1//===-- DWARFUnit.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 "DWARFUnit.h"
10
11#include "lldb/Core/Module.h"
12#include "lldb/Symbol/ObjectFile.h"
13#include "lldb/Utility/LLDBAssert.h"
14#include "lldb/Utility/StreamString.h"
15#include "lldb/Utility/Timer.h"
16#include "llvm/DebugInfo/DWARF/DWARFDebugAbbrev.h"
17#include "llvm/DebugInfo/DWARF/DWARFDebugLoc.h"
18#include "llvm/Object/Error.h"
19
20#include "DWARFCompileUnit.h"
21#include "DWARFDebugAranges.h"
22#include "DWARFDebugInfo.h"
23#include "DWARFTypeUnit.h"
24#include "LogChannelDWARF.h"
25#include "SymbolFileDWARFDwo.h"
26#include <optional>
27
28using namespace lldb;
29using namespace lldb_private;
30using namespace lldb_private::dwarf;
31using namespace lldb_private::plugin::dwarf;
32
33extern int g_verbose;
34
35DWARFUnit::DWARFUnit(SymbolFileDWARF &dwarf, lldb::user_id_t uid,
36                     const DWARFUnitHeader &header,
37                     const llvm::DWARFAbbreviationDeclarationSet &abbrevs,
38                     DIERef::Section section, bool is_dwo)
39    : UserID(uid), m_dwarf(dwarf), m_header(header), m_abbrevs(&abbrevs),
40      m_cancel_scopes(false), m_section(section), m_is_dwo(is_dwo),
41      m_has_parsed_non_skeleton_unit(false), m_dwo_id(header.GetDWOId()) {}
42
43DWARFUnit::~DWARFUnit() = default;
44
45// Parses first DIE of a compile unit, excluding DWO.
46void DWARFUnit::ExtractUnitDIENoDwoIfNeeded() {
47  {
48    llvm::sys::ScopedReader lock(m_first_die_mutex);
49    if (m_first_die)
50      return; // Already parsed
51  }
52  llvm::sys::ScopedWriter lock(m_first_die_mutex);
53  if (m_first_die)
54    return; // Already parsed
55
56  ElapsedTime elapsed(m_dwarf.GetDebugInfoParseTimeRef());
57
58  // Set the offset to that of the first DIE and calculate the start of the
59  // next compilation unit header.
60  lldb::offset_t offset = GetFirstDIEOffset();
61
62  // We are in our compile unit, parse starting at the offset we were told to
63  // parse
64  const DWARFDataExtractor &data = GetData();
65  if (offset < GetNextUnitOffset() &&
66      m_first_die.Extract(data, this, &offset)) {
67    AddUnitDIE(m_first_die);
68    return;
69  }
70}
71
72// Parses first DIE of a compile unit including DWO.
73void DWARFUnit::ExtractUnitDIEIfNeeded() {
74  ExtractUnitDIENoDwoIfNeeded();
75
76  if (m_has_parsed_non_skeleton_unit)
77    return;
78
79  m_has_parsed_non_skeleton_unit = true;
80  m_dwo_error.Clear();
81
82  if (!m_dwo_id)
83    return; // No DWO file.
84
85  std::shared_ptr<SymbolFileDWARFDwo> dwo_symbol_file =
86      m_dwarf.GetDwoSymbolFileForCompileUnit(*this, m_first_die);
87  if (!dwo_symbol_file)
88    return;
89
90  DWARFUnit *dwo_cu = dwo_symbol_file->GetDWOCompileUnitForHash(*m_dwo_id);
91
92  if (!dwo_cu) {
93    SetDwoError(Status::createWithFormat(
94        "unable to load .dwo file from \"{0}\" due to ID ({1:x16}) mismatch "
95        "for skeleton DIE at {2:x8}",
96        dwo_symbol_file->GetObjectFile()->GetFileSpec().GetPath().c_str(),
97        *m_dwo_id, m_first_die.GetOffset()));
98    return; // Can't fetch the compile unit from the dwo file.
99  }
100  dwo_cu->SetUserData(this);
101
102  DWARFBaseDIE dwo_cu_die = dwo_cu->GetUnitDIEOnly();
103  if (!dwo_cu_die.IsValid()) {
104    // Can't fetch the compile unit DIE from the dwo file.
105    SetDwoError(Status::createWithFormat(
106        "unable to extract compile unit DIE from .dwo file for skeleton "
107        "DIE at {0:x16}",
108        m_first_die.GetOffset()));
109    return;
110  }
111
112  // Here for DWO CU we want to use the address base set in the skeleton unit
113  // (DW_AT_addr_base) if it is available and use the DW_AT_GNU_addr_base
114  // otherwise. We do that because pre-DWARF v5 could use the DW_AT_GNU_*
115  // attributes which were applicable to the DWO units. The corresponding
116  // DW_AT_* attributes standardized in DWARF v5 are also applicable to the
117  // main unit in contrast.
118  if (m_addr_base)
119    dwo_cu->SetAddrBase(*m_addr_base);
120  else if (m_gnu_addr_base)
121    dwo_cu->SetAddrBase(*m_gnu_addr_base);
122
123  if (GetVersion() <= 4 && m_gnu_ranges_base)
124    dwo_cu->SetRangesBase(*m_gnu_ranges_base);
125  else if (dwo_symbol_file->GetDWARFContext()
126               .getOrLoadRngListsData()
127               .GetByteSize() > 0)
128    dwo_cu->SetRangesBase(llvm::DWARFListTableHeader::getHeaderSize(DWARF32));
129
130  if (GetVersion() >= 5 &&
131      dwo_symbol_file->GetDWARFContext().getOrLoadLocListsData().GetByteSize() >
132          0)
133    dwo_cu->SetLoclistsBase(llvm::DWARFListTableHeader::getHeaderSize(DWARF32));
134
135  dwo_cu->SetBaseAddress(GetBaseAddress());
136
137  m_dwo = std::shared_ptr<DWARFUnit>(std::move(dwo_symbol_file), dwo_cu);
138}
139
140// Parses a compile unit and indexes its DIEs if it hasn't already been done.
