xref: /freebsd/contrib/llvm-project/lldb/source/Plugins/SymbolFile/DWARF/DWARFUnit.cpp (revision 6966ac055c3b7a39266fb982493330df7a097997)
1 //===-- DWARFUnit.cpp -------------------------------------------*- C++ -*-===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
9 #include "DWARFUnit.h"
10 
11 #include "lldb/Core/Module.h"
12 #include "lldb/Host/StringConvert.h"
13 #include "lldb/Symbol/ObjectFile.h"
14 #include "lldb/Utility/LLDBAssert.h"
15 #include "lldb/Utility/StreamString.h"
16 #include "lldb/Utility/Timer.h"
17 #include "llvm/Object/Error.h"
18 
19 #include "DWARFCompileUnit.h"
20 #include "DWARFDebugAranges.h"
21 #include "DWARFDebugInfo.h"
22 #include "DWARFTypeUnit.h"
23 #include "LogChannelDWARF.h"
24 #include "SymbolFileDWARFDwo.h"
25 
26 using namespace lldb;
27 using namespace lldb_private;
28 using namespace std;
29 
30 extern int g_verbose;
31 
32 DWARFUnit::DWARFUnit(SymbolFileDWARF &dwarf, lldb::user_id_t uid,
33                      const DWARFUnitHeader &header,
34                      const DWARFAbbreviationDeclarationSet &abbrevs,
35                      DIERef::Section section)
36     : UserID(uid), m_dwarf(dwarf), m_header(header), m_abbrevs(&abbrevs),
37       m_cancel_scopes(false), m_section(section) {}
38 
39 DWARFUnit::~DWARFUnit() = default;
40 
41 // Parses first DIE of a compile unit.
42 void DWARFUnit::ExtractUnitDIEIfNeeded() {
43   {
44     llvm::sys::ScopedReader lock(m_first_die_mutex);
45     if (m_first_die)
46       return; // Already parsed
47   }
48   llvm::sys::ScopedWriter lock(m_first_die_mutex);
49   if (m_first_die)
50     return; // Already parsed
51 
52   static Timer::Category func_cat(LLVM_PRETTY_FUNCTION);
53   Timer scoped_timer(func_cat, "%8.8x: DWARFUnit::ExtractUnitDIEIfNeeded()",
54                      GetOffset());
55 
56   // Set the offset to that of the first DIE and calculate the start of the
57   // next compilation unit header.
58   lldb::offset_t offset = GetFirstDIEOffset();
59 
60   // We are in our compile unit, parse starting at the offset we were told to
61   // parse
62   const DWARFDataExtractor &data = GetData();
63   if (offset < GetNextUnitOffset() &&
64       m_first_die.Extract(data, this, &offset)) {
65     AddUnitDIE(m_first_die);
66     return;
67   }
68 }
69 
70 // Parses a compile unit and indexes its DIEs if it hasn't already been done.
71 // It will leave this compile unit extracted forever.
72 void DWARFUnit::ExtractDIEsIfNeeded() {
73   m_cancel_scopes = true;
74 
75   {
76     llvm::sys::ScopedReader lock(m_die_array_mutex);
77     if (!m_die_array.empty())
78       return; // Already parsed
79   }
80   llvm::sys::ScopedWriter lock(m_die_array_mutex);
81   if (!m_die_array.empty())
82     return; // Already parsed
83 
84   ExtractDIEsRWLocked();
85 }
86 
87 // Parses a compile unit and indexes its DIEs if it hasn't already been done.
88 // It will clear this compile unit after returned instance gets out of scope,
89 // no other ScopedExtractDIEs instance is running for this compile unit
90 // and no ExtractDIEsIfNeeded() has been executed during this ScopedExtractDIEs
91 // lifetime.
92 DWARFUnit::ScopedExtractDIEs DWARFUnit::ExtractDIEsScoped() {
93   ScopedExtractDIEs scoped(*this);
94 
95   {
96     llvm::sys::ScopedReader lock(m_die_array_mutex);
97     if (!m_die_array.empty())
98       return scoped; // Already parsed
99   }
100   llvm::sys::ScopedWriter lock(m_die_array_mutex);
101   if (!m_die_array.empty())
102     return scoped; // Already parsed
103 
104   // Otherwise m_die_array would be already populated.
105   lldbassert(!m_cancel_scopes);
106 
107   ExtractDIEsRWLocked();
108   scoped.m_clear_dies = true;
109   return scoped;
110 }
111 
112 DWARFUnit::ScopedExtractDIEs::ScopedExtractDIEs(DWARFUnit &cu) : m_cu(&cu) {
113   m_cu->m_die_array_scoped_mutex.lock_shared();
114 }
115 
116 DWARFUnit::ScopedExtractDIEs::~ScopedExtractDIEs() {
117   if (!m_cu)
118     return;
119   m_cu->m_die_array_scoped_mutex.unlock_shared();
120   if (!m_clear_dies || m_cu->m_cancel_scopes)
121     return;
122   // Be sure no other ScopedExtractDIEs is running anymore.
