!dynsym)
920     return 0;
921   assert(dynsym->GetObjectFile() == this);
922 
923   const ELFSectionHeaderInfo *header = GetSectionHeaderByIndex(dynsym->GetID());
924   if (!header)
925     return 0;
926   // sh_link: section header index of string table used by entries in the
927   // section.
928   Section *dynstr = section_list->FindSectionByID(header->sh_link).get();
929   if (!dynstr)
930     return 0;
931 
932   DataExtractor dynsym_data;
933   DataExtractor dynstr_data;
934   if (ReadSectionData(dynsym, dynsym_data) &&
935       ReadSectionData(dynstr, dynstr_data)) {
936     ELFDynamic symbol;
937     const lldb::offset_t section_size = dynsym_data.GetByteSize();
938     lldb::offset_t offset = 0;
939 
940     // The only type of entries we are concerned with are tagged DT_NEEDED,
941     // yielding the name of a required library.
942     while (offset < section_size) {
943       if (!symbol.Parse(dynsym_data, &offset))
944         break;
945 
946       if (symbol.d_tag != DT_NEEDED)
947         continue;
948 
949       uint32_t str_index = static_cast<uint32_t>(symbol.d_val);
950       const char *lib_name = dynstr_data.PeekCStr(str_index);
951       FileSpec file_spec(lib_name);
952       FileSystem::Instance().Resolve(file_spec);
953       m_filespec_up->Append(file_spec);
954     }
955   }
956 
957   return m_filespec_up->GetSize();
958 }
959 
960 // GetProgramHeaderInfo
GetProgramHeaderInfo(ProgramHeaderColl & program_headers,DataExtractor & object_data,const ELFHeader & header)961 size_t ObjectFileELF::GetProgramHeaderInfo(ProgramHeaderColl &program_headers,
962                                            DataExtractor &object_data,
963                                            const ELFHeader &header) {
964   // We have already parsed the program headers
965   if (!program_headers.empty())
966     return program_headers.size();
967 
968   // If there are no program headers to read we are done.
969   if (header.e_phnum == 0)
970     return 0;
971 
972   program_headers.resize(header.e_phnum);
973   if (program_headers.size() != header.e_phnum)
974     return 0;
975 
976   const size_t ph_size = header.e_phnum * header.e_phentsize;
977   const elf_off ph_offset = header.e_phoff;
978   DataExtractor data;
979   if (data.SetData(object_data, ph_offset, ph_size) != ph_size)
980     return 0;
981 
982   uint32_t idx;
983   lldb::offset_t offset;
984   for (idx = 0, offset = 0; idx < header.e_phnum; ++idx) {
985     if (!program_headers[idx].Parse(data, &offset))
986       break;
987   }
988 
989   if (idx < program_headers.size())
990     program_headers.resize(idx);
991 
992   return program_headers.size();
993 }
994 
995 // ParseProgramHeaders
ParseProgramHeaders()996 bool ObjectFileELF::ParseProgramHeaders() {
997   return GetProgramHeaderInfo(m_program_headers, m_data, m_header) != 0;
998 }
999 
1000 lldb_private::Status
RefineModuleDetailsFromNote(lldb_private::DataExtractor & data,lldb_private::ArchSpec & arch_spec,lldb_private::UUID & uuid)1001 ObjectFileELF::RefineModuleDetailsFromNote(lldb_private::DataExtractor &data,
1002                                            lldb_private::ArchSpec &arch_spec,
1003                                            lldb_private::UUID &uuid) {
1004   Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_MODULES));
1005   Status error;
1006 
1007   lldb::offset_t offset = 0;
1008 
1009   while (true) {
1010     // Parse the note header.  If this fails, bail out.
1011     const lldb::offset_t note_offset = offset;
1012     ELFNote note = ELFNote();
1013     if (!note.Parse(data, &offset)) {
1014       // We're done.
1015       return error;
1016     }
1017 
1018     LLDB_LOGF(log, "ObjectFileELF::%s parsing note name='%s', type=%" PRIu32,
1019               __FUNCTION__, note.n_name.c_str(), note.n_type);
1020 
1021     // Process FreeBSD ELF notes.
1022     if ((note.n_name == LLDB_NT_OWNER_FREEBSD) &&
1023         (note.n_type == LLDB_NT_FREEBSD_ABI_TAG) &&
1024         (note.n_descsz == LLDB_NT_FREEBSD_ABI_SIZE)) {
1025       // Pull out the min version info.
1026       uint32_t version_info;
1027       if (data.GetU32(&offset, &version_info, 1) == nullptr) {
1028         error.SetErrorString("failed to read FreeBSD ABI note payload");
1029         return error;
1030       }
1031 
1032       // Convert the version info into a major/minor number.
1033       const uint32_t version_major = version_info / 100000;
1034       const uint32_t version_minor = (version_info / 1000) % 100;
1035 
1036       char os_name[32];
1037       snprintf(os_name, sizeof(os_name), "freebsd%" PRIu32 ".%" PRIu32,
1038                version_major, version_minor);
1039 
1040       // Set the elf OS version to FreeBSD.  Also clear the vendor.
1041       arch_spec.GetTriple().setOSName(os_name);
1042       arch_spec.GetTriple().setVendor(llvm::Triple::VendorType::UnknownVendor);
1043 
1044       LLDB_LOGF(log,
1045                 "ObjectFileELF::%s detected FreeBSD %" PRIu32 ".%" PRIu32
1046                 ".%" PRIu32,
1047                 __FUNCTION__, version_major, version_minor,
1048                 static_cast<uint32_t>(version_info % 1000));
1049     }
1050     // Process GNU ELF notes.
1051     else if (note.n_name == LLDB_NT_OWNER_GNU) {
1052       switch (note.n_type) {
1053       case LLDB_NT_GNU_ABI_TAG:
1054         if (note.n_descsz == LLDB_NT_GNU_ABI_SIZE) {
1055           // Pull out the min OS version supporting the ABI.
1056           uint32_t version_info[4];
1057           if (data.GetU32(&offset, &version_info[0], note.n_descsz / 4) ==
1058               nullptr) {
1059             error.SetErrorString("failed to read GNU ABI note payload");
1060             return error;
1061           }
1062 
1063           // Set the OS per the OS field.
1064           switch (version_info[0]) {
1065           case LLDB_NT_GNU_ABI_OS_LINUX:
1066             arch_spec.GetTriple().setOS(llvm::Triple::OSType::Linux);
1067             arch_spec.GetTriple().setVendor(
1068                 llvm::Triple::VendorType::UnknownVendor);
1069             LLDB_LOGF(log,
1070                       "ObjectFileELF::%s detected Linux, min version %" PRIu32
1071                       ".%" PRIu32 ".%" PRIu32,
1072                       __FUNCTION__, version_info[1], version_info[2],
1073                       version_info[3]);
1074             // FIXME we have the minimal version number, we could be propagating
1075             // that.  version_info[1] = OS Major, version_info[2] = OS Minor,
1076             // version_info[3] = Revision.
1077             break;
1078           case LLDB_NT_GNU_ABI_OS_HURD:
1079             arch_spec.GetTriple().setOS(llvm::Triple::OSType::UnknownOS);
1080             arch_spec.GetTriple().setVendor(
1081                 llvm::Triple::VendorType::UnknownVendor);
1082             LLDB_LOGF(log,
1083                       "ObjectFileELF::%s detected Hurd (unsupported), min "
1084                       "version %" PRIu32 ".%" PRIu32 ".%" PRIu32,
1085                       __FUNCTION__, version_info[1], version_info[2],
1086                       version_info[3]);
1087             break;
1088           case LLDB_NT_GNU_ABI_OS_SOLARIS:
1089             arch_spec.GetTriple().setOS(llvm::Triple::OSType::Solaris);
1090             arch_spec.GetTriple().setVendor(
1091                 llvm::Triple::VendorType::UnknownVendor);
1092             LLDB_LOGF(log,
1093                       "ObjectFileELF::%s detected Solaris, min version %" PRIu32
1094                       ".%" PRIu32 ".%" PRIu32,
1095                       __FUNCTION__, version_info[1], version_info[2],
1096                       version_info[3]);
1097             break;
1098           default:
1099             LLDB_LOGF(log,
1100                       "ObjectFileELF::%s unrecognized OS in note, id %" PRIu32
1101                       ", min version %" PRIu32 ".%" PRIu32 ".%" PRIu32,
1102                       __FUNCTION__, version_info[0], version_info[1],
1103                       version_info[2], version_info[3]);
1104             break;
1105           }
1106         }
1107         break;
1108 
1109       case LLDB_NT_GNU_BUILD_ID_TAG:
1110         // Only bother processing this if we don't already have the uuid set.
1111         if (!uuid.IsValid()) {
1112           // 16 bytes is UUID|MD5, 20 bytes is SHA1. Other linkers may produce a
1113           // build-id of a different length. Accept it as long as it's at least
1114           // 4 bytes as it will be better than our own crc32.
1115           if (note.n_descsz >= 4) {
1116             if (const uint8_t *buf = data.PeekData(offset, note.n_descsz)) {
1117               // Save the build id as the UUID for the module.
1118               uuid = UUID::fromData(buf, note.n_descsz);
1119             } else {
1120               error.SetErrorString("failed to read GNU_BUILD_ID note payload");
1121               return error;
1122             }
1123           }
1124         }
1125         break;
1126       }
1127       if (arch_spec.IsMIPS() &&
1128           arch_spec.GetTriple().getOS() == llvm::Triple::OSType::UnknownOS)
1129         // The note.n_name == LLDB_NT_OWNER_GNU is valid for Linux platform
1130         arch_spec.GetTriple().setOS(llvm::Triple::OSType::Linux);
1131     }
1132     // Process NetBSD ELF executables and shared libraries
1133     else if ((note.n_name == LLDB_NT_OWNER_NETBSD) &&
1134              (note.n_type == LLDB_NT_NETBSD_IDENT_TAG) &&
1135              (note.n_descsz == LLDB_NT_NETBSD_IDENT_DESCSZ) &&
1136              (note.n_namesz == LLDB_NT_NETBSD_IDENT_NAMESZ)) {
1137       // Pull out the version info.
1138       uint32_t version_info;
1139       if (data.GetU32(&offset, &version_info, 1) == nullptr) {
1140         error.SetErrorString("failed to read NetBSD ABI note payload");
1141         return error;
1142       }
1143       // Convert the version info into a major/minor/patch number.
1144       //     #define __NetBSD_Version__ MMmmrrpp00
1145       //
1146       //     M = major version
1147       //     m = minor version; a minor number of 99 indicates current.
1148       //     r = 0 (since NetBSD 3.0 not used)
1149       //     p = patchlevel
1150       const uint32_t version_major = version_info / 100000000;
1151       const uint32_t version_minor = (version_info % 100000000) / 1000000;
1152       const uint32_t version_patch = (version_info % 10000) / 100;
1153       // Set the elf OS version to NetBSD.  Also clear the vendor.
1154       arch_spec.GetTriple().setOSName(
1155           llvm::formatv("netbsd{0}.{1}.{2}", version_major, version_minor,
1156                         version_patch).str());
1157       arch_spec.GetTriple().setVendor(llvm::Triple::VendorType::UnknownVendor);
1158     }
1159     // Process NetBSD ELF core(5) notes
1160     else if ((note.n_name == LLDB_NT_OWNER_NETBSDCORE) &&
1161              (note.n_type == LLDB_NT_NETBSD_PROCINFO)) {
1162       // Set the elf OS version to NetBSD.  Also clear the vendor.
1163       arch_spec.GetTriple().setOS(llvm::Triple::OSType::NetBSD);
1164       arch_spec.GetTriple().setVendor(llvm::Triple::VendorType::UnknownVendor);
1165     }
1166     // Process OpenBSD ELF notes.
1167     else if (note.n_name == LLDB_NT_OWNER_OPENBSD) {
1168       // Set the elf OS version to OpenBSD.  Also clear the vendor.
1169       arch_spec.GetTriple().setOS(llvm::Triple::OSType::OpenBSD);
1170       arch_spec.GetTriple().setVendor(llvm::Triple::VendorType::UnknownVendor);
1171     } else if (note.n_name == LLDB_NT_OWNER_ANDROID) {
1172       arch_spec.GetTriple().setOS(llvm::Triple::OSType::Linux);
1173       arch_spec.GetTriple().setEnvironment(
1174           llvm::Triple::EnvironmentType::Android);
1175     } else if (note.n_name == LLDB_NT_OWNER_LINUX) {
1176       // This is sometimes found in core files and usually contains extended
1177       // register info
1178       arch_spec.GetTriple().setOS(llvm::Triple::OSType::Linux);
1179     } else if (note.n_name == LLDB_NT_OWNER_CORE) {
1180       // Parse the NT_FILE to look for stuff in paths to shared libraries As
1181       // the contents look like this in a 64 bit ELF core file: count     =
1182       // 0x000000000000000a (10) page_size = 0x0000000000001000 (4096) Index
1183       // start              end                file_ofs           path =====
1184       // 0x0000000000401000 0x0000000000000000 /tmp/a.out [  1]
1185       // 0x0000000000600000 0x0000000000601000 0x0000000000000000 /tmp/a.out [
1186       // 2] 0x0000000000601000 0x0000000000602000 0x0000000000000001 /tmp/a.out
1187       // [  3] 0x00007fa79c9ed000 0x00007fa79cba8000 0x0000000000000000
1188       // /lib/x86_64-linux-gnu/libc-2.19.so [  4] 0x00007fa79cba8000
1189       // 0x00007fa79cda7000 0x00000000000001bb /lib/x86_64-linux-
1190       // gnu/libc-2.19.so [  5] 0x00007fa79cda7000 0x00007fa79cdab000
1191       // 0x00000000000001ba /lib/x86_64-linux-gnu/libc-2.19.so [  6]
1192       // 0x00007fa79cdab000 0x00007fa79cdad000 0x00000000000001be /lib/x86_64
1193       // -linux-gnu/libc-2.19.so [  7] 0x00007fa79cdb2000 0x00007fa79cdd5000
1194       // 0x0000000000000000 /lib/x86_64-linux-gnu/ld-2.19.so [  8]
1195       // 0x00007fa79cfd4000 0x00007fa79cfd5000 0x0000000000000022 /lib/x86_64
1196       // -linux-gnu/ld-2.19.so [  9] 0x00007fa79cfd5000 0x00007fa79cfd6000
1197       // 0x0000000000000023 /lib/x86_64-linux-gnu/ld-2.19.so In the 32 bit ELFs
1198       // the count, page_size, start, end, file_ofs are uint32_t For reference:
1199       // see readelf source code (in binutils).
