| // Copyright 2015 the V8 project authors. All rights reserved. |
| // Use of this source code is governed by a BSD-style license that can be |
| // found in the LICENSE file. |
| |
| #ifndef V8_WASM_DECODER_H_ |
| #define V8_WASM_DECODER_H_ |
| |
| #include <cinttypes> |
| #include <cstdarg> |
| #include <memory> |
| |
| #include "src/base/compiler-specific.h" |
| #include "src/base/memory.h" |
| #include "src/codegen/signature.h" |
| #include "src/flags/flags.h" |
| #include "src/utils/vector.h" |
| #include "src/wasm/wasm-opcodes.h" |
| #include "src/wasm/wasm-result.h" |
| #include "src/zone/zone-containers.h" |
| |
| namespace v8 { |
| namespace internal { |
| namespace wasm { |
| |
| #define TRACE(...) \ |
| do { \ |
| if (FLAG_trace_wasm_decoder) PrintF(__VA_ARGS__); \ |
| } while (false) |
| #define TRACE_IF(cond, ...) \ |
| do { \ |
| if (FLAG_trace_wasm_decoder && (cond)) PrintF(__VA_ARGS__); \ |
| } while (false) |
| |
| // A {DecodeResult} only stores the failure / success status, but no data. |
| using DecodeResult = VoidResult; |
| |
| // A helper utility to decode bytes, integers, fields, varints, etc, from |
| // a buffer of bytes. |
| class Decoder { |
| public: |
| // {ValidateFlag} can be used in a boolean manner ({if (!validate) ...}). |
| enum ValidateFlag : int8_t { |
| kNoValidation = 0, // Don't run validation, assume valid input. |
| kBooleanValidation, // Run validation but only store a generic error. |
| kFullValidation // Run full validation with error message and location. |
| }; |
| |
| enum TraceFlag : bool { kTrace = true, kNoTrace = false }; |
| |
| Decoder(const byte* start, const byte* end, uint32_t buffer_offset = 0) |
| : Decoder(start, start, end, buffer_offset) {} |
| explicit Decoder(const Vector<const byte> bytes, uint32_t buffer_offset = 0) |
| : Decoder(bytes.begin(), bytes.begin() + bytes.length(), buffer_offset) {} |
| Decoder(const byte* start, const byte* pc, const byte* end, |
| uint32_t buffer_offset = 0) |
| : start_(start), pc_(pc), end_(end), buffer_offset_(buffer_offset) { |
| DCHECK_LE(start, pc); |
| DCHECK_LE(pc, end); |
| DCHECK_EQ(static_cast<uint32_t>(end - start), end - start); |
| } |
| |
| virtual ~Decoder() = default; |
| |
| bool validate_size(const byte* pc, uint32_t length, const char* msg) { |
| DCHECK_LE(start_, pc); |
| if (V8_UNLIKELY(pc > end_ || length > static_cast<uint32_t>(end_ - pc))) { |
| error(pc, msg); |
| return false; |
| } |
| return true; |
| } |
| |
| // Reads an 8-bit unsigned integer. |
| template <ValidateFlag validate> |
| uint8_t read_u8(const byte* pc, const char* msg = "expected 1 byte") { |
| return read_little_endian<uint8_t, validate>(pc, msg); |
| } |
| |
| // Reads a 16-bit unsigned integer (little endian). |
| template <ValidateFlag validate> |
| uint16_t read_u16(const byte* pc, const char* msg = "expected 2 bytes") { |
| return read_little_endian<uint16_t, validate>(pc, msg); |
| } |
| |
| // Reads a 32-bit unsigned integer (little endian). |
| template <ValidateFlag validate> |
| uint32_t read_u32(const byte* pc, const char* msg = "expected 4 bytes") { |
| return read_little_endian<uint32_t, validate>(pc, msg); |
| } |
| |
| // Reads a 64-bit unsigned integer (little endian). |
| template <ValidateFlag validate> |
| uint64_t read_u64(const byte* pc, const char* msg = "expected 8 bytes") { |
| return read_little_endian<uint64_t, validate>(pc, msg); |
| } |
| |
| // Reads a variable-length unsigned integer (little endian). |
| template <ValidateFlag validate> |
| uint32_t read_u32v(const byte* pc, uint32_t* length, |
| const char* name = "LEB32") { |
