| OLD | NEW |
| 1 // Copyright (c) 2011 The Chromium Authors. All rights reserved. | 1 // Copyright (c) 2011 The Chromium Authors. All rights reserved. |
| 2 // Use of this source code is governed by a BSD-style license that can be | 2 // Use of this source code is governed by a BSD-style license that can be |
| 3 // found in the LICENSE file. | 3 // found in the LICENSE file. |
| 4 | 4 |
| 5 #include "courgette/disassembler_win32_x86.h" | 5 #include "courgette/disassembler_win32_x86.h" |
| 6 | 6 |
| 7 #include <stddef.h> | 7 #include <stddef.h> |
| 8 #include <stdint.h> | 8 #include <stdint.h> |
| 9 | 9 |
| 10 #include <algorithm> | 10 #include <algorithm> |
| 11 #include <string> | |
| 12 #include <vector> | |
| 13 | 11 |
| 14 #include "base/logging.h" | 12 #include "base/logging.h" |
| 15 | |
| 16 #include "courgette/assembly_program.h" | 13 #include "courgette/assembly_program.h" |
| 17 #include "courgette/courgette.h" | 14 #include "courgette/courgette.h" |
| 18 #include "courgette/encoded_program.h" | |
| 19 #include "courgette/rel32_finder_win32_x86.h" | 15 #include "courgette/rel32_finder_win32_x86.h" |
| 20 | 16 |
| 17 #if COURGETTE_HISTOGRAM_TARGETS |
| 18 #include <iostream> |
| 19 #endif |
| 20 |
| 21 namespace courgette { | 21 namespace courgette { |
| 22 | 22 |
| 23 DisassemblerWin32X86::DisassemblerWin32X86(const void* start, size_t length) | 23 DisassemblerWin32X86::DisassemblerWin32X86(const void* start, size_t length) |
| 24 : Disassembler(start, length), | 24 : Disassembler(start, length), |
| 25 incomplete_disassembly_(false), | 25 incomplete_disassembly_(false), |
| 26 is_PE32_plus_(false), | 26 is_PE32_plus_(false), |
| 27 optional_header_(NULL), | 27 optional_header_(nullptr), |
| 28 size_of_optional_header_(0), | 28 size_of_optional_header_(0), |
| 29 offset_of_data_directories_(0), | 29 offset_of_data_directories_(0), |
| 30 machine_type_(0), | 30 machine_type_(0), |
| 31 number_of_sections_(0), | 31 number_of_sections_(0), |
| 32 sections_(NULL), | 32 sections_(nullptr), |
| 33 has_text_section_(false), | 33 has_text_section_(false), |
| 34 size_of_code_(0), | 34 size_of_code_(0), |
| 35 size_of_initialized_data_(0), | 35 size_of_initialized_data_(0), |
| 36 size_of_uninitialized_data_(0), | 36 size_of_uninitialized_data_(0), |
| 37 base_of_code_(0), | 37 base_of_code_(0), |
| 38 base_of_data_(0), | 38 base_of_data_(0), |
| 39 image_base_(0), | 39 image_base_(0), |
| 40 size_of_image_(0), | 40 size_of_image_(0), |
| 41 number_of_data_directories_(0) { | 41 number_of_data_directories_(0) { |
| 42 } |
| 43 |
| 44 FileOffset DisassemblerWin32X86::RVAToFileOffset(RVA rva) const { |
| 45 const Section* section = RVAToSection(rva); |
| 46 if (section != nullptr) { |
| 47 FileOffset offset_in_section = rva - section->virtual_address; |
| 48 // Need this extra check, since an |rva| may be valid for a section, but is |
| 49 // non-existent in an image (e.g. uninit data). |
| 50 if (offset_in_section >= section->size_of_raw_data) |
| 51 return kNoFileOffset; |
| 52 |
| 53 return static_cast<FileOffset>(section->file_offset_of_raw_data + |
| 54 offset_in_section); |
| 55 } |
| 56 |
| 57 // Small RVA values point into the file header in the loaded image. |
| 58 // RVA 0 is the module load address which Windows uses as the module handle. |
| 59 // RVA 2 sometimes occurs, I'm not sure what it is, but it would map into the |
| 60 // DOS header. |
| 61 if (rva == 0 || rva == 2) |
| 62 return static_cast<FileOffset>(rva); |
| 63 |
| 64 NOTREACHED(); |
| 65 return kNoFileOffset; |
| 66 } |
| 67 |
| 68 RVA DisassemblerWin32X86::FileOffsetToRVA(FileOffset file_offset) const { |
| 69 for (int i = 0; i < number_of_sections_; ++i) { |
| 70 const Section* section = §ions_[i]; |