1200       if (note.n_type == NT_FILE) {
1201         uint64_t count = data.GetAddress(&offset);
1202         const char *cstr;
1203         data.GetAddress(&offset); // Skip page size
1204         offset += count * 3 *
1205                   data.GetAddressByteSize(); // Skip all start/end/file_ofs
1206         for (size_t i = 0; i < count; ++i) {
1207           cstr = data.GetCStr(&offset);
1208           if (cstr == nullptr) {
1209             error.SetErrorStringWithFormat("ObjectFileELF::%s trying to read "
1210                                            "at an offset after the end "
1211                                            "(GetCStr returned nullptr)",
1212                                            __FUNCTION__);
1213             return error;
1214           }
1215           llvm::StringRef path(cstr);
1216           if (path.contains("/lib/x86_64-linux-gnu") || path.contains("/lib/i386-linux-gnu")) {
1217             arch_spec.GetTriple().setOS(llvm::Triple::OSType::Linux);
1218             break;
1219           }
1220         }
1221         if (arch_spec.IsMIPS() &&
1222             arch_spec.GetTriple().getOS() == llvm::Triple::OSType::UnknownOS)
1223           // In case of MIPSR6, the LLDB_NT_OWNER_GNU note is missing for some
1224           // cases (e.g. compile with -nostdlib) Hence set OS to Linux
1225           arch_spec.GetTriple().setOS(llvm::Triple::OSType::Linux);
1226       }
1227     }
1228 
1229     // Calculate the offset of the next note just in case "offset" has been
1230     // used to poke at the contents of the note data
1231     offset = note_offset + note.GetByteSize();
1232   }
1233 
1234   return error;
1235 }
1236 
ParseARMAttributes(DataExtractor & data,uint64_t length,ArchSpec & arch_spec)1237 void ObjectFileELF::ParseARMAttributes(DataExtractor &data, uint64_t length,
1238                                        ArchSpec &arch_spec) {
1239   lldb::offset_t Offset = 0;
1240 
1241   uint8_t FormatVersion = data.GetU8(&Offset);
1242   if (FormatVersion != llvm::ELFAttrs::Format_Version)
1243     return;
1244 
1245   Offset = Offset + sizeof(uint32_t); // Section Length
1246   llvm::StringRef VendorName = data.GetCStr(&Offset);
1247 
1248   if (VendorName != "aeabi")
1249     return;
1250 
1251   if (arch_spec.GetTriple().getEnvironment() ==
1252       llvm::Triple::UnknownEnvironment)
1253     arch_spec.GetTriple().setEnvironment(llvm::Triple::EABI);
1254 
1255   while (Offset < length) {
1256     uint8_t Tag = data.GetU8(&Offset);
1257     uint32_t Size = data.GetU32(&Offset);
1258 
1259     if (Tag != llvm::ARMBuildAttrs::File || Size == 0)
1260       continue;
1261 
1262     while (Offset < length) {
1263       uint64_t Tag = data.GetULEB128(&Offset);
1264       switch (Tag) {
1265       default:
1266         if (Tag < 32)
1267           data.GetULEB128(&Offset);
1268         else if (Tag % 2 == 0)
1269           data.GetULEB128(&Offset);
1270         else
1271           data.GetCStr(&Offset);
1272 
1273         break;
1274 
1275       case llvm::ARMBuildAttrs::CPU_raw_name:
1276       case llvm::ARMBuildAttrs::CPU_name:
1277         data.GetCStr(&Offset);
1278 
1279         break;
1280 
1281       case llvm::ARMBuildAttrs::ABI_VFP_args: {
1282         uint64_t VFPArgs = data.GetULEB128(&Offset);
1283 
1284         if (VFPArgs == llvm::ARMBuildAttrs::BaseAAPCS) {
1285           if (arch_spec.GetTriple().getEnvironment() ==
1286                   llvm::Triple::UnknownEnvironment ||
1287               arch_spec.GetTriple().getEnvironment() == llvm::Triple::EABIHF)
1288             arch_spec.GetTriple().setEnvironment(llvm::Triple::EABI);
1289 
1290           arch_spec.SetFlags(ArchSpec::eARM_abi_soft_float);
1291         } else if (VFPArgs == llvm::ARMBuildAttrs::HardFPAAPCS) {
1292           if (arch_spec.GetTriple().getEnvironment() ==
1293                   llvm::Triple::UnknownEnvironment ||
1294               arch_spec.GetTriple().getEnvironment() == llvm::Triple::EABI)
1295             arch_spec.GetTriple().setEnvironment(llvm::Triple::EABIHF);
1296 
1297           arch_spec.SetFlags(ArchSpec::eARM_abi_hard_float);
1298         }
1299 
1300         break;
1301       }
1302       }
1303     }
1304   }
1305 }
1306 
1307 // GetSectionHeaderInfo
GetSectionHeaderInfo(SectionHeaderColl & section_headers,DataExtractor & object_data,const elf::ELFHeader & header,lldb_private::UUID & uuid,std::string & gnu_debuglink_file,uint32_t & gnu_debuglink_crc,ArchSpec & arch_spec)1308 size_t ObjectFileELF::GetSectionHeaderInfo(SectionHeaderColl &section_headers,
1309                                            DataExtractor &object_data,
1310                                            const elf::ELFHeader &header,
1311                                            lldb_private::UUID &uuid,
1312                                            std::string &gnu_debuglink_file,
1313                                            uint32_t &gnu_debuglink_crc,
1314                                            ArchSpec &arch_spec) {
1315   // Don't reparse the section headers if we already did that.
1316   if (!section_headers.empty())
1317     return section_headers.size();
1318 
1319   // Only initialize the arch_spec to okay defaults if they're not already set.
1320   // We'll refine this with note data as we parse the notes.
1321   if (arch_spec.GetTriple().getOS() == llvm::Triple::OSType::UnknownOS) {
1322     llvm::Triple::OSType ostype;
1323     llvm::Triple::OSType spec_ostype;
1324     const uint32_t sub_type = subTypeFromElfHeader(header);
1325     arch_spec.SetArchitecture(eArchTypeELF, header.e_machine, sub_type,
1326                               header.e_ident[EI_OSABI]);
1327 
1328     // Validate if it is ok to remove GetOsFromOSABI. Note, that now the OS is
1329     // determined based on EI_OSABI flag and the info extracted from ELF notes
1330     // (see RefineModuleDetailsFromNote). However in some cases that still
1331     // might be not enough: for example a shared library might not have any
1332     // notes at all and have EI_OSABI flag set to System V, as result the OS
1333     // will be set to UnknownOS.
1334     GetOsFromOSABI(header.e_ident[EI_OSABI], ostype);
1335     spec_ostype = arch_spec.GetTriple().getOS();
1336     assert(spec_ostype == ostype);
1337     UNUSED_IF_ASSERT_DISABLED(spec_ostype);
1338   }
1339 
1340   if (arch_spec.GetMachine() == llvm::Triple::mips ||
1341       arch_spec.GetMachine() == llvm::Triple::mipsel ||
1342       arch_spec.GetMachine() == llvm::Triple::mips64 ||
1343       arch_spec.GetMachine() == llvm::Triple::mips64el) {
1344     switch (header.e_flags & llvm::ELF::EF_MIPS_ARCH_ASE) {
1345     case llvm::ELF::EF_MIPS_MICROMIPS:
1346       arch_spec.SetFlags(ArchSpec::eMIPSAse_micromips);
1347       break;
1348     case llvm::ELF::EF_MIPS_ARCH_ASE_M16:
1349       arch_spec.SetFlags(ArchSpec::eMIPSAse_mips16);
1350       break;
1351     case llvm::ELF::EF_MIPS_ARCH_ASE_MDMX:
1352       arch_spec.SetFlags(ArchSpec::eMIPSAse_mdmx);
1353       break;
1354     default:
1355       break;
1356     }
1357   }
1358 
1359   if (arch_spec.GetMachine() == llvm::Triple::arm ||
1360       arch_spec.GetMachine() == llvm::Triple::thumb) {
1361     if (header.e_flags & llvm::ELF::EF_ARM_SOFT_FLOAT)
1362       arch_spec.SetFlags(ArchSpec::eARM_abi_soft_float);
1363     else if (header.e_flags & llvm::ELF::EF_ARM_VFP_FLOAT)
1364       arch_spec.SetFlags(ArchSpec::eARM_abi_hard_float);
1365   }
1366 
1367   // If there are no section headers we are done.
1368   if (header.e_shnum == 0)
1369     return 0;
1370 
1371   Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_MODULES));
1372 
1373   section_headers.resize(header.e_shnum);
1374   if (section_headers.size() != header.e_shnum)
1375     return 0;
1376 
1377   const size_t sh_size = header.e_shnum * header.e_shentsize;
1378   const elf_off sh_offset = header.e_shoff;
1379   DataExtractor sh_data;
1380   if (sh_data.SetData(object_data, sh_offset, sh_size) != sh_size)
1381     return 0;
1382 
1383   uint32_t idx;
1384   lldb::offset_t offset;
1385   for (idx = 0, offset = 0; idx < header.e_shnum; ++idx) {
1386     if (!section_headers[idx].Parse(sh_data, &offset))
1387       break;
1388   }
1389   if (idx < section_headers.size())
1390     section_headers.resize(idx);
1391 
1392   const unsigned strtab_idx = header.e_shstrndx;
1393   if (strtab_idx && strtab_idx < section_headers.size()) {
1394     const ELFSectionHeaderInfo &sheader = section_headers[strtab_idx];
1395     const size_t byte_size = sheader.sh_size;
1396     const Elf64_Off offset = sheader.sh_offset;
1397     lldb_private::DataExtractor shstr_data;
1398 
1399     if (shstr_data.SetData(object_data, offset, byte_size) == byte_size) {
1400       for (SectionHeaderCollIter I = section_headers.begin();
1401            I != section_headers.end(); ++I) {
1402         static ConstString g_sect_name_gnu_debuglink(".gnu_debuglink");
1403         const ELFSectionHeaderInfo &sheader = *I;
1404         const uint64_t section_size =
1405             sheader.sh_type == SHT_NOBITS ? 0 : sheader.sh_size;
1406         ConstString name(shstr_data.PeekCStr(I->sh_name));
1407 
1408         I->section_name = name;
1409 
1410         if (arch_spec.IsMIPS()) {
1411           uint32_t arch_flags = arch_spec.GetFlags();
1412           DataExtractor data;
1413           if (sheader.sh_type == SHT_MIPS_ABIFLAGS) {
1414 
1415             if (section_size && (data.SetData(object_data, sheader.sh_offset,
1416                                               section_size) == section_size)) {
1417               // MIPS ASE Mask is at offset 12 in MIPS.abiflags section
1418               lldb::offset_t offset = 12; // MIPS ABI Flags Version: 0
1419               arch_flags |= data.GetU32(&offset);
1420 
1421               // The floating point ABI is at offset 7
1422               offset = 7;
1423               switch (data.GetU8(&offset)) {
1424               case llvm::Mips::Val_GNU_MIPS_ABI_FP_ANY:
1425                 arch_flags |= lldb_private::ArchSpec::eMIPS_ABI_FP_ANY;
1426                 break;
1427               case llvm::Mips::Val_GNU_MIPS_ABI_FP_DOUBLE:
1428                 arch_flags |= lldb_private::ArchSpec::eMIPS_ABI_FP_DOUBLE;
1429                 break;
1430               case llvm::Mips::Val_GNU_MIPS_ABI_FP_SINGLE:
1431                 arch_flags |= lldb_private::ArchSpec::eMIPS_ABI_FP_SINGLE;
1432                 break;
1433               case llvm::Mips::Val_GNU_MIPS_ABI_FP_SOFT:
1434                 arch_flags |= lldb_private::ArchSpec::eMIPS_ABI_FP_SOFT;
1435                 break;
1436               case llvm::Mips::Val_GNU_MIPS_ABI_FP_OLD_64:
1437                 arch_flags |= lldb_private::ArchSpec::eMIPS_ABI_FP_OLD_64;
1438                 break;
1439               case llvm::Mips::Val_GNU_MIPS_ABI_FP_XX:
1440                 arch_flags |= lldb_private::ArchSpec::eMIPS_ABI_FP_XX;
1441                 break;
1442               case llvm::Mips::Val_GNU_MIPS_ABI_FP_64:
1443                 arch_flags |= lldb_private::ArchSpec::eMIPS_ABI_FP_64;
1444                 break;
1445               case llvm::Mips::Val_GNU_MIPS_ABI_FP_64A:
1446                 arch_flags |= lldb_private::ArchSpec::eMIPS_ABI_FP_64A;
1447                 break;
1448               }
1449             }
1450           }
1451           // Settings appropriate ArchSpec ABI Flags
1452           switch (header.e_flags & llvm::ELF::EF_MIPS_ABI) {
1453           case llvm::ELF::EF_MIPS_ABI_O32:
1454             arch_flags |= lldb_private::ArchSpec::eMIPSABI_O32;
1455             break;
1456           case EF_MIPS_ABI_O64:
1457             arch_flags |= lldb_private::ArchSpec::eMIPSABI_O64;
1458             break;
1459           case EF_MIPS_ABI_EABI32:
1460             arch_flags |= lldb_private::ArchSpec::eMIPSABI_EABI32;
1461             break;
1462           case EF_MIPS_ABI_EABI64:
1463             arch_flags |= lldb_private::ArchSpec::eMIPSABI_EABI64;
1464             break;
1465           default:
1466             // ABI Mask doesn't cover N32 and N64 ABI.
1467             if (header.e_ident[EI_CLASS] == llvm::ELF::ELFCLASS64)
1468               arch_flags |= lldb_private::ArchSpec::eMIPSABI_N64;
1469             else if (header.e_flags & llvm::ELF::EF_MIPS_ABI2)
1470               arch_flags |= lldb_private::ArchSpec::eMIPSABI_N32;
1471             break;
1472           }
1473           arch_spec.SetFlags(arch_flags);
1474         }
1475 
1476         if (arch_spec.GetMachine() == llvm::Triple::arm ||
1477             arch_spec.GetMachine() == llvm::Triple::thumb) {
1478           DataExtractor data;
1479 
1480           if (sheader.sh_type == SHT_ARM_ATTRIBUTES && section_size != 0 &&
1481               data.SetData(object_data, sheader.sh_offset, section_size) == section_size)
1482             ParseARMAttributes(data, section_size, arch_spec);
1483         }
1484 
1485         if (name == g_sect_name_gnu_debuglink) {
1486           DataExtractor data;
1487           if (section_size && (data.SetData(object_data, sheader.sh_offset,
1488                                             section_size) == section_size)) {
1489             lldb::offset_t gnu_debuglink_offset = 0;
1490             gnu_debuglink_file = data.GetCStr(&gnu_debuglink_offset);
1491             gnu_debuglink_offset = llvm::alignTo(gnu_debuglink_offset, 4);
1492             data.GetU32(&gnu_debuglink_offset, &gnu_debuglink_crc, 1);
1493           }
1494         }
1495 
1496         // Process ELF note section entries.
1497         bool is_note_header = (sheader.sh_type == SHT_NOTE);
1498 
1499         // The section header ".note.android.ident" is stored as a
1500         // PROGBITS type header but it is actually a note header.
1501         static ConstString g_sect_name_android_ident(".note.android.ident");
1502         if (!is_note_header && name == g_sect_name_android_ident)
1503           is_note_header = true;
1504 
1505         if (is_note_header) {
1506           // Allow notes to refine module info.