| return read_leb<uint32_t, validate, kNoTrace>(pc, length, name); |
| } |
| |
| // Reads a variable-length signed integer (little endian). |
| template <ValidateFlag validate> |
| int32_t read_i32v(const byte* pc, uint32_t* length, |
| const char* name = "signed LEB32") { |
| return read_leb<int32_t, validate, kNoTrace>(pc, length, name); |
| } |
| |
| // Reads a variable-length unsigned integer (little endian). |
| template <ValidateFlag validate> |
| uint64_t read_u64v(const byte* pc, uint32_t* length, |
| const char* name = "LEB64") { |
| return read_leb<uint64_t, validate, kNoTrace>(pc, length, name); |
| } |
| |
| // Reads a variable-length signed integer (little endian). |
| template <ValidateFlag validate> |
| int64_t read_i64v(const byte* pc, uint32_t* length, |
| const char* name = "signed LEB64") { |
| return read_leb<int64_t, validate, kNoTrace>(pc, length, name); |
| } |
| |
| // Reads a variable-length 33-bit signed integer (little endian). |
| template <ValidateFlag validate> |
| int64_t read_i33v(const byte* pc, uint32_t* length, |
| const char* name = "signed LEB33") { |
| return read_leb<int64_t, validate, kNoTrace, 33>(pc, length, name); |
| } |
| |
| // Convenient overload for callers who don't care about length. |
| template <ValidateFlag validate> |
| WasmOpcode read_prefixed_opcode(const byte* pc) { |
| uint32_t len; |
| return read_prefixed_opcode<validate>(pc, &len); |
| } |
| |
| // Reads a prefixed-opcode, possibly with variable-length index. |
| // `length` is set to the number of bytes that make up this opcode, |
| // *including* the prefix byte. For most opcodes, it will be 2. |
| template <ValidateFlag validate> |
| WasmOpcode read_prefixed_opcode(const byte* pc, uint32_t* length, |
| const char* name = "prefixed opcode") { |
| uint32_t index; |
| |
| // Prefixed opcodes all use LEB128 encoding. |
| index = read_u32v<validate>(pc + 1, length, "prefixed opcode index"); |
| *length += 1; // Prefix byte. |
| // Only support opcodes that go up to 0xFF (when decoded). Anything |
| // bigger will need 1 more byte, and the '<< 8' below will be wrong. |
| if (validate && V8_UNLIKELY(index > 0xff)) { |
| errorf(pc, "Invalid prefixed opcode %d", index); |
| // If size validation fails. |
| index = 0; |
| *length = 0; |
| } |
| |
| return static_cast<WasmOpcode>((*pc) << 8 | index); |
| } |
| |
| // Reads a 8-bit unsigned integer (byte) and advances {pc_}. |
| uint8_t consume_u8(const char* name = "uint8_t") { |
| return consume_little_endian<uint8_t>(name); |
| } |
| |
| // Reads a 16-bit unsigned integer (little endian) and advances {pc_}. |
| uint16_t consume_u16(const char* name = "uint16_t") { |
| return consume_little_endian<uint16_t>(name); |
| } |
| |
| // Reads a single 32-bit unsigned integer (little endian) and advances {pc_}. |
| uint32_t consume_u32(const char* name = "uint32_t") { |
| return consume_little_endian<uint32_t>(name); |
| } |
| |
| // Reads a LEB128 variable-length unsigned 32-bit integer and advances {pc_}. |
| uint32_t consume_u32v(const char* name = nullptr) { |
| uint32_t length = 0; |
| uint32_t result = |
| read_leb<uint32_t, kFullValidation, kTrace>(pc_, &length, name); |
| pc_ += length; |
| return result; |
| } |
| |
| // Reads a LEB128 variable-length signed 32-bit integer and advances {pc_}. |
| int32_t consume_i32v(const char* name = nullptr) { |
| uint32_t length = 0; |
| int32_t result = |
| read_leb<int32_t, kFullValidation, kTrace>(pc_, &length, name); |
| pc_ += length; |
| return result; |
| } |
| |
| // Reads a LEB128 variable-length unsigned 64-bit integer and advances {pc_}. |