123   llvm::sys::ScopedWriter lock_scoped(m_cu->m_die_array_scoped_mutex);
124   llvm::sys::ScopedWriter lock(m_cu->m_die_array_mutex);
125   if (m_cu->m_cancel_scopes)
126     return;
127   m_cu->ClearDIEsRWLocked();
128 }
129 
130 DWARFUnit::ScopedExtractDIEs::ScopedExtractDIEs(ScopedExtractDIEs &&rhs)
131     : m_cu(rhs.m_cu), m_clear_dies(rhs.m_clear_dies) {
132   rhs.m_cu = nullptr;
133 }
134 
135 DWARFUnit::ScopedExtractDIEs &DWARFUnit::ScopedExtractDIEs::operator=(
136     DWARFUnit::ScopedExtractDIEs &&rhs) {
137   m_cu = rhs.m_cu;
138   rhs.m_cu = nullptr;
139   m_clear_dies = rhs.m_clear_dies;
140   return *this;
141 }
142 
143 // Parses a compile unit and indexes its DIEs, m_die_array_mutex must be
144 // held R/W and m_die_array must be empty.
145 void DWARFUnit::ExtractDIEsRWLocked() {
146   llvm::sys::ScopedWriter first_die_lock(m_first_die_mutex);
147 
148   static Timer::Category func_cat(LLVM_PRETTY_FUNCTION);
149   Timer scoped_timer(func_cat, "%8.8x: DWARFUnit::ExtractDIEsIfNeeded()",
150                      GetOffset());
151 
152   // Set the offset to that of the first DIE and calculate the start of the
153   // next compilation unit header.
154   lldb::offset_t offset = GetFirstDIEOffset();
155   lldb::offset_t next_cu_offset = GetNextUnitOffset();
156 
157   DWARFDebugInfoEntry die;
158 
159   uint32_t depth = 0;
160   // We are in our compile unit, parse starting at the offset we were told to
161   // parse
162   const DWARFDataExtractor &data = GetData();
163   std::vector<uint32_t> die_index_stack;
164   die_index_stack.reserve(32);
165   die_index_stack.push_back(0);
166   bool prev_die_had_children = false;
167   while (offset < next_cu_offset && die.Extract(data, this, &offset)) {
168     const bool null_die = die.IsNULL();
169     if (depth == 0) {
170       assert(m_die_array.empty() && "Compile unit DIE already added");
171 
172       // The average bytes per DIE entry has been seen to be around 14-20 so
173       // lets pre-reserve half of that since we are now stripping the NULL
174       // tags.
175 
176       // Only reserve the memory if we are adding children of the main
177       // compile unit DIE. The compile unit DIE is always the first entry, so
178       // if our size is 1, then we are adding the first compile unit child
179       // DIE and should reserve the memory.
180       m_die_array.reserve(GetDebugInfoSize() / 24);
181       m_die_array.push_back(die);
182 
183       if (!m_first_die)
184         AddUnitDIE(m_die_array.front());
185 
186       // With -fsplit-dwarf-inlining, clang will emit non-empty skeleton compile
187       // units. We are not able to access these DIE *and* the dwo file
188       // simultaneously. We also don't need to do that as the dwo file will
189       // contain a superset of information. So, we don't even attempt to parse
190       // any remaining DIEs.
191       if (m_dwo_symbol_file) {
192         m_die_array.front().SetHasChildren(false);
193         break;
194       }
195 
196     } else {
197       if (null_die) {
198         if (prev_die_had_children) {
199           // This will only happen if a DIE says is has children but all it
200           // contains is a NULL tag. Since we are removing the NULL DIEs from
201           // the list (saves up to 25% in C++ code), we need a way to let the
202           // DIE know that it actually doesn't have children.
203           if (!m_die_array.empty())
204             m_die_array.back().SetHasChildren(false);
205         }
206       } else {
207         die.SetParentIndex(m_die_array.size() - die_index_stack[depth - 1]);
208 
209         if (die_index_stack.back())
210           m_die_array[die_index_stack.back()].SetSiblingIndex(
211               m_die_array.size() - die_index_stack.back());
212 
213         // Only push the DIE if it isn't a NULL DIE
214         m_die_array.push_back(die);
215       }
216     }
217 
218     if (null_die) {
219       // NULL DIE.