| 71 if (file_offset >= section->file_offset_of_raw_data) { |
| 72 FileOffset offset_in_section = |
| 73 file_offset - section->file_offset_of_raw_data; |
| 74 if (offset_in_section < section->size_of_raw_data) |
| 75 return static_cast<RVA>(section->virtual_address + offset_in_section); |
| 76 } |
| 77 } |
| 78 |
| 79 NOTREACHED(); |
| 80 return kNoRVA; |
| 42 } | 81 } |
| 43 | 82 |
| 44 // ParseHeader attempts to match up the buffer with the Windows data | 83 // ParseHeader attempts to match up the buffer with the Windows data |
| 45 // structures that exist within a Windows 'Portable Executable' format file. | 84 // structures that exist within a Windows 'Portable Executable' format file. |
| 46 // Returns 'true' if the buffer matches, and 'false' if the data looks | 85 // Returns 'true' if the buffer matches, and 'false' if the data looks |
| 47 // suspicious. Rather than try to 'map' the buffer to the numerous windows | 86 // suspicious. Rather than try to 'map' the buffer to the numerous windows |
| 48 // structures, we extract the information we need into the courgette::PEInfo | 87 // structures, we extract the information we need into the courgette::PEInfo |
| 49 // structure. | 88 // structure. |
| 50 // | 89 // |
| 51 bool DisassemblerWin32X86::ParseHeader() { | 90 bool DisassemblerWin32X86::ParseHeader() { |
| 52 if (length() < kOffsetOfFileAddressOfNewExeHeader + 4 /*size*/) | 91 if (length() < kOffsetOfFileAddressOfNewExeHeader + 4 /*size*/) |
| 53 return Bad("Too small"); | 92 return Bad("Too small"); |
| 54 | 93 |
| 55 // Have 'MZ' magic for a DOS header? | 94 // Have 'MZ' magic for a DOS header? |
| 56 if (start()[0] != 'M' || start()[1] != 'Z') | 95 if (start()[0] != 'M' || start()[1] != 'Z') |
| 57 return Bad("Not MZ"); | 96 return Bad("Not MZ"); |
| 58 | 97 |
| 59 // offset from DOS header to PE header is stored in DOS header. | 98 // offset from DOS header to PE header is stored in DOS header. |
| 60 uint32_t offset = ReadU32(start(), kOffsetOfFileAddressOfNewExeHeader); | 99 FileOffset file_offset = static_cast<FileOffset>( |
| 100 ReadU32(start(), kOffsetOfFileAddressOfNewExeHeader)); |
| 61 | 101 |
| 62 if (offset >= length()) | 102 if (file_offset >= length()) |
| 63 return Bad("Bad offset to PE header"); | 103 return Bad("Bad offset to PE header"); |
| 64 | 104 |
| 65 const uint8_t* const pe_header = OffsetToPointer(offset); | 105 const uint8_t* const pe_header = FileOffsetToPointer(file_offset); |
| 66 const size_t kMinPEHeaderSize = 4 /*signature*/ + kSizeOfCoffHeader; | 106 const size_t kMinPEHeaderSize = 4 /*signature*/ + kSizeOfCoffHeader; |
| 67 if (pe_header <= start() || | 107 if (pe_header <= start() || |
| 68 pe_header >= end() - kMinPEHeaderSize) | 108 pe_header >= end() - kMinPEHeaderSize) |
| 69 return Bad("Bad offset to PE header"); | 109 return Bad("Bad file offset to PE header"); |
| 70 | 110 |
| 71 if (offset % 8 != 0) | 111 if (file_offset % 8 != 0) |
| 72 return Bad("Misaligned PE header"); | 112 return Bad("Misaligned PE header"); |
| 73 | 113 |
| 74 // The 'PE' header is an IMAGE_NT_HEADERS structure as defined in WINNT.H. | 114 // The 'PE' header is an IMAGE_NT_HEADERS structure as defined in WINNT.H. |
| 75 // See http://msdn.microsoft.com/en-us/library/ms680336(VS.85).aspx | 115 // See http://msdn.microsoft.com/en-us/library/ms680336(VS.85).aspx |
| 76 // | 116 // |
| 77 // The first field of the IMAGE_NT_HEADERS is the signature. | 117 // The first field of the IMAGE_NT_HEADERS is the signature. |
| 78 if (!(pe_header[0] == 'P' && | 118 if (!(pe_header[0] == 'P' && |
| 79 pe_header[1] == 'E' && | 119 pe_header[1] == 'E' && |
| 80 pe_header[2] == 0 && | 120 pe_header[2] == 0 && |
