keyfunc/internal/bip39/bip39.go

// Package bip39 is the Golang implementation of the BIP39 spec.
//
// The official BIP39 spec can be found at
// https://github.com/bitcoin/bips/blob/master/bip-0039.mediawiki
//
// It is a copy of github.com/tyler-smith/go-bip39 v1.1.0, trimmed to what
// keyfunc uses.
//
//nolint:mnd // the numbers are BIP-39's own, written as upstream writes them
package bip39

import (
	"crypto/sha256"
	"crypto/sha512"
	"encoding/binary"
	"errors"
	"fmt"
	"math/big"
	"strings"

	"golang.org/x/crypto/pbkdf2"
)

var (
	// ErrInvalidMnemonic is returned when trying to use a malformed mnemonic.
	ErrInvalidMnemonic = errors.New("invalid mnenomic")

	// ErrEntropyLengthInvalid is returned when trying to use an entropy set with
	// an invalid size.
	ErrEntropyLengthInvalid = errors.New(
		"entropy length must be [128, 256] and a multiple of 32",
	)

	// ErrChecksumIncorrect is returned when entropy has the incorrect checksum.
	ErrChecksumIncorrect = errors.New("checksum incorrect")
)

// EntropyFromMnemonic takes a mnemonic generated by this library,
// and returns the input entropy used to generate the given mnemonic.
// An error is returned if the given mnemonic is invalid.
func EntropyFromMnemonic(mnemonic string) ([]byte, error) {
	mnemonicSlice, isValid := splitMnemonicWords(mnemonic)
	if !isValid {
		return nil, ErrInvalidMnemonic
	}

	// Some bitwise operands for working with big.Ints
	shift11BitsMask := big.NewInt(2048)
	bigOne := big.NewInt(1)

	// used to isolate the checksum bits from the entropy+checksum byte array
	wordLengthChecksumMasksMapping := map[int]*big.Int{
		12: big.NewInt(15),
		15: big.NewInt(31),
		18: big.NewInt(63),
		21: big.NewInt(127),
		24: big.NewInt(255),
	}
	// used to use only the desired x of 8 available checksum bits.
	// 256 bit (word length 24) requires all 8 bits of the checksum,
	// and thus no shifting is needed for it (we would get a divByZero crash if we did)
	wordLengthChecksumShiftMapping := map[int]*big.Int{
		12: big.NewInt(16),
		15: big.NewInt(8),
		18: big.NewInt(4),
		21: big.NewInt(2),
	}

	// wordMap is a reverse lookup map for the word list
	wordMap := map[string]int{}
	for i, v := range English() {
		wordMap[v] = i
	}

	// Decode the words into a big.Int.
	b := big.NewInt(0)

	for _, v := range mnemonicSlice {
		index, found := wordMap[v]
		if !found {
			return nil, fmt.Errorf(
				"%w: word `%v` not found in reverse map", ErrInvalidMnemonic, v,
			)
		}

		var wordBytes [2]byte

		//nolint:gosec // the index of a word in the list is below 2048
		binary.BigEndian.PutUint16(wordBytes[:], uint16(index))

		b = b.Mul(b, shift11BitsMask)
		b = b.Or(b, big.NewInt(0).SetBytes(wordBytes[:]))
	}

	// Build and add the checksum to the big.Int.
	checksum := big.NewInt(0)
	checksumMask := wordLengthChecksumMasksMapping[len(mnemonicSlice)]
	checksum = checksum.And(b, checksumMask)

	b.Div(b, big.NewInt(0).Add(checksumMask, bigOne))

	// The entropy is the underlying bytes of the big.Int. Any upper bytes of
	// all 0's are not returned so we pad the beginning of the slice with empty
	// bytes if necessary.
	entropy := b.Bytes()
	entropy = padByteSlice(entropy, len(mnemonicSlice)/3*4)