220       if (!die_index_stack.empty())
221         die_index_stack.pop_back();
222 
223       if (depth > 0)
224         --depth;
225       prev_die_had_children = false;
226     } else {
227       die_index_stack.back() = m_die_array.size() - 1;
228       // Normal DIE
229       const bool die_has_children = die.HasChildren();
230       if (die_has_children) {
231         die_index_stack.push_back(0);
232         ++depth;
233       }
234       prev_die_had_children = die_has_children;
235     }
236 
237     if (depth == 0)
238       break; // We are done with this compile unit!
239   }
240 
241   if (!m_die_array.empty()) {
242     if (m_first_die) {
243       // Only needed for the assertion.
244       m_first_die.SetHasChildren(m_die_array.front().HasChildren());
245       lldbassert(m_first_die == m_die_array.front());
246     }
247     m_first_die = m_die_array.front();
248   }
249 
250   m_die_array.shrink_to_fit();
251 
252   if (m_dwo_symbol_file) {
253     DWARFUnit *dwo_cu = m_dwo_symbol_file->GetCompileUnit();
254     dwo_cu->ExtractDIEsIfNeeded();
255   }
256 }
257 
258 // This is used when a split dwarf is enabled.
259 // A skeleton compilation unit may contain the DW_AT_str_offsets_base attribute
260 // that points to the first string offset of the CU contribution to the
261 // .debug_str_offsets. At the same time, the corresponding split debug unit also
262 // may use DW_FORM_strx* forms pointing to its own .debug_str_offsets.dwo and
263 // for that case, we should find the offset (skip the section header).
264 static void SetDwoStrOffsetsBase(DWARFUnit *dwo_cu) {
265   lldb::offset_t baseOffset = 0;
266 
267   const DWARFDataExtractor &strOffsets =
268       dwo_cu->GetSymbolFileDWARF().GetDWARFContext().getOrLoadStrOffsetsData();
269   uint64_t length = strOffsets.GetU32(&baseOffset);
270   if (length == 0xffffffff)
271     length = strOffsets.GetU64(&baseOffset);
272 
273   // Check version.
274   if (strOffsets.GetU16(&baseOffset) < 5)
275     return;
276 
277   // Skip padding.
278   baseOffset += 2;
279 
280   dwo_cu->SetStrOffsetsBase(baseOffset);
281 }
282 
283 // m_die_array_mutex must be already held as read/write.
284 void DWARFUnit::AddUnitDIE(const DWARFDebugInfoEntry &cu_die) {
285   llvm::Optional<uint64_t> addr_base, gnu_addr_base, ranges_base,
286       gnu_ranges_base;
287 
288   DWARFAttributes attributes;
289   size_t num_attributes = cu_die.GetAttributes(this, attributes);
290   for (size_t i = 0; i < num_attributes; ++i) {
291     dw_attr_t attr = attributes.AttributeAtIndex(i);
292     DWARFFormValue form_value;
293     if (!attributes.ExtractFormValueAtIndex(i, form_value))
294       continue;
295     switch (attr) {
296     case DW_AT_addr_base:
297       addr_base = form_value.Unsigned();
298       SetAddrBase(*addr_base);
299       break;
300     case DW_AT_rnglists_base:
301       ranges_base = form_value.Unsigned();
302       SetRangesBase(*ranges_base);
303       break;
304     case DW_AT_str_offsets_base:
305       SetStrOffsetsBase(form_value.Unsigned());
306       break;
307     case DW_AT_low_pc:
308       SetBaseAddress(form_value.Address());
309       break;
310     case DW_AT_entry_pc:
311       // If the value was already set by DW_AT_low_pc, don't update it.
312       if (m_base_addr == LLDB_INVALID_ADDRESS)
313         SetBaseAddress(form_value.Address());
314       break;
315     case DW_AT_stmt_list:
316       m_line_table_offset = form_value.Unsigned();
317       break;
318     case DW_AT_GNU_addr_base:
319       gnu_addr_base = form_value.Unsigned();
320       break;
321     case DW_AT_GNU_ranges_base:
322       gnu_ranges_base = form_value.Unsigned();
323       break;
324     }
325   }
326 
327   std::unique_ptr<SymbolFileDWARFDwo> dwo_symbol_file =
328       m_dwarf.GetDwoSymbolFileForCompileUnit(*this, cu_die);
329   if (!dwo_symbol_file)
330     return;
331 
332   DWARFUnit *dwo_cu = dwo_symbol_file->GetCompileUnit();
333   if (!dwo_cu)
334     return; // Can't fetch the compile unit from the dwo file.
335 
336   DWARFBaseDIE dwo_cu_die = dwo_cu->GetUnitDIEOnly();
337   if (!dwo_cu_die.IsValid())
338     return; // Can't fetch the compile unit DIE from the dwo file.