| 81 pe_header[3] == 0)) | 121 pe_header[3] == 0)) |
| (...skipping 80 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
| 162 if (!b) { | 202 if (!b) { |
| 163 return Bad("malformed data directory"); | 203 return Bad("malformed data directory"); |
| 164 } | 204 } |
| 165 | 205 |
| 166 // Sections follow the optional header. | 206 // Sections follow the optional header. |
| 167 sections_ = | 207 sections_ = |
| 168 reinterpret_cast<const Section*>(optional_header + | 208 reinterpret_cast<const Section*>(optional_header + |
| 169 size_of_optional_header_); | 209 size_of_optional_header_); |
| 170 size_t detected_length = 0; | 210 size_t detected_length = 0; |
| 171 | 211 |
| 172 for (int i = 0; i < number_of_sections_; ++i) { | 212 for (int i = 0; i < number_of_sections_; ++i) { |
| 173 const Section* section = §ions_[i]; | 213 const Section* section = §ions_[i]; |
| 174 | 214 |
| 175 // TODO(sra): consider using the 'characteristics' field of the section | 215 // TODO(sra): consider using the 'characteristics' field of the section |
| 176 // header to see if the section contains instructions. | 216 // header to see if the section contains instructions. |
| 177 if (memcmp(section->name, ".text", 6) == 0) | 217 if (memcmp(section->name, ".text", 6) == 0) |
| 178 has_text_section_ = true; | 218 has_text_section_ = true; |
| 179 | 219 |
| 180 uint32_t section_end = | 220 uint32_t section_end = |
| 181 section->file_offset_of_raw_data + section->size_of_raw_data; | 221 section->file_offset_of_raw_data + section->size_of_raw_data; |
| 182 if (section_end > detected_length) | 222 if (section_end > detected_length) |
| (...skipping 103 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
| 286 block += size; | 326 block += size; |
| 287 } | 327 } |
| 288 | 328 |
| 289 std::sort(relocs->begin(), relocs->end()); | 329 std::sort(relocs->begin(), relocs->end()); |
| 290 DCHECK(relocs->empty() || relocs->back() != kUnassignedRVA); | 330 DCHECK(relocs->empty() || relocs->back() != kUnassignedRVA); |
| 291 | 331 |
| 292 return true; | 332 return true; |
| 293 } | 333 } |
| 294 | 334 |
| 295 const Section* DisassemblerWin32X86::RVAToSection(RVA rva) const { | 335 const Section* DisassemblerWin32X86::RVAToSection(RVA rva) const { |
| 296 for (int i = 0; i < number_of_sections_; i++) { | 336 for (int i = 0; i < number_of_sections_; ++i) { |
| 297 const Section* section = §ions_[i]; | 337 const Section* section = §ions_[i]; |
| 298 uint32_t offset = rva - section->virtual_address; | 338 if (rva >= section->virtual_address) { |
| 299 if (offset < section->virtual_size) { | 339 FileOffset offset_in_section = rva - section->virtual_address; |
| 300 return section; | 340 if (offset_in_section < section->virtual_size) |
| 341 return section; |
| 301 } | 342 } |
| 302 } | 343 } |
| 303 return NULL; | 344 return nullptr; |
| 304 } | |
| 305 | |
| 306 int DisassemblerWin32X86::RVAToFileOffset(RVA rva) const { | |
| 307 const Section* section = RVAToSection(rva); | |
| 308 if (section) { | |
| 309 uint32_t offset = rva - section->virtual_address; | |
| 310 if (offset < section->size_of_raw_data) { | |
| 311 return section->file_offset_of_raw_data + offset; | |
| 312 } else { | |
| 313 return kNoOffset; // In section but not in file (e.g. uninit data). | |
| 314 } | |
| 315 } | |
| 316 | |
| 317 // Small RVA values point into the file header in the loaded image. | |
| 318 // RVA 0 is the module load address which Windows uses as the module handle. | |
| 319 // RVA 2 sometimes occurs, I'm not sure what it is, but it would map into the | |
| 320 // DOS header. | |
| 321 if (rva == 0 || rva == 2) | |
| 322 return rva; | |
| 323 | |
| 324 NOTREACHED(); | |
| 325 return kNoOffset; | |
| 326 } | |
