Revert of [Courgette] Clean up Disassembler; fix ELF Memory leaks.

courgette/disassembler_win32_x64.cc

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "courgette/disassembler_win32_x64.h"
 
 #include <stddef.h>
 #include <stdint.h>
 
 #include <algorithm>
-#include <iostream>
+#include <string>
+#include <vector>
 
 #include "base/logging.h"
 #include "base/numerics/safe_conversions.h"
+
 #include "courgette/assembly_program.h"
 #include "courgette/courgette.h"
+#include "courgette/encoded_program.h"
 
 namespace courgette {
 
 DisassemblerWin32X64::DisassemblerWin32X64(const void* start, size_t length)
-    : Disassembler(start, length),
-      incomplete_disassembly_(false),
-      is_PE32_plus_(false),
-      optional_header_(nullptr),
-      size_of_optional_header_(0),
-      offset_of_data_directories_(0),
-      machine_type_(0),
-      number_of_sections_(0),
-      sections_(nullptr),
-      has_text_section_(false),
-      size_of_code_(0),
-      size_of_initialized_data_(0),
-      size_of_uninitialized_data_(0),
-      base_of_code_(0),
-      base_of_data_(0),
-      image_base_(0),
-      size_of_image_(0),
-      number_of_data_directories_(0) {
-}
-
-FileOffset DisassemblerWin32X64::RVAToFileOffset(RVA rva) const {
-  const Section* section = RVAToSection(rva);
-  if (section != nullptr) {
-    FileOffset offset_in_section = rva - section->virtual_address;
-    // Need this extra check, since an |rva| may be valid for a section, but is
-    // non-existent in an image (e.g. uninit data).
-    if (offset_in_section >= section->size_of_raw_data)
-      return kNoFileOffset;
-
-    return static_cast<FileOffset>(section->file_offset_of_raw_data +
-                                   offset_in_section);
-  }
-
-  // Small RVA values point into the file header in the loaded image.
-  // RVA 0 is the module load address which Windows uses as the module handle.
-  // RVA 2 sometimes occurs, I'm not sure what it is, but it would map into the
-  // DOS header.
-  if (rva == 0 || rva == 2)
-    return static_cast<FileOffset>(rva);
-
-  NOTREACHED();
-  return kNoFileOffset;
-}
-
-RVA DisassemblerWin32X64::FileOffsetToRVA(FileOffset file_offset) const {
-  for (int i = 0; i < number_of_sections_; ++i) {
-    const Section* section = &sections_[i];
-    if (file_offset >= section->file_offset_of_raw_data) {
-      FileOffset offset_in_section =
-          file_offset - section->file_offset_of_raw_data;
-      if (offset_in_section < section->size_of_raw_data)
-        return static_cast<RVA>(section->virtual_address + offset_in_section);
-    }
-  }
-
-  NOTREACHED();
-  return kNoRVA;
+  : Disassembler(start, length),
+    incomplete_disassembly_(false),
+    is_PE32_plus_(false),
+    optional_header_(NULL),
+    size_of_optional_header_(0),
+    offset_of_data_directories_(0),
+    machine_type_(0),
+    number_of_sections_(0),
+    sections_(NULL),
+    has_text_section_(false),
+    size_of_code_(0),
+    size_of_initialized_data_(0),
+    size_of_uninitialized_data_(0),
+    base_of_code_(0),
+    base_of_data_(0),
+    image_base_(0),
+    size_of_image_(0),
+    number_of_data_directories_(0) {
 }
 
 // ParseHeader attempts to match up the buffer with the Windows data
 // structures that exist within a Windows 'Portable Executable' format file.
 // Returns 'true' if the buffer matches, and 'false' if the data looks
 // suspicious.  Rather than try to 'map' the buffer to the numerous windows
 // structures, we extract the information we need into the courgette::PEInfo
 // structure.
 //
 bool DisassemblerWin32X64::ParseHeader() {
   if (length() < kOffsetOfFileAddressOfNewExeHeader + 4 /*size*/)
     return Bad("Too small");
 
   // Have 'MZ' magic for a DOS header?
   if (start()[0] != 'M' || start()[1] != 'Z')
     return Bad("Not MZ");
 
   // offset from DOS header to PE header is stored in DOS header.
-  FileOffset file_offset = static_cast<FileOffset>(
-      ReadU32(start(), kOffsetOfFileAddressOfNewExeHeader));
+  uint32_t offset = ReadU32(start(), kOffsetOfFileAddressOfNewExeHeader);
 
