Merge alignment of common GlobalValue.
[oota-llvm.git] / lib / Object / MachOObjectFile.cpp
index 8e19eb5631d8ac4845ead2c7d923e759e24ffc6a..c6815ea27d24cfd02f8289a5a872cfc10c6cf2b5 100644 (file)
@@ -19,6 +19,8 @@
 #include "llvm/Support/DataExtractor.h"
 #include "llvm/Support/Format.h"
 #include "llvm/Support/Host.h"
+#include "llvm/Support/LEB128.h"
+#include "llvm/Support/MachO.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include "llvm/Support/raw_ostream.h"
 #include <cctype>
@@ -222,12 +224,11 @@ static uint32_t getSectionFlags(const MachOObjectFile *O,
   return Sect.flags;
 }
 
-MachOObjectFile::MachOObjectFile(std::unique_ptr<MemoryBuffer> Object,
-                                 bool IsLittleEndian, bool Is64bits,
-                                 std::error_code &EC)
-    : ObjectFile(getMachOType(IsLittleEndian, Is64bits), std::move(Object)),
+MachOObjectFile::MachOObjectFile(MemoryBufferRef Object, bool IsLittleEndian,
+                                 bool Is64bits, std::error_code &EC)
+    : ObjectFile(getMachOType(IsLittleEndian, Is64bits), Object),
       SymtabLoadCmd(nullptr), DysymtabLoadCmd(nullptr),
-      DataInCodeLoadCmd(nullptr) {
+      DataInCodeLoadCmd(nullptr), DyldInfoLoadCmd(nullptr) {
   uint32_t LoadCommandCount = this->getHeader().ncmds;
   MachO::LoadCommandType SegmentLoadType = is64Bit() ?
     MachO::LC_SEGMENT_64 : MachO::LC_SEGMENT;
@@ -243,6 +244,10 @@ MachOObjectFile::MachOObjectFile(std::unique_ptr<MemoryBuffer> Object,
     } else if (Load.C.cmd == MachO::LC_DATA_IN_CODE) {
       assert(!DataInCodeLoadCmd && "Multiple data in code tables");
       DataInCodeLoadCmd = Load.Ptr;
+    } else if (Load.C.cmd == MachO::LC_DYLD_INFO || 
+               Load.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
+      assert(!DyldInfoLoadCmd && "Multiple dyldinfo load commands");
+      DyldInfoLoadCmd = Load.Ptr;
     } else if (Load.C.cmd == SegmentLoadType) {
       uint32_t NumSections = getSegmentLoadCommandNumSections(this, Load);
       for (unsigned J = 0; J < NumSections; ++J) {
@@ -439,6 +444,9 @@ uint32_t MachOObjectFile::getSymbolFlags(DataRefImpl DRI) const {
   if (MachOFlags & (MachO::N_WEAK_REF | MachO::N_WEAK_DEF))
     Result |= SymbolRef::SF_Weak;
 
+  if (MachOFlags & (MachO::N_ARM_THUMB_DEF))
+    Result |= SymbolRef::SF_Thumb;
+
   if ((MachOType & MachO::N_TYPE) == MachO::N_ABS)
     Result |= SymbolRef::SF_Absolute;
 
@@ -1157,7 +1165,7 @@ guess_qtx:
 // It is passed the index (0 - based) of the library as translated from
 // GET_LIBRARY_ORDINAL (1 - based).
 std::error_code MachOObjectFile::getLibraryShortNameByIndex(unsigned Index,
-                                                            StringRef &Res) {
+                                                         StringRef &Res) const {
   if (Index >= Libraries.size())
     return object_error::parse_failed;
 
@@ -1293,7 +1301,11 @@ Triple::ArchType MachOObjectFile::getArch(uint32_t CPUType) {
   }
 }
 