| 327 | |
| 328 const uint8_t* DisassemblerWin32X86::RVAToPointer(RVA rva) const { | |
| 329 int file_offset = RVAToFileOffset(rva); | |
| 330 if (file_offset == kNoOffset) | |
| 331 return NULL; | |
| 332 else | |
| 333 return OffsetToPointer(file_offset); | |
| 334 } | 345 } |
| 335 | 346 |
| 336 std::string DisassemblerWin32X86::SectionName(const Section* section) { | 347 std::string DisassemblerWin32X86::SectionName(const Section* section) { |
| 337 if (section == NULL) | 348 if (section == nullptr) |
| 338 return "<none>"; | 349 return "<none>"; |
| 339 char name[9]; | 350 char name[9]; |
| 340 memcpy(name, section->name, 8); | 351 memcpy(name, section->name, 8); |
| 341 name[8] = '\0'; // Ensure termination. | 352 name[8] = '\0'; // Ensure termination. |
| 342 return name; | 353 return name; |
| 343 } | 354 } |
| 344 | 355 |
| 345 CheckBool DisassemblerWin32X86::ParseFile(AssemblyProgram* program) { | 356 CheckBool DisassemblerWin32X86::ParseFile(AssemblyProgram* program) { |
| 346 // Walk all the bytes in the file, whether or not in a section. | 357 // Walk all the bytes in the file, whether or not in a section. |
| 347 uint32_t file_offset = 0; | 358 FileOffset file_offset = 0; |
| 348 while (file_offset < length()) { | 359 while (file_offset < length()) { |
| 349 const Section* section = FindNextSection(file_offset); | 360 const Section* section = FindNextSection(file_offset); |
| 350 if (section == NULL) { | 361 if (section == nullptr) { |
| 351 // No more sections. There should not be extra stuff following last | 362 // No more sections. There should not be extra stuff following last |
| 352 // section. | 363 // section. |
| 353 // ParseNonSectionFileRegion(file_offset, pe_info().length(), program); | 364 // ParseNonSectionFileRegion(file_offset, pe_info().length(), program); |
| 354 break; | 365 break; |
| 355 } | 366 } |
| 356 if (file_offset < section->file_offset_of_raw_data) { | 367 if (file_offset < section->file_offset_of_raw_data) { |
| 357 uint32_t section_start_offset = section->file_offset_of_raw_data; | 368 FileOffset section_start_offset = section->file_offset_of_raw_data; |
| 358 if(!ParseNonSectionFileRegion(file_offset, section_start_offset, | 369 if (!ParseNonSectionFileRegion(file_offset, section_start_offset, |
| 359 program)) | 370 program)) { |
| 360 return false; | 371 return false; |
| 372 } |
| 361 | 373 |
| 362 file_offset = section_start_offset; | 374 file_offset = section_start_offset; |
| 363 } | 375 } |
| 364 uint32_t end = file_offset + section->size_of_raw_data; | 376 FileOffset end = file_offset + section->size_of_raw_data; |
| 365 if (!ParseFileRegion(section, file_offset, end, program)) | 377 if (!ParseFileRegion(section, file_offset, end, program)) |
| 366 return false; | 378 return false; |
| 367 file_offset = end; | 379 file_offset = end; |
| 368 } | 380 } |
| 369 | 381 |
| 370 #if COURGETTE_HISTOGRAM_TARGETS | 382 #if COURGETTE_HISTOGRAM_TARGETS |
| 371 HistogramTargets("abs32 relocs", abs32_target_rvas_); | 383 HistogramTargets("abs32 relocs", abs32_target_rvas_); |
| 372 HistogramTargets("rel32 relocs", rel32_target_rvas_); | 384 HistogramTargets("rel32 relocs", rel32_target_rvas_); |
| 373 #endif | 385 #endif |
| 374 | 386 |
| 375 return true; | 387 return true; |
| 376 } | 388 } |
| 377 | 389 |
| 378 bool DisassemblerWin32X86::ParseAbs32Relocs() { | 390 bool DisassemblerWin32X86::ParseAbs32Relocs() { |
| 379 abs32_locations_.clear(); | 391 abs32_locations_.clear(); |
| 380 if (!ParseRelocs(&abs32_locations_)) | 392 if (!ParseRelocs(&abs32_locations_)) |
| 381 return false; | 393 return false; |
| 382 | 394 |
| 383 #if COURGETTE_HISTOGRAM_TARGETS | 395 #if COURGETTE_HISTOGRAM_TARGETS |