-  if (file_offset >= length())
+  if (offset >= length())
     return Bad("Bad offset to PE header");
 
-  const uint8_t* const pe_header = FileOffsetToPointer(file_offset);
+  const uint8_t* const pe_header = OffsetToPointer(offset);
   const size_t kMinPEHeaderSize = 4 /*signature*/ + kSizeOfCoffHeader;
   if (pe_header <= start() ||
       pe_header >= end() - kMinPEHeaderSize)
-    return Bad("Bad file offset to PE header");
+    return Bad("Bad offset to PE header");
 
-  if (file_offset % 8 != 0)
+  if (offset % 8 != 0)
     return Bad("Misaligned PE header");
 
   // The 'PE' header is an IMAGE_NT_HEADERS structure as defined in WINNT.H.
   // See http://msdn.microsoft.com/en-us/library/ms680336(VS.85).aspx
   //
   // The first field of the IMAGE_NT_HEADERS is the signature.
   if (!(pe_header[0] == 'P' &&
         pe_header[1] == 'E' &&
         pe_header[2] == 0 &&
         pe_header[3] == 0))
@@ skipping 80 matching lines @@
   if (!b) {
     return Bad("malformed data directory");
   }
 
   // Sections follow the optional header.
   sections_ =
       reinterpret_cast<const Section*>(optional_header +
                                        size_of_optional_header_);
   size_t detected_length = 0;
 
-  for (int i = 0; i < number_of_sections_; ++i) {
+  for (int i = 0;  i < number_of_sections_;  ++i) {
     const Section* section = &sections_[i];
 
     // TODO(sra): consider using the 'characteristics' field of the section
     // header to see if the section contains instructions.
     if (memcmp(section->name, ".text", 6) == 0)
       has_text_section_ = true;
 
     uint32_t section_end =
         section->file_offset_of_raw_data + section->size_of_raw_data;
     if (section_end > detected_length)
@@ skipping 77 matching lines @@
 
     // Walk through the two-byte entries.
     for (const uint8_t* p = block + 8; p < end_entries; p += 2) {
       uint16_t entry = ReadU16(p, 0);
       int type = entry >> 12;
       int offset = entry & 0xFFF;
 
       RVA rva = page_rva + offset;
       // TODO(sebmarchand): Skip the relocs that live outside of the image. See
       //   the version of this function in disassembler_win32_x86.cc.
-      if (type == 10) {  // IMAGE_REL_BASED_DIR64
+      if (type == 10) {         // IMAGE_REL_BASED_DIR64
         relocs->push_back(rva);
       } else if (type == 0) {  // IMAGE_REL_BASED_ABSOLUTE
         // Ignore, used as padding.
       } else {
         // Does not occur in Windows x64 executables.
         return Bad("unknown type of reloc");
       }
     }
 
     block += size;
   }
 
   std::sort(relocs->begin(), relocs->end());
   DCHECK(relocs->empty() || relocs->back() != kUnassignedRVA);
 
   return true;
 }
 
 const Section* DisassemblerWin32X64::RVAToSection(RVA rva) const {
-  for (int i = 0; i < number_of_sections_; ++i) {
+  for (int i = 0; i < number_of_sections_; i++) {
     const Section* section = &sections_[i];
-    if (rva >= section->virtual_address) {
-      FileOffset offset_in_section = rva - section->virtual_address;
-      if (offset_in_section < section->virtual_size)
-        return section;
+    uint32_t offset = rva - section->virtual_address;
+    if (offset < section->virtual_size) {
+      return section;
     }
   }
-  return nullptr;
+  return NULL;
+}
+
+int DisassemblerWin32X64::RVAToFileOffset(RVA rva) const {
+  const Section* section = RVAToSection(rva);
+  if (section) {
+    uint32_t offset = rva - section->virtual_address;
+    if (offset < section->size_of_raw_data) {
+      return section->file_offset_of_raw_data + offset;
+    } else {
+      return kNoOffset;  // In section but not in file (e.g. uninit data).
+    }
+  }
+
+  // Small RVA values point into the file header in the loaded image.
+  // RVA 0 is the module load address which Windows uses as the module handle.
+  // RVA 2 sometimes occurs, I'm not sure what it is, but it would map into the
+  // DOS header.
+  if (rva == 0 || rva == 2)
+    return rva;
+
+  NOTREACHED();
+  return kNoOffset;
+}
+
+const uint8_t* DisassemblerWin32X64::RVAToPointer(RVA rva) const {
+  int file_offset = RVAToFileOffset(rva);
+  if (file_offset == kNoOffset)
+    return NULL;
+  else
+    return OffsetToPointer(file_offset);
 }
 
 std::string DisassemblerWin32X64::SectionName(const Section* section) {
-  if (section == nullptr)
+  if (section == NULL)
     return "<none>";
   char name[9];
   memcpy(name, section->name, 8);
   name[8] = '\0';  // Ensure termination.
   return name;
 }
 
