Factored out common ELF processing into a common superclass, an x86 subclass, and an initial ARM subclass.

Created a new disassembler virtual class, DisassemblerElf32, that
implements the common functions for processing ELF 32-bit files.  This
class is defined in disassembler_elf_32.{h|cc}, and has nearly all of
the methods and data that used to be in
disassembler_elf_32_x86.{h|cc}.  This class has two subclasses, one
for x86 and one for arm, which implement the architecture-specific
parts for generating courgette patches, namely extracting relative and
absolute addresses.  The ARM subclass is just a shell, but still
yields correct courgette patches.

The rest of the files add new enums for ARM and use the ARM subclasses
for the courgette command-line tool.

BUG=

Review URL: https://chromiumcodereview.appspot.com/17325003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@207388 0039d316-1c4b-4281-b951-d872f2087c98

1 files changed, 1 insertions, 439 deletions

diff --git a/courgette/disassembler_elf_32_x86.cc b/courgette/disassembler_elf_32_x86.cc
index a857c96..00bb650 100644
--- a/courgette/disassembler_elf_32_x86.cc
+++ b/courgette/disassembler_elf_32_x86.cc
@@ -18,143 +18,7 @@
 namespace courgette {
 
 DisassemblerElf32X86::DisassemblerElf32X86(const void* start, size_t length)
-  : Disassembler(start, length),
-    header_(NULL),
-    section_header_table_(NULL),
-    section_header_table_size_(0),
-    program_header_table_(NULL),
-    program_header_table_size_(0),
-    default_string_section_(NULL) {
-}
-
-bool DisassemblerElf32X86::ParseHeader() {
-  if (length() < sizeof(Elf32_Ehdr))
-    return Bad("Too small");
-
-  header_ = (Elf32_Ehdr *)start();
-
-  // Have magic for elf header?
-  if (header_->e_ident[0] != 0x7f ||
-      header_->e_ident[1] != 'E' ||
-      header_->e_ident[2] != 'L' ||
-      header_->e_ident[3] != 'F')
-    return Bad("No Magic Number");
-
-  if (header_->e_type != ET_EXEC &&
-      header_->e_type != ET_DYN)
-    return Bad("Not an executable file or shared library");
-
-  if (header_->e_machine != EM_386)
-    return Bad("Not a supported architecture");
-
-  if (header_->e_version != 1)
-    return Bad("Unknown file version");
-
-  if (header_->e_shentsize != sizeof(Elf32_Shdr))
-    return Bad("Unexpected section header size");
-
-  if (header_->e_shoff >= length())
-    return Bad("Out of bounds section header table offset");
-
-  section_header_table_ = (Elf32_Shdr *)OffsetToPointer(header_->e_shoff);
-  section_header_table_size_ = header_->e_shnum;
-
-  if ((header_->e_shoff + header_->e_shnum ) >= length())
-    return Bad("Out of bounds section header table");