| 384 for (size_t i = 0; i < abs32_locations_.size(); ++i) { | 396 for (size_t i = 0; i < abs32_locations_.size(); ++i) { |
| 385 RVA rva = abs32_locations_[i]; | 397 RVA rva = abs32_locations_[i]; |
| 386 // The 4 bytes at the relocation are a reference to some address. | 398 // The 4 bytes at the relocation are a reference to some address. |
| 387 uint32_t target_address = Read32LittleEndian(RVAToPointer(rva)); | 399 uint32_t target_address = Read32LittleEndian(RVAToPointer(rva)); |
| 388 ++abs32_target_rvas_[target_address - image_base()]; | 400 ++abs32_target_rvas_[target_address - image_base()]; |
| 389 } | 401 } |
| 390 #endif | 402 #endif |
| 391 return true; | 403 return true; |
| 392 } | 404 } |
| 393 | 405 |
| 394 void DisassemblerWin32X86::ParseRel32RelocsFromSections() { | 406 void DisassemblerWin32X86::ParseRel32RelocsFromSections() { |
| 395 uint32_t file_offset = 0; | 407 FileOffset file_offset = 0; |
| 396 while (file_offset < length()) { | 408 while (file_offset < length()) { |
| 397 const Section* section = FindNextSection(file_offset); | 409 const Section* section = FindNextSection(file_offset); |
| 398 if (section == NULL) | 410 if (section == nullptr) |
| 399 break; | 411 break; |
| 400 if (file_offset < section->file_offset_of_raw_data) | 412 if (file_offset < section->file_offset_of_raw_data) |
| 401 file_offset = section->file_offset_of_raw_data; | 413 file_offset = section->file_offset_of_raw_data; |
| 402 ParseRel32RelocsFromSection(section); | 414 ParseRel32RelocsFromSection(section); |
| 403 file_offset += section->size_of_raw_data; | 415 file_offset += section->size_of_raw_data; |
| 404 } | 416 } |
| 405 std::sort(rel32_locations_.begin(), rel32_locations_.end()); | 417 std::sort(rel32_locations_.begin(), rel32_locations_.end()); |
| 406 DCHECK(rel32_locations_.empty() || | 418 DCHECK(rel32_locations_.empty() || |
| 407 rel32_locations_.back() != kUnassignedRVA); | 419 rel32_locations_.back() != kUnassignedRVA); |
| 408 | 420 |
| 409 #if COURGETTE_HISTOGRAM_TARGETS | 421 #if COURGETTE_HISTOGRAM_TARGETS |
| 410 VLOG(1) << "abs32_locations_ " << abs32_locations_.size() | 422 VLOG(1) << "abs32_locations_ " << abs32_locations_.size() |
| 411 << "\nrel32_locations_ " << rel32_locations_.size() | 423 << "\nrel32_locations_ " << rel32_locations_.size() |
| 412 << "\nabs32_target_rvas_ " << abs32_target_rvas_.size() | 424 << "\nabs32_target_rvas_ " << abs32_target_rvas_.size() |
| 413 << "\nrel32_target_rvas_ " << rel32_target_rvas_.size(); | 425 << "\nrel32_target_rvas_ " << rel32_target_rvas_.size(); |
| 414 | 426 |
| 415 int common = 0; | 427 int common = 0; |
| 416 std::map<RVA, int>::iterator abs32_iter = abs32_target_rvas_.begin(); | 428 std::map<RVA, int>::iterator abs32_iter = abs32_target_rvas_.begin(); |
| 417 std::map<RVA, int>::iterator rel32_iter = rel32_target_rvas_.begin(); | 429 std::map<RVA, int>::iterator rel32_iter = rel32_target_rvas_.begin(); |
| 418 while (abs32_iter != abs32_target_rvas_.end() && | 430 while (abs32_iter != abs32_target_rvas_.end() && |
| 419 rel32_iter != rel32_target_rvas_.end()) { | 431 rel32_iter != rel32_target_rvas_.end()) { |
| 420 if (abs32_iter->first < rel32_iter->first) | 432 if (abs32_iter->first < rel32_iter->first) { |
| 421 ++abs32_iter; | 433 ++abs32_iter; |
| 422 else if (rel32_iter->first < abs32_iter->first) | 434 } else if (rel32_iter->first < abs32_iter->first) { |
| 423 ++rel32_iter; | 435 ++rel32_iter; |
| 424 else { | 436 } else { |
| 425 ++common; | 437 ++common; |
| 426 ++abs32_iter; | 438 ++abs32_iter; |
| 427 ++rel32_iter; | 439 ++rel32_iter; |
| 428 } | 440 } |
| 429 } | 441 } |
| 430 VLOG(1) << "common " << common; | 442 VLOG(1) << "common " << common; |
| 431 #endif | 443 #endif |
| 432 } | 444 } |