| uint64_t consume_u64v(const char* name = nullptr) { |
| uint32_t length = 0; |
| uint64_t result = |
| read_leb<uint64_t, kFullValidation, kTrace>(pc_, &length, name); |
| pc_ += length; |
| return result; |
| } |
| |
| // Consume {size} bytes and send them to the bit bucket, advancing {pc_}. |
| void consume_bytes(uint32_t size, const char* name = "skip") { |
| // Only trace if the name is not null. |
| TRACE_IF(name, " +%u %-20s: %u bytes\n", pc_offset(), name, size); |
| if (checkAvailable(size)) { |
| pc_ += size; |
| } else { |
| pc_ = end_; |
| } |
| } |
| |
| // Check that at least {size} bytes exist between {pc_} and {end_}. |
| bool checkAvailable(uint32_t size) { |
| DCHECK_LE(pc_, end_); |
| if (V8_UNLIKELY(size > static_cast<uint32_t>(end_ - pc_))) { |
| errorf(pc_, "expected %u bytes, fell off end", size); |
| return false; |
| } |
| return true; |
| } |
| |
| // Use this for "boolean validation", i.e. if the error message is not used |
| // anyway. |
| void V8_NOINLINE MarkError() { |
| if (!ok()) return; |
| error_ = {0, "validation failed"}; |
| onFirstError(); |
| } |
| |
| // Do not inline error methods. This has measurable impact on validation time, |
| // see https://crbug.com/910432. |
| void V8_NOINLINE error(const char* msg) { errorf(pc_offset(), "%s", msg); } |
| void V8_NOINLINE error(const uint8_t* pc, const char* msg) { |
| errorf(pc_offset(pc), "%s", msg); |
| } |
| void V8_NOINLINE error(uint32_t offset, const char* msg) { |
| errorf(offset, "%s", msg); |
| } |
| |
| void V8_NOINLINE PRINTF_FORMAT(2, 3) errorf(const char* format, ...) { |
| va_list args; |
| va_start(args, format); |
| verrorf(pc_offset(), format, args); |
| va_end(args); |
| } |
| |
| void V8_NOINLINE PRINTF_FORMAT(3, 4) |
| errorf(uint32_t offset, const char* format, ...) { |
| va_list args; |
| va_start(args, format); |
| verrorf(offset, format, args); |
| va_end(args); |
| } |
| |
| void V8_NOINLINE PRINTF_FORMAT(3, 4) |
| errorf(const uint8_t* pc, const char* format, ...) { |
| va_list args; |
| va_start(args, format); |
| verrorf(pc_offset(pc), format, args); |
| va_end(args); |
| } |
| |
| // Behavior triggered on first error, overridden in subclasses. |
| virtual void onFirstError() {} |
| |
| // Debugging helper to print a bytes range as hex bytes. |
| void traceByteRange(const byte* start, const byte* end) { |
| DCHECK_LE(start, end); |
| for (const byte* p = start; p < end; ++p) TRACE("%02x ", *p); |
| } |
| |
| // Debugging helper to print bytes up to the end. |
| void traceOffEnd() { |
| traceByteRange(pc_, end_); |
| TRACE("<end>\n"); |
| } |
| |
| // Converts the given value to a {Result}, copying the error if necessary. |
| template <typename T, typename U = typename std::remove_reference<T>::type> |
| Result<U> toResult(T&& val) { |
| if (failed()) { |
| TRACE("Result error: %s\n", error_.message().c_str()); |
| return Result<U>{error_}; |
| } |
| return Result<U>{std::forward<T>(val)}; |
| } |
| |
| // Resets the boundaries of this decoder. |
| void Reset(const byte* start, const byte* end, uint32_t buffer_offset = 0) { |
| DCHECK_LE(start, end); |
| DCHECK_EQ(static_cast<uint32_t>(end - start), end - start); |
| start_ = start; |
| pc_ = start; |
| end_ = end; |
| buffer_offset_ = buffer_offset; |
| error_ = {}; |
| } |
| |
| void Reset(Vector<const uint8_t> bytes, uint32_t buffer_offset = 0) { |
| Reset(bytes.begin(), bytes.end(), buffer_offset); |
| } |
| |
| bool ok() const { return error_.empty(); } |
| bool failed() const { return !ok(); } |
| bool more() const { return pc_ < end_; } |
| const WasmError& error() const { return error_; } |