141// It will leave this compile unit extracted forever.
142void DWARFUnit::ExtractDIEsIfNeeded() {
143  m_cancel_scopes = true;
144
145  {
146    llvm::sys::ScopedReader lock(m_die_array_mutex);
147    if (!m_die_array.empty())
148      return; // Already parsed
149  }
150  llvm::sys::ScopedWriter lock(m_die_array_mutex);
151  if (!m_die_array.empty())
152    return; // Already parsed
153
154  ExtractDIEsRWLocked();
155}
156
157// Parses a compile unit and indexes its DIEs if it hasn't already been done.
158// It will clear this compile unit after returned instance gets out of scope,
159// no other ScopedExtractDIEs instance is running for this compile unit
160// and no ExtractDIEsIfNeeded() has been executed during this ScopedExtractDIEs
161// lifetime.
162DWARFUnit::ScopedExtractDIEs DWARFUnit::ExtractDIEsScoped() {
163  ScopedExtractDIEs scoped(*this);
164
165  {
166    llvm::sys::ScopedReader lock(m_die_array_mutex);
167    if (!m_die_array.empty())
168      return scoped; // Already parsed
169  }
170  llvm::sys::ScopedWriter lock(m_die_array_mutex);
171  if (!m_die_array.empty())
172    return scoped; // Already parsed
173
174  // Otherwise m_die_array would be already populated.
175  lldbassert(!m_cancel_scopes);
176
177  ExtractDIEsRWLocked();
178  scoped.m_clear_dies = true;
179  return scoped;
180}
181
182DWARFUnit::ScopedExtractDIEs::ScopedExtractDIEs(DWARFUnit &cu) : m_cu(&cu) {
183  m_cu->m_die_array_scoped_mutex.lock_shared();
184}
185
186DWARFUnit::ScopedExtractDIEs::~ScopedExtractDIEs() {
187  if (!m_cu)
188    return;
189  m_cu->m_die_array_scoped_mutex.unlock_shared();
190  if (!m_clear_dies || m_cu->m_cancel_scopes)
191    return;
192  // Be sure no other ScopedExtractDIEs is running anymore.
193  llvm::sys::ScopedWriter lock_scoped(m_cu->m_die_array_scoped_mutex);
194  llvm::sys::ScopedWriter lock(m_cu->m_die_array_mutex);
195  if (m_cu->m_cancel_scopes)
196    return;
197  m_cu->ClearDIEsRWLocked();
198}
199
200DWARFUnit::ScopedExtractDIEs::ScopedExtractDIEs(ScopedExtractDIEs &&rhs)
201    : m_cu(rhs.m_cu), m_clear_dies(rhs.m_clear_dies) {
202  rhs.m_cu = nullptr;
203}
204
205DWARFUnit::ScopedExtractDIEs &
206DWARFUnit::ScopedExtractDIEs::operator=(DWARFUnit::ScopedExtractDIEs &&rhs) {
207  m_cu = rhs.m_cu;
208  rhs.m_cu = nullptr;
209  m_clear_dies = rhs.m_clear_dies;
210  return *this;
211}
212
213// Parses a compile unit and indexes its DIEs, m_die_array_mutex must be
214// held R/W and m_die_array must be empty.
215void DWARFUnit::ExtractDIEsRWLocked() {
216  llvm::sys::ScopedWriter first_die_lock(m_first_die_mutex);
217
218  ElapsedTime elapsed(m_dwarf.GetDebugInfoParseTimeRef());
219  LLDB_SCOPED_TIMERF(
220      "%s",
221      llvm::formatv("{0:x16}: DWARFUnit::ExtractDIEsIfNeeded()", GetOffset())
222          .str()
223          .c_str());
224
225  // Set the offset to that of the first DIE and calculate the start of the
226  // next compilation unit header.
227  lldb::offset_t offset = GetFirstDIEOffset();
228  lldb::offset_t next_cu_offset = GetNextUnitOffset();
229
230  DWARFDebugInfoEntry die;
231
232  uint32_t depth = 0;
233  // We are in our compile unit, parse starting at the offset we were told to
234  // parse
235  const DWARFDataExtractor &data = GetData();
236  std::vector<uint32_t> die_index_stack;
237  die_index_stack.reserve(32);
238  die_index_stack.push_back(0);
239  bool prev_die_had_children = false;
240  while (offset < next_cu_offset && die.Extract(data, this, &offset)) {
241    const bool null_die = die.IsNULL();
242    if (depth == 0) {
243      assert(m_die_array.empty() && "Compile unit DIE already added");
244
245      // The average bytes per DIE entry has been seen to be around 14-20 so
246      // lets pre-reserve half of that since we are now stripping the NULL
247      // tags.
248
249      // Only reserve the memory if we are adding children of the main
250      // compile unit DIE. The compile unit DIE is always the first entry, so
251      // if our size is 1, then we are adding the first compile unit child
252      // DIE and should reserve the memory.
253      m_die_array.reserve(GetDebugInfoSize() / 24);
254      m_die_array.push_back(die);
255
256      if (!m_first_die)
257        AddUnitDIE(m_die_array.front());
258
259      // With -fsplit-dwarf-inlining, clang will emit non-empty skeleton compile
260      // units. We are not able to access these DIE *and* the dwo file
261      // simultaneously. We also don't need to do that as the dwo file will
262      // contain a superset of information. So, we don't even attempt to parse
263      // any remaining DIEs.
264      if (m_dwo) {
265        m_die_array.front().SetHasChildren(false);
266        break;
267      }
268
269    } else {
270      if (null_die) {
271        if (prev_die_had_children) {
272          // This will only happen if a DIE says is has children but all it
273          // contains is a NULL tag. Since we are removing the NULL DIEs from
274          // the list (saves up to 25% in C++ code), we need a way to let the
275          // DIE know that it actually doesn't have children.