1507           DataExtractor data;
1508           if (section_size && (data.SetData(object_data, sheader.sh_offset,
1509                                             section_size) == section_size)) {
1510             Status error = RefineModuleDetailsFromNote(data, arch_spec, uuid);
1511             if (error.Fail()) {
1512               LLDB_LOGF(log, "ObjectFileELF::%s ELF note processing failed: %s",
1513                         __FUNCTION__, error.AsCString());
1514             }
1515           }
1516         }
1517       }
1518 
1519       // Make any unknown triple components to be unspecified unknowns.
1520       if (arch_spec.GetTriple().getVendor() == llvm::Triple::UnknownVendor)
1521         arch_spec.GetTriple().setVendorName(llvm::StringRef());
1522       if (arch_spec.GetTriple().getOS() == llvm::Triple::UnknownOS)
1523         arch_spec.GetTriple().setOSName(llvm::StringRef());
1524 
1525       return section_headers.size();
1526     }
1527   }
1528 
1529   section_headers.clear();
1530   return 0;
1531 }
1532 
1533 llvm::StringRef
StripLinkerSymbolAnnotations(llvm::StringRef symbol_name) const1534 ObjectFileELF::StripLinkerSymbolAnnotations(llvm::StringRef symbol_name) const {
1535   size_t pos = symbol_name.find('@');
1536   return symbol_name.substr(0, pos);
1537 }
1538 
1539 // ParseSectionHeaders
ParseSectionHeaders()1540 size_t ObjectFileELF::ParseSectionHeaders() {
1541   return GetSectionHeaderInfo(m_section_headers, m_data, m_header, m_uuid,
1542                               m_gnu_debuglink_file, m_gnu_debuglink_crc,
1543                               m_arch_spec);
1544 }
1545 
1546 const ObjectFileELF::ELFSectionHeaderInfo *
GetSectionHeaderByIndex(lldb::user_id_t id)1547 ObjectFileELF::GetSectionHeaderByIndex(lldb::user_id_t id) {
1548   if (!ParseSectionHeaders())
1549     return nullptr;
1550 
1551   if (id < m_section_headers.size())
1552     return &m_section_headers[id];
1553 
1554   return nullptr;
1555 }
1556 
GetSectionIndexByName(const char * name)1557 lldb::user_id_t ObjectFileELF::GetSectionIndexByName(const char *name) {
1558   if (!name || !name[0] || !ParseSectionHeaders())
1559     return 0;
1560   for (size_t i = 1; i < m_section_headers.size(); ++i)
1561     if (m_section_headers[i].section_name == ConstString(name))
1562       return i;
1563   return 0;
1564 }
1565 
GetSectionTypeFromName(llvm::StringRef Name)1566 static SectionType GetSectionTypeFromName(llvm::StringRef Name) {
1567   if (Name.consume_front(".debug_") || Name.consume_front(".zdebug_")) {
1568     return llvm::StringSwitch<SectionType>(Name)
1569         .Case("abbrev", eSectionTypeDWARFDebugAbbrev)
1570         .Case("abbrev.dwo", eSectionTypeDWARFDebugAbbrevDwo)
1571         .Case("addr", eSectionTypeDWARFDebugAddr)
1572         .Case("aranges", eSectionTypeDWARFDebugAranges)
1573         .Case("cu_index", eSectionTypeDWARFDebugCuIndex)
1574         .Case("frame", eSectionTypeDWARFDebugFrame)
1575         .Case("info", eSectionTypeDWARFDebugInfo)
1576         .Case("info.dwo", eSectionTypeDWARFDebugInfoDwo)
1577         .Cases("line", "line.dwo", eSectionTypeDWARFDebugLine)
1578         .Cases("line_str", "line_str.dwo", eSectionTypeDWARFDebugLineStr)
1579         .Case("loc", eSectionTypeDWARFDebugLoc)
1580         .Case("loc.dwo", eSectionTypeDWARFDebugLocDwo)
1581         .Case("loclists", eSectionTypeDWARFDebugLocLists)
1582         .Case("loclists.dwo", eSectionTypeDWARFDebugLocListsDwo)
1583         .Case("macinfo", eSectionTypeDWARFDebugMacInfo)
1584         .Cases("macro", "macro.dwo", eSectionTypeDWARFDebugMacro)
1585         .Case("names", eSectionTypeDWARFDebugNames)
1586         .Case("pubnames", eSectionTypeDWARFDebugPubNames)
1587         .Case("pubtypes", eSectionTypeDWARFDebugPubTypes)
1588         .Case("ranges", eSectionTypeDWARFDebugRanges)
1589         .Case("rnglists", eSectionTypeDWARFDebugRngLists)
1590         .Case("rnglists.dwo", eSectionTypeDWARFDebugRngListsDwo)
1591         .Case("str", eSectionTypeDWARFDebugStr)
1592         .Case("str.dwo", eSectionTypeDWARFDebugStrDwo)
1593         .Case("str_offsets", eSectionTypeDWARFDebugStrOffsets)
1594         .Case("str_offsets.dwo", eSectionTypeDWARFDebugStrOffsetsDwo)
1595         .Case("tu_index", eSectionTypeDWARFDebugTuIndex)
1596         .Case("types", eSectionTypeDWARFDebugTypes)
1597         .Case("types.dwo", eSectionTypeDWARFDebugTypesDwo)
1598         .Default(eSectionTypeOther);
1599   }
1600   return llvm::StringSwitch<SectionType>(Name)
1601       .Case(".ARM.exidx", eSectionTypeARMexidx)
1602       .Case(".ARM.extab", eSectionTypeARMextab)
1603       .Cases(".bss", ".tbss", eSectionTypeZeroFill)
1604       .Cases(".data", ".tdata", eSectionTypeData)
1605       .Case(".eh_frame", eSectionTypeEHFrame)
1606       .Case(".gnu_debugaltlink", eSectionTypeDWARFGNUDebugAltLink)
1607       .Case(".gosymtab", eSectionTypeGoSymtab)
1608       .Case(".text", eSectionTypeCode)
1609       .Default(eSectionTypeOther);
1610 }
1611 
GetSectionType(const ELFSectionHeaderInfo & H) const1612 SectionType ObjectFileELF::GetSectionType(const ELFSectionHeaderInfo &H) const {
1613   switch (H.sh_type) {
1614   case SHT_PROGBITS:
1615     if (H.sh_flags & SHF_EXECINSTR)
1616       return eSectionTypeCode;
1617     break;
1618   case SHT_SYMTAB:
1619     return eSectionTypeELFSymbolTable;
1620   case SHT_DYNSYM:
1621     return eSectionTypeELFDynamicSymbols;
1622   case SHT_RELA:
1623   case SHT_REL:
1624     return eSectionTypeELFRelocationEntries;
1625   case SHT_DYNAMIC:
1626     return eSectionTypeELFDynamicLinkInfo;
1627   }
1628   return GetSectionTypeFromName(H.section_name.GetStringRef());
1629 }
1630 
GetTargetByteSize(SectionType Type,const ArchSpec & arch)1631 static uint32_t GetTargetByteSize(SectionType Type, const ArchSpec &arch) {
1632   switch (Type) {
1633   case eSectionTypeData:
1634   case eSectionTypeZeroFill:
1635     return arch.GetDataByteSize();
1636   case eSectionTypeCode:
1637     return arch.GetCodeByteSize();
1638   default:
1639     return 1;
1640   }
1641 }
1642 
GetPermissions(const ELFSectionHeader & H)1643 static Permissions GetPermissions(const ELFSectionHeader &H) {
1644   Permissions Perm = Permissions(0);
1645   if (H.sh_flags & SHF_ALLOC)
1646     Perm |= ePermissionsReadable;
1647   if (H.sh_flags & SHF_WRITE)
1648     Perm |= ePermissionsWritable;
1649   if (H.sh_flags & SHF_EXECINSTR)
1650     Perm |= ePermissionsExecutable;
1651   return Perm;
1652 }
1653 
GetPermissions(const ELFProgramHeader & H)1654 static Permissions GetPermissions(const ELFProgramHeader &H) {
1655   Permissions Perm = Permissions(0);
1656   if (H.p_flags & PF_R)
1657     Perm |= ePermissionsReadable;
1658   if (H.p_flags & PF_W)
1659     Perm |= ePermissionsWritable;
1660   if (H.p_flags & PF_X)
1661     Perm |= ePermissionsExecutable;
1662   return Perm;
1663 }
1664 
1665 namespace {
1666 
1667 using VMRange = lldb_private::Range<addr_t, addr_t>;
1668 
1669 struct SectionAddressInfo {
1670   SectionSP Segment;
1671   VMRange Range;
1672 };
1673 
1674 // (Unlinked) ELF object files usually have 0 for every section address, meaning
1675 // we need to compute synthetic addresses in order for "file addresses" from
1676 // different sections to not overlap. This class handles that logic.
1677 class VMAddressProvider {
1678   using VMMap = llvm::IntervalMap<addr_t, SectionSP, 4,
1679                                        llvm::IntervalMapHalfOpenInfo<addr_t>>;
1680 
1681   ObjectFile::Type ObjectType;
1682   addr_t NextVMAddress = 0;
1683   VMMap::Allocator Alloc;
1684   VMMap Segments = VMMap(Alloc);
1685   VMMap Sections = VMMap(Alloc);
1686   lldb_private::Log *Log = GetLogIfAllCategoriesSet(LIBLLDB_LOG_MODULES);
1687   size_t SegmentCount = 0;
1688   std::string SegmentName;
1689 
GetVMRange(const ELFSectionHeader & H)1690   VMRange GetVMRange(const ELFSectionHeader &H) {
1691     addr_t Address = H.sh_addr;
1692     addr_t Size = H.sh_flags & SHF_ALLOC ? H.sh_size : 0;
1693     if (ObjectType == ObjectFile::Type::eTypeObjectFile && Segments.empty() && (H.sh_flags & SHF_ALLOC)) {
1694       NextVMAddress =
1695           llvm::alignTo(NextVMAddress, std::max<addr_t>(H.sh_addralign, 1));
1696       Address = NextVMAddress;
1697       NextVMAddress += Size;
1698     }
1699     return VMRange(Address, Size);
1700   }
1701 
1702 public:
VMAddressProvider(ObjectFile::Type Type,llvm::StringRef SegmentName)1703   VMAddressProvider(ObjectFile::Type Type, llvm::StringRef SegmentName)
1704       : ObjectType(Type), SegmentName(std::string(SegmentName)) {}
1705 
GetNextSegmentName() const1706   std::string GetNextSegmentName() const {
1707     return llvm::formatv("{0}[{1}]", SegmentName, SegmentCount).str();
1708   }
1709 
GetAddressInfo(const ELFProgramHeader & H)1710   llvm::Optional<VMRange> GetAddressInfo(const ELFProgramHeader &H) {
1711     if (H.p_memsz == 0) {
1712       LLDB_LOG(Log, "Ignoring zero-sized {0} segment. Corrupt object file?",
1713                SegmentName);
1714       return llvm::None;
1715     }
1716 
1717     if (Segments.overlaps(H.p_vaddr, H.p_vaddr + H.p_memsz)) {
1718       LLDB_LOG(Log, "Ignoring overlapping {0} segment. Corrupt object file?",
1719                SegmentName);
1720       return llvm::None;
1721     }
1722     return VMRange(H.p_vaddr, H.p_memsz);
1723   }
1724 
GetAddressInfo(const ELFSectionHeader & H)1725   llvm::Optional<SectionAddressInfo> GetAddressInfo(const ELFSectionHeader &H) {
1726     VMRange Range = GetVMRange(H);
1727     SectionSP Segment;
1728     auto It = Segments.find(Range.GetRangeBase());
1729     if ((H.sh_flags & SHF_ALLOC) && It.valid()) {
1730       addr_t MaxSize;
1731       if (It.start() <= Range.GetRangeBase()) {
1732         MaxSize = It.stop() - Range.GetRangeBase();
1733         Segment = *It;
1734       } else
1735         MaxSize = It.start() - Range.GetRangeBase();
1736       if (Range.GetByteSize() > MaxSize) {
1737         LLDB_LOG(Log, "Shortening section crossing segment boundaries. "
1738                       "Corrupt object file?");
1739         Range.SetByteSize(MaxSize);
1740       }
1741     }
1742     if (Range.GetByteSize() > 0 &&
1743         Sections.overlaps(Range.GetRangeBase(), Range.GetRangeEnd())) {
1744       LLDB_LOG(Log, "Ignoring overlapping section. Corrupt object file?");
1745       return llvm::None;
1746     }
1747     if (Segment)
1748       Range.Slide(-Segment->GetFileAddress());
1749     return SectionAddressInfo{Segment, Range};
1750   }
1751 
AddSegment(const VMRange & Range,SectionSP Seg)1752   void AddSegment(const VMRange &Range, SectionSP Seg) {
1753     Segments.insert(Range.GetRangeBase(), Range.GetRangeEnd(), std::move(Seg));
1754     ++SegmentCount;
1755   }
1756 
AddSection(SectionAddressInfo Info,SectionSP Sect)1757   void AddSection(SectionAddressInfo Info, SectionSP Sect) {
1758     if (Info.Range.GetByteSize() == 0)
1759       return;
1760     if (Info.Segment)
1761       Info.Range.Slide(Info.Segment->GetFileAddress());
1762     Sections.insert(Info.Range.GetRangeBase(), Info.Range.GetRangeEnd(),
1763                     std::move(Sect));
1764   }
1765 };
1766 }
1767 
CreateSections(SectionList & unified_section_list)1768 void ObjectFileELF::CreateSections(SectionList &unified_section_list) {
1769   if (m_sections_up)
1770     return;
1771 
1772   m_sections_up = std::make_unique<SectionList>();
1773   VMAddressProvider regular_provider(GetType(), "PT_LOAD");
1774   VMAddressProvider tls_provider(GetType(), "PT_TLS");
1775 
1776   for (const auto &EnumPHdr : llvm::enumerate(ProgramHeaders())) {
1777     const ELFProgramHeader &PHdr = EnumPHdr.value();
1778     if (PHdr.p_type != PT_LOAD && PHdr.p_type != PT_TLS)
1779       continue;
1780 
1781     VMAddressProvider &provider =
1782         PHdr.p_type == PT_TLS ? tls_provider : regular_provider;
1783     auto InfoOr = provider.GetAddressInfo(PHdr);
1784     if (!InfoOr)
1785       continue;
1786 
1787     uint32_t Log2Align = llvm::Log2_64(std::max<elf_xword>(PHdr.p_align, 1));
1788     SectionSP Segment = std::make_shared<Section>(
1789         GetModule(), this, SegmentID(EnumPHdr.index()),
1790         ConstString(provider.GetNextSegmentName()), eSectionTypeContainer,
1791         InfoOr->GetRangeBase(), InfoOr->GetByteSize(), PHdr.p_offset,
1792         PHdr.p_filesz, Log2Align, /*flags*/ 0);
1793     Segment->SetPermissions(GetPermissions(PHdr));
1794     Segment->SetIsThreadSpecific(PHdr.p_type == PT_TLS);
1795     m_sections_up->AddSection(Segment);
1796 
1797     provider.AddSegment(*InfoOr, std::move(Segment));
1798   }
1799 
1800   ParseSectionHeaders();
1801   if (m_section_headers.empty())
1802     return;
1803 
1804   for (SectionHeaderCollIter I = std::next(m_section_headers.begin());
1805        I != m_section_headers.end(); ++I) {
1806     const ELFSectionHeaderInfo &header = *I;
1807 
1808     ConstString &name = I->section_name;
1809     const uint64_t file_size =
1810         header.sh_type == SHT_NOBITS ? 0 : header.sh_size;
1811 
1812     VMAddressProvider &provider =
1813         header.sh_flags & SHF_TLS ? tls_provider : regular_provider;
1814     auto InfoOr = provider.GetAddressInfo(header);
1815     if (!InfoOr)
1816       continue;
1817 
1818     SectionType sect_type = GetSectionType(header);
1819 
1820     const uint32_t target_bytes_size =
1821         GetTargetByteSize(sect_type, m_arch_spec);
1822 
1823     elf::elf_xword log2align =
1824         (header.sh_addralign == 0) ? 0 : llvm::Log2_64(header.sh_addralign);
1825 
1826     SectionSP section_sp(new Section(
1827         InfoOr->Segment, GetModule(), // Module to which this section belongs.
1828         this,            // ObjectFile to which this section belongs and should
1829                          // read section data from.
