// Copyright 2016 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 "net/http2/hpack/huffman/hpack_huffman_decoder.h" #include #include #include "base/logging.h" // Terminology: // // Symbol - a plain text (unencoded) character (uint8), or the End-of-String // (EOS) symbol, 256. // // Code - the sequence of bits used to encode a symbol, varying in length from // 5 bits for the most common symbols (e.g. '0', '1', and 'a'), to // 30 bits for the least common (e.g. the EOS symbol). // For those symbols whose codes have the same length, their code values // are sorted such that the lower symbol value has a lower code value. // // Canonical - a symbol's cardinal value when sorted first by code length, and // then by symbol value. For example, canonical 0 is for ASCII '0' // (uint8 value 0x30), which is the first of the symbols whose code // is 5 bits long, and the last canonical is EOS, which is the last // of the symbols whose code is 30 bits long. // TODO(jamessynge): Remove use of binary literals, that is a C++ 14 feature. namespace net { namespace { // HuffmanCode is used to store the codes associated with symbols (a pattern of // from 5 to 30 bits). typedef uint32_t HuffmanCode; // HuffmanCodeBitCount is used to store a count of bits in a code. typedef uint16_t HuffmanCodeBitCount; // HuffmanCodeBitSet is used for producing a string version of a code because // std::bitset logs nicely. typedef std::bitset<32> HuffmanCodeBitSet; typedef std::bitset<64> HuffmanAccumulatorBitSet; static constexpr HuffmanCodeBitCount kMinCodeBitCount = 5; static constexpr HuffmanCodeBitCount kMaxCodeBitCount = 30; static constexpr HuffmanCodeBitCount kHuffmanCodeBitCount = std::numeric_limits::digits; static_assert(std::numeric_limits::digits >= kMaxCodeBitCount, "HuffmanCode isn't big enough."); static_assert(std::numeric_limits::digits >= kMaxCodeBitCount, "HuffmanAccumulator isn't big enough."); static constexpr HuffmanAccumulatorBitCount kHuffmanAccumulatorBitCount = std::numeric_limits::digits; static constexpr HuffmanAccumulatorBitCount kExtraAccumulatorBitCount = kHuffmanAccumulatorBitCount - kHuffmanCodeBitCount; // PrefixInfo holds info about a group of codes that are all of the same length. struct PrefixInfo { // Given the leading bits (32 in this case) of the encoded string, and that // they start with a code of length |code_length|, return the corresponding // canonical for that leading code. uint32_t DecodeToCanonical(HuffmanCode bits) const { // What is the position of the canonical symbol being decoded within // the canonical symbols of |length|? HuffmanCode ordinal_in_length = ((bits - first_code) >> (kHuffmanCodeBitCount - code_length)); // Combined with |canonical| to produce the position of the canonical symbol // being decoded within all of the canonical symbols. return first_canonical + ordinal_in_length; } const HuffmanCode first_code; // First code of this length, left justified in // the field (i.e. the first bit of the code is // the high-order bit). const uint16_t code_length; // Length of the prefix code |base|. const uint16_t