276          if (!m_die_array.empty())
277            m_die_array.back().SetHasChildren(false);
278        }
279      } else {
280        die.SetParentIndex(m_die_array.size() - die_index_stack[depth - 1]);
281
282        if (die_index_stack.back())
283          m_die_array[die_index_stack.back()].SetSiblingIndex(
284              m_die_array.size() - die_index_stack.back());
285
286        // Only push the DIE if it isn't a NULL DIE
287        m_die_array.push_back(die);
288      }
289    }
290
291    if (null_die) {
292      // NULL DIE.
293      if (!die_index_stack.empty())
294        die_index_stack.pop_back();
295
296      if (depth > 0)
297        --depth;
298      prev_die_had_children = false;
299    } else {
300      die_index_stack.back() = m_die_array.size() - 1;
301      // Normal DIE
302      const bool die_has_children = die.HasChildren();
303      if (die_has_children) {
304        die_index_stack.push_back(0);
305        ++depth;
306      }
307      prev_die_had_children = die_has_children;
308    }
309
310    if (depth == 0)
311      break; // We are done with this compile unit!
312  }
313
314  if (!m_die_array.empty()) {
315    // The last die cannot have children (if it did, it wouldn't be the last
316    // one). This only makes a difference for malformed dwarf that does not have
317    // a terminating null die.
318    m_die_array.back().SetHasChildren(false);
319
320    if (m_first_die) {
321      // Only needed for the assertion.
322      m_first_die.SetHasChildren(m_die_array.front().HasChildren());
323      lldbassert(m_first_die == m_die_array.front());
324    }
325    m_first_die = m_die_array.front();
326  }
327
328  m_die_array.shrink_to_fit();
329
330  if (m_dwo)
331    m_dwo->ExtractDIEsIfNeeded();
332}
333
334// This is used when a split dwarf is enabled.
335// A skeleton compilation unit may contain the DW_AT_str_offsets_base attribute
336// that points to the first string offset of the CU contribution to the
337// .debug_str_offsets. At the same time, the corresponding split debug unit also
338// may use DW_FORM_strx* forms pointing to its own .debug_str_offsets.dwo and
339// for that case, we should find the offset (skip the section header).
340void DWARFUnit::SetDwoStrOffsetsBase() {
341  lldb::offset_t baseOffset = 0;
342
343  if (const llvm::DWARFUnitIndex::Entry *entry = m_header.GetIndexEntry()) {
344    if (const auto *contribution =
345            entry->getContribution(llvm::DW_SECT_STR_OFFSETS))
346      baseOffset = contribution->getOffset();
347    else
348      return;
349  }
350
351  if (GetVersion() >= 5) {
352    const DWARFDataExtractor &strOffsets =
353        GetSymbolFileDWARF().GetDWARFContext().getOrLoadStrOffsetsData();
354    uint64_t length = strOffsets.GetU32(&baseOffset);
355    if (length == 0xffffffff)
356      length = strOffsets.GetU64(&baseOffset);
357
358    // Check version.
359    if (strOffsets.GetU16(&baseOffset) < 5)
360      return;
361
362    // Skip padding.
363    baseOffset += 2;
364  }
365
366  SetStrOffsetsBase(baseOffset);
367}
368
369std::optional<uint64_t> DWARFUnit::GetDWOId() {
370  ExtractUnitDIENoDwoIfNeeded();
371  return m_dwo_id;
372}
373
374// m_die_array_mutex must be already held as read/write.
375void DWARFUnit::AddUnitDIE(const DWARFDebugInfoEntry &cu_die) {
376  DWARFAttributes attributes = cu_die.GetAttributes(this);
377
378  // Extract DW_AT_addr_base first, as other attributes may need it.
379  for (size_t i = 0; i < attributes.Size(); ++i) {
380    if (attributes.AttributeAtIndex(i) != DW_AT_addr_base)
381      continue;
382    DWARFFormValue form_value;
383    if (attributes.ExtractFormValueAtIndex(i, form_value)) {
384      SetAddrBase(form_value.Unsigned());
385      break;
386    }
387  }
388
389  for (size_t i = 0; i < attributes.Size(); ++i) {
390    dw_attr_t attr = attributes.AttributeAtIndex(i);
391    DWARFFormValue form_value;
392    if (!attributes.ExtractFormValueAtIndex(i, form_value))
393      continue;
394    switch (attr) {
395    default:
396      break;
397    case DW_AT_loclists_base:
398      SetLoclistsBase(form_value.Unsigned());
399      break;
400    case DW_AT_rnglists_base:
401      SetRangesBase(form_value.Unsigned());
402      break;
403    case DW_AT_str_offsets_base:
404      SetStrOffsetsBase(form_value.Unsigned());
405      break;
406    case DW_AT_low_pc:
407      SetBaseAddress(form_value.Address());
408      break;
409    case DW_AT_entry_pc:
410      // If the value was already set by DW_AT_low_pc, don't update it.
411      if (m_base_addr == LLDB_INVALID_ADDRESS)
412        SetBaseAddress(form_value.Address());
413      break;
414    case DW_AT_stmt_list:
415      m_line_table_offset = form_value.Unsigned();
416      break;
417    case DW_AT_GNU_addr_base:
418      m_gnu_addr_base = form_value.Unsigned();
419      break;
420    case DW_AT_GNU_ranges_base:
421      m_gnu_ranges_base = form_value.Unsigned();
422      break;
423    case DW_AT_GNU_dwo_id:
424      m_dwo_id = form_value.Unsigned();
425      break;
426    }
427  }
428
429  if (m_is_dwo) {
430    m_has_parsed_non_skeleton_unit = true;
431    SetDwoStrOffsetsBase();
432    return;
433  }
434}
435
436size_t DWARFUnit::GetDebugInfoSize() const {
437  return GetLengthByteSize() + GetLength() - GetHeaderByteSize();
438}
439
440const llvm::DWARFAbbreviationDeclarationSet *
441DWARFUnit::GetAbbreviations() const {
442  return m_abbrevs;
443}
444
445dw_offset_t DWARFUnit::GetAbbrevOffset() const {
446  return m_abbrevs ? m_abbrevs->getOffset() : DW_INVALID_OFFSET;
447}
448
449dw_offset_t DWARFUnit::GetLineTableOffset() {
450  ExtractUnitDIENoDwoIfNeeded();
451  return m_line_table_offset;
452}
453
454void DWARFUnit::SetAddrBase(dw_addr_t addr_base) { m_addr_base = addr_base; }
455
456// Parse the rangelist table header, including the optional array of offsets
457// following it (DWARF v5 and later).