1830         SectionIndex(I), // Section ID.
1831         name,            // Section name.
1832         sect_type,       // Section type.
1833         InfoOr->Range.GetRangeBase(), // VM address.
1834         InfoOr->Range.GetByteSize(),  // VM size in bytes of this section.
1835         header.sh_offset,             // Offset of this section in the file.
1836         file_size,           // Size of the section as found in the file.
1837         log2align,           // Alignment of the section
1838         header.sh_flags,     // Flags for this section.
1839         target_bytes_size)); // Number of host bytes per target byte
1840 
1841     section_sp->SetPermissions(GetPermissions(header));
1842     section_sp->SetIsThreadSpecific(header.sh_flags & SHF_TLS);
1843     (InfoOr->Segment ? InfoOr->Segment->GetChildren() : *m_sections_up)
1844         .AddSection(section_sp);
1845     provider.AddSection(std::move(*InfoOr), std::move(section_sp));
1846   }
1847 
1848   // For eTypeDebugInfo files, the Symbol Vendor will take care of updating the
1849   // unified section list.
1850   if (GetType() != eTypeDebugInfo)
1851     unified_section_list = *m_sections_up;
1852 
1853   // If there's a .gnu_debugdata section, we'll try to read the .symtab that's
1854   // embedded in there and replace the one in the original object file (if any).
1855   // If there's none in the orignal object file, we add it to it.
1856   if (auto gdd_obj_file = GetGnuDebugDataObjectFile()) {
1857     if (auto gdd_objfile_section_list = gdd_obj_file->GetSectionList()) {
1858       if (SectionSP symtab_section_sp =
1859               gdd_objfile_section_list->FindSectionByType(
1860                   eSectionTypeELFSymbolTable, true)) {
1861         SectionSP module_section_sp = unified_section_list.FindSectionByType(
1862             eSectionTypeELFSymbolTable, true);
1863         if (module_section_sp)
1864           unified_section_list.ReplaceSection(module_section_sp->GetID(),
1865                                               symtab_section_sp);
1866         else
1867           unified_section_list.AddSection(symtab_section_sp);
1868       }
1869     }
1870   }
1871 }
1872 
GetGnuDebugDataObjectFile()1873 std::shared_ptr<ObjectFileELF> ObjectFileELF::GetGnuDebugDataObjectFile() {
1874   if (m_gnu_debug_data_object_file != nullptr)
1875     return m_gnu_debug_data_object_file;
1876 
1877   SectionSP section =
1878       GetSectionList()->FindSectionByName(ConstString(".gnu_debugdata"));
1879   if (!section)
1880     return nullptr;
1881 
1882   if (!lldb_private::lzma::isAvailable()) {
1883     GetModule()->ReportWarning(
1884         "No LZMA support found for reading .gnu_debugdata section");
1885     return nullptr;
1886   }
1887 
1888   // Uncompress the data
1889   DataExtractor data;
1890   section->GetSectionData(data);
1891   llvm::SmallVector<uint8_t, 0> uncompressedData;
1892   auto err = lldb_private::lzma::uncompress(data.GetData(), uncompressedData);
1893   if (err) {
1894     GetModule()->ReportWarning(
1895         "An error occurred while decompression the section %s: %s",
1896         section->GetName().AsCString(), llvm::toString(std::move(err)).c_str());
1897     return nullptr;
1898   }
1899 
1900   // Construct ObjectFileELF object from decompressed buffer
1901   DataBufferSP gdd_data_buf(
1902       new DataBufferHeap(uncompressedData.data(), uncompressedData.size()));
1903   auto fspec = GetFileSpec().CopyByAppendingPathComponent(
1904       llvm::StringRef("gnu_debugdata"));
1905   m_gnu_debug_data_object_file.reset(new ObjectFileELF(
1906       GetModule(), gdd_data_buf, 0, &fspec, 0, gdd_data_buf->GetByteSize()));
1907 
1908   // This line is essential; otherwise a breakpoint can be set but not hit.
1909   m_gnu_debug_data_object_file->SetType(ObjectFile::eTypeDebugInfo);
1910 
1911   ArchSpec spec = m_gnu_debug_data_object_file->GetArchitecture();
1912   if (spec && m_gnu_debug_data_object_file->SetModulesArchitecture(spec))
1913     return m_gnu_debug_data_object_file;
1914 
1915   return nullptr;
1916 }
1917 
1918 // Find the arm/aarch64 mapping symbol character in the given symbol name.
1919 // Mapping symbols have the form of "$<char>[.<any>]*". Additionally we
1920 // recognize cases when the mapping symbol prefixed by an arbitrary string
1921 // because if a symbol prefix added to each symbol in the object file with
1922 // objcopy then the mapping symbols are also prefixed.
FindArmAarch64MappingSymbol(const char * symbol_name)1923 static char FindArmAarch64MappingSymbol(const char *symbol_name) {
1924   if (!symbol_name)
1925     return '\0';
1926 
1927   const char *dollar_pos = ::strchr(symbol_name, '$');
1928   if (!dollar_pos || dollar_pos[1] == '\0')
1929     return '\0';
1930 
1931   if (dollar_pos[2] == '\0' || dollar_pos[2] == '.')
1932     return dollar_pos[1];
1933   return '\0';
1934 }
1935 
1936 #define STO_MIPS_ISA (3 << 6)
1937 #define STO_MICROMIPS (2 << 6)
1938 #define IS_MICROMIPS(ST_OTHER) (((ST_OTHER)&STO_MIPS_ISA) == STO_MICROMIPS)
1939 
1940 // private
ParseSymbols(Symtab * symtab,user_id_t start_id,SectionList * section_list,const size_t num_symbols,const DataExtractor & symtab_data,const DataExtractor & strtab_data)1941 unsigned ObjectFileELF::ParseSymbols(Symtab *symtab, user_id_t start_id,
1942                                      SectionList *section_list,
1943                                      const size_t num_symbols,
1944                                      const DataExtractor &symtab_data,
1945                                      const DataExtractor &strtab_data) {
1946   ELFSymbol symbol;
1947   lldb::offset_t offset = 0;
1948 
1949   static ConstString text_section_name(".text");
1950   static ConstString init_section_name(".init");
1951   static ConstString fini_section_name(".fini");
1952   static ConstString ctors_section_name(".ctors");
1953   static ConstString dtors_section_name(".dtors");
1954 
1955   static ConstString data_section_name(".data");
1956   static ConstString rodata_section_name(".rodata");
1957   static ConstString rodata1_section_name(".rodata1");
1958   static ConstString data2_section_name(".data1");
1959   static ConstString bss_section_name(".bss");
1960   static ConstString opd_section_name(".opd"); // For ppc64
1961 
1962   // On Android the oatdata and the oatexec symbols in the oat and odex files
1963   // covers the full .text section what causes issues with displaying unusable
1964   // symbol name to the user and very slow unwinding speed because the
1965   // instruction emulation based unwind plans try to emulate all instructions
1966   // in these symbols. Don't add these symbols to the symbol list as they have
1967   // no use for the debugger and they are causing a lot of trouble. Filtering
1968   // can't be restricted to Android because this special object file don't
1969   // contain the note section specifying the environment to Android but the
1970   // custom extension and file name makes it highly unlikely that this will
1971   // collide with anything else.
1972   ConstString file_extension = m_file.GetFileNameExtension();
1973   bool skip_oatdata_oatexec =
1974       file_extension == ".oat" || file_extension == ".odex";
1975 
1976   ArchSpec arch = GetArchitecture();
1977   ModuleSP module_sp(GetModule());
1978   SectionList *module_section_list =
1979       module_sp ? module_sp->GetSectionList() : nullptr;
1980 
1981   // Local cache to avoid doing a FindSectionByName for each symbol. The "const
1982   // char*" key must came from a ConstString object so they can be compared by
1983   // pointer
1984   std::unordered_map<const char *, lldb::SectionSP> section_name_to_section;
1985 
1986   unsigned i;
1987   for (i = 0; i < num_symbols; ++i) {
1988     if (!symbol.Parse(symtab_data, &offset))
1989       break;
1990 
1991     const char *symbol_name = strtab_data.PeekCStr(symbol.st_name);
1992     if (!symbol_name)
1993       symbol_name = "";
1994 
1995     // No need to add non-section symbols that have no names
1996     if (symbol.getType() != STT_SECTION &&
1997         (symbol_name == nullptr || symbol_name[0] == '\0'))
1998       continue;
1999 
2000     // Skipping oatdata and oatexec sections if it is requested. See details
2001     // above the definition of skip_oatdata_oatexec for the reasons.
2002     if (skip_oatdata_oatexec && (::strcmp(symbol_name, "oatdata") == 0 ||
2003                                  ::strcmp(symbol_name, "oatexec") == 0))
2004       continue;
2005 
2006     SectionSP symbol_section_sp;
2007     SymbolType symbol_type = eSymbolTypeInvalid;
2008     Elf64_Half shndx = symbol.st_shndx;
2009 
2010     switch (shndx) {
2011     case SHN_ABS:
2012       symbol_type = eSymbolTypeAbsolute;
2013       break;
2014     case SHN_UNDEF:
2015       symbol_type = eSymbolTypeUndefined;
2016       break;
2017     default:
2018       symbol_section_sp = section_list->FindSectionByID(shndx);
2019       break;
2020     }
2021 
2022     // If a symbol is undefined do not process it further even if it has a STT
2023     // type
2024     if (symbol_type != eSymbolTypeUndefined) {
2025       switch (symbol.getType()) {
2026       default:
2027       case STT_NOTYPE:
2028         // The symbol's type is not specified.
2029         break;
2030 
2031       case STT_OBJECT:
2032         // The symbol is associated with a data object, such as a variable, an
2033         // array, etc.
2034         symbol_type = eSymbolTypeData;
2035         break;
2036 
2037       case STT_FUNC:
2038         // The symbol is associated with a function or other executable code.
2039         symbol_type = eSymbolTypeCode;
2040         break;
2041 
2042       case STT_SECTION:
2043         // The symbol is associated with a section. Symbol table entries of
2044         // this type exist primarily for relocation and normally have STB_LOCAL
2045         // binding.
2046         break;
2047 
2048       case STT_FILE:
2049         // Conventionally, the symbol's name gives the name of the source file
2050         // associated with the object file. A file symbol has STB_LOCAL
2051         // binding, its section index is SHN_ABS, and it precedes the other
2052         // STB_LOCAL symbols for the file, if it is present.
2053         symbol_type = eSymbolTypeSourceFile;
2054         break;
2055 
2056       case STT_GNU_IFUNC:
2057         // The symbol is associated with an indirect function. The actual
2058         // function will be resolved if it is referenced.
2059         symbol_type = eSymbolTypeResolver;
2060         break;
2061       }
2062     }
2063 
2064     if (symbol_type == eSymbolTypeInvalid && symbol.getType() != STT_SECTION) {
2065       if (symbol_section_sp) {
2066         ConstString sect_name = symbol_section_sp->GetName();
2067         if (sect_name == text_section_name || sect_name == init_section_name ||
2068             sect_name == fini_section_name || sect_name == ctors_section_name ||
2069             sect_name == dtors_section_name) {
2070           symbol_type = eSymbolTypeCode;
2071         } else if (sect_name == data_section_name ||
2072                    sect_name == data2_section_name ||
2073                    sect_name == rodata_section_name ||
2074                    sect_name == rodata1_section_name ||
2075                    sect_name == bss_section_name) {
2076           symbol_type = eSymbolTypeData;
2077         }
2078       }
2079     }
2080 
2081     int64_t symbol_value_offset = 0;
2082     uint32_t additional_flags = 0;
2083 
2084     if (arch.IsValid()) {
2085       if (arch.GetMachine() == llvm::Triple::arm) {
2086         if (symbol.getBinding() == STB_LOCAL) {
2087           char mapping_symbol = FindArmAarch64MappingSymbol(symbol_name);
2088           if (symbol_type == eSymbolTypeCode) {
2089             switch (mapping_symbol) {
2090             case 'a':
2091               // $a[.<any>]* - marks an ARM instruction sequence
2092               m_address_class_map[symbol.st_value] = AddressClass::eCode;
2093               break;
2094             case 'b':
2095             case 't':
2096               // $b[.<any>]* - marks a THUMB BL instruction sequence
2097               // $t[.<any>]* - marks a THUMB instruction sequence
2098               m_address_class_map[symbol.st_value] =
2099                   AddressClass::eCodeAlternateISA;
2100               break;
2101             case 'd':
2102               // $d[.<any>]* - marks a data item sequence (e.g. lit pool)
2103               m_address_class_map[symbol.st_value] = AddressClass::eData;
2104               break;
2105             }
2106           }
2107           if (mapping_symbol)
2108             continue;
2109         }
2110       } else if (arch.GetMachine() == llvm::Triple::aarch64) {
2111         if (symbol.getBinding() == STB_LOCAL) {
2112           char mapping_symbol = FindArmAarch64MappingSymbol(symbol_name);
2113           if (symbol_type == eSymbolTypeCode) {
2114             switch (mapping_symbol) {
2115             case 'x':
2116               // $x[.<any>]* - marks an A64 instruction sequence
2117               m_address_class_map[symbol.st_value] = AddressClass::eCode;
2118               break;
2119             case 'd':
2120               // $d[.<any>]* - marks a data item sequence (e.g. lit pool)
2121               m_address_class_map[symbol.st_value] = AddressClass::eData;
2122               break;
2123             }
2124           }
2125           if (mapping_symbol)
2126             continue;
2127         }
2128       }
2129 
2130       if (arch.GetMachine() == llvm::Triple::arm) {
2131         if (symbol_type == eSymbolTypeCode) {
2132           if (symbol.st_value & 1) {
2133             // Subtracting 1 from the address effectively unsets the low order
2134             // bit, which results in the address actually pointing to the
2135             // beginning of the symbol. This delta will be used below in
2136             // conjunction with symbol.st_value to produce the final
2137             // symbol_value that we store in the symtab.