 CheckBool DisassemblerWin32X64::ParseFile(AssemblyProgram* program) {
   // Walk all the bytes in the file, whether or not in a section.
-  FileOffset file_offset = 0;
+  uint32_t file_offset = 0;
   while (file_offset < length()) {
     const Section* section = FindNextSection(file_offset);
-    if (section == nullptr) {
-      // No more sections. There should not be extra stuff following last
+    if (section == NULL) {
+      // No more sections.  There should not be extra stuff following last
       // section.
       //   ParseNonSectionFileRegion(file_offset, pe_info().length(), program);
       break;
     }
     if (file_offset < section->file_offset_of_raw_data) {
-      FileOffset section_start_offset = section->file_offset_of_raw_data;
-      if (!ParseNonSectionFileRegion(file_offset, section_start_offset,
-                                     program)) {
+      uint32_t section_start_offset = section->file_offset_of_raw_data;
+      if(!ParseNonSectionFileRegion(file_offset, section_start_offset,
+                                    program))
         return false;
-      }
 
       file_offset = section_start_offset;
     }
-    FileOffset end = file_offset + section->size_of_raw_data;
+    uint32_t end = file_offset + section->size_of_raw_data;
     if (!ParseFileRegion(section, file_offset, end, program))
       return false;
     file_offset = end;
   }
 
 #if COURGETTE_HISTOGRAM_TARGETS
   HistogramTargets("abs32 relocs", abs32_target_rvas_);
   HistogramTargets("rel32 relocs", rel32_target_rvas_);
 #endif
 
   return true;
 }
 
 bool DisassemblerWin32X64::ParseAbs32Relocs() {
   abs32_locations_.clear();
   if (!ParseRelocs(&abs32_locations_))
     return false;
 
 #if COURGETTE_HISTOGRAM_TARGETS
-  for (size_t i = 0; i < abs32_locations_.size(); ++i) {
+  for (size_t i = 0;  i < abs32_locations_.size(); ++i) {
     RVA rva = abs32_locations_[i];
     // The 4 bytes at the relocation are a reference to some address.
     uint32_t target_address = Read32LittleEndian(RVAToPointer(rva));
     ++abs32_target_rvas_[target_address - image_base()];
   }
 #endif
   return true;
 }
 
 void DisassemblerWin32X64::ParseRel32RelocsFromSections() {
-  FileOffset file_offset = 0;
+  uint32_t file_offset = 0;
   while (file_offset < length()) {
     const Section* section = FindNextSection(file_offset);
-    if (section == nullptr)
+    if (section == NULL)
       break;
     if (file_offset < section->file_offset_of_raw_data)
       file_offset = section->file_offset_of_raw_data;
     ParseRel32RelocsFromSection(section);
     file_offset += section->size_of_raw_data;
   }
   std::sort(rel32_locations_.begin(), rel32_locations_.end());
   DCHECK(rel32_locations_.empty() ||
          rel32_locations_.back() != kUnassignedRVA);
 
 #if COURGETTE_HISTOGRAM_TARGETS
   VLOG(1) << "abs32_locations_ " << abs32_locations_.size()
           << "\nrel32_locations_ " << rel32_locations_.size()
           << "\nabs32_target_rvas_ " << abs32_target_rvas_.size()
           << "\nrel32_target_rvas_ " << rel32_target_rvas_.size();
 
   int common = 0;
   std::map<RVA, int>::iterator abs32_iter = abs32_target_rvas_.begin();
   std::map<RVA, int>::iterator rel32_iter = rel32_target_rvas_.begin();
   while (abs32_iter != abs32_target_rvas_.end() &&
          rel32_iter != rel32_target_rvas_.end()) {
-    if (abs32_iter->first < rel32_iter->first) {
+    if (abs32_iter->first < rel32_iter->first)
       ++abs32_iter;
-    } else if (rel32_iter->first < abs32_iter->first) {
+    else if (rel32_iter->first < abs32_iter->first)
       ++rel32_iter;
-    } else {
+    else {
       ++common;
       ++abs32_iter;
       ++rel32_iter;
     }
   }
   VLOG(1) << "common " << common;
 #endif
 }
 
 void DisassemblerWin32X64::ParseRel32RelocsFromSection(const Section* section) {
   // TODO(sra): use characteristic.
   bool isCode = strcmp(section->name, ".text") == 0;
   if (!isCode)
     return;
 