| |
| const byte* start() const { return start_; } |
| const byte* pc() const { return pc_; } |
| uint32_t V8_INLINE position() const { |
| return static_cast<uint32_t>(pc_ - start_); |
| } |
| // This needs to be inlined for performance (see https://crbug.com/910432). |
| uint32_t V8_INLINE pc_offset(const uint8_t* pc) const { |
| DCHECK_LE(start_, pc); |
| DCHECK_GE(kMaxUInt32 - buffer_offset_, pc - start_); |
| return static_cast<uint32_t>(pc - start_) + buffer_offset_; |
| } |
| uint32_t pc_offset() const { return pc_offset(pc_); } |
| uint32_t buffer_offset() const { return buffer_offset_; } |
| // Takes an offset relative to the module start and returns an offset relative |
| // to the current buffer of the decoder. |
| uint32_t GetBufferRelativeOffset(uint32_t offset) const { |
| DCHECK_LE(buffer_offset_, offset); |
| return offset - buffer_offset_; |
| } |
| const byte* end() const { return end_; } |
| void set_end(const byte* end) { end_ = end; } |
| |
| // Check if the byte at {offset} from the current pc equals {expected}. |
| bool lookahead(int offset, byte expected) { |
| DCHECK_LE(pc_, end_); |
| return end_ - pc_ > offset && pc_[offset] == expected; |
| } |
| |
| protected: |
| const byte* start_; |
| const byte* pc_; |
| const byte* end_; |
| // The offset of the current buffer in the module. Needed for streaming. |
| uint32_t buffer_offset_; |
| WasmError error_; |
| |
| private: |
| void verrorf(uint32_t offset, const char* format, va_list args) { |
| // Only report the first error. |
| if (!ok()) return; |
| constexpr int kMaxErrorMsg = 256; |
| EmbeddedVector<char, kMaxErrorMsg> buffer; |
| int len = VSNPrintF(buffer, format, args); |
| CHECK_LT(0, len); |
| error_ = {offset, {buffer.begin(), static_cast<size_t>(len)}}; |
| onFirstError(); |
| } |
| |
| template <typename IntType, ValidateFlag validate> |
| IntType read_little_endian(const byte* pc, const char* msg) { |
| if (!validate) { |
| DCHECK(validate_size(pc, sizeof(IntType), msg)); |
| } else if (!validate_size(pc, sizeof(IntType), msg)) { |
| return IntType{0}; |
| } |
| return base::ReadLittleEndianValue<IntType>(reinterpret_cast<Address>(pc)); |
| } |
| |
| template <typename IntType> |
| IntType consume_little_endian(const char* name) { |
| TRACE(" +%u %-20s: ", pc_offset(), name); |
| if (!checkAvailable(sizeof(IntType))) { |
| traceOffEnd(); |
| pc_ = end_; |
| return IntType{0}; |
| } |
| IntType val = read_little_endian<IntType, kNoValidation>(pc_, name); |
| traceByteRange(pc_, pc_ + sizeof(IntType)); |
| TRACE("= %d\n", val); |
| pc_ += sizeof(IntType); |
| return val; |
| } |
| |
| template <typename IntType, ValidateFlag validate, TraceFlag trace, |
| size_t size_in_bits = 8 * sizeof(IntType)> |
| V8_INLINE IntType read_leb(const byte* pc, uint32_t* length, |
| const char* name = "varint") { |
| static_assert(size_in_bits <= 8 * sizeof(IntType), |
| "leb does not fit in type"); |
| TRACE_IF(trace, " +%u %-20s: ", pc_offset(), name); |
| // Fast path for single-byte integers. |
| if ((!validate || V8_LIKELY(pc < end_)) && !(*pc & 0x80)) { |
| TRACE_IF(trace, "%02x ", *pc); |
| *length = 1; |
| IntType result = *pc; |
| if (std::is_signed<IntType>::value) { |
| // Perform sign extension. |
| constexpr int sign_ext_shift = int{8 * sizeof(IntType)} - 7; |
| result = (result << sign_ext_shift) >> sign_ext_shift; |
| TRACE_IF(trace, "= %" PRIi64 "\n", static_cast<int64_t>(result)); |
| } else { |
| TRACE_IF(trace, "= %" PRIu64 "\n", static_cast<uint64_t>(result)); |
| } |
| return result; |
| } |
| return read_leb_slowpath<IntType, validate, trace, size_in_bits>(pc, length, |