2138             symbol_value_offset = -1;
2139             m_address_class_map[symbol.st_value ^ 1] =
2140                 AddressClass::eCodeAlternateISA;
2141           } else {
2142             // This address is ARM
2143             m_address_class_map[symbol.st_value] = AddressClass::eCode;
2144           }
2145         }
2146       }
2147 
2148       /*
2149        * MIPS:
2150        * The bit #0 of an address is used for ISA mode (1 for microMIPS, 0 for
2151        * MIPS).
2152        * This allows processor to switch between microMIPS and MIPS without any
2153        * need
2154        * for special mode-control register. However, apart from .debug_line,
2155        * none of
2156        * the ELF/DWARF sections set the ISA bit (for symbol or section). Use
2157        * st_other
2158        * flag to check whether the symbol is microMIPS and then set the address
2159        * class
2160        * accordingly.
2161       */
2162       if (arch.IsMIPS()) {
2163         if (IS_MICROMIPS(symbol.st_other))
2164           m_address_class_map[symbol.st_value] = AddressClass::eCodeAlternateISA;
2165         else if ((symbol.st_value & 1) && (symbol_type == eSymbolTypeCode)) {
2166           symbol.st_value = symbol.st_value & (~1ull);
2167           m_address_class_map[symbol.st_value] = AddressClass::eCodeAlternateISA;
2168         } else {
2169           if (symbol_type == eSymbolTypeCode)
2170             m_address_class_map[symbol.st_value] = AddressClass::eCode;
2171           else if (symbol_type == eSymbolTypeData)
2172             m_address_class_map[symbol.st_value] = AddressClass::eData;
2173           else
2174             m_address_class_map[symbol.st_value] = AddressClass::eUnknown;
2175         }
2176       }
2177     }
2178 
2179     // symbol_value_offset may contain 0 for ARM symbols or -1 for THUMB
2180     // symbols. See above for more details.
2181     uint64_t symbol_value = symbol.st_value + symbol_value_offset;
2182 
2183     if (symbol_section_sp == nullptr && shndx == SHN_ABS &&
2184         symbol.st_size != 0) {
2185       // We don't have a section for a symbol with non-zero size. Create a new
2186       // section for it so the address range covered by the symbol is also
2187       // covered by the module (represented through the section list). It is
2188       // needed so module lookup for the addresses covered by this symbol will
2189       // be successfull. This case happens for absolute symbols.
2190       ConstString fake_section_name(std::string(".absolute.") + symbol_name);
2191       symbol_section_sp =
2192           std::make_shared<Section>(module_sp, this, SHN_ABS, fake_section_name,
2193                                     eSectionTypeAbsoluteAddress, symbol_value,
2194                                     symbol.st_size, 0, 0, 0, SHF_ALLOC);
2195 
2196       module_section_list->AddSection(symbol_section_sp);
2197       section_list->AddSection(symbol_section_sp);
2198     }
2199 
2200     if (symbol_section_sp &&
2201         CalculateType() != ObjectFile::Type::eTypeObjectFile)
2202       symbol_value -= symbol_section_sp->GetFileAddress();
2203 
2204     if (symbol_section_sp && module_section_list &&
2205         module_section_list != section_list) {
2206       ConstString sect_name = symbol_section_sp->GetName();
2207       auto section_it = section_name_to_section.find(sect_name.GetCString());
2208       if (section_it == section_name_to_section.end())
2209         section_it =
2210             section_name_to_section
2211                 .emplace(sect_name.GetCString(),
2212                          module_section_list->FindSectionByName(sect_name))
2213                 .first;
2214       if (section_it->second)
2215         symbol_section_sp = section_it->second;
2216     }
2217 
2218     bool is_global = symbol.getBinding() == STB_GLOBAL;
2219     uint32_t flags = symbol.st_other << 8 | symbol.st_info | additional_flags;
2220     llvm::StringRef symbol_ref(symbol_name);
2221 
2222     // Symbol names may contain @VERSION suffixes. Find those and strip them
2223     // temporarily.
2224     size_t version_pos = symbol_ref.find('@');
2225     bool has_suffix = version_pos != llvm::StringRef::npos;
2226     llvm::StringRef symbol_bare = symbol_ref.substr(0, version_pos);
2227     Mangled mangled(symbol_bare);
2228 
2229     // Now append the suffix back to mangled and unmangled names. Only do it if
2230     // the demangling was successful (string is not empty).
2231     if (has_suffix) {
2232       llvm::StringRef suffix = symbol_ref.substr(version_pos);
2233 
2234       llvm::StringRef mangled_name = mangled.GetMangledName().GetStringRef();
2235       if (!mangled_name.empty())
2236         mangled.SetMangledName(ConstString((mangled_name + suffix).str()));
2237 
2238       ConstString demangled = mangled.GetDemangledName();
2239       llvm::StringRef demangled_name = demangled.GetStringRef();
2240       if (!demangled_name.empty())
2241         mangled.SetDemangledName(ConstString((demangled_name + suffix).str()));
2242     }
2243 
2244     // In ELF all symbol should have a valid size but it is not true for some
2245     // function symbols coming from hand written assembly. As none of the
2246     // function symbol should have 0 size we try to calculate the size for
2247     // these symbols in the symtab with saying that their original size is not
2248     // valid.
2249     bool symbol_size_valid =
2250         symbol.st_size != 0 || symbol.getType() != STT_FUNC;
2251 
2252     Symbol dc_symbol(
2253         i + start_id, // ID is the original symbol table index.
2254         mangled,
2255         symbol_type,                    // Type of this symbol
2256         is_global,                      // Is this globally visible?
2257         false,                          // Is this symbol debug info?
2258         false,                          // Is this symbol a trampoline?
2259         false,                          // Is this symbol artificial?
2260         AddressRange(symbol_section_sp, // Section in which this symbol is
2261                                         // defined or null.
2262                      symbol_value,      // Offset in section or symbol value.
2263                      symbol.st_size),   // Size in bytes of this symbol.
2264         symbol_size_valid,              // Symbol size is valid
2265         has_suffix,                     // Contains linker annotations?
2266         flags);                         // Symbol flags.
2267     if (symbol.getBinding() == STB_WEAK)
2268       dc_symbol.SetIsWeak(true);
2269     symtab->AddSymbol(dc_symbol);
2270   }
2271   return i;
2272 }
2273 
ParseSymbolTable(Symtab * symbol_table,user_id_t start_id,lldb_private::Section * symtab)2274 unsigned ObjectFileELF::ParseSymbolTable(Symtab *symbol_table,
2275                                          user_id_t start_id,
2276                                          lldb_private::Section *symtab) {
2277   if (symtab->GetObjectFile() != this) {
2278     // If the symbol table section is owned by a different object file, have it
2279     // do the parsing.
2280     ObjectFileELF *obj_file_elf =
2281         static_cast<ObjectFileELF *>(symtab->GetObjectFile());
2282     return obj_file_elf->ParseSymbolTable(symbol_table, start_id, symtab);
2283   }
2284 
2285   // Get section list for this object file.
2286   SectionList *section_list = m_sections_up.get();
2287   if (!section_list)
2288     return 0;
2289 
2290   user_id_t symtab_id = symtab->GetID();
2291   const ELFSectionHeaderInfo *symtab_hdr = GetSectionHeaderByIndex(symtab_id);
2292   assert(symtab_hdr->sh_type == SHT_SYMTAB ||
2293          symtab_hdr->sh_type == SHT_DYNSYM);
2294 
2295   // sh_link: section header index of associated string table.
2296   user_id_t strtab_id = symtab_hdr->sh_link;
2297   Section *strtab = section_list->FindSectionByID(strtab_id).get();
2298 
2299   if (symtab && strtab) {
2300     assert(symtab->GetObjectFile() == this);
2301     assert(strtab->GetObjectFile() == this);
2302 
2303     DataExtractor symtab_data;
2304     DataExtractor strtab_data;
2305     if (ReadSectionData(symtab, symtab_data) &&
2306         ReadSectionData(strtab, strtab_data)) {
2307       size_t num_symbols = symtab_data.GetByteSize() / symtab_hdr->sh_entsize;
2308 
2309       return ParseSymbols(symbol_table, start_id, section_list, num_symbols,
2310                           symtab_data, strtab_data);
2311     }
2312   }
2313 
2314   return 0;
2315 }
2316 
ParseDynamicSymbols()2317 size_t ObjectFileELF::ParseDynamicSymbols() {
2318   if (m_dynamic_symbols.size())
2319     return m_dynamic_symbols.size();
2320 
2321   SectionList *section_list = GetSectionList();
2322   if (!section_list)
2323     return 0;
2324 
2325   // Find the SHT_DYNAMIC section.
2326   Section *dynsym =
2327       section_list->FindSectionByType(eSectionTypeELFDynamicLinkInfo, true)
2328           .get();
2329   if (!dynsym)
2330     return 0;
2331   assert(dynsym->GetObjectFile() == this);
2332 
2333   ELFDynamic symbol;
2334   DataExtractor dynsym_data;
2335   if (ReadSectionData(dynsym, dynsym_data)) {
2336     const lldb::offset_t section_size = dynsym_data.GetByteSize();
2337     lldb::offset_t cursor = 0;
2338 
2339     while (cursor < section_size) {
2340       if (!symbol.Parse(dynsym_data, &cursor))
2341         break;
2342 
2343       m_dynamic_symbols.push_back(symbol);
2344     }
2345   }
2346 
2347   return m_dynamic_symbols.size();
2348 }
2349 
FindDynamicSymbol(unsigned tag)2350 const ELFDynamic *ObjectFileELF::FindDynamicSymbol(unsigned tag) {
2351   if (!ParseDynamicSymbols())
2352     return nullptr;
2353 
2354   DynamicSymbolCollIter I = m_dynamic_symbols.begin();
2355   DynamicSymbolCollIter E = m_dynamic_symbols.end();
2356   for (; I != E; ++I) {
2357     ELFDynamic *symbol = &*I;
2358 
2359     if (symbol->d_tag == tag)
2360       return symbol;
2361   }
2362 
2363   return nullptr;
2364 }
2365 
PLTRelocationType()2366 unsigned ObjectFileELF::PLTRelocationType() {
2367   // DT_PLTREL
2368   //  This member specifies the type of relocation entry to which the
2369   //  procedure linkage table refers. The d_val member holds DT_REL or
2370   //  DT_RELA, as appropriate. All relocations in a procedure linkage table
2371   //  must use the same relocation.
2372   const ELFDynamic *symbol = FindDynamicSymbol(DT_PLTREL);
2373 
2374   if (symbol)
2375     return symbol->d_val;
2376 
2377   return 0;
2378 }
2379 
2380 // Returns the size of the normal plt entries and the offset of the first
2381 // normal plt entry. The 0th entry in the plt table is usually a resolution
2382 // entry which have different size in some architectures then the rest of the
2383 // plt entries.
2384 static std::pair<uint64_t, uint64_t>
GetPltEntrySizeAndOffset(const ELFSectionHeader * rel_hdr,const ELFSectionHeader * plt_hdr)2385 GetPltEntrySizeAndOffset(const ELFSectionHeader *rel_hdr,
2386                          const ELFSectionHeader *plt_hdr) {
2387   const elf_xword num_relocations = rel_hdr->sh_size / rel_hdr->sh_entsize;
2388 
2389   // Clang 3.3 sets entsize to 4 for 32-bit binaries, but the plt entries are
2390   // 16 bytes. So round the entsize up by the alignment if addralign is set.
2391   elf_xword plt_entsize =
2392       plt_hdr->sh_addralign
2393           ? llvm::alignTo(plt_hdr->sh_entsize, plt_hdr->sh_addralign)
2394           : plt_hdr->sh_entsize;
2395 
2396   // Some linkers e.g ld for arm, fill plt_hdr->sh_entsize field incorrectly.
2397   // PLT entries relocation code in general requires multiple instruction and
2398   // should be greater than 4 bytes in most cases. Try to guess correct size
2399   // just in case.
2400   if (plt_entsize <= 4) {
2401     // The linker haven't set the plt_hdr->sh_entsize field. Try to guess the
2402     // size of the plt entries based on the number of entries and the size of
2403     // the plt section with the assumption that the size of the 0th entry is at
2404     // least as big as the size of the normal entries and it isn't much bigger
2405     // then that.
2406     if (plt_hdr->sh_addralign)
2407       plt_entsize = plt_hdr->sh_size / plt_hdr->sh_addralign /
2408                     (num_relocations + 1) * plt_hdr->sh_addralign;
2409     else
2410       plt_entsize = plt_hdr->sh_size / (num_relocations + 1);
2411   }
2412 
2413   elf_xword plt_offset = plt_hdr->sh_size - num_relocations * plt_entsize;
2414 
2415   return std::make_pair(plt_entsize, plt_offset);
2416 }
2417 
ParsePLTRelocations(Symtab * symbol_table,user_id_t start_id,unsigned rel_type,const ELFHeader * hdr,const ELFSectionHeader * rel_hdr,const ELFSectionHeader * plt_hdr,const ELFSectionHeader * sym_hdr,const lldb::SectionSP & plt_section_sp,DataExtractor & rel_data,DataExtractor & symtab_data,DataExtractor & strtab_data)2418 static unsigned ParsePLTRelocations(
2419     Symtab *symbol_table, user_id_t start_id, unsigned rel_type,
2420     const ELFHeader *hdr, const ELFSectionHeader *rel_hdr,
2421     const ELFSectionHeader *plt_hdr, const ELFSectionHeader *sym_hdr,
2422     const lldb::SectionSP &plt_section_sp, DataExtractor &rel_data,
2423     DataExtractor &symtab_data, DataExtractor &strtab_data) {
2424   ELFRelocation rel(rel_type);
2425   ELFSymbol symbol;
2426   lldb::offset_t offset = 0;
2427 
2428   uint64_t plt_offset, plt_entsize;
2429   std::tie(plt_entsize, plt_offset) =
2430       GetPltEntrySizeAndOffset(rel_hdr, plt_hdr);
2431   const elf_xword num_relocations = rel_hdr->sh_size / rel_hdr->sh_entsize;
2432 
2433   typedef unsigned (*reloc_info_fn)(const ELFRelocation &rel);
2434   reloc_info_fn reloc_type;
2435   reloc_info_fn reloc_symbol;
2436 
2437   if (hdr->Is32Bit()) {
2438     reloc_type = ELFRelocation::RelocType32;
2439     reloc_symbol = ELFRelocation::RelocSymbol32;
2440   } else {
2441     reloc_type = ELFRelocation::RelocType64;
2442     reloc_symbol = ELFRelocation::RelocSymbol64;
2443   }
2444 
2445   unsigned slot_type = hdr->GetRelocationJumpSlotType();
2446   unsigned i;
2447   for (i = 0; i < num_relocations; ++i) {
2448     if (!rel.Parse(rel_data, &offset))
2449       break;
2450 
2451     if (reloc_type(rel) != slot_type)
2452       continue;
2453 
2454     lldb::offset_t symbol_offset = reloc_symbol(rel) * sym_hdr->sh_entsize;
2455     if (!symbol.Parse(symtab_data, &symbol_offset))
2456       break;
2457 
2458     const char *symbol_name = strtab_data.PeekCStr(symbol.st_name);
2459     uint64_t plt_index = plt_offset + i * plt_entsize;
2460 
2461     Symbol jump_symbol(
2462         i + start_id,          // Symbol table index
2463         symbol_name,           // symbol name.