458template <typename ListTableType>
459static llvm::Expected<ListTableType>
460ParseListTableHeader(const llvm::DWARFDataExtractor &data, uint64_t offset,
461                     DwarfFormat format) {
462  // We are expected to be called with Offset 0 or pointing just past the table
463  // header. Correct Offset in the latter case so that it points to the start
464  // of the header.
465  if (offset == 0) {
466    // This means DW_AT_rnglists_base is missing and therefore DW_FORM_rnglistx
467    // cannot be handled. Returning a default-constructed ListTableType allows
468    // DW_FORM_sec_offset to be supported.
469    return ListTableType();
470  }
471
472  uint64_t HeaderSize = llvm::DWARFListTableHeader::getHeaderSize(format);
473  if (offset < HeaderSize)
474    return llvm::createStringError(std::errc::invalid_argument,
475                                   "did not detect a valid"
476                                   " list table with base = 0x%" PRIx64 "\n",
477                                   offset);
478  offset -= HeaderSize;
479  ListTableType Table;
480  if (llvm::Error E = Table.extractHeaderAndOffsets(data, &offset))
481    return std::move(E);
482  return Table;
483}
484
485void DWARFUnit::SetLoclistsBase(dw_addr_t loclists_base) {
486  uint64_t offset = 0;
487  if (const llvm::DWARFUnitIndex::Entry *entry = m_header.GetIndexEntry()) {
488    const auto *contribution = entry->getContribution(llvm::DW_SECT_LOCLISTS);
489    if (!contribution) {
490      GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
491          "Failed to find location list contribution for CU with DWO Id "
492          "{0:x16}",
493          *GetDWOId());
494      return;
495    }
496    offset += contribution->getOffset();
497  }
498  m_loclists_base = loclists_base;
499
500  uint64_t header_size = llvm::DWARFListTableHeader::getHeaderSize(DWARF32);
501  if (loclists_base < header_size)
502    return;
503
504  m_loclist_table_header.emplace(".debug_loclists", "locations");
505  offset += loclists_base - header_size;
506  if (llvm::Error E = m_loclist_table_header->extract(
507          m_dwarf.GetDWARFContext().getOrLoadLocListsData().GetAsLLVMDWARF(),
508          &offset)) {
509    GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
510        "Failed to extract location list table at offset {0:x16} (location "
511        "list base: {1:x16}): {2}",
512        offset, loclists_base, toString(std::move(E)).c_str());
513  }
514}
515
516std::unique_ptr<llvm::DWARFLocationTable>
517DWARFUnit::GetLocationTable(const DataExtractor &data) const {
518  llvm::DWARFDataExtractor llvm_data(
519      data.GetData(), data.GetByteOrder() == lldb::eByteOrderLittle,
520      data.GetAddressByteSize());
521
522  if (m_is_dwo || GetVersion() >= 5)
523    return std::make_unique<llvm::DWARFDebugLoclists>(llvm_data, GetVersion());
524  return std::make_unique<llvm::DWARFDebugLoc>(llvm_data);
525}
526
527DWARFDataExtractor DWARFUnit::GetLocationData() const {
528  DWARFContext &Ctx = GetSymbolFileDWARF().GetDWARFContext();
529  const DWARFDataExtractor &data =
530      GetVersion() >= 5 ? Ctx.getOrLoadLocListsData() : Ctx.getOrLoadLocData();
531  if (const llvm::DWARFUnitIndex::Entry *entry = m_header.GetIndexEntry()) {
532    if (const auto *contribution = entry->getContribution(
533            GetVersion() >= 5 ? llvm::DW_SECT_LOCLISTS : llvm::DW_SECT_EXT_LOC))
534      return DWARFDataExtractor(data, contribution->getOffset(),
535                                contribution->getLength32());
536    return DWARFDataExtractor();
537  }
538  return data;
539}
540
541DWARFDataExtractor DWARFUnit::GetRnglistData() const {
542  DWARFContext &Ctx = GetSymbolFileDWARF().GetDWARFContext();
543  const DWARFDataExtractor &data = Ctx.getOrLoadRngListsData();
544  if (const llvm::DWARFUnitIndex::Entry *entry = m_header.GetIndexEntry()) {
545    if (const auto *contribution =
546            entry->getContribution(llvm::DW_SECT_RNGLISTS))
547      return DWARFDataExtractor(data, contribution->getOffset(),
548                                contribution->getLength32());
549    GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
550        "Failed to find range list contribution for CU with signature {0:x16}",
551        entry->getSignature());
552
553    return DWARFDataExtractor();
554  }
555  return data;
556}
557
558void DWARFUnit::SetRangesBase(dw_addr_t ranges_base) {
559  lldbassert(!m_rnglist_table_done);
560
561  m_ranges_base = ranges_base;
562}
563
564const std::optional<llvm::DWARFDebugRnglistTable> &
565DWARFUnit::GetRnglistTable() {
566  if (GetVersion() >= 5 && !m_rnglist_table_done) {
567    m_rnglist_table_done = true;
568    if (auto table_or_error =
569            ParseListTableHeader<llvm::DWARFDebugRnglistTable>(
570                GetRnglistData().GetAsLLVMDWARF(), m_ranges_base, DWARF32))
571      m_rnglist_table = std::move(table_or_error.get());
572    else
573      GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
574          "Failed to extract range list table at offset {0:x16}: {1}",
575          m_ranges_base, toString(table_or_error.takeError()).c_str());
576  }
577  return m_rnglist_table;
578}
579
580// This function is called only for DW_FORM_rnglistx.