-  FileOffset start_file_offset = section->file_offset_of_raw_data;
-  FileOffset end_file_offset = start_file_offset + section->size_of_raw_data;
+  uint32_t start_file_offset = section->file_offset_of_raw_data;
+  uint32_t end_file_offset = start_file_offset + section->size_of_raw_data;
   RVA relocs_start_rva = base_relocation_table().address_;
 
-  const uint8_t* start_pointer = FileOffsetToPointer(start_file_offset);
-  const uint8_t* end_pointer = FileOffsetToPointer(end_file_offset);
+  const uint8_t* start_pointer = OffsetToPointer(start_file_offset);
+  const uint8_t* end_pointer = OffsetToPointer(end_file_offset);
 
   RVA start_rva = FileOffsetToRVA(start_file_offset);
   RVA end_rva = start_rva + section->virtual_size;
 
   // Quick way to convert from Pointer to RVA within a single Section is to
-  // subtract |pointer_to_rva|.
+  // subtract 'pointer_to_rva'.
   const uint8_t* const adjust_pointer_to_rva = start_pointer - start_rva;
 
   std::vector<RVA>::iterator abs32_pos = abs32_locations_.begin();
 
   // Find the rel32 relocations.
   const uint8_t* p = start_pointer;
   while (p < end_pointer) {
     RVA current_rva = static_cast<RVA>(p - adjust_pointer_to_rva);
     if (current_rva == relocs_start_rva) {
       uint32_t relocs_size = base_relocation_table().size_;
       if (relocs_size) {
         p += relocs_size;
         continue;
       }
     }
 
+    //while (abs32_pos != abs32_locations_.end() && *abs32_pos < current_rva)
+    //  ++abs32_pos;
+
     // Heuristic discovery of rel32 locations in instruction stream: are the
     // next few bytes the start of an instruction containing a rel32
     // addressing mode?
-    const uint8_t* rel32 = nullptr;
+    const uint8_t* rel32 = NULL;
     bool is_rip_relative = false;
 
     if (p + 5 <= end_pointer) {
       if (*p == 0xE8 || *p == 0xE9)  // jmp rel32 and call rel32
         rel32 = p + 1;
     }
     if (p + 6 <= end_pointer) {
       if (*p == 0x0F && (*(p + 1) & 0xF0) == 0x80) {  // Jcc long form
         if (p[1] != 0x8A && p[1] != 0x8B)  // JPE/JPO unlikely
           rel32 = p + 2;
@@ skipping 30 matching lines @@
           // Beginning of abs32 reloc is before end of rel32 reloc so they
           // overlap.  Skip four bytes past the abs32 reloc.
           p += (*abs32_pos + 4) - current_rva;
           continue;
         }
       }
 
       RVA target_rva = rel32_rva + 4 + Read32LittleEndian(rel32);
       // To be valid, rel32 target must be within image, and within this
       // section.
-      if (target_rva < size_of_image_ &&  // Subsumes rva != kUnassignedRVA.
+      if (IsValidRVA(target_rva) &&
           (is_rip_relative ||
            (start_rva <= target_rva && target_rva < end_rva))) {
         rel32_locations_.push_back(rel32_rva);
 #if COURGETTE_HISTOGRAM_TARGETS
         ++rel32_target_rvas_[target_rva];
 #endif
         p = rel32 + 4;
         continue;
       }
     }
     p += 1;
   }
 }
 
 CheckBool DisassemblerWin32X64::ParseNonSectionFileRegion(
-    FileOffset start_file_offset,
-    FileOffset end_file_offset,
+    uint32_t start_file_offset,
+    uint32_t end_file_offset,
     AssemblyProgram* program) {
   if (incomplete_disassembly_)
     return true;
 
   if (end_file_offset > start_file_offset) {
-    if (!program->EmitBytesInstruction(FileOffsetToPointer(start_file_offset),
+    if (!program->EmitBytesInstruction(OffsetToPointer(start_file_offset),
                                        end_file_offset - start_file_offset)) {
       return false;
     }
   }
 
   return true;
 }
 
 CheckBool DisassemblerWin32X64::ParseFileRegion(const Section* section,
-                                                FileOffset start_file_offset,
-                                                FileOffset end_file_offset,
+                                                uint32_t start_file_offset,
+                                                uint32_t end_file_offset,
                                                 AssemblyProgram* program) {
   RVA relocs_start_rva = base_relocation_table().address_;
 