-
-  if (header_->e_phoff >= length())
-    return Bad("Out of bounds program header table offset");
-
-  program_header_table_ = (Elf32_Phdr *)OffsetToPointer(header_->e_phoff);
-  program_header_table_size_ = header_->e_phnum;
-
-  if ((header_->e_phoff + header_->e_phnum) >= length())
-    return Bad("Out of bounds program header table");
-
-  default_string_section_ = (const char *)SectionBody((int)header_->e_shstrndx);
-
-  ReduceLength(DiscoverLength());
-
-  return Good();
-}
-
-bool DisassemblerElf32X86::Disassemble(AssemblyProgram* target) {
-  if (!ok())
-    return false;
-
-  // The Image Base is always 0 for ELF Executables
-  target->set_image_base(0);
-
-  if (!ParseAbs32Relocs())
-    return false;
-
-  if (!ParseRel32RelocsFromSections())
-    return false;
-
-  if (!ParseFile(target))
-    return false;
-
-  target->DefaultAssignIndexes();
-
-  return true;
-}
-
-uint32 DisassemblerElf32X86::DiscoverLength() {
-  uint32 result = 0;
-
-  // Find the end of the last section
-  for (int section_id = 0; section_id < SectionHeaderCount(); section_id++) {
-    const Elf32_Shdr *section_header = SectionHeader(section_id);
-
-    if (section_header->sh_type == SHT_NOBITS)
-      continue;
-
-    uint32 section_end = section_header->sh_offset + section_header->sh_size;
-
-    if (section_end > result)
-      result = section_end;
-  }
-
-  // Find the end of the last segment
-  for (int i = 0; i < ProgramSegmentHeaderCount(); i++) {
-    const Elf32_Phdr *segment_header = ProgramSegmentHeader(i);
-
-    uint32 segment_end = segment_header->p_offset + segment_header->p_filesz;
-
-    if (segment_end > result)
-      result = segment_end;
-  }
-
-  uint32 section_table_end = header_->e_shoff +
-                             (header_->e_shnum * sizeof(Elf32_Shdr));
-  if (section_table_end > result)
-    result = section_table_end;
-
-  uint32 segment_table_end = header_->e_phoff +
-                             (header_->e_phnum * sizeof(Elf32_Phdr));
-  if (segment_table_end > result)
-    result = segment_table_end;
-
-  return result;
-}
-
-CheckBool DisassemblerElf32X86::IsValidRVA(RVA rva) const {
-
-  // It's valid if it's contained in any program segment
-  for (int i = 0; i < ProgramSegmentHeaderCount(); i++) {
-    const Elf32_Phdr *segment_header = ProgramSegmentHeader(i);
-
-    if (segment_header->p_type != PT_LOAD)
-      continue;
-
-    Elf32_Addr begin = segment_header->p_vaddr;
-    Elf32_Addr end = segment_header->p_vaddr + segment_header->p_memsz;
-
-    if (rva >= begin && rva < end)
-      return true;
-  }
-
-  return false;
+  : DisassemblerElf32(start, length) {
 }
 