581llvm::Expected<uint64_t> DWARFUnit::GetRnglistOffset(uint32_t Index) {
582  if (!GetRnglistTable())
583    return llvm::createStringError(std::errc::invalid_argument,
584                                   "missing or invalid range list table");
585  if (!m_ranges_base)
586    return llvm::createStringError(
587        std::errc::invalid_argument,
588        llvm::formatv("DW_FORM_rnglistx cannot be used without "
589                      "DW_AT_rnglists_base for CU at {0:x16}",
590                      GetOffset())
591            .str()
592            .c_str());
593  if (std::optional<uint64_t> off = GetRnglistTable()->getOffsetEntry(
594          GetRnglistData().GetAsLLVM(), Index))
595    return *off + m_ranges_base;
596  return llvm::createStringError(
597      std::errc::invalid_argument,
598      "invalid range list table index %u; OffsetEntryCount is %u, "
599      "DW_AT_rnglists_base is %" PRIu64,
600      Index, GetRnglistTable()->getOffsetEntryCount(), m_ranges_base);
601}
602
603void DWARFUnit::SetStrOffsetsBase(dw_offset_t str_offsets_base) {
604  m_str_offsets_base = str_offsets_base;
605}
606
607dw_addr_t DWARFUnit::ReadAddressFromDebugAddrSection(uint32_t index) const {
608  uint32_t index_size = GetAddressByteSize();
609  dw_offset_t addr_base = GetAddrBase();
610  dw_addr_t offset = addr_base + static_cast<dw_addr_t>(index) * index_size;
611  const DWARFDataExtractor &data =
612      m_dwarf.GetDWARFContext().getOrLoadAddrData();
613  if (data.ValidOffsetForDataOfSize(offset, index_size))
614    return data.GetMaxU64_unchecked(&offset, index_size);
615  return LLDB_INVALID_ADDRESS;
616}
617
618// It may be called only with m_die_array_mutex held R/W.
619void DWARFUnit::ClearDIEsRWLocked() {
620  m_die_array.clear();
621  m_die_array.shrink_to_fit();
622
623  if (m_dwo && !m_dwo->m_cancel_scopes)
624    m_dwo->ClearDIEsRWLocked();
625}
626
627lldb::ByteOrder DWARFUnit::GetByteOrder() const {
628  return m_dwarf.GetObjectFile()->GetByteOrder();
629}
630
631void DWARFUnit::SetBaseAddress(dw_addr_t base_addr) { m_base_addr = base_addr; }
632
633// Compare function DWARFDebugAranges::Range structures
634static bool CompareDIEOffset(const DWARFDebugInfoEntry &die,
635                             const dw_offset_t die_offset) {
636  return die.GetOffset() < die_offset;
637}
638
639// GetDIE()
640//
641// Get the DIE (Debug Information Entry) with the specified offset by first
642// checking if the DIE is contained within this compile unit and grabbing the
643// DIE from this compile unit. Otherwise we grab the DIE from the DWARF file.
644DWARFDIE
645DWARFUnit::GetDIE(dw_offset_t die_offset) {
646  if (die_offset == DW_INVALID_OFFSET)
647    return DWARFDIE(); // Not found
648
649  if (!ContainsDIEOffset(die_offset)) {
650    GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
651        "GetDIE for DIE {0:x16} is outside of its CU {0:x16}", die_offset,
652        GetOffset());
653    return DWARFDIE(); // Not found
654  }
655
656  ExtractDIEsIfNeeded();
657  DWARFDebugInfoEntry::const_iterator end = m_die_array.cend();
658  DWARFDebugInfoEntry::const_iterator pos =
659      lower_bound(m_die_array.cbegin(), end, die_offset, CompareDIEOffset);
660
661  if (pos != end && die_offset == (*pos).GetOffset())
662    return DWARFDIE(this, &(*pos));
663  return DWARFDIE(); // Not found
664}
665
666llvm::StringRef DWARFUnit::PeekDIEName(dw_offset_t die_offset) {
667  DWARFDebugInfoEntry die;
668  if (!die.Extract(GetData(), this, &die_offset))
669    return llvm::StringRef();
670
671  // Does die contain a DW_AT_Name?
672  if (const char *name =
673          die.GetAttributeValueAsString(this, DW_AT_name, nullptr))
674    return name;
675
676  // Does its DW_AT_specification or DW_AT_abstract_origin contain an AT_Name?
677  for (auto attr : {DW_AT_specification, DW_AT_abstract_origin}) {
678    DWARFFormValue form_value;
679    if (!die.GetAttributeValue(this, attr, form_value))
680      continue;
681    auto [unit, offset] = form_value.ReferencedUnitAndOffset();
682    if (unit)
683      if (auto name = unit->PeekDIEName(offset); !name.empty())
684        return name;
685  }
686
687  return llvm::StringRef();
688}
689
690DWARFUnit &DWARFUnit::GetNonSkeletonUnit() {
691  ExtractUnitDIEIfNeeded();
692  if (m_dwo)
693    return *m_dwo;
694  return *this;
695}
696
697uint8_t DWARFUnit::GetAddressByteSize(const DWARFUnit *cu) {
698  if (cu)
699    return cu->GetAddressByteSize();
700  return DWARFUnit::GetDefaultAddressSize();
701}
702
703uint8_t DWARFUnit::GetDefaultAddressSize() { return 4; }
704
705void *DWARFUnit::GetUserData() const { return m_user_data; }
706
707void DWARFUnit::SetUserData(void *d) { m_user_data = d; }
708
709bool DWARFUnit::Supports_DW_AT_APPLE_objc_complete_type() {
710  return GetProducer() != eProducerLLVMGCC;
711}
712
713bool DWARFUnit::DW_AT_decl_file_attributes_are_invalid() {
714  // llvm-gcc makes completely invalid decl file attributes and won't ever be
715  // fixed, so we need to know to ignore these.