| name); |
| } |
| |
| template <typename IntType, ValidateFlag validate, TraceFlag trace, |
| size_t size_in_bits = 8 * sizeof(IntType)> |
| V8_NOINLINE IntType read_leb_slowpath(const byte* pc, uint32_t* length, |
| const char* name) { |
| // Create an unrolled LEB decoding function per integer type. |
| return read_leb_tail<IntType, validate, trace, size_in_bits, 0>(pc, length, |
| name, 0); |
| } |
| |
| template <typename IntType, ValidateFlag validate, TraceFlag trace, |
| size_t size_in_bits, int byte_index> |
| V8_INLINE IntType read_leb_tail(const byte* pc, uint32_t* length, |
| const char* name, IntType result) { |
| constexpr bool is_signed = std::is_signed<IntType>::value; |
| constexpr int kMaxLength = (size_in_bits + 6) / 7; |
| static_assert(byte_index < kMaxLength, "invalid template instantiation"); |
| constexpr int shift = byte_index * 7; |
| constexpr bool is_last_byte = byte_index == kMaxLength - 1; |
| const bool at_end = validate && pc >= end_; |
| byte b = 0; |
| if (V8_LIKELY(!at_end)) { |
| DCHECK_LT(pc, end_); |
| b = *pc; |
| TRACE_IF(trace, "%02x ", b); |
| using Unsigned = typename std::make_unsigned<IntType>::type; |
| result = result | |
| (static_cast<Unsigned>(static_cast<IntType>(b) & 0x7f) << shift); |
| } |
| if (!is_last_byte && (b & 0x80)) { |
| // Make sure that we only instantiate the template for valid byte indexes. |
| // Compilers are not smart enough to figure out statically that the |
| // following call is unreachable if is_last_byte is false. |
| constexpr int next_byte_index = byte_index + (is_last_byte ? 0 : 1); |
| return read_leb_tail<IntType, validate, trace, size_in_bits, |
| next_byte_index>(pc + 1, length, name, result); |
| } |
| *length = byte_index + (at_end ? 0 : 1); |
| if (validate && V8_UNLIKELY(at_end || (b & 0x80))) { |
| TRACE_IF(trace, at_end ? "<end> " : "<length overflow> "); |
| if (validate == kFullValidation) { |
| errorf(pc, "expected %s", name); |
| } else { |
| MarkError(); |
| } |
| result = 0; |
| *length = 0; |
| } |
| if (is_last_byte) { |
| // A signed-LEB128 must sign-extend the final byte, excluding its |
| // most-significant bit; e.g. for a 32-bit LEB128: |
| // kExtraBits = 4 (== 32 - (5-1) * 7) |
| // For unsigned values, the extra bits must be all zero. |
| // For signed values, the extra bits *plus* the most significant bit must |
| // either be 0, or all ones. |
| constexpr int kExtraBits = size_in_bits - ((kMaxLength - 1) * 7); |
| constexpr int kSignExtBits = kExtraBits - (is_signed ? 1 : 0); |
| const byte checked_bits = b & (0xFF << kSignExtBits); |
| constexpr byte kSignExtendedExtraBits = 0x7f & (0xFF << kSignExtBits); |
| const bool valid_extra_bits = |
| checked_bits == 0 || |
| (is_signed && checked_bits == kSignExtendedExtraBits); |
| if (!validate) { |
| DCHECK(valid_extra_bits); |
| } else if (V8_UNLIKELY(!valid_extra_bits)) { |
| if (validate == kFullValidation) { |
| error(pc, "extra bits in varint"); |
| } else { |
| MarkError(); |
| } |
| result = 0; |
| *length = 0; |
| } |
| } |
| constexpr int sign_ext_shift = |
| is_signed ? std::max(0, int{8 * sizeof(IntType)} - shift - 7) : 0; |
| // Perform sign extension. |
| result = (result << sign_ext_shift) >> sign_ext_shift; |
| if (trace && is_signed) { |
| TRACE("= %" PRIi64 "\n", static_cast<int64_t>(result)); |
| } else if (trace) { |
| TRACE("= %" PRIu64 "\n", static_cast<uint64_t>(result)); |
| } |
| return result; |
| } |
| }; |
| |
| #undef TRACE |
| } // namespace wasm |
| } // namespace internal |
| } // namespace v8 |
| |
| #endif // V8_WASM_DECODER_H_ |