 // Convert an ELF relocation struction into an RVA
@@ -196,151 +60,6 @@ CheckBool DisassemblerElf32X86::RelToRVA(Elf32_Rel rel, RVA* result) const {
   return false;
 }
 
-// Returns RVA for an in memory address, or NULL.
-CheckBool DisassemblerElf32X86::RVAToFileOffset(Elf32_Addr addr,
-                                                size_t* result) const {
-
-  for (int i = 0; i < ProgramSegmentHeaderCount(); i++) {
-    Elf32_Addr begin = ProgramSegmentMemoryBegin(i);
-    Elf32_Addr end = begin + ProgramSegmentMemorySize(i);
-
-    if (addr >= begin  && addr < end) {
-      Elf32_Addr offset = addr - begin;
-
-      if (offset < ProgramSegmentFileSize(i)) {
-        *result = ProgramSegmentFileOffset(i) + offset;
-        return true;
-      }
-    }
-  }
-
-  return false;
-}
-
-RVA DisassemblerElf32X86::FileOffsetToRVA(size_t offset) const {
-  // File offsets can be 64 bit values, but we are dealing with 32
-  // bit executables and so only need to support 32bit file sizes.
-  uint32 offset32 = (uint32)offset;
-
-  for (int i = 0; i < SectionHeaderCount(); i++) {
-
-    const Elf32_Shdr *section_header = SectionHeader(i);
-
-    // These can appear to have a size in the file, but don't.
-    if (section_header->sh_type == SHT_NOBITS)
-      continue;
-
-    Elf32_Off section_begin = section_header->sh_offset;
-    Elf32_Off section_end = section_begin + section_header->sh_size;
-
-    if (offset32 >= section_begin && offset32 < section_end) {
-      return section_header->sh_addr + (offset32 - section_begin);
-    }
-  }
-
-  return 0;
-}
-
-CheckBool DisassemblerElf32X86::RVAsToOffsets(std::vector<RVA>* rvas,
-                                              std::vector<size_t>* offsets) {
-  offsets->clear();
-
-  for (std::vector<RVA>::iterator rva = rvas->begin();
-       rva != rvas->end();
-       rva++) {
-
-    size_t offset;
-
-    if (!RVAToFileOffset(*rva, &offset))
-      return false;
-
-    offsets->push_back(offset);
-  }
-
-  return true;
-}
-
-CheckBool DisassemblerElf32X86::ParseFile(AssemblyProgram* program) {
-  // Walk all the bytes in the file, whether or not in a section.
-  uint32 file_offset = 0;
-
-  std::vector<size_t> abs_offsets;
-  std::vector<size_t> rel_offsets;
-
-  if (!RVAsToOffsets(&abs32_locations_, &abs_offsets))
-    return false;
-
-  if (!RVAsToOffsets(&rel32_locations_, &rel_offsets))
-    return false;
-
-  std::vector<size_t>::iterator current_abs_offset = abs_offsets.begin();
-  std::vector<size_t>::iterator current_rel_offset = rel_offsets.begin();
-
-  std::vector<size_t>::iterator end_abs_offset = abs_offsets.end();
-  std::vector<size_t>::iterator end_rel_offset = rel_offsets.end();
-
-  for (int section_id = 0;
-       section_id < SectionHeaderCount();
-       section_id++) {
-
-    const Elf32_Shdr *section_header = SectionHeader(section_id);
-
-    if (!ParseSimpleRegion(file_offset,
-                           section_header->sh_offset,
-                           program))
-      return false;
-    file_offset = section_header->sh_offset;
-
-    switch (section_header->sh_type) {
-      case SHT_REL:
-        if (!ParseRelocationSection(section_header, program))
-          return false;
-        file_offset = section_header->sh_offset + section_header->sh_size;
-        break;
-      case SHT_PROGBITS:
-        if (!ParseProgbitsSection(section_header,
-                                  &current_abs_offset, end_abs_offset,
-                                  &current_rel_offset, end_rel_offset,
-                                  program))
-          return false;
-        file_offset = section_header->sh_offset + section_header->sh_size;
-        break;
-      case SHT_NOBITS:
-        // Fall through
-      case SHT_INIT_ARRAY:
-        // Fall through
-      case SHT_FINI_ARRAY:
-        while (current_abs_offset != end_abs_offset &&
-               *current_abs_offset >= section_header->sh_offset &&
-               *current_abs_offset <
-               (section_header->sh_offset + section_header->sh_size)) {
-          // Skip any abs_offsets appear in the unsupported INIT_ARRAY section
-          VLOG(1) << "Skipping relocation entry for unsupported section: " <<
-            section_header->sh_type;
-          current_abs_offset++;
-        }
-        break;
-      default:
-        if (current_abs_offset != end_abs_offset &&
-               *current_abs_offset >= section_header->sh_offset &&
-               *current_abs_offset <
-               (section_header->sh_offset + section_header->sh_size))
-          VLOG(1) << "Relocation address in unrecognized ELF section: " << \
-            section_header->sh_type;
-      break;
-    }
-  }
-
-  // Rest of the file past the last section
-  if (!ParseSimpleRegion(file_offset,
-                         length(),
-                         program))
-    return false;
-
-  // Make certain we consume all of the relocations as expected
-  return (current_abs_offset == end_abs_offset);
-}
-
 CheckBool DisassemblerElf32X86::ParseRelocationSection(
     const Elf32_Shdr *section_header,
       AssemblyProgram* program) {
@@ -395,163 +114,6 @@ CheckBool DisassemblerElf32X86::ParseRelocationSection(
   return ParseSimpleRegion(file_offset, section_end, program);
 }
 