716  return GetProducer() == eProducerLLVMGCC;
717}
718
719bool DWARFUnit::Supports_unnamed_objc_bitfields() {
720  if (GetProducer() == eProducerClang)
721    return GetProducerVersion() >= llvm::VersionTuple(425, 0, 13);
722  // Assume all other compilers didn't have incorrect ObjC bitfield info.
723  return true;
724}
725
726void DWARFUnit::ParseProducerInfo() {
727  m_producer = eProducerOther;
728  const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
729  if (!die)
730    return;
731
732  llvm::StringRef producer(
733      die->GetAttributeValueAsString(this, DW_AT_producer, nullptr));
734  if (producer.empty())
735    return;
736
737  static const RegularExpression g_swiftlang_version_regex(
738      llvm::StringRef(R"(swiftlang-([0-9]+\.[0-9]+\.[0-9]+(\.[0-9]+)?))"));
739  static const RegularExpression g_clang_version_regex(
740      llvm::StringRef(R"(clang-([0-9]+\.[0-9]+\.[0-9]+(\.[0-9]+)?))"));
741  static const RegularExpression g_llvm_gcc_regex(
742      llvm::StringRef(R"(4\.[012]\.[01] )"
743                      R"(\(Based on Apple Inc\. build [0-9]+\) )"
744                      R"(\(LLVM build [\.0-9]+\)$)"));
745
746  llvm::SmallVector<llvm::StringRef, 3> matches;
747  if (g_swiftlang_version_regex.Execute(producer, &matches)) {
748    m_producer_version.tryParse(matches[1]);
749    m_producer = eProducerSwift;
750  } else if (producer.contains("clang")) {
751    if (g_clang_version_regex.Execute(producer, &matches))
752      m_producer_version.tryParse(matches[1]);
753    m_producer = eProducerClang;
754  } else if (producer.contains("GNU")) {
755    m_producer = eProducerGCC;
756  } else if (g_llvm_gcc_regex.Execute(producer)) {
757    m_producer = eProducerLLVMGCC;
758  }
759}
760
761DWARFProducer DWARFUnit::GetProducer() {
762  if (m_producer == eProducerInvalid)
763    ParseProducerInfo();
764  return m_producer;
765}
766
767llvm::VersionTuple DWARFUnit::GetProducerVersion() {
768  if (m_producer_version.empty())
769    ParseProducerInfo();
770  return m_producer_version;
771}
772
773uint64_t DWARFUnit::GetDWARFLanguageType() {
774  if (m_language_type)
775    return *m_language_type;
776
777  const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
778  if (!die)
779    m_language_type = 0;
780  else
781    m_language_type = die->GetAttributeValueAsUnsigned(this, DW_AT_language, 0);
782  return *m_language_type;
783}
784
785bool DWARFUnit::GetIsOptimized() {
786  if (m_is_optimized == eLazyBoolCalculate) {
787    const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
788    if (die) {
789      m_is_optimized = eLazyBoolNo;
790      if (die->GetAttributeValueAsUnsigned(this, DW_AT_APPLE_optimized, 0) ==
791          1) {
792        m_is_optimized = eLazyBoolYes;
793      }
794    }
795  }
796  return m_is_optimized == eLazyBoolYes;
797}
798
799FileSpec::Style DWARFUnit::GetPathStyle() {
800  if (!m_comp_dir)
801    ComputeCompDirAndGuessPathStyle();
802  return m_comp_dir->GetPathStyle();
803}
804
805const FileSpec &DWARFUnit::GetCompilationDirectory() {
806  if (!m_comp_dir)
807    ComputeCompDirAndGuessPathStyle();
808  return *m_comp_dir;
809}
810
811const FileSpec &DWARFUnit::GetAbsolutePath() {
812  if (!m_file_spec)
813    ComputeAbsolutePath();
814  return *m_file_spec;
815}
816
817FileSpec DWARFUnit::GetFile(size_t file_idx) {
818  return m_dwarf.GetFile(*this, file_idx);
819}
820
821// DWARF2/3 suggests the form hostname:pathname for compilation directory.
822// Remove the host part if present.
823static llvm::StringRef
824removeHostnameFromPathname(llvm::StringRef path_from_dwarf) {
825  if (!path_from_dwarf.contains(':'))
826    return path_from_dwarf;
827  llvm::StringRef host, path;
828  std::tie(host, path) = path_from_dwarf.split(':');
829
830  if (host.contains('/'))
831    return path_from_dwarf;
832
833  // check whether we have a windows path, and so the first character is a
834  // drive-letter not a hostname.
835  if (host.size() == 1 && llvm::isAlpha(host[0]) &&
836      (path.starts_with("\\") || path.starts_with("/")))
837    return path_from_dwarf;
838
839  return path;
840}
841
842void DWARFUnit::ComputeCompDirAndGuessPathStyle() {
843  m_comp_dir = FileSpec();
844  const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
845  if (!die)
846    return;
847
848  llvm::StringRef comp_dir = removeHostnameFromPathname(
849      die->GetAttributeValueAsString(this, DW_AT_comp_dir, nullptr));
850  if (!comp_dir.empty()) {
851    FileSpec::Style comp_dir_style =
852        FileSpec::GuessPathStyle(comp_dir).value_or(FileSpec::Style::native);
853    m_comp_dir = FileSpec(comp_dir, comp_dir_style);
854  } else {
855    // Try to detect the style based on the DW_AT_name attribute, but just store
856    // the detected style in the m_comp_dir field.