| 433 | 445 |
| 434 void DisassemblerWin32X86::ParseRel32RelocsFromSection(const Section* section) { | 446 void DisassemblerWin32X86::ParseRel32RelocsFromSection(const Section* section) { |
| 435 // TODO(sra): use characteristic. | 447 // TODO(sra): use characteristic. |
| 436 bool isCode = strcmp(section->name, ".text") == 0; | 448 bool isCode = strcmp(section->name, ".text") == 0; |
| 437 if (!isCode) | 449 if (!isCode) |
| 438 return; | 450 return; |
| 439 | 451 |
| 440 uint32_t start_file_offset = section->file_offset_of_raw_data; | 452 FileOffset start_file_offset = section->file_offset_of_raw_data; |
| 441 uint32_t end_file_offset = start_file_offset + section->size_of_raw_data; | 453 FileOffset end_file_offset = start_file_offset + section->size_of_raw_data; |
| 442 | 454 |
| 443 const uint8_t* start_pointer = OffsetToPointer(start_file_offset); | 455 const uint8_t* start_pointer = FileOffsetToPointer(start_file_offset); |
| 444 const uint8_t* end_pointer = OffsetToPointer(end_file_offset); | 456 const uint8_t* end_pointer = FileOffsetToPointer(end_file_offset); |
| 445 | 457 |
| 446 RVA start_rva = FileOffsetToRVA(start_file_offset); | 458 RVA start_rva = FileOffsetToRVA(start_file_offset); |
| 447 RVA end_rva = start_rva + section->virtual_size; | 459 RVA end_rva = start_rva + section->virtual_size; |
| 448 | 460 |
| 449 Rel32FinderWin32X86_Basic finder( | 461 Rel32FinderWin32X86_Basic finder( |
| 450 base_relocation_table().address_, | 462 base_relocation_table().address_, |
| 451 base_relocation_table().address_ + base_relocation_table().size_, | 463 base_relocation_table().address_ + base_relocation_table().size_); |
| 452 size_of_image_); | |
| 453 finder.Find(start_pointer, end_pointer, start_rva, end_rva, abs32_locations_); | 464 finder.Find(start_pointer, end_pointer, start_rva, end_rva, abs32_locations_); |
| 454 finder.SwapRel32Locations(&rel32_locations_); | 465 finder.SwapRel32Locations(&rel32_locations_); |
| 455 | 466 |
| 456 #if COURGETTE_HISTOGRAM_TARGETS | 467 #if COURGETTE_HISTOGRAM_TARGETS |
| 457 DCHECK(rel32_target_rvas_.empty()); | 468 DCHECK(rel32_target_rvas_.empty()); |
| 458 finder.SwapRel32TargetRVAs(&rel32_target_rvas_); | 469 finder.SwapRel32TargetRVAs(&rel32_target_rvas_); |
| 459 #endif | 470 #endif |
| 460 } | 471 } |
| 461 | 472 |
| 462 CheckBool DisassemblerWin32X86::ParseNonSectionFileRegion( | 473 CheckBool DisassemblerWin32X86::ParseNonSectionFileRegion( |
| 463 uint32_t start_file_offset, | 474 FileOffset start_file_offset, |
| 464 uint32_t end_file_offset, | 475 FileOffset end_file_offset, |
| 465 AssemblyProgram* program) { | 476 AssemblyProgram* program) { |
| 466 if (incomplete_disassembly_) | 477 if (incomplete_disassembly_) |
| 467 return true; | 478 return true; |
| 468 | 479 |
| 469 if (end_file_offset > start_file_offset) { | 480 if (end_file_offset > start_file_offset) { |
| 470 if (!program->EmitBytesInstruction(OffsetToPointer(start_file_offset), | 481 if (!program->EmitBytesInstruction(FileOffsetToPointer(start_file_offset), |
| 471 end_file_offset - start_file_offset)) { | 482 end_file_offset - start_file_offset)) { |
| 472 return false; | 483 return false; |
| 473 } | 484 } |
| 474 } | 485 } |
| 475 | 486 |
| 476 return true; | 487 return true; |
| 477 } | 488 } |
| 478 | 489 |
| 479 CheckBool DisassemblerWin32X86::ParseFileRegion(const Section* section, | 490 CheckBool DisassemblerWin32X86::ParseFileRegion(const Section* section, |
| 480 uint32_t start_file_offset, | 491 FileOffset start_file_offset, |
| 481 uint32_t end_file_offset, | 492 FileOffset end_file_offset, |
| 482 AssemblyProgram* program) { | 493 AssemblyProgram* program) { |