-CheckBool DisassemblerElf32X86::ParseProgbitsSection(
-    const Elf32_Shdr *section_header,
-    std::vector<size_t>::iterator* current_abs_offset,
-    std::vector<size_t>::iterator end_abs_offset,
-    std::vector<size_t>::iterator* current_rel_offset,
-    std::vector<size_t>::iterator end_rel_offset,
-    AssemblyProgram* program) {
-
-  // Walk all the bytes in the file, whether or not in a section.
-  size_t file_offset = section_header->sh_offset;
-  size_t section_end = section_header->sh_offset + section_header->sh_size;
-
-  Elf32_Addr origin = section_header->sh_addr;
-  size_t origin_offset = section_header->sh_offset;
-  if (!program->EmitOriginInstruction(origin))
-    return false;
-
-  while (file_offset < section_end) {
-
-    if (*current_abs_offset != end_abs_offset &&
-        file_offset > **current_abs_offset)
-      return false;
-
-    while (*current_rel_offset != end_rel_offset &&
-           file_offset > **current_rel_offset) {
-      (*current_rel_offset)++;
-    }
-
-    size_t next_relocation = section_end;
-
-    if (*current_abs_offset != end_abs_offset &&
-        next_relocation > **current_abs_offset)
-      next_relocation = **current_abs_offset;
-
-    // Rel offsets are heuristically derived, and might (incorrectly) overlap
-    // an Abs value, or the end of the section, so +3 to make sure there is
-    // room for the full 4 byte value.
-    if (*current_rel_offset != end_rel_offset &&
-        next_relocation > (**current_rel_offset + 3))
-      next_relocation = **current_rel_offset;
-
-    if (next_relocation > file_offset) {
-      if (!ParseSimpleRegion(file_offset, next_relocation, program))
-        return false;
-
-      file_offset = next_relocation;
-      continue;
-    }
-
-    if (*current_abs_offset != end_abs_offset &&
-        file_offset == **current_abs_offset) {
-
-      const uint8* p = OffsetToPointer(file_offset);
-      RVA target_rva = Read32LittleEndian(p);
-
-      if (!program->EmitAbs32(program->FindOrMakeAbs32Label(target_rva)))
-        return false;
-      file_offset += sizeof(RVA);
-      (*current_abs_offset)++;
-      continue;
-    }
-
-    if (*current_rel_offset != end_rel_offset &&
-        file_offset == **current_rel_offset) {
-
-      const uint8* p = OffsetToPointer(file_offset);
-      uint32 relative_target = Read32LittleEndian(p);
-      // This cast is for 64 bit systems, and is only safe because we
-      // are working on 32 bit executables.
-      RVA target_rva = (RVA)(origin + (file_offset - origin_offset) +
-                             4 + relative_target);
-
-      if (!program->EmitRel32(program->FindOrMakeRel32Label(target_rva)))
-        return false;
-      file_offset += sizeof(RVA);
-      (*current_rel_offset)++;
-      continue;
-    }
-  }
-
-  // Rest of the section (if any)
-  return ParseSimpleRegion(file_offset, section_end, program);
-}
-
-CheckBool DisassemblerElf32X86::ParseSimpleRegion(
-    size_t start_file_offset,
-    size_t end_file_offset,
-    AssemblyProgram* program) {
-
-  const uint8* start = OffsetToPointer(start_file_offset);
-  const uint8* end = OffsetToPointer(end_file_offset);
-
-  const uint8* p = start;
-
-  while (p < end) {
-    if (!program->EmitByteInstruction(*p))
-      return false;
-    ++p;
-  }
-
-  return true;
-}
-
-CheckBool DisassemblerElf32X86::ParseAbs32Relocs() {
-  abs32_locations_.clear();
-
-  // Loop through sections for relocation sections
-  for (int section_id = 0; section_id < SectionHeaderCount(); section_id++) {
-    const Elf32_Shdr *section_header = SectionHeader(section_id);
-
-    if (section_header->sh_type == SHT_REL) {
-
-      Elf32_Rel *relocs_table = (Elf32_Rel *)SectionBody(section_id);
-
-      int relocs_table_count = section_header->sh_size /
-                               section_header->sh_entsize;
-
-      // Elf32_Word relocation_section_id = section_header->sh_info;
-
-      // Loop through relocation objects in the relocation section
-      for (int rel_id = 0; rel_id < relocs_table_count; rel_id++) {
-        RVA rva;
-
-        // Quite a few of these conversions fail, and we simply skip
-        // them, that's okay.
-        if (RelToRVA(relocs_table[rel_id], &rva))
-          abs32_locations_.push_back(rva);
-      }
-    }
-  }
-
-  std::sort(abs32_locations_.begin(), abs32_locations_.end());
-  return true;
-}
-
-CheckBool DisassemblerElf32X86::ParseRel32RelocsFromSections() {
-
-  rel32_locations_.clear();
-
-  // Loop through sections for relocation sections
-  for (int section_id = 0;
-       section_id < SectionHeaderCount();
-       section_id++) {
-
-    const Elf32_Shdr *section_header = SectionHeader(section_id);
-
-    if (section_header->sh_type != SHT_PROGBITS)
-      continue;
-
-    if (!ParseRel32RelocsFromSection(section_header))
-      return false;
-  }
-
-  std::sort(rel32_locations_.begin(), rel32_locations_.end());
-  return true;
-}
-
 CheckBool DisassemblerElf32X86::ParseRel32RelocsFromSection(
     const Elf32_Shdr* section_header) {