857    const char *name =
858        die->GetAttributeValueAsString(this, DW_AT_name, nullptr);
859    m_comp_dir = FileSpec(
860        "", FileSpec::GuessPathStyle(name).value_or(FileSpec::Style::native));
861  }
862}
863
864void DWARFUnit::ComputeAbsolutePath() {
865  m_file_spec = FileSpec();
866  const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
867  if (!die)
868    return;
869
870  m_file_spec =
871      FileSpec(die->GetAttributeValueAsString(this, DW_AT_name, nullptr),
872               GetPathStyle());
873
874  if (m_file_spec->IsRelative())
875    m_file_spec->MakeAbsolute(GetCompilationDirectory());
876}
877
878SymbolFileDWARFDwo *DWARFUnit::GetDwoSymbolFile() {
879  ExtractUnitDIEIfNeeded();
880  if (m_dwo)
881    return &llvm::cast<SymbolFileDWARFDwo>(m_dwo->GetSymbolFileDWARF());
882  return nullptr;
883}
884
885const DWARFDebugAranges &DWARFUnit::GetFunctionAranges() {
886  if (m_func_aranges_up == nullptr) {
887    m_func_aranges_up = std::make_unique<DWARFDebugAranges>();
888    const DWARFDebugInfoEntry *die = DIEPtr();
889    if (die)
890      die->BuildFunctionAddressRangeTable(this, m_func_aranges_up.get());
891
892    if (m_dwo) {
893      const DWARFDebugInfoEntry *dwo_die = m_dwo->DIEPtr();
894      if (dwo_die)
895        dwo_die->BuildFunctionAddressRangeTable(m_dwo.get(),
896                                                m_func_aranges_up.get());
897    }
898
899    const bool minimize = false;
900    m_func_aranges_up->Sort(minimize);
901  }
902  return *m_func_aranges_up;
903}
904
905llvm::Error DWARFUnitHeader::ApplyIndexEntry(
906    const llvm::DWARFUnitIndex::Entry *index_entry) {
907  // We should only be calling this function when the index entry is not set and
908  // we have a valid one to set it to.
909  assert(index_entry);
910  assert(!m_index_entry);
911
912  if (m_abbr_offset)
913    return llvm::createStringError(
914        llvm::inconvertibleErrorCode(),
915        "Package unit with a non-zero abbreviation offset");
916
917  auto *unit_contrib = index_entry->getContribution();
918  if (!unit_contrib || unit_contrib->getLength32() != m_length + 4)
919    return llvm::createStringError(llvm::inconvertibleErrorCode(),
920                                   "Inconsistent DWARF package unit index");
921
922  auto *abbr_entry = index_entry->getContribution(llvm::DW_SECT_ABBREV);
923  if (!abbr_entry)
924    return llvm::createStringError(
925        llvm::inconvertibleErrorCode(),
926        "DWARF package index missing abbreviation column");
927
928  m_abbr_offset = abbr_entry->getOffset();
929  m_index_entry = index_entry;
930  return llvm::Error::success();
931}
932
933llvm::Expected<DWARFUnitHeader>
934DWARFUnitHeader::extract(const DWARFDataExtractor &data,
935                         DIERef::Section section, DWARFContext &context,
936                         lldb::offset_t *offset_ptr) {
937  DWARFUnitHeader header;
938  header.m_offset = *offset_ptr;
939  header.m_length = data.GetDWARFInitialLength(offset_ptr);
940  header.m_version = data.GetU16(offset_ptr);
941  if (header.m_version == 5) {
942    header.m_unit_type = data.GetU8(offset_ptr);
943    header.m_addr_size = data.GetU8(offset_ptr);
944    header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
945    if (header.m_unit_type == llvm::dwarf::DW_UT_skeleton ||
946        header.m_unit_type == llvm::dwarf::DW_UT_split_compile)
947      header.m_dwo_id = data.GetU64(offset_ptr);
948  } else {
949    header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
950    header.m_addr_size = data.GetU8(offset_ptr);
951    header.m_unit_type =
952        section == DIERef::Section::DebugTypes ? DW_UT_type : DW_UT_compile;
953  }
954
955  if (header.IsTypeUnit()) {
956    header.m_type_hash = data.GetU64(offset_ptr);
957    header.m_type_offset = data.GetDWARFOffset(offset_ptr);
958  }
959
960  bool length_OK = data.ValidOffset(header.GetNextUnitOffset() - 1);
961  bool version_OK = SymbolFileDWARF::SupportedVersion(header.m_version);
962  bool addr_size_OK = (header.m_addr_size == 2) || (header.m_addr_size == 4) ||
963                      (header.m_addr_size == 8);
964  bool type_offset_OK =
965      !header.IsTypeUnit() || (header.m_type_offset <= header.GetLength());
966
967  if (!length_OK)
968    return llvm::make_error<llvm::object::GenericBinaryError>(
969        "Invalid unit length");
970  if (!version_OK)
971    return llvm::make_error<llvm::object::GenericBinaryError>(
972        "Unsupported unit version");
973  if (!addr_size_OK)
974    return llvm::make_error<llvm::object::GenericBinaryError>(
975        "Invalid unit address size");
976  if (!type_offset_OK)
977    return llvm::make_error<llvm::object::GenericBinaryError>(
978        "Type offset out of range");
979
980  return header;
981}
982
983llvm::Expected<DWARFUnitSP>
984DWARFUnit::extract(SymbolFileDWARF &dwarf, user_id_t uid,
985                   const DWARFDataExtractor &debug_info,
986                   DIERef::Section section, lldb::offset_t *offset_ptr) {
987  assert(debug_info.ValidOffset(*offset_ptr));
988
989  DWARFContext &context = dwarf.GetDWARFContext();
990  auto expected_header =
991      DWARFUnitHeader::extract(debug_info, section, context, offset_ptr);
992  if (!expected_header)
993    return expected_header.takeError();
994
995  if (context.isDwo()) {
996    const llvm::DWARFUnitIndex::Entry *entry = nullptr;
997    const llvm::DWARFUnitIndex &index = expected_header->IsTypeUnit()
998                                            ? context.GetAsLLVM().getTUIndex()
