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

courgette/disassembler_win32_x64.cc

Index: courgette/disassembler_win32_x64.cc
diff --git a/courgette/disassembler_win32_x64.cc b/courgette/disassembler_win32_x64.cc
index 74b0fe48a7fe7800abe9ffebcf1637daabc292db..a6a90118867c4dae2058c6204b3b396a30119077 100644
--- a/courgette/disassembler_win32_x64.cc
+++ b/courgette/disassembler_win32_x64.cc
@@ -8,28 +8,25 @@
 #include <stdint.h>
 
 #include <algorithm>
-#include <string>
-#include <vector>
+#include <iostream>
 
 #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),
+    : Disassembler(start, length),
     incomplete_disassembly_(false),
     is_PE32_plus_(false),
-    optional_header_(NULL),
+    optional_header_(nullptr),
     size_of_optional_header_(0),
     offset_of_data_directories_(0),
     machine_type_(0),
     number_of_sections_(0),
-    sections_(NULL),
+    sections_(nullptr),
     has_text_section_(false),
     size_of_code_(0),
     size_of_initialized_data_(0),
@@ -41,6 +38,45 @@ DisassemblerWin32X64::DisassemblerWin32X64(const void* start, size_t length)
     number_of_data_directories_(0) {
 }
 
+FileOffset DisassemblerWin32X64::RVAToFileOffset(RVA rva) const {
+  const Section* section = RVAToSection(rva);
+  if (section != nullptr) {
+    uint32_t 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) {
+      uint32_t 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;
+}
+
 // 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
@@ -57,18 +93,19 @@ bool DisassemblerWin32X64::ParseHeader() {
     return Bad("Not MZ");
 
   // offset from DOS header to PE header is stored in DOS header.
-  uint32_t offset = ReadU32(start(), kOffsetOfFileAddressOfNewExeHeader);
+  FileOffset file_offset = static_cast<FileOffset>(
+      ReadU32(start(), kOffsetOfFileAddressOfNewExeHeader));
 
-  if (offset >= length())
+  if (file_offset >= length())
     return Bad("Bad offset to PE header");
 
-  const uint8_t* const pe_header = OffsetToPointer(offset);
+  const uint8_t* const pe_header = FileOffsetToPointer(file_offset);
   const size_t kMinPEHeaderSize = 4 /*signature*/ + kSizeOfCoffHeader;
   if (pe_header <= start() ||
       pe_header >= end() - kMinPEHeaderSize)
-    return Bad("Bad offset to PE header");
+    return Bad("Bad file offset to PE header");
 
-  if (offset % 8 != 0)
+  if (file_offset % 8 != 0)
     return Bad("Misaligned PE header");
 
   // The 'PE' header is an IMAGE_NT_HEADERS structure as defined in WINNT.H.
@@ -169,7 +206,7 @@ bool DisassemblerWin32X64::ParseHeader() {
                                        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
@@ -287,48 +324,19 @@ bool DisassemblerWin32X64::ParseRelocs(std::vector<RVA> *relocs) {
 }
 
 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];
-    uint32_t offset = rva - section->virtual_address;
-    if (offset < section->virtual_size) {
-      return section;
+    if (rva >= section->virtual_address) {
+      uint32_t offset_in_section = rva - section->virtual_address;
+      if (offset_in_section < section->virtual_size)
+        return section;
     }
   }
-  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);
+  return nullptr;
 }
 
 std::string DisassemblerWin32X64::SectionName(const Section* section) {
-  if (section == NULL)
+  if (section == nullptr)
     return "<none>";
   char name[9];
   memcpy(name, section->name, 8);
@@ -338,24 +346,25 @@ std::string DisassemblerWin32X64::SectionName(const Section* section) {
 
 CheckBool DisassemblerWin32X64::ParseFile(AssemblyProgram* program) {
   // Walk all the bytes in the file, whether or not in a section.
-  uint32_t file_offset = 0;
+  FileOffset file_offset = 0;
   while (file_offset < length()) {
     const Section* section = FindNextSection(file_offset);
-    if (section == NULL) {
-      // No more sections.  There should not be extra stuff following last
+    if (section == nullptr) {
+      // 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) {
-      uint32_t section_start_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))
+                                    program)) {
         return false;
+      }
 
       file_offset = section_start_offset;
     }
-    uint32_t end = file_offset + section->size_of_raw_data;
+    FileOffset end = file_offset + section->size_of_raw_data;
     if (!ParseFileRegion(section, file_offset, end, program))
       return false;
     file_offset = end;
@@ -389,7 +398,7 @@ void DisassemblerWin32X64::ParseRel32RelocsFromSections() {
   uint32_t file_offset = 0;
   while (file_offset < length()) {
     const Section* section = FindNextSection(file_offset);
-    if (section == NULL)
+    if (section == nullptr)
       break;
     if (file_offset < section->file_offset_of_raw_data)
       file_offset = section->file_offset_of_raw_data;
@@ -411,11 +420,11 @@ void DisassemblerWin32X64::ParseRel32RelocsFromSections() {
   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;
@@ -431,18 +440,18 @@ void DisassemblerWin32X64::ParseRel32RelocsFromSection(const Section* section) {
   if (!isCode)
     return;
 