339 
340   uint64_t main_dwo_id =
341       cu_die.GetAttributeValueAsUnsigned(this, DW_AT_GNU_dwo_id, 0);
342   uint64_t sub_dwo_id =
343       dwo_cu_die.GetAttributeValueAsUnsigned(DW_AT_GNU_dwo_id, 0);
344   if (main_dwo_id != sub_dwo_id)
345     return; // The 2 dwo ID isn't match. Don't use the dwo file as it belongs to
346   // a differectn compilation.
347 
348   m_dwo_symbol_file = std::move(dwo_symbol_file);
349 
350   // Here for DWO CU we want to use the address base set in the skeleton unit
351   // (DW_AT_addr_base) if it is available and use the DW_AT_GNU_addr_base
352   // otherwise. We do that because pre-DWARF v5 could use the DW_AT_GNU_*
353   // attributes which were applicable to the DWO units. The corresponding
354   // DW_AT_* attributes standardized in DWARF v5 are also applicable to the main
355   // unit in contrast.
356   if (addr_base)
357     dwo_cu->SetAddrBase(*addr_base);
358   else if (gnu_addr_base)
359     dwo_cu->SetAddrBase(*gnu_addr_base);
360 
361   if (ranges_base)
362     dwo_cu->SetRangesBase(*ranges_base);
363   else if (gnu_ranges_base)
364     dwo_cu->SetRangesBase(*gnu_ranges_base);
365 
366   for (size_t i = 0; i < m_dwo_symbol_file->DebugInfo()->GetNumUnits(); ++i) {
367     DWARFUnit *unit = m_dwo_symbol_file->DebugInfo()->GetUnitAtIndex(i);
368     SetDwoStrOffsetsBase(unit);
369   }
370 }
371 
372 DWARFDIE DWARFUnit::LookupAddress(const dw_addr_t address) {
373   if (DIE()) {
374     const DWARFDebugAranges &func_aranges = GetFunctionAranges();
375 
376     // Re-check the aranges auto pointer contents in case it was created above
377     if (!func_aranges.IsEmpty())
378       return GetDIE(func_aranges.FindAddress(address));
379   }
380   return DWARFDIE();
381 }
382 
383 size_t DWARFUnit::AppendDIEsWithTag(const dw_tag_t tag,
384                                     std::vector<DWARFDIE> &dies,
385                                     uint32_t depth) const {
386   size_t old_size = dies.size();
387   {
388     llvm::sys::ScopedReader lock(m_die_array_mutex);
389     DWARFDebugInfoEntry::const_iterator pos;
390     DWARFDebugInfoEntry::const_iterator end = m_die_array.end();
391     for (pos = m_die_array.begin(); pos != end; ++pos) {
392       if (pos->Tag() == tag)
393         dies.emplace_back(this, &(*pos));
394     }
395   }
396 
397   // Return the number of DIEs added to the collection
398   return dies.size() - old_size;
399 }
400 
401 size_t DWARFUnit::GetDebugInfoSize() const {
402   return GetLengthByteSize() + GetLength() - GetHeaderByteSize();
403 }
404 
405 const DWARFAbbreviationDeclarationSet *DWARFUnit::GetAbbreviations() const {
406   return m_abbrevs;
407 }
408 
409 dw_offset_t DWARFUnit::GetAbbrevOffset() const {
410   return m_abbrevs ? m_abbrevs->GetOffset() : DW_INVALID_OFFSET;
411 }
412 
413 dw_offset_t DWARFUnit::GetLineTableOffset() {
414   ExtractUnitDIEIfNeeded();
415   return m_line_table_offset;
416 }
417 
418 void DWARFUnit::SetAddrBase(dw_addr_t addr_base) { m_addr_base = addr_base; }
419 
420 void DWARFUnit::SetRangesBase(dw_addr_t ranges_base) {
421   m_ranges_base = ranges_base;
422 }
423 
424 void DWARFUnit::SetStrOffsetsBase(dw_offset_t str_offsets_base) {
425   m_str_offsets_base = str_offsets_base;
426 }
427 
428 // It may be called only with m_die_array_mutex held R/W.