| 483 RVA relocs_start_rva = base_relocation_table().address_; | 494 RVA relocs_start_rva = base_relocation_table().address_; |
| 484 | 495 |
| 485 const uint8_t* start_pointer = OffsetToPointer(start_file_offset); | 496 const uint8_t* start_pointer = FileOffsetToPointer(start_file_offset); |
| 486 const uint8_t* end_pointer = OffsetToPointer(end_file_offset); | 497 const uint8_t* end_pointer = FileOffsetToPointer(end_file_offset); |
| 487 | 498 |
| 488 RVA start_rva = FileOffsetToRVA(start_file_offset); | 499 RVA start_rva = FileOffsetToRVA(start_file_offset); |
| 489 RVA end_rva = start_rva + section->virtual_size; | 500 RVA end_rva = start_rva + section->virtual_size; |
| 490 | 501 |
| 491 // Quick way to convert from Pointer to RVA within a single Section is to | 502 // Quick way to convert from Pointer to RVA within a single Section is to |
| 492 // subtract 'pointer_to_rva'. | 503 // subtract 'pointer_to_rva'. |
| 493 const uint8_t* const adjust_pointer_to_rva = start_pointer - start_rva; | 504 const uint8_t* const adjust_pointer_to_rva = start_pointer - start_rva; |
| 494 | 505 |
| 495 std::vector<RVA>::iterator rel32_pos = rel32_locations_.begin(); | 506 std::vector<RVA>::iterator rel32_pos = rel32_locations_.begin(); |
| 496 std::vector<RVA>::iterator abs32_pos = abs32_locations_.begin(); | 507 std::vector<RVA>::iterator abs32_pos = abs32_locations_.begin(); |
| (...skipping 88 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
| 585 p != h.rend(); | 596 p != h.rend(); |
| 586 ++p) { | 597 ++p) { |
| 587 ++index; | 598 ++index; |
| 588 if (index <= kFirstN || p->first <= 3) { | 599 if (index <= kFirstN || p->first <= 3) { |
| 589 if (someSkipped) { | 600 if (someSkipped) { |
| 590 std::cout << "..." << std::endl; | 601 std::cout << "..." << std::endl; |
| 591 } | 602 } |
| 592 size_t count = p->second.size(); | 603 size_t count = p->second.size(); |
| 593 std::cout << std::dec << p->first << ": " << count; | 604 std::cout << std::dec << p->first << ": " << count; |
| 594 if (count <= 2) { | 605 if (count <= 2) { |
| 595 for (size_t i = 0; i < count; ++i) | 606 for (size_t i = 0; i < count; ++i) |
| 596 std::cout << " " << DescribeRVA(p->second[i]); | 607 std::cout << " " << DescribeRVA(p->second[i]); |
| 597 } | 608 } |
| 598 std::cout << std::endl; | 609 std::cout << std::endl; |
| 599 someSkipped = false; | 610 someSkipped = false; |
| 600 } else { | 611 } else { |
| 601 someSkipped = true; | 612 someSkipped = true; |
| 602 } | 613 } |
| 603 } | 614 } |
| 604 } | 615 } |
| 605 #endif // COURGETTE_HISTOGRAM_TARGETS | 616 #endif // COURGETTE_HISTOGRAM_TARGETS |
| 606 | 617 |
| 607 | |
| 608 // DescribeRVA is for debugging only. I would put it under #ifdef DEBUG except | 618 // DescribeRVA is for debugging only. I would put it under #ifdef DEBUG except |
| 609 // that during development I'm finding I need to call it when compiled in | 619 // that during development I'm finding I need to call it when compiled in |
| 610 // Release mode. Hence: | 620 // Release mode. Hence: |
| 611 // TODO(sra): make this compile only for debug mode. | 621 // TODO(sra): make this compile only for debug mode. |
| 612 std::string DisassemblerWin32X86::DescribeRVA(RVA rva) const { | 622 std::string DisassemblerWin32X86::DescribeRVA(RVA rva) const { |
| 613 const Section* section = RVAToSection(rva); | 623 const Section* section = RVAToSection(rva); |
| 614 std::ostringstream s; | 624 std::ostringstream s; |
| 615 s << std::hex << rva; | 625 s << std::hex << rva; |
| 616 if (section) { | 626 if (section) { |
| 617 s << " ("; | 627 s << " ("; |
| 618 s << SectionName(section) << "+" | 628 s << SectionName(section) << "+" |
| 619 << std::hex << (rva - section->virtual_address) | 629 << std::hex << (rva - section->virtual_address) |