-  uint32_t start_file_offset = section->file_offset_of_raw_data;
-  uint32_t end_file_offset = start_file_offset + section->size_of_raw_data;
+  FileOffset start_file_offset = section->file_offset_of_raw_data;
+  FileOffset end_file_offset = start_file_offset + section->size_of_raw_data;
   RVA relocs_start_rva = base_relocation_table().address_;
 
-  const uint8_t* start_pointer = OffsetToPointer(start_file_offset);
-  const uint8_t* end_pointer = OffsetToPointer(end_file_offset);
+  const uint8_t* start_pointer = FileOffsetToPointer(start_file_offset);
+  const uint8_t* end_pointer = FileOffsetToPointer(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();
@@ -459,13 +468,10 @@ void DisassemblerWin32X64::ParseRel32RelocsFromSection(const Section* section) {
       }
     }
 
-    //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 = NULL;
+    const uint8_t* rel32 = nullptr;
     bool is_rip_relative = false;
 
     if (p + 5 <= end_pointer) {
@@ -516,9 +522,9 @@ void DisassemblerWin32X64::ParseRel32RelocsFromSection(const Section* section) {
       RVA target_rva = rel32_rva + 4 + Read32LittleEndian(rel32);
       // To be valid, rel32 target must be within image, and within this
       // section.
-      if (IsValidRVA(target_rva) &&
+      if (target_rva < size_of_image_ &&  // Subsumes rva != kUnassignedRVA.
           (is_rip_relative ||

[Will Harris, 2016/03/08 19:09:07]

is git cl format okay with this indenting on line 526?

[huangs, 2016/03/09 17:53:29]

"git cl format" recommends single relative indent.  Going to apply some other
recommendations as well.

-           (start_rva <= target_rva && target_rva < end_rva))) {
+              (start_rva <= target_rva && target_rva < end_rva))) {
         rel32_locations_.push_back(rel32_rva);
 #if COURGETTE_HISTOGRAM_TARGETS
         ++rel32_target_rvas_[target_rva];
@@ -532,14 +538,14 @@ void DisassemblerWin32X64::ParseRel32RelocsFromSection(const Section* section) {
 }
 
 CheckBool DisassemblerWin32X64::ParseNonSectionFileRegion(
-    uint32_t start_file_offset,
-    uint32_t end_file_offset,
+    FileOffset start_file_offset,
+    FileOffset end_file_offset,
     AssemblyProgram* program) {
   if (incomplete_disassembly_)
     return true;
 
   if (end_file_offset > start_file_offset) {
-    if (!program->EmitBytesInstruction(OffsetToPointer(start_file_offset),
+    if (!program->EmitBytesInstruction(FileOffsetToPointer(start_file_offset),
                                        end_file_offset - start_file_offset)) {
       return false;
     }
@@ -549,13 +555,13 @@ CheckBool DisassemblerWin32X64::ParseNonSectionFileRegion(
 }
 
 CheckBool DisassemblerWin32X64::ParseFileRegion(const Section* section,
-                                                uint32_t start_file_offset,
-                                                uint32_t end_file_offset,
+                                                FileOffset start_file_offset,
+                                                FileOffset end_file_offset,
                                                 AssemblyProgram* program) {
   RVA relocs_start_rva = base_relocation_table().address_;
 
-  const uint8_t* start_pointer = OffsetToPointer(start_file_offset);
-  const uint8_t* end_pointer = OffsetToPointer(end_file_offset);
+  const uint8_t* start_pointer = FileOffsetToPointer(start_file_offset);
+  const uint8_t* end_pointer = FileOffsetToPointer(end_file_offset);
 
   RVA start_rva = FileOffsetToRVA(start_file_offset);
   RVA end_rva = start_rva + section->virtual_size;
@@ -664,7 +670,7 @@ void DisassemblerWin32X64::HistogramTargets(const char* kind,
       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;
@@ -676,7 +682,6 @@ void DisassemblerWin32X64::HistogramTargets(const char* kind,
 }
 #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:
@@ -695,12 +700,12 @@ std::string DisassemblerWin32X64::DescribeRVA(RVA rva) const {
 }
 
 const Section* DisassemblerWin32X64::FindNextSection(
-    uint32_t fileOffset) const {
+    FileOffset file_offset) 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 (fileOffset <= section->file_offset_of_raw_data) {
+      if (file_offset <= section->file_offset_of_raw_data) {
         if (best == 0 ||
             section->file_offset_of_raw_data < best->file_offset_of_raw_data) {
           best = section;
@@ -711,26 +716,15 @@ const Section* DisassemblerWin32X64::FindNextSection(
   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_) {
-    size_t offset = index * 8 + offset_of_data_directories_;
-    if (offset >= size_of_optional_header_)
+    FileOffset file_offset = index * 8 + offset_of_data_directories_;
+    if (file_offset >= size_of_optional_header_)
       return Bad("number of data directories inconsistent");
-    const uint8_t* data_directory = optional_header_ + offset;
+    const uint8_t* data_directory = optional_header_ + file_offset;
     if (data_directory < start() ||
         data_directory + 8 >= end())
       return Bad("data directory outside image");