999                                            : context.GetAsLLVM().getCUIndex();
1000    if (index) {
1001      if (expected_header->IsTypeUnit())
1002        entry = index.getFromHash(expected_header->GetTypeHash());
1003      else if (auto dwo_id = expected_header->GetDWOId())
1004        entry = index.getFromHash(*dwo_id);
1005    }
1006    if (!entry)
1007      entry = index.getFromOffset(expected_header->GetOffset());
1008    if (entry)
1009      if (llvm::Error err = expected_header->ApplyIndexEntry(entry))
1010        return std::move(err);
1011  }
1012
1013  const llvm::DWARFDebugAbbrev *abbr = dwarf.DebugAbbrev();
1014  if (!abbr)
1015    return llvm::make_error<llvm::object::GenericBinaryError>(
1016        "No debug_abbrev data");
1017
1018  bool abbr_offset_OK =
1019      dwarf.GetDWARFContext().getOrLoadAbbrevData().ValidOffset(
1020          expected_header->GetAbbrOffset());
1021  if (!abbr_offset_OK)
1022    return llvm::make_error<llvm::object::GenericBinaryError>(
1023        "Abbreviation offset for unit is not valid");
1024
1025  llvm::Expected<const llvm::DWARFAbbreviationDeclarationSet *> abbrevs_or_err =
1026      abbr->getAbbreviationDeclarationSet(expected_header->GetAbbrOffset());
1027  if (!abbrevs_or_err)
1028    return abbrevs_or_err.takeError();
1029
1030  const llvm::DWARFAbbreviationDeclarationSet *abbrevs = *abbrevs_or_err;
1031  if (!abbrevs)
1032    return llvm::make_error<llvm::object::GenericBinaryError>(
1033        "No abbrev exists at the specified offset.");
1034
1035  bool is_dwo = dwarf.GetDWARFContext().isDwo();
1036  if (expected_header->IsTypeUnit())
1037    return DWARFUnitSP(new DWARFTypeUnit(dwarf, uid, *expected_header, *abbrevs,
1038                                         section, is_dwo));
1039  return DWARFUnitSP(new DWARFCompileUnit(dwarf, uid, *expected_header,
1040                                          *abbrevs, section, is_dwo));
1041}
1042
1043const lldb_private::DWARFDataExtractor &DWARFUnit::GetData() const {
1044  return m_section == DIERef::Section::DebugTypes
1045             ? m_dwarf.GetDWARFContext().getOrLoadDebugTypesData()
1046             : m_dwarf.GetDWARFContext().getOrLoadDebugInfoData();
1047}
1048
1049uint32_t DWARFUnit::GetHeaderByteSize() const {
1050  switch (m_header.GetUnitType()) {
1051  case llvm::dwarf::DW_UT_compile:
1052  case llvm::dwarf::DW_UT_partial:
1053    return GetVersion() < 5 ? 11 : 12;
1054  case llvm::dwarf::DW_UT_skeleton:
1055  case llvm::dwarf::DW_UT_split_compile:
1056    return 20;
1057  case llvm::dwarf::DW_UT_type:
1058  case llvm::dwarf::DW_UT_split_type:
1059    return GetVersion() < 5 ? 23 : 24;
1060  }
1061  llvm_unreachable("invalid UnitType.");
1062}
1063
1064std::optional<uint64_t>
1065DWARFUnit::GetStringOffsetSectionItem(uint32_t index) const {
1066  offset_t offset = GetStrOffsetsBase() + index * 4;
1067  return m_dwarf.GetDWARFContext().getOrLoadStrOffsetsData().GetU32(&offset);
1068}
1069
1070llvm::Expected<DWARFRangeList>
1071DWARFUnit::FindRnglistFromOffset(dw_offset_t offset) {
1072  if (GetVersion() <= 4) {
1073    const DWARFDebugRanges *debug_ranges = m_dwarf.GetDebugRanges();
1074    if (!debug_ranges)
1075      return llvm::make_error<llvm::object::GenericBinaryError>(
1076          "No debug_ranges section");
1077    return debug_ranges->FindRanges(this, offset);
1078  }
1079
1080  if (!GetRnglistTable())
1081    return llvm::createStringError(std::errc::invalid_argument,
1082                                   "missing or invalid range list table");
1083
1084  llvm::DWARFDataExtractor data = GetRnglistData().GetAsLLVMDWARF();
1085
1086  // As DW_AT_rnglists_base may be missing we need to call setAddressSize.
1087  data.setAddressSize(m_header.GetAddressByteSize());
1088  auto range_list_or_error = GetRnglistTable()->findList(data, offset);
1089  if (!range_list_or_error)
1090    return range_list_or_error.takeError();
1091
1092  llvm::Expected<llvm::DWARFAddressRangesVector> llvm_ranges =
1093      range_list_or_error->getAbsoluteRanges(
1094          llvm::object::SectionedAddress{GetBaseAddress()},
1095          GetAddressByteSize(), [&](uint32_t index) {
1096            uint32_t index_size = GetAddressByteSize();
1097            dw_offset_t addr_base = GetAddrBase();
1098            lldb::offset_t offset =
1099                addr_base + static_cast<lldb::offset_t>(index) * index_size;
1100            return llvm::object::SectionedAddress{
1101                m_dwarf.GetDWARFContext().getOrLoadAddrData().GetMaxU64(
1102                    &offset, index_size)};
1103          });
1104  if (!llvm_ranges)
1105    return llvm_ranges.takeError();
1106
1107  DWARFRangeList ranges;
1108  for (const llvm::DWARFAddressRange &llvm_range : *llvm_ranges) {
1109    ranges.Append(DWARFRangeList::Entry(llvm_range.LowPC,
1110                                        llvm_range.HighPC - llvm_range.LowPC));
1111  }
1112  return ranges;
1113}
1114
1115llvm::Expected<DWARFRangeList> DWARFUnit::FindRnglistFromIndex(uint32_t index) {
1116  llvm::Expected<uint64_t> maybe_offset = GetRnglistOffset(index);
1117  if (!maybe_offset)
1118    return maybe_offset.takeError();
1119  return FindRnglistFromOffset(*maybe_offset);
1120}
1121
1122bool DWARFUnit::HasAny(llvm::ArrayRef<dw_tag_t> tags) {
1123  ExtractUnitDIEIfNeeded();
1124  if (m_dwo)
1125    return m_dwo->HasAny(tags);
1126
1127  for (const auto &die : m_die_array) {
1128    for (const auto tag : tags) {
1129      if (tag == die.Tag())
1130        return true;
1131    }
1132  }
1133  return false;
1134}
1135