-Triple MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) {
+Triple MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType,
+                                const char **McpuDefault) {
+  if (McpuDefault)
+    *McpuDefault = nullptr;
+
   switch (CPUType) {
   case MachO::CPU_TYPE_I386:
     switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
@@ -1322,14 +1334,20 @@ Triple MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) {
     case MachO::CPU_SUBTYPE_ARM_V6:
       return Triple("armv6-apple-darwin");
     case MachO::CPU_SUBTYPE_ARM_V6M:
+      if (McpuDefault)
+        *McpuDefault = "cortex-m0";
       return Triple("armv6m-apple-darwin");
     case MachO::CPU_SUBTYPE_ARM_V7:
       return Triple("armv7-apple-darwin");
     case MachO::CPU_SUBTYPE_ARM_V7EM:
+      if (McpuDefault)
+        *McpuDefault = "cortex-m4";
       return Triple("armv7em-apple-darwin");
     case MachO::CPU_SUBTYPE_ARM_V7K:
       return Triple("armv7k-apple-darwin");
     case MachO::CPU_SUBTYPE_ARM_V7M:
+      if (McpuDefault)
+        *McpuDefault = "cortex-m3";
       return Triple("armv7m-apple-darwin");
     case MachO::CPU_SUBTYPE_ARM_V7S:
       return Triple("armv7s-apple-darwin");
@@ -1362,6 +1380,57 @@ Triple MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) {
   }
 }
 
+Triple MachOObjectFile::getThumbArch(uint32_t CPUType, uint32_t CPUSubType,
+                                     const char **McpuDefault) {
+  if (McpuDefault)
+    *McpuDefault = nullptr;
+
+  switch (CPUType) {
+  case MachO::CPU_TYPE_ARM:
+    switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
+    case MachO::CPU_SUBTYPE_ARM_V4T:
+      return Triple("thumbv4t-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_V5TEJ:
+      return Triple("thumbv5e-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_XSCALE:
+      return Triple("xscale-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_V6:
+      return Triple("thumbv6-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_V6M:
+      if (McpuDefault)
+        *McpuDefault = "cortex-m0";
+      return Triple("thumbv6m-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_V7:
+      return Triple("thumbv7-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_V7EM:
+      if (McpuDefault)
+        *McpuDefault = "cortex-m4";
+      return Triple("thumbv7em-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_V7K:
+      return Triple("thumbv7k-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_V7M:
+      if (McpuDefault)
+        *McpuDefault = "cortex-m3";
+      return Triple("thumbv7m-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM_V7S:
+      return Triple("thumbv7s-apple-darwin");
+    default:
+      return Triple();
+    }
+  default:
+    return Triple();
+  }
+}
+
+Triple MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType,
+                                const char **McpuDefault,
+                               Triple *ThumbTriple) {
+  Triple T = MachOObjectFile::getArch(CPUType, CPUSubType, McpuDefault);
+  *ThumbTriple = MachOObjectFile::getThumbArch(CPUType, CPUSubType,
+                                               McpuDefault);
+  return T;
+}
+
 Triple MachOObjectFile::getHostArch() {
   return Triple(sys::getDefaultTargetTriple());
 }
@@ -1390,6 +1459,25 @@ unsigned MachOObjectFile::getArch() const {
   return getArch(getCPUType(this));
 }
 
+Triple MachOObjectFile::getArch(const char **McpuDefault,
+                                Triple *ThumbTriple) const {
+  Triple T;
+  if (is64Bit()) {
+    MachO::mach_header_64 H_64;
+    H_64 = getHeader64();
+    T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype, McpuDefault);
+    *ThumbTriple = MachOObjectFile::getThumbArch(H_64.cputype, H_64.cpusubtype,
+                                                 McpuDefault);
+  } else {
+    MachO::mach_header H;
+    H = getHeader();
+    T = MachOObjectFile::getArch(H.cputype, H.cpusubtype, McpuDefault);
+    *ThumbTriple = MachOObjectFile::getThumbArch(H.cputype, H.cpusubtype,
+                                                 McpuDefault);
+  }
+  return T;
+}
+
 relocation_iterator MachOObjectFile::section_rel_begin(unsigned Index) const {
   DataRefImpl DRI;
   DRI.d.a = Index;
@@ -1423,6 +1511,183 @@ dice_iterator MachOObjectFile::end_dices() const {
   return dice_iterator(DiceRef(DRI, this));
 }
 