429 void DWARFUnit::ClearDIEsRWLocked() {
430   m_die_array.clear();
431   m_die_array.shrink_to_fit();
432 
433   if (m_dwo_symbol_file)
434     m_dwo_symbol_file->GetCompileUnit()->ClearDIEsRWLocked();
435 }
436 
437 lldb::ByteOrder DWARFUnit::GetByteOrder() const {
438   return m_dwarf.GetObjectFile()->GetByteOrder();
439 }
440 
441 TypeSystem *DWARFUnit::GetTypeSystem() {
442   return m_dwarf.GetTypeSystemForLanguage(GetLanguageType());
443 }
444 
445 void DWARFUnit::SetBaseAddress(dw_addr_t base_addr) { m_base_addr = base_addr; }
446 
447 // Compare function DWARFDebugAranges::Range structures
448 static bool CompareDIEOffset(const DWARFDebugInfoEntry &die,
449                              const dw_offset_t die_offset) {
450   return die.GetOffset() < die_offset;
451 }
452 
453 // GetDIE()
454 //
455 // Get the DIE (Debug Information Entry) with the specified offset by first
456 // checking if the DIE is contained within this compile unit and grabbing the
457 // DIE from this compile unit. Otherwise we grab the DIE from the DWARF file.
458 DWARFDIE
459 DWARFUnit::GetDIE(dw_offset_t die_offset) {
460   if (die_offset != DW_INVALID_OFFSET) {
461     if (GetDwoSymbolFile())
462       return GetDwoSymbolFile()->GetCompileUnit()->GetDIE(die_offset);
463 
464     if (ContainsDIEOffset(die_offset)) {
465       ExtractDIEsIfNeeded();
466       DWARFDebugInfoEntry::const_iterator end = m_die_array.cend();
467       DWARFDebugInfoEntry::const_iterator pos =
468           lower_bound(m_die_array.cbegin(), end, die_offset, CompareDIEOffset);
469       if (pos != end) {
470         if (die_offset == (*pos).GetOffset())
471           return DWARFDIE(this, &(*pos));
472       }
473     } else
474       GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
475           "GetDIE for DIE 0x%" PRIx32 " is outside of its CU 0x%" PRIx32,
476           die_offset, GetOffset());
477   }
478   return DWARFDIE(); // Not found
479 }
480 
481 DWARFUnit &DWARFUnit::GetNonSkeletonUnit() {
482   if (SymbolFileDWARFDwo *dwo = GetDwoSymbolFile())
483     return *dwo->GetCompileUnit();
484   return *this;
485 }
486 
487 uint8_t DWARFUnit::GetAddressByteSize(const DWARFUnit *cu) {
488   if (cu)
489     return cu->GetAddressByteSize();
490   return DWARFUnit::GetDefaultAddressSize();
491 }
492 
493 uint8_t DWARFUnit::GetDefaultAddressSize() { return 4; }
494 
495 void *DWARFUnit::GetUserData() const { return m_user_data; }
496 
497 void DWARFUnit::SetUserData(void *d) {
498   m_user_data = d;
499   if (m_dwo_symbol_file)
500     m_dwo_symbol_file->GetCompileUnit()->SetUserData(d);
501 }
502 
503 bool DWARFUnit::Supports_DW_AT_APPLE_objc_complete_type() {
504   return GetProducer() != eProducerLLVMGCC;
505 }
506 
507 bool DWARFUnit::DW_AT_decl_file_attributes_are_invalid() {
508   // llvm-gcc makes completely invalid decl file attributes and won't ever be
509   // fixed, so we need to know to ignore these.
510   return GetProducer() == eProducerLLVMGCC;
511 }
512 
513 bool DWARFUnit::Supports_unnamed_objc_bitfields() {
514   if (GetProducer() == eProducerClang) {
515     const uint32_t major_version = GetProducerVersionMajor();
516     return major_version > 425 ||
517            (major_version == 425 && GetProducerVersionUpdate() >= 13);
518   }
519   return true; // Assume all other compilers didn't have incorrect ObjC bitfield
520                // info
521 }
522 
523 void DWARFUnit::ParseProducerInfo() {
524   m_producer_version_major = UINT32_MAX;
525   m_producer_version_minor = UINT32_MAX;
526   m_producer_version_update = UINT32_MAX;
527 
528   const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
529   if (die) {
530 
531     const char *producer_cstr =
532         die->GetAttributeValueAsString(this, DW_AT_producer, nullptr);
533     if (producer_cstr) {
534       RegularExpression llvm_gcc_regex(
535           llvm::StringRef("^4\\.[012]\\.[01] \\(Based on Apple "
536                           "Inc\\. build [0-9]+\\) \\(LLVM build "
537                           "[\\.0-9]+\\)$"));
538       if (llvm_gcc_regex.Execute(llvm::StringRef(producer_cstr))) {
539         m_producer = eProducerLLVMGCC;
540       } else if (strstr(producer_cstr, "clang")) {