| 620 << ")"; | 630 << ")"; |
| 621 } | 631 } |
| 622 return s.str(); | 632 return s.str(); |
| 623 } | 633 } |
| 624 | 634 |
| 625 const Section* DisassemblerWin32X86::FindNextSection( | 635 const Section* DisassemblerWin32X86::FindNextSection( |
| 626 uint32_t fileOffset) const { | 636 FileOffset file_offset) const { |
| 627 const Section* best = 0; | 637 const Section* best = 0; |
| 628 for (int i = 0; i < number_of_sections_; i++) { | 638 for (int i = 0; i < number_of_sections_; ++i) { |
| 629 const Section* section = §ions_[i]; | 639 const Section* section = §ions_[i]; |
| 630 if (section->size_of_raw_data > 0) { // i.e. has data in file. | 640 if (section->size_of_raw_data > 0) { // i.e. has data in file. |
| 631 if (fileOffset <= section->file_offset_of_raw_data) { | 641 if (file_offset <= section->file_offset_of_raw_data) { |
| 632 if (best == 0 || | 642 if (best == 0 || |
| 633 section->file_offset_of_raw_data < best->file_offset_of_raw_data) { | 643 section->file_offset_of_raw_data < best->file_offset_of_raw_data) { |
| 634 best = section; | 644 best = section; |
| 635 } | 645 } |
| 636 } | 646 } |
| 637 } | 647 } |
| 638 } | 648 } |
| 639 return best; | 649 return best; |
| 640 } | 650 } |
| 641 | 651 |
| 642 RVA DisassemblerWin32X86::FileOffsetToRVA(uint32_t file_offset) const { | |
| 643 for (int i = 0; i < number_of_sections_; i++) { | |
| 644 const Section* section = §ions_[i]; | |
| 645 uint32_t offset = file_offset - section->file_offset_of_raw_data; | |
| 646 if (offset < section->size_of_raw_data) { | |
| 647 return section->virtual_address + offset; | |
| 648 } | |
| 649 } | |
| 650 return 0; | |
| 651 } | |
| 652 | |
| 653 bool DisassemblerWin32X86::ReadDataDirectory( | 652 bool DisassemblerWin32X86::ReadDataDirectory( |
| 654 int index, | 653 int index, |
| 655 ImageDataDirectory* directory) { | 654 ImageDataDirectory* directory) { |
| 656 | 655 |
| 657 if (index < number_of_data_directories_) { | 656 if (index < number_of_data_directories_) { |
| 658 size_t offset = index * 8 + offset_of_data_directories_; | 657 FileOffset file_offset = index * 8 + offset_of_data_directories_; |
| 659 if (offset >= size_of_optional_header_) | 658 if (file_offset >= size_of_optional_header_) |
| 660 return Bad("number of data directories inconsistent"); | 659 return Bad("number of data directories inconsistent"); |
| 661 const uint8_t* data_directory = optional_header_ + offset; | 660 const uint8_t* data_directory = optional_header_ + file_offset; |
| 662 if (data_directory < start() || | 661 if (data_directory < start() || |
| 663 data_directory + 8 >= end()) | 662 data_directory + 8 >= end()) |
| 664 return Bad("data directory outside image"); | 663 return Bad("data directory outside image"); |
| 665 RVA rva = ReadU32(data_directory, 0); | 664 RVA rva = ReadU32(data_directory, 0); |
| 666 size_t size = ReadU32(data_directory, 4); | 665 size_t size = ReadU32(data_directory, 4); |
| 667 if (size > size_of_image_) | 666 if (size > size_of_image_) |
| 668 return Bad("data directory size too big"); | 667 return Bad("data directory size too big"); |
| 669 | 668 |
| 670 // TODO(sra): validate RVA. | 669 // TODO(sra): validate RVA. |
| 671 directory->address_ = rva; | 670 directory->address_ = rva; |
| 672 directory->size_ = static_cast<uint32_t>(size); | 671 directory->size_ = static_cast<uint32_t>(size); |
| 673 return true; | 672 return true; |
| 674 } else { | 673 } else { |
| 675 directory->address_ = 0; | 674 directory->address_ = 0; |
| 676 directory->size_ = 0; | 675 directory->size_ = 0; |
| 677 return true; | 676 return true; |
| 678 } | 677 } |
| 679 } | 678 } |
| 680 | 679 |
| 681 } // namespace courgette | 680 } // namespace courgette |
| OLD | NEW |