-  const uint8_t* start_pointer = FileOffsetToPointer(start_file_offset);
-  const uint8_t* end_pointer = FileOffsetToPointer(end_file_offset);
+  const uint8_t* start_pointer = OffsetToPointer(start_file_offset);
+  const uint8_t* end_pointer = OffsetToPointer(end_file_offset);
 
   RVA start_rva = FileOffsetToRVA(start_file_offset);
   RVA end_rva = start_rva + section->virtual_size;
 
   // Quick way to convert from Pointer to RVA within a single Section is to
   // subtract 'pointer_to_rva'.
   const uint8_t* const adjust_pointer_to_rva = start_pointer - start_rva;
 
   std::vector<RVA>::iterator rel32_pos = rel32_locations_.begin();
   std::vector<RVA>::iterator abs32_pos = abs32_locations_.begin();
@@ skipping 88 matching lines @@
        p != h.rend();
        ++p) {
     ++index;
     if (index <= kFirstN || p->first <= 3) {
       if (someSkipped) {
         std::cout << "..." << std::endl;
       }
       size_t count = p->second.size();
       std::cout << std::dec << p->first << ": " << count;
       if (count <= 2) {
-        for (size_t i = 0; i < count; ++i)
+        for (size_t i = 0;  i < count;  ++i)
           std::cout << "  " << DescribeRVA(p->second[i]);
       }
       std::cout << std::endl;
       someSkipped = false;
     } else {
       someSkipped = true;
     }
   }
 }
 #endif  // COURGETTE_HISTOGRAM_TARGETS
 
+
 // DescribeRVA is for debugging only.  I would put it under #ifdef DEBUG except
 // that during development I'm finding I need to call it when compiled in
 // Release mode.  Hence:
 // TODO(sra): make this compile only for debug mode.
 std::string DisassemblerWin32X64::DescribeRVA(RVA rva) const {
   const Section* section = RVAToSection(rva);
   std::ostringstream s;
   s << std::hex << rva;
   if (section) {
     s << " (";
     s << SectionName(section) << "+"
       << std::hex << (rva - section->virtual_address)
       << ")";
   }
   return s.str();
 }
 
 const Section* DisassemblerWin32X64::FindNextSection(
-    FileOffset file_offset) const {
+    uint32_t fileOffset) const {
   const Section* best = 0;
-  for (int i = 0; i < number_of_sections_; ++i) {
+  for (int i = 0; i < number_of_sections_; i++) {
     const Section* section = &sections_[i];
     if (section->size_of_raw_data > 0) {  // i.e. has data in file.
-      if (file_offset <= section->file_offset_of_raw_data) {
+      if (fileOffset <= section->file_offset_of_raw_data) {
         if (best == 0 ||
             section->file_offset_of_raw_data < best->file_offset_of_raw_data) {
           best = section;
         }
       }
     }
   }
   return best;
 }
 
+RVA DisassemblerWin32X64::FileOffsetToRVA(uint32_t file_offset) const {
+  for (int i = 0; i < number_of_sections_; i++) {
+    const Section* section = &sections_[i];
+    uint32_t offset = file_offset - section->file_offset_of_raw_data;
+    if (offset < section->size_of_raw_data) {
+      return section->virtual_address + offset;
+    }
+  }
+  return 0;
+}
+
 bool DisassemblerWin32X64::ReadDataDirectory(
     int index,
     ImageDataDirectory* directory) {
 
   if (index < number_of_data_directories_) {
-    FileOffset file_offset = index * 8 + offset_of_data_directories_;
-    if (file_offset >= size_of_optional_header_)
+    size_t offset = index * 8 + offset_of_data_directories_;
+    if (offset >= size_of_optional_header_)
       return Bad("number of data directories inconsistent");
-    const uint8_t* data_directory = optional_header_ + file_offset;
+    const uint8_t* data_directory = optional_header_ + offset;
     if (data_directory < start() ||
         data_directory + 8 >= end())
       return Bad("data directory outside image");
     RVA rva = ReadU32(data_directory, 0);
     size_t size  = ReadU32(data_directory, 4);
     if (size > size_of_image_)
       return Bad("data directory size too big");
 
     // TODO(sra): validate RVA.
     directory->address_ = rva;
     directory->size_ = static_cast<uint32_t>(size);
     return true;
   } else {
     directory->address_ = 0;
     directory->size_ = 0;
     return true;
   }
 }
 
 }  // namespace courgette