541         static RegularExpression g_clang_version_regex(
542             llvm::StringRef("clang-([0-9]+)\\.([0-9]+)\\.([0-9]+)"));
543         RegularExpression::Match regex_match(3);
544         if (g_clang_version_regex.Execute(llvm::StringRef(producer_cstr),
545                                           &regex_match)) {
546           std::string str;
547           if (regex_match.GetMatchAtIndex(producer_cstr, 1, str))
548             m_producer_version_major =
549                 StringConvert::ToUInt32(str.c_str(), UINT32_MAX, 10);
550           if (regex_match.GetMatchAtIndex(producer_cstr, 2, str))
551             m_producer_version_minor =
552                 StringConvert::ToUInt32(str.c_str(), UINT32_MAX, 10);
553           if (regex_match.GetMatchAtIndex(producer_cstr, 3, str))
554             m_producer_version_update =
555                 StringConvert::ToUInt32(str.c_str(), UINT32_MAX, 10);
556         }
557         m_producer = eProducerClang;
558       } else if (strstr(producer_cstr, "GNU"))
559         m_producer = eProducerGCC;
560     }
561   }
562   if (m_producer == eProducerInvalid)
563     m_producer = eProcucerOther;
564 }
565 
566 DWARFProducer DWARFUnit::GetProducer() {
567   if (m_producer == eProducerInvalid)
568     ParseProducerInfo();
569   return m_producer;
570 }
571 
572 uint32_t DWARFUnit::GetProducerVersionMajor() {
573   if (m_producer_version_major == 0)
574     ParseProducerInfo();
575   return m_producer_version_major;
576 }
577 
578 uint32_t DWARFUnit::GetProducerVersionMinor() {
579   if (m_producer_version_minor == 0)
580     ParseProducerInfo();
581   return m_producer_version_minor;
582 }
583 
584 uint32_t DWARFUnit::GetProducerVersionUpdate() {
585   if (m_producer_version_update == 0)
586     ParseProducerInfo();
587   return m_producer_version_update;
588 }
589 LanguageType DWARFUnit::LanguageTypeFromDWARF(uint64_t val) {
590   // Note: user languages between lo_user and hi_user must be handled
591   // explicitly here.
592   switch (val) {
593   case DW_LANG_Mips_Assembler:
594     return eLanguageTypeMipsAssembler;
595   case DW_LANG_GOOGLE_RenderScript:
596     return eLanguageTypeExtRenderScript;
597   default:
598     return static_cast<LanguageType>(val);
599   }
600 }
601 
602 LanguageType DWARFUnit::GetLanguageType() {
603   if (m_language_type != eLanguageTypeUnknown)
604     return m_language_type;
605 
606   const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
607   if (die)
608     m_language_type = LanguageTypeFromDWARF(
609         die->GetAttributeValueAsUnsigned(this, DW_AT_language, 0));
610   return m_language_type;
611 }
612 
613 bool DWARFUnit::GetIsOptimized() {
614   if (m_is_optimized == eLazyBoolCalculate) {
615     const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
616     if (die) {
617       m_is_optimized = eLazyBoolNo;
618       if (die->GetAttributeValueAsUnsigned(this, DW_AT_APPLE_optimized, 0) ==
619           1) {
620         m_is_optimized = eLazyBoolYes;
621       }
622     }
623   }
624   return m_is_optimized == eLazyBoolYes;
625 }
626 
627 FileSpec::Style DWARFUnit::GetPathStyle() {
628   if (!m_comp_dir)
629     ComputeCompDirAndGuessPathStyle();
630   return m_comp_dir->GetPathStyle();
631 }
632 
633 const FileSpec &DWARFUnit::GetCompilationDirectory() {
634   if (!m_comp_dir)
635     ComputeCompDirAndGuessPathStyle();
636   return *m_comp_dir;
637 }
638 
639 const FileSpec &DWARFUnit::GetAbsolutePath() {
640   if (!m_file_spec)
641     ComputeAbsolutePath();
642   return *m_file_spec;
643 }
644 
645 FileSpec DWARFUnit::GetFile(size_t file_idx) {
646   return m_dwarf.GetFile(*this, file_idx);
647 }
648 
649 // DWARF2/3 suggests the form hostname:pathname for compilation directory.
650 // Remove the host part if present.
651 static llvm::StringRef
652 removeHostnameFromPathname(llvm::StringRef path_from_dwarf) {
653   llvm::StringRef host, path;
654   std::tie(host, path) = path_from_dwarf.split(':');
655 
656   if (host.contains('/'))
657     return path_from_dwarf;
658 
659   // check whether we have a windows path, and so the first character is a
660   // drive-letter not a hostname.