+ExportEntry::ExportEntry(ArrayRef<uint8_t> T) 
+  : Trie(T), Malformed(false), Done(false) { }
+
+void ExportEntry::moveToFirst() {
+  pushNode(0);
+  pushDownUntilBottom();
+}
+
+void ExportEntry::moveToEnd() {
+  Stack.clear();
+  Done = true;
+}
+
+bool ExportEntry::operator==(const ExportEntry &Other) const {
+  // Common case, one at end, other iterating from begin. 
+  if (Done || Other.Done)
+    return (Done == Other.Done);
+  // Not equal if different stack sizes.
+  if (Stack.size() != Other.Stack.size())
+    return false;
+  // Not equal if different cumulative strings.
+  if (!CumulativeString.str().equals(Other.CumulativeString.str()))
+    return false;
+  // Equal if all nodes in both stacks match.
+  for (unsigned i=0; i < Stack.size(); ++i) {
+    if (Stack[i].Start != Other.Stack[i].Start)
+      return false;
+  }
+  return true;  
+}
+
+uint64_t ExportEntry::readULEB128(const uint8_t *&Ptr) {
+  unsigned Count;
+  uint64_t Result = decodeULEB128(Ptr, &Count);
+  Ptr += Count;
+  if (Ptr > Trie.end()) {
+    Ptr = Trie.end();
+    Malformed = true;
+  }
+  return Result;
+}
+
+StringRef ExportEntry::name() const {
+  return CumulativeString.str();
+}
+
+uint64_t ExportEntry::flags() const {
+  return Stack.back().Flags;
+}
+
+uint64_t ExportEntry::address() const {
+  return Stack.back().Address;
+}
+
+uint64_t ExportEntry::other() const {
+  return Stack.back().Other;
+}
+
+StringRef ExportEntry::otherName() const {
+  const char* ImportName = Stack.back().ImportName;
+  if (ImportName)
+    return StringRef(ImportName);
+  return StringRef();
+}
+
+uint32_t ExportEntry::nodeOffset() const {
+  return Stack.back().Start - Trie.begin();
+}
+
+ExportEntry::NodeState::NodeState(const uint8_t *Ptr) 
+  : Start(Ptr), Current(Ptr), Flags(0), Address(0), Other(0), 
+    ImportName(nullptr), ChildCount(0), NextChildIndex(0),  
+    ParentStringLength(0), IsExportNode(false) {
+}
+
+void ExportEntry::pushNode(uint64_t offset) {
+  const uint8_t *Ptr = Trie.begin() + offset;
+  NodeState State(Ptr);
+  uint64_t ExportInfoSize = readULEB128(State.Current);
+  State.IsExportNode = (ExportInfoSize != 0);
+  const uint8_t* Children = State.Current + ExportInfoSize;
+  if (State.IsExportNode) {
+    State.Flags = readULEB128(State.Current);
+    if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT) {
+      State.Address = 0;
+      State.Other = readULEB128(State.Current); // dylib ordinal
+      State.ImportName = reinterpret_cast<const char*>(State.Current);
+    } else {
+      State.Address = readULEB128(State.Current);
+      if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER)
+        State.Other = readULEB128(State.Current); 
+    }
+  }
+  State.ChildCount = *Children;
+  State.Current = Children + 1;
+  State.NextChildIndex = 0;
+  State.ParentStringLength = CumulativeString.size();
+  Stack.push_back(State);
+}
+
+void ExportEntry::pushDownUntilBottom() {
+  while (Stack.back().NextChildIndex < Stack.back().ChildCount) {
+    NodeState &Top = Stack.back();
+    CumulativeString.resize(Top.ParentStringLength);
+    for (;*Top.Current != 0; Top.Current++) {
+      char C = *Top.Current;
+      CumulativeString.push_back(C);
+    }
+    Top.Current += 1;
+    uint64_t childNodeIndex = readULEB128(Top.Current);
+    Top.NextChildIndex += 1;
+    pushNode(childNodeIndex);
+  }
+  if (!Stack.back().IsExportNode) {
+    Malformed = true;
+    moveToEnd();
+  }
+}
+
+// We have a trie data structure and need a way to walk it that is compatible
+// with the C++ iterator model. The solution is a non-recursive depth first
+// traversal where the iterator contains a stack of parent nodes along with a
+// string that is the accumulation of all edge strings along the parent chain
+// to this point.
+//
+// There is one “export” node for each exported symbol.  But because some
+// symbols may be a prefix of another symbol (e.g. _dup and _dup2), an export
+// node may have child nodes too.  
+//
+// The algorithm for moveNext() is to keep moving down the leftmost unvisited
+// child until hitting a node with no children (which is an export node or
+// else the trie is malformed). On the way down, each node is pushed on the
+// stack ivar.  If there is no more ways down, it pops up one and tries to go
+// down a sibling path until a childless node is reached.
+void ExportEntry::moveNext() {
+  if (Stack.empty() || !Stack.back().IsExportNode) {
+    Malformed = true;
+    moveToEnd();
+    return;
+  }
+
+  Stack.pop_back();
+  while (!Stack.empty()) {
+    NodeState &Top = Stack.back();
+    if (Top.NextChildIndex < Top.ChildCount) {
+      pushDownUntilBottom();
+      // Now at the next export node.
+      return;
+    } else {
+      if (Top.IsExportNode) {
+        // This node has no children but is itself an export node.
+        CumulativeString.resize(Top.ParentStringLength);
+        return;
+      }
+      Stack.pop_back();
+    }
+  }
+  Done = true;
+}
+
+iterator_range<export_iterator> 
+MachOObjectFile::exports(ArrayRef<uint8_t> Trie) {
+  ExportEntry Start(Trie);
+  Start.moveToFirst();
+
+  ExportEntry Finish(Trie);
+  Finish.moveToEnd();
+
+  return iterator_range<export_iterator>(export_iterator(Start), 
+                                         export_iterator(Finish));
+}
+
+iterator_range<export_iterator> MachOObjectFile::exports() const {
+  return exports(getDyldInfoExportsTrie());
+}
+
+
 StringRef
 MachOObjectFile::getSectionFinalSegmentName(DataRefImpl Sec) const {
   ArrayRef<char> Raw = getSectionRawFinalSegmentName(Sec);
@@ -1433,14 +1698,14 @@ ArrayRef<char>
 MachOObjectFile::getSectionRawName(DataRefImpl Sec) const {
   const section_base *Base =
     reinterpret_cast<const section_base *>(Sections[Sec.d.a]);
-  return ArrayRef<char>(Base->sectname);
+  return makeArrayRef(Base->sectname);
 }
 