first_canonical; // First canonical symbol of this length. }; inline std::ostream& operator<<(std::ostream& out, const PrefixInfo& v) { return out << "{first_code: " << HuffmanCodeBitSet(v.first_code) << ", code_length: " << v.code_length << ", first_canonical: " << v.first_canonical << "}"; } // Given |value|, a sequence of the leading bits remaining to be decoded, // figure out which group of canonicals (by code length) that value starts // with. This function was generated. PrefixInfo PrefixToInfo(HuffmanCode value) { if (value < 0b10111000000000000000000000000000) { if (value < 0b01010000000000000000000000000000) { return {0b00000000000000000000000000000000, 5, 0}; } else { return {0b01010000000000000000000000000000, 6, 10}; } } else { if (value < 0b11111110000000000000000000000000) { if (value < 0b11111000000000000000000000000000) { return {0b10111000000000000000000000000000, 7, 36}; } else { return {0b11111000000000000000000000000000, 8, 68}; } } else { if (value < 0b11111111110000000000000000000000) { if (value < 0b11111111101000000000000000000000) { if (value < 0b11111111010000000000000000000000) { return {0b11111110000000000000000000000000, 10, 74}; } else { return {0b11111111010000000000000000000000, 11, 79}; } } else { return {0b11111111101000000000000000000000, 12, 82}; } } else { if (value < 0b11111111111111100000000000000000) { if (value < 0b11111111111110000000000000000000) { if (value < 0b11111111111100000000000000000000) { return {0b11111111110000000000000000000000, 13, 84}; } else { return {0b11111111111100000000000000000000, 14, 90}; } } else { return {0b11111111111110000000000000000000, 15, 92}; } } else { if (value < 0b11111111111111110100100000000000) { if (value < 0b11111111111111101110000000000000) { if (value < 0b11111111111111100110000000000000) { return {0b11111111111111100000000000000000, 19, 95}; } else { return {0b11111111111111100110000000000000, 20, 98}; } } else { return {0b11111111111111101110000000000000, 21, 106}; } } else { if (value < 0b11111111111111111110101000000000) { if (value < 0b11111111111111111011000000000000) { return {0b11111111111111110100100000000000, 22, 119}; } else { return {0b11111111111111111011000000000000, 23, 145}; } } else { if (value < 0b11111111111111111111101111000000) { if (value < 0b11111111111111111111100000000000) { if (value < 0b11111111111111111111011000000000) { return {0b11111111111111111110101000000000, 24, 174}; } else { return {0b11111111111111111111011000000000, 25, 186}; } } else { return {0b11111111111111111111100000000000, 26, 190}; } } else { if (value < 0b11111111111111111111111111110000) { if (value < 0b11111111111111111111111000100000) { return {0b11111111111111111111101111000000, 27, 205}; } else { return {0b11111111111111111111111000100000, 28, 224}; } } else { return {0b11111111111111111111111111110000, 30, 253}; } } } } } } } } } // Mapping from canonical symbol (0 to 255) to actual symbol. // clang-format off constexpr unsigned char kCanonicalToSymbol[] = { '0', '1', '2', 'a', 'c', 'e', 'i', 'o', 's', 't', 0x20, '%', '-', '.', '/', '3', '4', '5', '6', '7', '8', '9', '=', 