661   if (host.size() == 1 && llvm::isAlpha(host[0]) && path.startswith("\\"))
662     return path_from_dwarf;
663 
664   return path;
665 }
666 
667 static FileSpec resolveCompDir(const FileSpec &path) {
668   bool is_symlink = SymbolFileDWARF::GetSymlinkPaths().FindFileIndex(
669                         0, path, /*full*/ true) != UINT32_MAX;
670 
671   if (!is_symlink)
672     return path;
673 
674   namespace fs = llvm::sys::fs;
675   if (fs::get_file_type(path.GetPath(), false) != fs::file_type::symlink_file)
676     return path;
677 
678   FileSpec resolved_symlink;
679   const auto error = FileSystem::Instance().Readlink(path, resolved_symlink);
680   if (error.Success())
681     return resolved_symlink;
682 
683   return path;
684 }
685 
686 void DWARFUnit::ComputeCompDirAndGuessPathStyle() {
687   m_comp_dir = FileSpec();
688   const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
689   if (!die)
690     return;
691 
692   llvm::StringRef comp_dir = removeHostnameFromPathname(
693       die->GetAttributeValueAsString(this, DW_AT_comp_dir, nullptr));
694   if (!comp_dir.empty()) {
695     FileSpec::Style comp_dir_style =
696         FileSpec::GuessPathStyle(comp_dir).getValueOr(FileSpec::Style::native);
697     m_comp_dir = resolveCompDir(FileSpec(comp_dir, comp_dir_style));
698   } else {
699     // Try to detect the style based on the DW_AT_name attribute, but just store
700     // the detected style in the m_comp_dir field.
701     const char *name =
702         die->GetAttributeValueAsString(this, DW_AT_name, nullptr);
703     m_comp_dir = FileSpec(
704         "", FileSpec::GuessPathStyle(name).getValueOr(FileSpec::Style::native));
705   }
706 }
707 
708 void DWARFUnit::ComputeAbsolutePath() {
709   m_file_spec = FileSpec();
710   const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
711   if (!die)
712     return;
713 
714   m_file_spec =
715       FileSpec(die->GetAttributeValueAsString(this, DW_AT_name, nullptr),
716                GetPathStyle());
717 
718   if (m_file_spec->IsRelative())
719     m_file_spec->MakeAbsolute(GetCompilationDirectory());
720 }
721 
722 SymbolFileDWARFDwo *DWARFUnit::GetDwoSymbolFile() const {
723   return m_dwo_symbol_file.get();
724 }
725 
726 const DWARFDebugAranges &DWARFUnit::GetFunctionAranges() {
727   if (m_func_aranges_up == nullptr) {
728     m_func_aranges_up.reset(new DWARFDebugAranges());
729     const DWARFDebugInfoEntry *die = DIEPtr();
730     if (die)
731       die->BuildFunctionAddressRangeTable(this, m_func_aranges_up.get());
732 
733     if (m_dwo_symbol_file) {
734       DWARFUnit *dwo_cu = m_dwo_symbol_file->GetCompileUnit();
735       const DWARFDebugInfoEntry *dwo_die = dwo_cu->DIEPtr();
736       if (dwo_die)
737         dwo_die->BuildFunctionAddressRangeTable(dwo_cu,
738                                                 m_func_aranges_up.get());
739     }
740 
741     const bool minimize = false;
742     m_func_aranges_up->Sort(minimize);
743   }
744   return *m_func_aranges_up;
745 }
746 
747 llvm::Expected<DWARFUnitHeader>
748 DWARFUnitHeader::extract(const DWARFDataExtractor &data, DIERef::Section section,
749                          lldb::offset_t *offset_ptr) {
750   DWARFUnitHeader header;
751   header.m_offset = *offset_ptr;
752   header.m_length = data.GetDWARFInitialLength(offset_ptr);
753   header.m_version = data.GetU16(offset_ptr);
754   if (header.m_version == 5) {
755     header.m_unit_type = data.GetU8(offset_ptr);
756     header.m_addr_size = data.GetU8(offset_ptr);
757     header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
758     if (header.m_unit_type == llvm::dwarf::DW_UT_skeleton)
759       header.m_dwo_id = data.GetU64(offset_ptr);
760   } else {
761     header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
762     header.m_addr_size = data.GetU8(offset_ptr);
763     header.m_unit_type =
764         section == DIERef::Section::DebugTypes ? DW_UT_type : DW_UT_compile;
765   }
766 
767   if (header.IsTypeUnit()) {
768     header.m_type_hash = data.GetU64(offset_ptr);
769     header.m_type_offset = data.GetDWARFOffset(offset_ptr);
770   }
771 
772   bool length_OK = data.ValidOffset(header.GetNextUnitOffset() - 1);