 ArrayRef<char>
 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const {
   const section_base *Base =
     reinterpret_cast<const section_base *>(Sections[Sec.d.a]);
-  return ArrayRef<char>(Base->segname);
+  return makeArrayRef(Base->segname);
 }
 
 bool
@@ -1596,6 +1861,31 @@ MachOObjectFile::getDylibIDLoadCommand(const LoadCommandInfo &L) const {
   return getStruct<MachO::dylib_command>(this, L.Ptr);
 }
 
+MachO::dyld_info_command
+MachOObjectFile::getDyldInfoLoadCommand(const LoadCommandInfo &L) const {
+  return getStruct<MachO::dyld_info_command>(this, L.Ptr);
+}
+
+MachO::dylinker_command
+MachOObjectFile::getDylinkerCommand(const LoadCommandInfo &L) const {
+  return getStruct<MachO::dylinker_command>(this, L.Ptr);
+}
+
+MachO::uuid_command
+MachOObjectFile::getUuidCommand(const LoadCommandInfo &L) const {
+  return getStruct<MachO::uuid_command>(this, L.Ptr);
+}
+
+MachO::source_version_command
+MachOObjectFile::getSourceVersionCommand(const LoadCommandInfo &L) const {
+  return getStruct<MachO::source_version_command>(this, L.Ptr);
+}
+
+MachO::entry_point_command
+MachOObjectFile::getEntryPointCommand(const LoadCommandInfo &L) const {
+  return getStruct<MachO::entry_point_command>(this, L.Ptr);
+}
+
 
 MachO::any_relocation_info
 MachOObjectFile::getRelocation(DataRefImpl Rel) const {
@@ -1666,6 +1956,62 @@ MachOObjectFile::getDataInCodeLoadCommand() const {
   return Cmd;
 }
 