'A', '_', 'b', 'd', 'f', 'g', 'h', 'l', 'm', 'n', 'p', 'r', 'u', ':', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'Y', 'j', 'k', 'q', 'v', 'w', 'x', 'y', 'z', '&', '*', ',', ';', 'X', 'Z', '!', '\"', '(', ')', '?', '\'', '+', '|', '#', '>', 0x00, '$', '@', '[', ']', '~', '^', '}', '<', '`', '{', '\\', 0xc3, 0xd0, 0x80, 0x82, 0x83, 0xa2, 0xb8, 0xc2, 0xe0, 0xe2, 0x99, 0xa1, 0xa7, 0xac, 0xb0, 0xb1, 0xb3, 0xd1, 0xd8, 0xd9, 0xe3, 0xe5, 0xe6, 0x81, 0x84, 0x85, 0x86, 0x88, 0x92, 0x9a, 0x9c, 0xa0, 0xa3, 0xa4, 0xa9, 0xaa, 0xad, 0xb2, 0xb5, 0xb9, 0xba, 0xbb, 0xbd, 0xbe, 0xc4, 0xc6, 0xe4, 0xe8, 0xe9, 0x01, 0x87, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8f, 0x93, 0x95, 0x96, 0x97, 0x98, 0x9b, 0x9d, 0x9e, 0xa5, 0xa6, 0xa8, 0xae, 0xaf, 0xb4, 0xb6, 0xb7, 0xbc, 0xbf, 0xc5, 0xe7, 0xef, 0x09, 0x8e, 0x90, 0x91, 0x94, 0x9f, 0xab, 0xce, 0xd7, 0xe1, 0xec, 0xed, 0xc7, 0xcf, 0xea, 0xeb, 0xc0, 0xc1, 0xc8, 0xc9, 0xca, 0xcd, 0xd2, 0xd5, 0xda, 0xdb, 0xee, 0xf0, 0xf2, 0xf3, 0xff, 0xcb, 0xcc, 0xd3, 0xd4, 0xd6, 0xdd, 0xde, 0xdf, 0xf1, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x0b, 0x0c, 0x0e, 0x0f, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x7f, 0xdc, 0xf9, 0x0a, 0x0d, 0x16, }; // clang-format on constexpr size_t kShortCodeTableSize = 124; struct ShortCodeInfo { uint8_t symbol; uint8_t length; } kShortCodeTable[kShortCodeTableSize] = { {0x30, 5}, // Match: 0b0000000, Symbol: 0 {0x30, 5}, // Match: 0b0000001, Symbol: 0 {0x30, 5}, // Match: 0b0000010, Symbol: 0 {0x30, 5}, // Match: 0b0000011, Symbol: 0 {0x31, 5}, // Match: 0b0000100, Symbol: 1 {0x31, 5}, // Match: 0b0000101, Symbol: 1 {0x31, 5}, // Match: 0b0000110, Symbol: 1 {0x31, 5}, // Match: 0b0000111, Symbol: 1 {0x32, 5}, // Match: 0b0001000, Symbol: 2 {0x32, 5}, // Match: 0b0001001, Symbol: 2 {0x32, 5}, // Match: 0b0001010, Symbol: 2 {0x32, 5}, // Match: 0b0001011, Symbol: 2 {0x61, 5}, // Match: 0b0001100, Symbol: a {0x61, 5}, // Match: 0b0001101, Symbol: a {0x61, 5}, // Match: 0b0001110, Symbol: a {0x61, 5}, // Match: 0b0001111, Symbol: a {0x63, 5}, // Match: 0b0010000, Symbol: c {0x63, 5}, // Match: 0b0010001, Symbol: c {0x63, 5}, // Match: 0b0010010, Symbol: c {0x63, 5}, // Match: 0b0010011, Symbol: c {0x65, 5}, // Match: 0b0010100, Symbol: e {0x65, 5}, // Match: 0b0010101, Symbol: e {0x65, 5}, // Match: 0b0010110, Symbol: e {0x65, 5}, // Match: 0b0010111, Symbol: e {0x69, 5}, // Match: 0b0011000, Symbol: i {0x69, 5}, // Match: 0b0011001, Symbol: i {0x69, 5}, // Match: 0b0011010, Symbol: i {0x69, 5}, // Match: 0b0011011, Symbol: i {0x6f, 5}, // Match: 0b0011100, Symbol: o {0x6f, 5}, // Match: 0b0011101, Symbol: o {0x6f, 5}, // Match: 0b0011110, Symbol: o {0x6f, 5}, // Match: 0b0011111, Symbol: o {0x73, 5}, // Match: 0b0100000, Symbol: s {0x73, 5}, // Match: 0b0100001, Symbol: s {0x73, 5}, // Match: 0b0100010, Symbol: s {0x73, 5}, // Match: 0b0100011, Symbol: s {0x74, 5}, // Match: 0b0100100, Symbol: t {0x74, 5}, // Match: 0b0100101, Symbol: t {0x74, 5}, // Match: 0b0100110, Symbol: t {0x74, 5}, // Match: 0b0100111, Symbol: t {0x20, 6}, // Match: 