773   bool version_OK = SymbolFileDWARF::SupportedVersion(header.m_version);
774   bool addr_size_OK = (header.m_addr_size == 4) || (header.m_addr_size == 8);
775   bool type_offset_OK =
776       !header.IsTypeUnit() || (header.m_type_offset <= header.GetLength());
777 
778   if (!length_OK)
779     return llvm::make_error<llvm::object::GenericBinaryError>(
780         "Invalid unit length");
781   if (!version_OK)
782     return llvm::make_error<llvm::object::GenericBinaryError>(
783         "Unsupported unit version");
784   if (!addr_size_OK)
785     return llvm::make_error<llvm::object::GenericBinaryError>(
786         "Invalid unit address size");
787   if (!type_offset_OK)
788     return llvm::make_error<llvm::object::GenericBinaryError>(
789         "Type offset out of range");
790 
791   return header;
792 }
793 
794 llvm::Expected<DWARFUnitSP>
795 DWARFUnit::extract(SymbolFileDWARF &dwarf, user_id_t uid,
796                    const DWARFDataExtractor &debug_info,
797                    DIERef::Section section, lldb::offset_t *offset_ptr) {
798   assert(debug_info.ValidOffset(*offset_ptr));
799 
800   auto expected_header =
801       DWARFUnitHeader::extract(debug_info, section, offset_ptr);
802   if (!expected_header)
803     return expected_header.takeError();
804 
805   const DWARFDebugAbbrev *abbr = dwarf.DebugAbbrev();
806   if (!abbr)
807     return llvm::make_error<llvm::object::GenericBinaryError>(
808         "No debug_abbrev data");
809 
810   bool abbr_offset_OK =
811       dwarf.GetDWARFContext().getOrLoadAbbrevData().ValidOffset(
812           expected_header->GetAbbrOffset());
813   if (!abbr_offset_OK)
814     return llvm::make_error<llvm::object::GenericBinaryError>(
815         "Abbreviation offset for unit is not valid");
816 
817   const DWARFAbbreviationDeclarationSet *abbrevs =
818       abbr->GetAbbreviationDeclarationSet(expected_header->GetAbbrOffset());
819   if (!abbrevs)
820     return llvm::make_error<llvm::object::GenericBinaryError>(
821         "No abbrev exists at the specified offset.");
822 
823   if (expected_header->IsTypeUnit())
824     return DWARFUnitSP(
825         new DWARFTypeUnit(dwarf, uid, *expected_header, *abbrevs, section));
826   return DWARFUnitSP(
827       new DWARFCompileUnit(dwarf, uid, *expected_header, *abbrevs, section));
828 }
829 
830 const lldb_private::DWARFDataExtractor &DWARFUnit::GetData() const {
831   return m_section == DIERef::Section::DebugTypes
832              ? m_dwarf.GetDWARFContext().getOrLoadDebugTypesData()
833              : m_dwarf.GetDWARFContext().getOrLoadDebugInfoData();
834 }
835 
836 uint32_t DWARFUnit::GetHeaderByteSize() const {
837   switch (m_header.GetUnitType()) {
838   case llvm::dwarf::DW_UT_compile:
839   case llvm::dwarf::DW_UT_partial:
840     return GetVersion() < 5 ? 11 : 12;
841   case llvm::dwarf::DW_UT_skeleton:
842   case llvm::dwarf::DW_UT_split_compile:
843     return 20;
844   case llvm::dwarf::DW_UT_type:
845   case llvm::dwarf::DW_UT_split_type:
846     return GetVersion() < 5 ? 23 : 24;
847   }
848   llvm_unreachable("invalid UnitType.");
849 }
850 
851 llvm::Expected<DWARFRangeList>
852 DWARFUnit::FindRnglistFromOffset(dw_offset_t offset) const {
853   const DWARFDebugRangesBase *debug_ranges;
854   llvm::StringRef section;
855   if (GetVersion() <= 4) {
856     debug_ranges = m_dwarf.GetDebugRanges();
857     section = "debug_ranges";
858   } else {
859     debug_ranges = m_dwarf.GetDebugRngLists();
860     section = "debug_rnglists";
861   }
862   if (!debug_ranges)
863     return llvm::make_error<llvm::object::GenericBinaryError>("No " + section +
864                                                               " section");
865 
866   DWARFRangeList ranges;
867   debug_ranges->FindRanges(this, offset, ranges);
868   return ranges;
869 }
870 
871 llvm::Expected<DWARFRangeList>
872 DWARFUnit::FindRnglistFromIndex(uint32_t index) const {
873   const DWARFDebugRangesBase *debug_rnglists = m_dwarf.GetDebugRngLists();
874   if (!debug_rnglists)
875     return llvm::make_error<llvm::object::GenericBinaryError>(
876         "No debug_rnglists section");
877   return FindRnglistFromOffset(debug_rnglists->GetOffset(index));
878 }
879