2464         eSymbolTypeTrampoline, // Type of this symbol
2465         false,                 // Is this globally visible?
2466         false,                 // Is this symbol debug info?
2467         true,                  // Is this symbol a trampoline?
2468         true,                  // Is this symbol artificial?
2469         plt_section_sp, // Section in which this symbol is defined or null.
2470         plt_index,      // Offset in section or symbol value.
2471         plt_entsize,    // Size in bytes of this symbol.
2472         true,           // Size is valid
2473         false,          // Contains linker annotations?
2474         0);             // Symbol flags.
2475 
2476     symbol_table->AddSymbol(jump_symbol);
2477   }
2478 
2479   return i;
2480 }
2481 
2482 unsigned
ParseTrampolineSymbols(Symtab * symbol_table,user_id_t start_id,const ELFSectionHeaderInfo * rel_hdr,user_id_t rel_id)2483 ObjectFileELF::ParseTrampolineSymbols(Symtab *symbol_table, user_id_t start_id,
2484                                       const ELFSectionHeaderInfo *rel_hdr,
2485                                       user_id_t rel_id) {
2486   assert(rel_hdr->sh_type == SHT_RELA || rel_hdr->sh_type == SHT_REL);
2487 
2488   // The link field points to the associated symbol table.
2489   user_id_t symtab_id = rel_hdr->sh_link;
2490 
2491   // If the link field doesn't point to the appropriate symbol name table then
2492   // try to find it by name as some compiler don't fill in the link fields.
2493   if (!symtab_id)
2494     symtab_id = GetSectionIndexByName(".dynsym");
2495 
2496   // Get PLT section.  We cannot use rel_hdr->sh_info, since current linkers
2497   // point that to the .got.plt or .got section instead of .plt.
2498   user_id_t plt_id = GetSectionIndexByName(".plt");
2499 
2500   if (!symtab_id || !plt_id)
2501     return 0;
2502 
2503   const ELFSectionHeaderInfo *plt_hdr = GetSectionHeaderByIndex(plt_id);
2504   if (!plt_hdr)
2505     return 0;
2506 
2507   const ELFSectionHeaderInfo *sym_hdr = GetSectionHeaderByIndex(symtab_id);
2508   if (!sym_hdr)
2509     return 0;
2510 
2511   SectionList *section_list = m_sections_up.get();
2512   if (!section_list)
2513     return 0;
2514 
2515   Section *rel_section = section_list->FindSectionByID(rel_id).get();
2516   if (!rel_section)
2517     return 0;
2518 
2519   SectionSP plt_section_sp(section_list->FindSectionByID(plt_id));
2520   if (!plt_section_sp)
2521     return 0;
2522 
2523   Section *symtab = section_list->FindSectionByID(symtab_id).get();
2524   if (!symtab)
2525     return 0;
2526 
2527   // sh_link points to associated string table.
2528   Section *strtab = section_list->FindSectionByID(sym_hdr->sh_link).get();
2529   if (!strtab)
2530     return 0;
2531 
2532   DataExtractor rel_data;
2533   if (!ReadSectionData(rel_section, rel_data))
2534     return 0;
2535 
2536   DataExtractor symtab_data;
2537   if (!ReadSectionData(symtab, symtab_data))
2538     return 0;
2539 
2540   DataExtractor strtab_data;
2541   if (!ReadSectionData(strtab, strtab_data))
2542     return 0;
2543 
2544   unsigned rel_type = PLTRelocationType();
2545   if (!rel_type)
2546     return 0;
2547 
2548   return ParsePLTRelocations(symbol_table, start_id, rel_type, &m_header,
2549                              rel_hdr, plt_hdr, sym_hdr, plt_section_sp,
2550                              rel_data, symtab_data, strtab_data);
2551 }
2552 
ApplyRelocations(Symtab * symtab,const ELFHeader * hdr,const ELFSectionHeader * rel_hdr,const ELFSectionHeader * symtab_hdr,const ELFSectionHeader * debug_hdr,DataExtractor & rel_data,DataExtractor & symtab_data,DataExtractor & debug_data,Section * rel_section)2553 unsigned ObjectFileELF::ApplyRelocations(
2554     Symtab *symtab, const ELFHeader *hdr, const ELFSectionHeader *rel_hdr,
2555     const ELFSectionHeader *symtab_hdr, const ELFSectionHeader *debug_hdr,
2556     DataExtractor &rel_data, DataExtractor &symtab_data,
2557     DataExtractor &debug_data, Section *rel_section) {
2558   ELFRelocation rel(rel_hdr->sh_type);
2559   lldb::addr_t offset = 0;
2560   const unsigned num_relocations = rel_hdr->sh_size / rel_hdr->sh_entsize;
2561   typedef unsigned (*reloc_info_fn)(const ELFRelocation &rel);
2562   reloc_info_fn reloc_type;
2563   reloc_info_fn reloc_symbol;
2564 
2565   if (hdr->Is32Bit()) {
2566     reloc_type = ELFRelocation::RelocType32;
2567     reloc_symbol = ELFRelocation::RelocSymbol32;
2568   } else {
2569     reloc_type = ELFRelocation::RelocType64;
2570     reloc_symbol = ELFRelocation::RelocSymbol64;
2571   }
2572 
2573   for (unsigned i = 0; i < num_relocations; ++i) {
2574     if (!rel.Parse(rel_data, &offset))
2575       break;
2576 
2577     Symbol *symbol = nullptr;
2578 
2579     if (hdr->Is32Bit()) {
2580       switch (reloc_type(rel)) {
2581       case R_386_32:
2582       case R_386_PC32:
2583       default:
2584         // FIXME: This asserts with this input:
2585         //
2586         // foo.cpp
2587         // int main(int argc, char **argv) { return 0; }
2588         //
2589         // clang++.exe --target=i686-unknown-linux-gnu -g -c foo.cpp -o foo.o
2590         //
2591         // and running this on the foo.o module.
2592         assert(false && "unexpected relocation type");
2593       }
2594     } else {
2595       switch (reloc_type(rel)) {
2596       case R_AARCH64_ABS64:
2597       case R_X86_64_64: {
2598         symbol = symtab->FindSymbolByID(reloc_symbol(rel));
2599         if (symbol) {
2600           addr_t value = symbol->GetAddressRef().GetFileAddress();
2601           DataBufferSP &data_buffer_sp = debug_data.GetSharedDataBuffer();
2602           uint64_t *dst = reinterpret_cast<uint64_t *>(
2603               data_buffer_sp->GetBytes() + rel_section->GetFileOffset() +
2604               ELFRelocation::RelocOffset64(rel));
2605           uint64_t val_offset = value + ELFRelocation::RelocAddend64(rel);
2606           memcpy(dst, &val_offset, sizeof(uint64_t));
2607         }
2608         break;
2609       }
2610       case R_X86_64_32:
2611       case R_X86_64_32S:
2612       case R_AARCH64_ABS32: {
2613         symbol = symtab->FindSymbolByID(reloc_symbol(rel));
2614         if (symbol) {
2615           addr_t value = symbol->GetAddressRef().GetFileAddress();
2616           value += ELFRelocation::RelocAddend32(rel);
2617           if ((reloc_type(rel) == R_X86_64_32 && (value > UINT32_MAX)) ||
2618               (reloc_type(rel) == R_X86_64_32S &&
2619                ((int64_t)value > INT32_MAX && (int64_t)value < INT32_MIN)) ||
2620               (reloc_type(rel) == R_AARCH64_ABS32 &&
2621                ((int64_t)value > INT32_MAX && (int64_t)value < INT32_MIN))) {
2622             Log *log =
2623                 lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_MODULES);
2624             LLDB_LOGF(log, "Failed to apply debug info relocations");
2625             break;
2626           }
2627           uint32_t truncated_addr = (value & 0xFFFFFFFF);
2628           DataBufferSP &data_buffer_sp = debug_data.GetSharedDataBuffer();
2629           uint32_t *dst = reinterpret_cast<uint32_t *>(
2630               data_buffer_sp->GetBytes() + rel_section->GetFileOffset() +
2631               ELFRelocation::RelocOffset32(rel));
2632           memcpy(dst, &truncated_addr, sizeof(uint32_t));
2633         }
2634         break;
2635       }
2636       case R_X86_64_PC32:
2637       default:
2638         assert(false && "unexpected relocation type");
2639       }
2640     }
2641   }
2642 
2643   return 0;
2644 }
2645 
RelocateDebugSections(const ELFSectionHeader * rel_hdr,user_id_t rel_id,lldb_private::Symtab * thetab)2646 unsigned ObjectFileELF::RelocateDebugSections(const ELFSectionHeader *rel_hdr,
2647                                               user_id_t rel_id,
2648                                               lldb_private::Symtab *thetab) {
2649   assert(rel_hdr->sh_type == SHT_RELA || rel_hdr->sh_type == SHT_REL);
2650 
2651   // Parse in the section list if needed.
2652   SectionList *section_list = GetSectionList();
2653   if (!section_list)
2654     return 0;
2655 
2656   user_id_t symtab_id = rel_hdr->sh_link;
2657   user_id_t debug_id = rel_hdr->sh_info;
2658 
2659   const ELFSectionHeader *symtab_hdr = GetSectionHeaderByIndex(symtab_id);
2660   if (!symtab_hdr)
2661     return 0;
2662 
2663   const ELFSectionHeader *debug_hdr = GetSectionHeaderByIndex(debug_id);
2664   if (!debug_hdr)
2665     return 0;
2666 
2667   Section *rel = section_list->FindSectionByID(rel_id).get();
2668   if (!rel)
2669     return 0;
2670 
2671   Section *symtab = section_list->FindSectionByID(symtab_id).get();
2672   if (!symtab)
2673     return 0;
2674 
2675   Section *debug = section_list->FindSectionByID(debug_id).get();
2676   if (!debug)
2677     return 0;
2678 
2679   DataExtractor rel_data;
2680   DataExtractor symtab_data;
2681   DataExtractor debug_data;
2682 
2683   if (GetData(rel->GetFileOffset(), rel->GetFileSize(), rel_data) &&
2684       GetData(symtab->GetFileOffset(), symtab->GetFileSize(), symtab_data) &&
2685       GetData(debug->GetFileOffset(), debug->GetFileSize(), debug_data)) {
2686     ApplyRelocations(thetab, &m_header, rel_hdr, symtab_hdr, debug_hdr,
2687                      rel_data, symtab_data, debug_data, debug);
2688   }
2689 
2690   return 0;
2691 }
2692 
GetSymtab()2693 Symtab *ObjectFileELF::GetSymtab() {
2694   ModuleSP module_sp(GetModule());
2695   if (!module_sp)
2696     return nullptr;
2697 
2698   // We always want to use the main object file so we (hopefully) only have one
2699   // cached copy of our symtab, dynamic sections, etc.
2700   ObjectFile *module_obj_file = module_sp->GetObjectFile();
2701   if (module_obj_file && module_obj_file != this)
2702     return module_obj_file->GetSymtab();
2703 
2704   if (m_symtab_up == nullptr) {
2705     SectionList *section_list = module_sp->GetSectionList();
2706     if (!section_list)
2707       return nullptr;
2708 
2709     uint64_t symbol_id = 0;
2710     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
2711 
2712     // Sharable objects and dynamic executables usually have 2 distinct symbol
2713     // tables, one named ".symtab", and the other ".dynsym". The dynsym is a
2714     // smaller version of the symtab that only contains global symbols. The
2715     // information found in the dynsym is therefore also found in the symtab,
2716     // while the reverse is not necessarily true.
2717     Section *symtab =
2718         section_list->FindSectionByType(eSectionTypeELFSymbolTable, true).get();
2719     if (symtab) {
2720       m_symtab_up = std::make_unique<Symtab>(symtab->GetObjectFile());
2721       symbol_id += ParseSymbolTable(m_symtab_up.get(), symbol_id, symtab);
2722     }
2723 
2724     // The symtab section is non-allocable and can be stripped, while the
2725     // .dynsym section which should always be always be there. To support the
2726     // minidebuginfo case we parse .dynsym when there's a .gnu_debuginfo
2727     // section, nomatter if .symtab was already parsed or not. This is because
2728     // minidebuginfo normally removes the .symtab symbols which have their
2729     // matching .dynsym counterparts.
2730     if (!symtab ||
2731         GetSectionList()->FindSectionByName(ConstString(".gnu_debugdata"))) {
2732       Section *dynsym =
2733           section_list->FindSectionByType(eSectionTypeELFDynamicSymbols, true)
2734               .get();
2735       if (dynsym) {
2736         if (!m_symtab_up)
2737           m_symtab_up = std::make_unique<Symtab>(dynsym->GetObjectFile());
2738         symbol_id += ParseSymbolTable(m_symtab_up.get(), symbol_id, dynsym);
2739       }
2740     }
2741 
2742     // DT_JMPREL
2743     //      If present, this entry's d_ptr member holds the address of
2744     //      relocation
2745     //      entries associated solely with the procedure linkage table.
2746     //      Separating
2747     //      these relocation entries lets the dynamic linker ignore them during
2748     //      process initialization, if lazy binding is enabled. If this entry is
2749     //      present, the related entries of types DT_PLTRELSZ and DT_PLTREL must
2750     //      also be present.
2751     const ELFDynamic *symbol = FindDynamicSymbol(DT_JMPREL);
2752     if (symbol) {
2753       // Synthesize trampoline symbols to help navigate the PLT.
2754       addr_t addr = symbol->d_ptr;
2755       Section *reloc_section =
2756           section_list->FindSectionContainingFileAddress(addr).get();
2757       if (reloc_section) {
2758         user_id_t reloc_id = reloc_section->GetID();
2759         const ELFSectionHeaderInfo *reloc_header =
2760             GetSectionHeaderByIndex(reloc_id);
2761         assert(reloc_header);
2762 
2763         if (m_symtab_up == nullptr)
2764           m_symtab_up =
2765               std::make_unique<Symtab>(reloc_section->GetObjectFile());
2766 
2767         ParseTrampolineSymbols(m_symtab_up.get(), symbol_id, reloc_header,
2768                                reloc_id);
2769       }
2770     }
2771 
2772     if (DWARFCallFrameInfo *eh_frame =
2773             GetModule()->GetUnwindTable().GetEHFrameInfo()) {
2774       if (m_symtab_up == nullptr)
2775         m_symtab_up = std::make_unique<Symtab>(this);
2776       ParseUnwindSymbols(m_symtab_up.get(), eh_frame);
2777     }
2778 
2779     // If we still don't have any symtab then create an empty instance to avoid
2780     // do the section lookup next time.