0b0101000, Symbol: (space) {0x20, 6}, // Match: 0b0101001, Symbol: (space) {0x25, 6}, // Match: 0b0101010, Symbol: % {0x25, 6}, // Match: 0b0101011, Symbol: % {0x2d, 6}, // Match: 0b0101100, Symbol: - {0x2d, 6}, // Match: 0b0101101, Symbol: - {0x2e, 6}, // Match: 0b0101110, Symbol: . {0x2e, 6}, // Match: 0b0101111, Symbol: . {0x2f, 6}, // Match: 0b0110000, Symbol: / {0x2f, 6}, // Match: 0b0110001, Symbol: / {0x33, 6}, // Match: 0b0110010, Symbol: 3 {0x33, 6}, // Match: 0b0110011, Symbol: 3 {0x34, 6}, // Match: 0b0110100, Symbol: 4 {0x34, 6}, // Match: 0b0110101, Symbol: 4 {0x35, 6}, // Match: 0b0110110, Symbol: 5 {0x35, 6}, // Match: 0b0110111, Symbol: 5 {0x36, 6}, // Match: 0b0111000, Symbol: 6 {0x36, 6}, // Match: 0b0111001, Symbol: 6 {0x37, 6}, // Match: 0b0111010, Symbol: 7 {0x37, 6}, // Match: 0b0111011, Symbol: 7 {0x38, 6}, // Match: 0b0111100, Symbol: 8 {0x38, 6}, // Match: 0b0111101, Symbol: 8 {0x39, 6}, // Match: 0b0111110, Symbol: 9 {0x39, 6}, // Match: 0b0111111, Symbol: 9 {0x3d, 6}, // Match: 0b1000000, Symbol: = {0x3d, 6}, // Match: 0b1000001, Symbol: = {0x41, 6}, // Match: 0b1000010, Symbol: A {0x41, 6}, // Match: 0b1000011, Symbol: A {0x5f, 6}, // Match: 0b1000100, Symbol: _ {0x5f, 6}, // Match: 0b1000101, Symbol: _ {0x62, 6}, // Match: 0b1000110, Symbol: b {0x62, 6}, // Match: 0b1000111, Symbol: b {0x64, 6}, // Match: 0b1001000, Symbol: d {0x64, 6}, // Match: 0b1001001, Symbol: d {0x66, 6}, // Match: 0b1001010, Symbol: f {0x66, 6}, // Match: 0b1001011, Symbol: f {0x67, 6}, // Match: 0b1001100, Symbol: g {0x67, 6}, // Match: 0b1001101, Symbol: g {0x68, 6}, // Match: 0b1001110, Symbol: h {0x68, 6}, // Match: 0b1001111, Symbol: h {0x6c, 6}, // Match: 0b1010000, Symbol: l {0x6c, 6}, // Match: 0b1010001, Symbol: l {0x6d, 6}, // Match: 0b1010010, Symbol: m {0x6d, 6}, // Match: 0b1010011, Symbol: m {0x6e, 6}, // Match: 0b1010100, Symbol: n {0x6e, 6}, // Match: 0b1010101, Symbol: n {0x70, 6}, // Match: 0b1010110, Symbol: p {0x70, 6}, // Match: 0b1010111, Symbol: p {0x72, 6}, // Match: 0b1011000, Symbol: r {0x72, 6}, // Match: 0b1011001, Symbol: r {0x75, 6}, // Match: 0b1011010, Symbol: u {0x75, 6}, // Match: 0b1011011, Symbol: u {0x3a, 7}, // Match: 0b1011100, Symbol: : {0x42, 7}, // Match: 0b1011101, Symbol: B {0x43, 7}, // Match: 0b1011110, Symbol: C {0x44, 7}, // Match: 0b1011111, Symbol: D {0x45, 7}, // Match: 0b1100000, Symbol: E {0x46, 7}, // Match: 0b1100001, Symbol: F {0x47, 7}, // Match: 0b1100010, Symbol: G {0x48, 7}, // Match: 0b1100011, Symbol: H {0x49, 7}, // Match: 0b1100100, Symbol: I {0x4a, 7}, // Match: 0b1100101, Symbol: J {0x4b, 7}, // Match: 0b1100110, Symbol: K {0x4c, 7}, // Match: 0b1100111, Symbol: L {0x4d, 7}, // Match: 0b1101000, Symbol: M {0x4e, 7}, // Match: 0b1101001, Symbol: N {0x4f, 7}, // Match: 0b1101010, Symbol: O {0x50, 7}, // Match: 0b1101011, Symbol: P {0x51, 7}, // Match: 0b1101100, Symbol: Q {0x52, 7}, // Match: 0b1101101, Symbol: R {0x53, 7}, // Match: 0b1101110, Symbol: S {0x54, 7}, // Match: 0b1101111, Symbol: T {0x55, 7}, // Match: 0b1110000, Symbol: U {0x56, 7}, // Match: 0b1110001, Symbol: V {0x57, 