+ArrayRef<uint8_t> MachOObjectFile::getDyldInfoRebaseOpcodes() const {
+  if (!DyldInfoLoadCmd) 
+    return ArrayRef<uint8_t>();
+
+  MachO::dyld_info_command DyldInfo 
+                   = getStruct<MachO::dyld_info_command>(this, DyldInfoLoadCmd);
+  const uint8_t *Ptr = reinterpret_cast<const uint8_t*>(
+                                             getPtr(this, DyldInfo.rebase_off));
+  return ArrayRef<uint8_t>(Ptr, DyldInfo.rebase_size);
+}
+
+ArrayRef<uint8_t> MachOObjectFile::getDyldInfoBindOpcodes() const {
+  if (!DyldInfoLoadCmd) 
+    return ArrayRef<uint8_t>();
+
+  MachO::dyld_info_command DyldInfo 
+                   = getStruct<MachO::dyld_info_command>(this, DyldInfoLoadCmd);
+  const uint8_t *Ptr = reinterpret_cast<const uint8_t*>(
+                                               getPtr(this, DyldInfo.bind_off));
+  return ArrayRef<uint8_t>(Ptr, DyldInfo.bind_size);
+}
+
+ArrayRef<uint8_t> MachOObjectFile::getDyldInfoWeakBindOpcodes() const {
+  if (!DyldInfoLoadCmd) 
+    return ArrayRef<uint8_t>();
+
+  MachO::dyld_info_command DyldInfo 
+                   = getStruct<MachO::dyld_info_command>(this, DyldInfoLoadCmd);
+  const uint8_t *Ptr = reinterpret_cast<const uint8_t*>(
+                                          getPtr(this, DyldInfo.weak_bind_off));
+  return ArrayRef<uint8_t>(Ptr, DyldInfo.weak_bind_size);
+}
+
+ArrayRef<uint8_t> MachOObjectFile::getDyldInfoLazyBindOpcodes() const {
+  if (!DyldInfoLoadCmd) 
+    return ArrayRef<uint8_t>();
+
+  MachO::dyld_info_command DyldInfo 
+                   = getStruct<MachO::dyld_info_command>(this, DyldInfoLoadCmd);
+  const uint8_t *Ptr = reinterpret_cast<const uint8_t*>(
+                                          getPtr(this, DyldInfo.lazy_bind_off));
+  return ArrayRef<uint8_t>(Ptr, DyldInfo.lazy_bind_size);
+}
+
+ArrayRef<uint8_t> MachOObjectFile::getDyldInfoExportsTrie() const {
+  if (!DyldInfoLoadCmd) 
+    return ArrayRef<uint8_t>();
+
+  MachO::dyld_info_command DyldInfo 
+                   = getStruct<MachO::dyld_info_command>(this, DyldInfoLoadCmd);
+  const uint8_t *Ptr = reinterpret_cast<const uint8_t*>(
+                                             getPtr(this, DyldInfo.export_off));
+  return ArrayRef<uint8_t>(Ptr, DyldInfo.export_size);
+}
+
+
 StringRef MachOObjectFile::getStringTableData() const {
   MachO::symtab_command S = getSymtabLoadCommand();
   return getData().substr(S.stroff, S.strsize);
@@ -1688,19 +2034,23 @@ void MachOObjectFile::ReadULEB128s(uint64_t Index,
   }
 }
 
+bool MachOObjectFile::isRelocatableObject() const {
+  return getHeader().filetype == MachO::MH_OBJECT;
+}
+
 ErrorOr<std::unique_ptr<MachOObjectFile>>
-ObjectFile::createMachOObjectFile(std::unique_ptr<MemoryBuffer> &Buffer) {
-  StringRef Magic = Buffer->getBuffer().slice(0, 4);
+ObjectFile::createMachOObjectFile(MemoryBufferRef Buffer) {
+  StringRef Magic = Buffer.getBuffer().slice(0, 4);
   std::error_code EC;
   std::unique_ptr<MachOObjectFile> Ret;
   if (Magic == "\xFE\xED\xFA\xCE")
-    Ret.reset(new MachOObjectFile(std::move(Buffer), false, false, EC));
+    Ret.reset(new MachOObjectFile(Buffer, false, false, EC));
   else if (Magic == "\xCE\xFA\xED\xFE")
-    Ret.reset(new MachOObjectFile(std::move(Buffer), true, false, EC));
+    Ret.reset(new MachOObjectFile(Buffer, true, false, EC));
   else if (Magic == "\xFE\xED\xFA\xCF")
-    Ret.reset(new MachOObjectFile(std::move(Buffer), false, true, EC));
+    Ret.reset(new MachOObjectFile(Buffer, false, true, EC));
   else if (Magic == "\xCF\xFA\xED\xFE")
-    Ret.reset(new MachOObjectFile(std::move(Buffer), true, true, EC));
+    Ret.reset(new MachOObjectFile(Buffer, true, true, EC));
   else
     return object_error::parse_failed;