2781     if (m_symtab_up == nullptr)
2782       m_symtab_up = std::make_unique<Symtab>(this);
2783 
2784     // In the event that there's no symbol entry for the entry point we'll
2785     // artificially create one. We delegate to the symtab object the figuring
2786     // out of the proper size, this will usually make it span til the next
2787     // symbol it finds in the section. This means that if there are missing
2788     // symbols the entry point might span beyond its function definition.
2789     // We're fine with this as it doesn't make it worse than not having a
2790     // symbol entry at all.
2791     if (CalculateType() == eTypeExecutable) {
2792       ArchSpec arch = GetArchitecture();
2793       auto entry_point_addr = GetEntryPointAddress();
2794       bool is_valid_entry_point =
2795           entry_point_addr.IsValid() && entry_point_addr.IsSectionOffset();
2796       addr_t entry_point_file_addr = entry_point_addr.GetFileAddress();
2797       if (is_valid_entry_point && !m_symtab_up->FindSymbolContainingFileAddress(
2798                                       entry_point_file_addr)) {
2799         uint64_t symbol_id = m_symtab_up->GetNumSymbols();
2800         Symbol symbol(symbol_id,
2801                       GetNextSyntheticSymbolName().GetCString(), // Symbol name.
2802                       eSymbolTypeCode, // Type of this symbol.
2803                       true,            // Is this globally visible?
2804                       false,           // Is this symbol debug info?
2805                       false,           // Is this symbol a trampoline?
2806                       true,            // Is this symbol artificial?
2807                       entry_point_addr.GetSection(), // Section where this
2808                                                      // symbol is defined.
2809                       0,     // Offset in section or symbol value.
2810                       0,     // Size.
2811                       false, // Size is valid.
2812                       false, // Contains linker annotations?
2813                       0);    // Symbol flags.
2814         m_symtab_up->AddSymbol(symbol);
2815         // When the entry point is arm thumb we need to explicitly set its
2816         // class address to reflect that. This is important because expression
2817         // evaluation relies on correctly setting a breakpoint at this
2818         // address.
2819         if (arch.GetMachine() == llvm::Triple::arm &&
2820             (entry_point_file_addr & 1))
2821           m_address_class_map[entry_point_file_addr ^ 1] =
2822               AddressClass::eCodeAlternateISA;
2823         else
2824           m_address_class_map[entry_point_file_addr] = AddressClass::eCode;
2825       }
2826     }
2827 
2828     m_symtab_up->CalculateSymbolSizes();
2829   }
2830 
2831   return m_symtab_up.get();
2832 }
2833 
RelocateSection(lldb_private::Section * section)2834 void ObjectFileELF::RelocateSection(lldb_private::Section *section)
2835 {
2836   static const char *debug_prefix = ".debug";
2837 
2838   // Set relocated bit so we stop getting called, regardless of whether we
2839   // actually relocate.
2840   section->SetIsRelocated(true);
2841 
2842   // We only relocate in ELF relocatable files
2843   if (CalculateType() != eTypeObjectFile)
2844     return;
2845 
2846   const char *section_name = section->GetName().GetCString();
2847   // Can't relocate that which can't be named
2848   if (section_name == nullptr)
2849     return;
2850 
2851   // We don't relocate non-debug sections at the moment
2852   if (strncmp(section_name, debug_prefix, strlen(debug_prefix)))
2853     return;
2854 
2855   // Relocation section names to look for
2856   std::string needle = std::string(".rel") + section_name;
2857   std::string needlea = std::string(".rela") + section_name;
2858 
2859   for (SectionHeaderCollIter I = m_section_headers.begin();
2860        I != m_section_headers.end(); ++I) {
2861     if (I->sh_type == SHT_RELA || I->sh_type == SHT_REL) {
2862       const char *hay_name = I->section_name.GetCString();
2863       if (hay_name == nullptr)
2864         continue;
2865       if (needle == hay_name || needlea == hay_name) {
2866         const ELFSectionHeader &reloc_header = *I;
2867         user_id_t reloc_id = SectionIndex(I);
2868         RelocateDebugSections(&reloc_header, reloc_id, GetSymtab());
2869         break;
2870       }
2871     }
2872   }
2873 }
2874 
ParseUnwindSymbols(Symtab * symbol_table,DWARFCallFrameInfo * eh_frame)2875 void ObjectFileELF::ParseUnwindSymbols(Symtab *symbol_table,
2876                                        DWARFCallFrameInfo *eh_frame) {
2877   SectionList *section_list = GetSectionList();
2878   if (!section_list)
2879     return;
2880 
2881   // First we save the new symbols into a separate list and add them to the
2882   // symbol table after we collected all symbols we want to add. This is
2883   // neccessary because adding a new symbol invalidates the internal index of
2884   // the symtab what causing the next lookup to be slow because it have to
2885   // recalculate the index first.
2886   std::vector<Symbol> new_symbols;
2887 
2888   eh_frame->ForEachFDEEntries([this, symbol_table, section_list, &new_symbols](
2889       lldb::addr_t file_addr, uint32_t size, dw_offset_t) {
2890     Symbol *symbol = symbol_table->FindSymbolAtFileAddress(file_addr);
2891     if (symbol) {
2892       if (!symbol->GetByteSizeIsValid()) {
2893         symbol->SetByteSize(size);
2894         symbol->SetSizeIsSynthesized(true);
2895       }
2896     } else {
2897       SectionSP section_sp =
2898           section_list->FindSectionContainingFileAddress(file_addr);
2899       if (section_sp) {
2900         addr_t offset = file_addr - section_sp->GetFileAddress();
2901         const char *symbol_name = GetNextSyntheticSymbolName().GetCString();
2902         uint64_t symbol_id = symbol_table->GetNumSymbols();
2903         Symbol eh_symbol(
2904             symbol_id,       // Symbol table index.
2905             symbol_name,     // Symbol name.
2906             eSymbolTypeCode, // Type of this symbol.
2907             true,            // Is this globally visible?
2908             false,           // Is this symbol debug info?
2909             false,           // Is this symbol a trampoline?
2910             true,            // Is this symbol artificial?
2911             section_sp,      // Section in which this symbol is defined or null.
2912             offset,          // Offset in section or symbol value.
2913             0,     // Size:          Don't specify the size as an FDE can
2914             false, // Size is valid: cover multiple symbols.
2915             false, // Contains linker annotations?
2916             0);    // Symbol flags.
2917         new_symbols.push_back(eh_symbol);
2918       }
2919     }
2920     return true;
2921   });
2922 
2923   for (const Symbol &s : new_symbols)
2924     symbol_table->AddSymbol(s);
2925 }
2926 
IsStripped()2927 bool ObjectFileELF::IsStripped() {
2928   // TODO: determine this for ELF
2929   return false;
2930 }
2931 
2932 //===----------------------------------------------------------------------===//
2933 // Dump
2934 //
2935 // Dump the specifics of the runtime file container (such as any headers
2936 // segments, sections, etc).
Dump(Stream * s)2937 void ObjectFileELF::Dump(Stream *s) {
2938   ModuleSP module_sp(GetModule());
2939   if (!module_sp) {
2940     return;
2941   }
2942 
2943   std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
2944   s->Printf("%p: ", static_cast<void *>(this));
2945   s->Indent();
2946   s->PutCString("ObjectFileELF");
2947 
2948   ArchSpec header_arch = GetArchitecture();
2949 
2950   *s << ", file = '" << m_file
2951      << "', arch = " << header_arch.GetArchitectureName() << "\n";
2952 
2953   DumpELFHeader(s, m_header);
2954   s->EOL();
2955   DumpELFProgramHeaders(s);
2956   s->EOL();
2957   DumpELFSectionHeaders(s);
2958   s->EOL();
2959   SectionList *section_list = GetSectionList();
2960   if (section_list)
2961     section_list->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true,
2962                        UINT32_MAX);
2963   Symtab *symtab = GetSymtab();
2964   if (symtab)
2965     symtab->Dump(s, nullptr, eSortOrderNone);
2966   s->EOL();
2967   DumpDependentModules(s);
2968   s->EOL();
2969 }
2970 
2971 // DumpELFHeader
2972 //
2973 // Dump the ELF header to the specified output stream
DumpELFHeader(Stream * s,const ELFHeader & header)2974 void ObjectFileELF::DumpELFHeader(Stream *s, const ELFHeader &header) {
2975   s->PutCString("ELF Header\n");
2976   s->Printf("e_ident[EI_MAG0   ] = 0x%2.2x\n", header.e_ident[EI_MAG0]);
2977   s->Printf("e_ident[EI_MAG1   ] = 0x%2.2x '%c'\n", header.e_ident[EI_MAG1],
2978             header.e_ident[EI_MAG1]);
2979   s->Printf("e_ident[EI_MAG2   ] = 0x%2.2x '%c'\n", header.e_ident[EI_MAG2],
2980             header.e_ident[EI_MAG2]);
2981   s->Printf("e_ident[EI_MAG3   ] = 0x%2.2x '%c'\n", header.e_ident[EI_MAG3],
2982             header.e_ident[EI_MAG3]);
2983 
2984   s->Printf("e_ident[EI_CLASS  ] = 0x%2.2x\n", header.e_ident[EI_CLASS]);
2985   s->Printf("e_ident[EI_DATA   ] = 0x%2.2x ", header.e_ident[EI_DATA]);
2986   DumpELFHeader_e_ident_EI_DATA(s, header.e_ident[EI_DATA]);
2987   s->Printf("\ne_ident[EI_VERSION] = 0x%2.2x\n", header.e_ident[EI_VERSION]);
2988   s->Printf("e_ident[EI_PAD    ] = 0x%2.2x\n", header.e_ident[EI_PAD]);
2989 
2990   s->Printf("e_type      = 0x%4.4x ", header.e_type);
2991   DumpELFHeader_e_type(s, header.e_type);
2992   s->Printf("\ne_machine   = 0x%4.4x\n", header.e_machine);
2993   s->Printf("e_version   = 0x%8.8x\n", header.e_version);
2994   s->Printf("e_entry     = 0x%8.8" PRIx64 "\n", header.e_entry);
2995   s->Printf("e_phoff     = 0x%8.8" PRIx64 "\n", header.e_phoff);
2996   s->Printf("e_shoff     = 0x%8.8" PRIx64 "\n", header.e_shoff);
2997   s->Printf("e_flags     = 0x%8.8x\n", header.e_flags);
2998   s->Printf("e_ehsize    = 0x%4.4x\n", header.e_ehsize);
2999   s->Printf("e_phentsize = 0x%4.4x\n", header.e_phentsize);
3000   s->Printf("e_phnum     = 0x%8.8x\n", header.e_phnum);
3001   s->Printf("e_shentsize = 0x%4.4x\n", header.e_shentsize);
3002   s->Printf("e_shnum     = 0x%8.8x\n", header.e_shnum);
3003   s->Printf("e_shstrndx  = 0x%8.8x\n", header.e_shstrndx);
3004 }
3005 
3006 // DumpELFHeader_e_type
3007 //
3008 // Dump an token value for the ELF header member e_type
DumpELFHeader_e_type(Stream * s,elf_half e_type)3009 void ObjectFileELF::DumpELFHeader_e_type(Stream *s, elf_half e_type) {
3010   switch (e_type) {
3011   case ET_NONE:
3012     *s << "ET_NONE";
3013     break;
3014   case ET_REL:
3015     *s << "ET_REL";
3016     break;
3017   case ET_EXEC:
3018     *s << "ET_EXEC";
3019     break;
3020   case ET_DYN:
3021     *s << "ET_DYN";
3022     break;
3023   case ET_CORE:
3024     *s << "ET_CORE";
3025     break;
3026   default:
3027     break;
3028   }
3029 }
3030 
3031 // DumpELFHeader_e_ident_EI_DATA
3032 //
3033 // Dump an token value for the ELF header member e_ident[EI_DATA]
DumpELFHeader_e_ident_EI_DATA(Stream * s,unsigned char ei_data)3034 void ObjectFileELF::DumpELFHeader_e_ident_EI_DATA(Stream *s,
3035                                                   unsigned char ei_data) {
3036   switch (ei_data) {
3037   case ELFDATANONE:
3038     *s << "ELFDATANONE";
3039     break;
3040   case ELFDATA2LSB:
3041     *s << "ELFDATA2LSB - Little Endian";
3042     break;
3043   case ELFDATA2MSB:
3044     *s << "ELFDATA2MSB - Big Endian";
3045     break;
3046   default:
3047     break;
3048   }
3049 }
3050 
3051 // DumpELFProgramHeader
3052 //
3053 // Dump a single ELF program header to the specified output stream
DumpELFProgramHeader(Stream * s,const ELFProgramHeader & ph)3054 void ObjectFileELF::DumpELFProgramHeader(Stream *s,
3055                                          const ELFProgramHeader &ph) {
3056   DumpELFProgramHeader_p_type(s, ph.p_type);
3057   s->Printf(" %8.8" PRIx64 " %8.8" PRIx64 " %8.8" PRIx64, ph.p_offset,
3058             ph.p_vaddr, ph.p_paddr);
3059   s->Printf(" %8.8" PRIx64 " %8.8" PRIx64 " %8.8x (", ph.p_filesz, ph.p_memsz,
3060             ph.p_flags);
3061 
3062   DumpELFProgramHeader_p_flags(s, ph.p_flags);
3063   s->Printf(") %8.8" PRIx64, ph.p_align);
3064 }
3065 
3066 // DumpELFProgramHeader_p_type
3067 //
3068 // Dump an token value for the ELF program header member p_type which describes
3069 // the type of the program header
DumpELFProgramHeader_p_type(Stream * s,elf_word p_type)3070 void ObjectFileELF::DumpELFProgramHeader_p_type(Stream *s, elf_word p_type) {
3071   const int kStrWidth = 15;
3072   switch (p_type) {
3073     CASE_AND_STREAM(s, PT_NULL, kStrWidth);
3074     CASE_AND_STREAM(s, PT_LOAD, kStrWidth);
3075     CASE_AND_STREAM(s, PT_DYNAMIC, kStrWidth);
3076     CASE_AND_STREAM(s, PT_INTERP, kStrWidth);
3077     CASE_AND_STREAM(s, PT_NOTE, kStrWidth);
3078     CASE_AND_STREAM(s, PT_SHLIB, kStrWidth);
3079     CASE_AND_STREAM(s, PT_PHDR, kStrWidth);
3080     CASE_AND_STREAM(s, PT_TLS, kStrWidth);
3081     CASE_AND_STREAM(s, PT_GNU_EH_FRAME, kStrWidth);
3082   default:
3083     s->Printf("0x%8.8x%*s", p_type, kStrWidth - 10, "");
3084     break;
3085   }
3086 }
3087 
3088 // DumpELFProgramHeader_p_flags
3089 //
3090 // Dump an token value for the ELF program header member p_flags
DumpELFProgramHeader_p_flags(Stream * s,elf_word p_flags)3091 void ObjectFileELF::DumpELFProgramHeader_p_flags(Stream *s, elf_word p_flags) {
3092   *s << ((p_flags & PF_X) ? "PF_X" : "    ")
3093      << (((p_flags & PF_X) && (p_flags & PF_W)) ? '+' : ' ')
3094      << ((p_flags & PF_W) ? "PF_W" : "    ")
3095      << (((p_flags & PF_W) && (p_flags & PF_R)) ? '+' : ' ')
3096      << ((p_flags & PF_R) ? "PF_R" : "    ");
3097 }
3098 
3099 // DumpELFProgramHeaders
3100 //
3101 // Dump all of the ELF program header to the specified output stream
DumpELFProgramHeaders(Stream * s)3102 void ObjectFileELF::DumpELFProgramHeaders(Stream *s) {
3103   if (!ParseProgramHeaders())
3104     return;
3105 
3106   s->PutCString("Program Headers\n");
3107   s->PutCString("IDX  p_type          p_offset p_vaddr  p_paddr  "
3108                 "p_filesz p_memsz  p_flags                   p_align\n");
3109   s->PutCString("==== --------------- -------- -------- -------- "
3110                 "-------- -------- ------------------------- --------\n");
3111 
3112   for (const auto &H : llvm::enumerate(m_program_headers)) {
3113     s->Format("[{0,2}] ", H.index());
3114     ObjectFileELF::DumpELFProgramHeader(s, H.value());
3115     s->EOL();
3116   }
3117 }
3118 
3119 // DumpELFSectionHeader
3120 //
3121 // Dump a single ELF section header to the specified output stream
DumpELFSectionHeader(Stream * s,const ELFSectionHeaderInfo & sh)3122 void ObjectFileELF::DumpELFSectionHeader(Stream *s,
3123                                          const ELFSectionHeaderInfo &sh) {
3124   s->Printf("%8.8x ", sh.sh_name);
3125   DumpELFSectionHeader_sh_type(s, sh.sh_type);
3126   s->Printf(" %8.8" PRIx64 " (", sh.sh_flags);
3127   DumpELFSectionHeader_sh_flags(s, sh.sh_flags);
3128   s->Printf(") %8.8" PRIx64 " %8.8" PRIx64 " %8.8" PRIx64, sh.sh_addr,
3129             sh.sh_offset, sh.sh_size);
3130   s->Printf(" %8.8x %8.8x", sh.sh_link, sh.sh_info);
3131   s->Printf(" %8.8" PRIx64 " %8.8" PRIx64, sh.sh_addralign, sh.sh_entsize);
3132 }
3133 
3134 // DumpELFSectionHeader_sh_type
3135 //
3136 // Dump an token value for the ELF section header member sh_type which
3137 // describes the type of the section
DumpELFSectionHeader_sh_type(Stream * s,elf_word sh_type)3138 void ObjectFileELF::DumpELFSectionHeader_sh_type(Stream *s, elf_word sh_type) {
3139   const int kStrWidth = 12;
3140   switch (sh_type) {
3141     CASE_AND_STREAM(s, SHT_NULL, kStrWidth);
3142     CASE_AND_STREAM(s, SHT_PROGBITS, kStrWidth);
3143     CASE_AND_STREAM(s, SHT_SYMTAB, kStrWidth);
3144     CASE_AND_STREAM(s, SHT_STRTAB, kStrWidth);
3145     CASE_AND_STREAM(s, SHT_RELA, kStrWidth);
3146     CASE_AND_STREAM(s, SHT_HASH, kStrWidth);
3147     CASE_AND_STREAM(s, SHT_DYNAMIC, kStrWidth);
3148     CASE_AND_STREAM(s, SHT_NOTE, kStrWidth);
3149     CASE_AND_STREAM(s, SHT_NOBITS, kStrWidth);
3150     CASE_AND_STREAM(s, SHT_REL, kStrWidth);
3151     CASE_AND_STREAM(s, SHT_SHLIB, kStrWidth);
3152     CASE_AND_STREAM(s, SHT_DYNSYM, kStrWidth);
3153     CASE_AND_STREAM(s, SHT_LOPROC, kStrWidth);
3154     CASE_AND_STREAM(s, SHT_HIPROC, kStrWidth);
3155     CASE_AND_STREAM(s, SHT_LOUSER, kStrWidth);
3156     CASE_AND_STREAM(s, SHT_HIUSER, kStrWidth);
3157   default:
3158     s->Printf("0x%8.8x%*s", sh_type, kStrWidth - 10, "");
3159     break;
3160   }
3161 }
3162 
3163 // DumpELFSectionHeader_sh_flags
3164 //
3165 // Dump an token value for the ELF section header member sh_flags
DumpELFSectionHeader_sh_flags(Stream * s,elf_xword sh_flags)3166 void ObjectFileELF::DumpELFSectionHeader_sh_flags(Stream *s,
3167                                                   elf_xword sh_flags) {
3168   *s << ((sh_flags & SHF_WRITE) ? "WRITE" : "     ")
3169      << (((sh_flags & SHF_WRITE) && (sh_flags & SHF_ALLOC)) ? '+' : ' ')
3170      << ((sh_flags & SHF_ALLOC) ? "ALLOC" : "     ")
3171      << (((sh_flags & SHF_ALLOC) && (sh_flags & SHF_EXECINSTR)) ? '+' : ' ')
3172      << ((sh_flags & SHF_EXECINSTR) ? "EXECINSTR" : "         ");
3173 }
3174 
3175 // DumpELFSectionHeaders
3176 //
3177 // Dump all of the ELF section header to the specified output stream
DumpELFSectionHeaders(Stream * s)3178 void ObjectFileELF::DumpELFSectionHeaders(Stream *s) {
3179   if (!ParseSectionHeaders())
3180     return;
3181 
3182   s->PutCString("Section Headers\n");
3183   s->PutCString("IDX  name     type         flags                            "
3184                 "addr     offset   size     link     info     addralgn "
3185                 "entsize  Name\n");
3186   s->PutCString("==== -------- ------------ -------------------------------- "
3187                 "-------- -------- -------- -------- -------- -------- "
3188                 "-------- ====================\n");
3189 
3190   uint32_t idx = 0;
3191   for (SectionHeaderCollConstIter I = m_section_headers.begin();
3192        I != m_section_headers.end(); ++I, ++idx) {
3193     s->Printf("[%2u] ", idx);
3194     ObjectFileELF::DumpELFSectionHeader(s, *I);
3195     const char *section_name = I->section_name.AsCString("");
3196     if (section_name)
3197       *s << ' ' << section_name << "\n";
3198   }
3199 }
3200 
DumpDependentModules(lldb_private::Stream * s)3201 void ObjectFileELF::DumpDependentModules(lldb_private::Stream *s) {
3202   size_t num_modules = ParseDependentModules();
3203 
3204   if (num_modules > 0) {
3205     s->PutCString("Dependent Modules:\n");
3206     for (unsigned i = 0; i < num_modules; ++i) {
3207       const FileSpec &spec = m_filespec_up->GetFileSpecAtIndex(i);
3208       s->Printf("   %s\n", spec.GetFilename().GetCString());
3209     }
3210   }
3211 }
3212 
GetArchitecture()3213 ArchSpec ObjectFileELF::GetArchitecture() {
3214   if (!ParseHeader())
3215     return ArchSpec();
3216 
3217   if (m_section_headers.empty()) {
3218     // Allow elf notes to be parsed which may affect the detected architecture.
3219     ParseSectionHeaders();
3220   }
3221 
3222   if (CalculateType() == eTypeCoreFile &&
3223       !m_arch_spec.TripleOSWasSpecified()) {
3224     // Core files don't have section headers yet they have PT_NOTE program
3225     // headers that might shed more light on the architecture
3226     for (const elf::ELFProgramHeader &H : ProgramHeaders()) {
3227       if (H.p_type != PT_NOTE || H.p_offset == 0 || H.p_filesz == 0)
3228         continue;
3229       DataExtractor data;
3230       if (data.SetData(m_data, H.p_offset, H.p_filesz) == H.p_filesz) {
3231         UUID uuid;
3232         RefineModuleDetailsFromNote(data, m_arch_spec, uuid);
3233       }
3234     }
3235   }
3236   return m_arch_spec;
3237 }
3238 
CalculateType()3239 ObjectFile::Type ObjectFileELF::CalculateType() {
3240   switch (m_header.e_type) {
3241   case llvm::ELF::ET_NONE:
3242     // 0 - No file type
3243     return eTypeUnknown;
3244 
3245   case llvm::ELF::ET_REL:
3246     // 1 - Relocatable file
3247     return eTypeObjectFile;
3248 
3249   case llvm::ELF::ET_EXEC:
3250     // 2 - Executable file
3251     return eTypeExecutable;
3252 
3253   case llvm::ELF::ET_DYN:
3254     // 3 - Shared object file
3255     return eTypeSharedLibrary;
3256 
3257   case ET_CORE:
3258     // 4 - Core file
3259     return eTypeCoreFile;
3260 
3261   default:
3262     break;
3263   }
3264   return eTypeUnknown;
3265 }
3266 
CalculateStrata()3267 ObjectFile::Strata ObjectFileELF::CalculateStrata() {
3268   switch (m_header.e_type) {
3269   case llvm::ELF::ET_NONE:
3270     // 0 - No file type
3271     return eStrataUnknown;
3272 
3273   case llvm::ELF::ET_REL:
3274     // 1 - Relocatable file
3275     return eStrataUnknown;
3276 
3277   case llvm::ELF::ET_EXEC:
3278     // 2 - Executable file
3279     // TODO: is there any way to detect that an executable is a kernel
3280     // related executable by inspecting the program headers, section headers,
3281     // symbols, or any other flag bits???
3282     return eStrataUser;
3283 
3284   case llvm::ELF::ET_DYN:
3285     // 3 - Shared object file
3286     // TODO: is there any way to detect that an shared library is a kernel
3287     // related executable by inspecting the program headers, section headers,
3288     // symbols, or any other flag bits???
3289     return eStrataUnknown;
3290 
3291   case ET_CORE:
3292     // 4 - Core file
3293     // TODO: is there any way to detect that an core file is a kernel
3294     // related executable by inspecting the program headers, section headers,
3295     // symbols, or any other flag bits???
3296     return eStrataUnknown;
3297 
3298   default:
3299     break;
3300   }
3301   return eStrataUnknown;
3302 }
3303 
ReadSectionData(Section * section,lldb::offset_t section_offset,void * dst,size_t dst_len)3304 size_t ObjectFileELF::ReadSectionData(Section *section,
3305                        lldb::offset_t section_offset, void *dst,
3306                        size_t dst_len) {
3307   // If some other objectfile owns this data, pass this to them.
3308   if (section->GetObjectFile() != this)
3309     return section->GetObjectFile()->ReadSectionData(section, section_offset,
3310                                                      dst, dst_len);
3311 
3312   if (!section->Test(SHF_COMPRESSED))
3313     return ObjectFile::ReadSectionData(section, section_offset, dst, dst_len);
3314 
3315   // For compressed sections we need to read to full data to be able to
3316   // decompress.
3317   DataExtractor data;
3318   ReadSectionData(section, data);
3319   return data.CopyData(section_offset, dst_len, dst);
3320 }
3321 
ReadSectionData(Section * section,DataExtractor & section_data)3322 size_t ObjectFileELF::ReadSectionData(Section *section,
3323                                       DataExtractor &section_data) {
3324   // If some other objectfile owns this data, pass this to them.
3325   if (section->GetObjectFile() != this)
3326     return section->GetObjectFile()->ReadSectionData(section, section_data);
3327 
3328   size_t result = ObjectFile::ReadSectionData(section, section_data);
3329   if (result == 0 || !llvm::object::Decompressor::isCompressedELFSection(
3330                          section->Get(), section->GetName().GetStringRef()))
3331     return result;
3332 
3333   auto Decompressor = llvm::object::Decompressor::create(
3334       section->GetName().GetStringRef(),
3335       {reinterpret_cast<const char *>(section_data.GetDataStart()),
3336        size_t(section_data.GetByteSize())},
3337       GetByteOrder() == eByteOrderLittle, GetAddressByteSize() == 8);
3338   if (!Decompressor) {
3339     GetModule()->ReportWarning(
3340         "Unable to initialize decompressor for section '%s': %s",
3341         section->GetName().GetCString(),
3342         llvm::toString(Decompressor.takeError()).c_str());
3343     section_data.Clear();
3344     return 0;
3345   }
3346 
3347   auto buffer_sp =
3348       std::make_shared<DataBufferHeap>(Decompressor->getDecompressedSize(), 0);
3349   if (auto error = Decompressor->decompress(
3350           {reinterpret_cast<char *>(buffer_sp->GetBytes()),
3351            size_t(buffer_sp->GetByteSize())})) {
3352     GetModule()->ReportWarning(
3353         "Decompression of section '%s' failed: %s",
3354         section->GetName().GetCString(),
3355         llvm::toString(std::move(error)).c_str());
3356     section_data.Clear();
3357     return 0;
3358   }
3359 
3360   section_data.SetData(buffer_sp);
3361   return buffer_sp->GetByteSize();
3362 }
3363 
ProgramHeaders()3364 llvm::ArrayRef<ELFProgramHeader> ObjectFileELF::ProgramHeaders() {
3365   ParseProgramHeaders();
3366   return m_program_headers;
3367 }
3368 
GetSegmentData(const ELFProgramHeader & H)3369 DataExtractor ObjectFileELF::GetSegmentData(const ELFProgramHeader &H) {
3370   return DataExtractor(m_data, H.p_offset, H.p_filesz);
3371 }
3372 
AnySegmentHasPhysicalAddress()3373 bool ObjectFileELF::AnySegmentHasPhysicalAddress() {
3374   for (const ELFProgramHeader &H : ProgramHeaders()) {
3375     if (H.p_paddr != 0)
3376       return true;
3377   }
3378   return false;
3379 }
3380 
3381 std::vector<ObjectFile::LoadableData>
GetLoadableData(Target & target)3382 ObjectFileELF::GetLoadableData(Target &target) {
3383   // Create a list of loadable data from loadable segments, using physical
3384   // addresses if they aren't all null
3385   std::vector<LoadableData> loadables;
3386   bool should_use_paddr = AnySegmentHasPhysicalAddress();
3387   for (const ELFProgramHeader &H : ProgramHeaders()) {
3388     LoadableData loadable;
3389     if (H.p_type != llvm::ELF::PT_LOAD)
3390       continue;
3391     loadable.Dest = should_use_paddr ? H.p_paddr : H.p_vaddr;
3392     if (loadable.Dest == LLDB_INVALID_ADDRESS)
3393       continue;
3394     if (H.p_filesz == 0)
3395       continue;
3396     auto segment_data = GetSegmentData(H);
3397     loadable.Contents = llvm::ArrayRef<uint8_t>(segment_data.GetDataStart(),
3398                                                 segment_data.GetByteSize());
3399     loadables.push_back(loadable);
3400   }
3401   return loadables;
3402 }
3403