7}, // Match: 0b1110010, Symbol: W {0x59, 7}, // Match: 0b1110011, Symbol: Y {0x6a, 7}, // Match: 0b1110100, Symbol: j {0x6b, 7}, // Match: 0b1110101, Symbol: k {0x71, 7}, // Match: 0b1110110, Symbol: q {0x76, 7}, // Match: 0b1110111, Symbol: v {0x77, 7}, // Match: 0b1111000, Symbol: w {0x78, 7}, // Match: 0b1111001, Symbol: x {0x79, 7}, // Match: 0b1111010, Symbol: y {0x7a, 7}, // Match: 0b1111011, Symbol: z }; } // namespace HuffmanBitBuffer::HuffmanBitBuffer() { Reset(); } void HuffmanBitBuffer::Reset() { accumulator_ = 0; count_ = 0; } size_t HuffmanBitBuffer::AppendBytes(Http2StringPiece input) { HuffmanAccumulatorBitCount free_cnt = free_count(); size_t bytes_available = input.size(); if (free_cnt < 8 || bytes_available == 0) { return 0; } // Top up |accumulator_| until there isn't room for a whole byte. size_t bytes_used = 0; auto* ptr = reinterpret_cast(input.data()); do { auto b = static_cast(*ptr++); free_cnt -= 8; accumulator_ |= (b << free_cnt); ++bytes_used; } while (free_cnt >= 8 && bytes_used < bytes_available); count_ += (bytes_used * 8); return bytes_used; } HuffmanAccumulatorBitCount HuffmanBitBuffer::free_count() const { return kHuffmanAccumulatorBitCount - count_; } void HuffmanBitBuffer::ConsumeBits(HuffmanAccumulatorBitCount code_length) { DCHECK_LE(code_length, count_); accumulator_ <<= code_length; count_ -= code_length; } bool HuffmanBitBuffer::InputProperlyTerminated() const { auto cnt = count(); if (cnt < 8) { if (cnt == 0) { return true; } HuffmanAccumulator expected = ~(~HuffmanAccumulator() >> cnt); // We expect all the bits below the high order |cnt| bits of accumulator_ // to be cleared as we perform left shift operations while decoding. DCHECK_EQ(accumulator_ & ~expected, 0u) << "\n expected: " << HuffmanAccumulatorBitSet(expected) << "\n " << *this; return accumulator_ == expected; } return false; } Http2String HuffmanBitBuffer::DebugString() const { std::stringstream ss; ss << "{accumulator: " << HuffmanAccumulatorBitSet(accumulator_) << "; count: " << count_ << "}"; return ss.str(); } HpackHuffmanDecoder::HpackHuffmanDecoder() {} HpackHuffmanDecoder::~HpackHuffmanDecoder() {} bool HpackHuffmanDecoder::Decode(Http2StringPiece input, Http2String* output) { return DecodeShortCodesFirst(input, output); } // "Legacy" decoder, used until cl/129771019 submitted, which added // DecodeShortCodesFirst() as primary decoder method. // TODO(jamessynge): Remove this once satisfied that there is no going back. bool HpackHuffmanDecoder::DecodeWithIfTreeAndStruct(Http2StringPiece input, Http2String* output) { DVLOG(1) << "HpackHuffmanDecoder::DecodeWithIfTreeAndStruct"; // Fill bit_buffer_ from input. input.remove_prefix(bit_buffer_.AppendBytes(input)); while (true) { DVLOG(3) << "Enter Decode Loop, bit_buffer_: " << bit_buffer_; HuffmanCode code_prefix = bit_buffer_.value() >> kExtraAccumulatorBitCount; DVLOG(3) << "code_prefix: " << HuffmanCodeBitSet(code_prefix); PrefixInfo prefix_info = PrefixToInfo(code_prefix); DVLOG(3) << "prefix_info: " << prefix_info; DCHECK_LE(kMinCodeBitCount, prefix_info.code_length); DCHECK_LE(prefix_info.code_length, kMaxCodeBitCount); if (prefix_info.code_length <= bit_buffer_.count()) { // We have enough bits for one code. uint32_t canonical = prefix_info.DecodeToCanonical(code_prefix); if (canonical < 256) { // Valid code. char c = kCanonicalToSymbol[canonical]; output->push_back(c); bit_buffer_.ConsumeBits(prefix_info.code_length); continue; } // Encoder is not supposed to explicity encode the EOS symbol. DLOG(ERROR) << "EOS explicitly encoded!\n " << bit_buffer_ << "\n " << prefix_info; return false; } // bit_buffer_ doesn't have enough bits in it to decode the next symbol. // Append to it as many bytes as are available AND fit. size_t byte_count = bit_buffer_.AppendBytes(input); if (byte_count == 0) { DCHECK_EQ(input.size(), 0u); return true; } input.remove_prefix(byte_count); } } bool HpackHuffmanDecoder::DecodeShortCodesFirst(Http2StringPiece input, Http2String* output) { DVLOG(1) << "HpackHuffmanDecoder::DecodeShortCodesFirst"; // Fill bit_buffer_ from input. input.remove_prefix(bit_buffer_.AppendBytes(input)); while (true) { DVLOG(3) << "Enter Decode Loop, bit_buffer_: " << bit_buffer_; if (bit_buffer_.count() >= 7) { // Get high 7 bits of the bit buffer, see if that contains a complete // code of 5, 6 or 7 bits. uint8_t short_code = bit_buffer_.value() >> (kHuffmanAccumulatorBitCount - 7); DCHECK_LT(short_code, 128); if (short_code < kShortCodeTableSize) { ShortCodeInfo info = kShortCodeTable[short_code]; bit_buffer_.ConsumeBits(info.length); output->push_back(static_cast(info.symbol)); continue; } // The code is more than 7 bits long. Use PrefixToInfo, etc. to decode // longer codes. } else { // We may have (mostly) drained bit_buffer_. If we can top it up, try // using the table decoder above. size_t byte_count = bit_buffer_.AppendBytes(input); if (byte_count > 0) { input.remove_prefix(byte_count); continue; } } HuffmanCode code_prefix = bit_buffer_.value() >> kExtraAccumulatorBitCount; DVLOG(3) << "code_prefix: " << HuffmanCodeBitSet(code_prefix); PrefixInfo prefix_info = PrefixToInfo(code_prefix); DVLOG(3) << "prefix_info: " << prefix_info; DCHECK_LE(kMinCodeBitCount, prefix_info.code_length); DCHECK_LE(prefix_info.code_length, kMaxCodeBitCount); if (prefix_info.code_length <= bit_buffer_.count()) { // We have enough bits for one code. uint32_t canonical = prefix_info.DecodeToCanonical(code_prefix); if (canonical < 256) { // Valid code. char c = kCanonicalToSymbol[canonical]; output->push_back(c); bit_buffer_.ConsumeBits(prefix_info.code_length); continue; } // Encoder is not supposed to explicity encode the EOS symbol. DLOG(ERROR) << "EOS explicitly encoded!\n " << bit_buffer_ << "\n " << prefix_info; return false; } // bit_buffer_ doesn't have enough bits in it to decode the next symbol. // Append to it as many bytes as are available AND fit. size_t byte_count = bit_buffer_.AppendBytes(input); if (byte_count == 0) { DCHECK_EQ(input.size(), 0u); return true; } input.remove_prefix(byte_count); } } Http2String HpackHuffmanDecoder::DebugString() const { return bit_buffer_.DebugString(); } } // namespace net