	// Generate the checksum and compare with the one we got from the mneomnic.
	entropyChecksumBytes := computeChecksum(entropy)
	entropyChecksum := big.NewInt(int64(entropyChecksumBytes[0]))

	if l := len(mnemonicSlice); l != 24 {
		checksumShift := wordLengthChecksumShiftMapping[l]
		entropyChecksum.Div(entropyChecksum, checksumShift)
	}

	if checksum.Cmp(entropyChecksum) != 0 {
		return nil, ErrChecksumIncorrect
	}

	return entropy, nil
}

// NewMnemonic will return a string consisting of the mnemonic words for
// the given entropy.
// If the provide entropy is invalid, an error will be returned.
func NewMnemonic(entropy []byte) (string, error) {
	// Compute some lengths for convenience.
	entropyBitLength := len(entropy) * 8
	checksumBitLength := entropyBitLength / 32
	sentenceLength := (entropyBitLength + checksumBitLength) / 11

	// Validate that the requested size is supported.
	err := validateEntropyBitSize(entropyBitLength)
	if err != nil {
		return "", err
	}

	// Some bitwise operands for working with big.Ints
	last11BitsMask := big.NewInt(2047)
	shift11BitsMask := big.NewInt(2048)

	// wordList is the set of words to use
	wordList := English()

	// Add checksum to entropy.
	entropy = addChecksum(entropy)

	// Break entropy up into sentenceLength chunks of 11 bits.
	// For each word AND mask the rightmost 11 bits and find the word at that index.
	// Then bitshift entropy 11 bits right and repeat.
	// Add to the last empty slot so we can work with LSBs instead of MSB.

	// Entropy as an int so we can bitmask without worrying about bytes slices.
	entropyInt := new(big.Int).SetBytes(entropy)

	// Slice to hold words in.
	words := make([]string, sentenceLength)

	// Throw away big.Int for AND masking.
	word := big.NewInt(0)

	for i := sentenceLength - 1; i >= 0; i-- {
		// Get 11 right most bits and bitshift 11 to the right for next time.
		word.And(entropyInt, last11BitsMask)
		entropyInt.Div(entropyInt, shift11BitsMask)

		// Get the bytes representing the 11 bits as a 2 byte slice.
		wordBytes := padByteSlice(word.Bytes(), 2)

		// Convert bytes to an index and add that word to the list.
		words[i] = wordList[binary.BigEndian.Uint16(wordBytes)]
	}

	return strings.Join(words, " "), nil
}

// NewSeed creates a hashed seed output given a provided string and password.
// No checking is performed to validate that the string provided is a valid mnemonic.
func NewSeed(mnemonic string, password string) []byte {
	return pbkdf2.Key([]byte(mnemonic), []byte("mnemonic"+password), 2048, 64, sha512.New)
}

// IsMnemonicValid attempts to verify that the provided mnemonic is valid.
// Validity is determined by both the number of words being appropriate,
// and that all the words in the mnemonic are present in the word list.
func IsMnemonicValid(mnemonic string) bool {
	_, err := EntropyFromMnemonic(mnemonic)

	return err == nil
}

// Appends to data the first (len(data) / 32)bits of the result of sha256(data)
// Currently only supports data up to 32 bytes
func addChecksum(data []byte) []byte {
	// Some bitwise operands for working with big.Ints
	bigOne := big.NewInt(1)
	bigTwo := big.NewInt(2)

	// Get first byte of sha256
	hash := computeChecksum(data)
	firstChecksumByte := hash[0]

	// len() is in bytes so we divide by 4
	checksumBitLength := uint(len(data) / 4)

	// For each bit of check sum we want we shift the data one the left
	// and then set the (new) right most bit equal to checksum bit at that index
	// staring from the left
	dataBigInt := new(big.Int).SetBytes(data)
	for i := range checksumBitLength {
		// Bitshift 1 left
		dataBigInt.Mul(dataBigInt, bigTwo)

		// Set rightmost bit if leftmost checksum bit is set
		if firstChecksumByte&(1<<(7-i)) > 0 {
			dataBigInt.Or(dataBigInt, bigOne)
		}
	}

	return dataBigInt.Bytes()
}

func computeChecksum(data []byte) []byte {
	hasher := sha256.New()
	hasher.Write(data)

	return hasher.Sum(nil)
}

// validateEntropyBitSize ensures that entropy is the correct size for being a
// mnemonic.
func validateEntropyBitSize(bitSize int) error {
	if (bitSize%32) != 0 || bitSize < 128 || bitSize > 256 {
		return ErrEntropyLengthInvalid
	}

	return nil
}

// padByteSlice returns a byte slice of the given size with contents of the
// given slice left padded and any empty spaces filled with 0's.
func padByteSlice(slice []byte, length int) []byte {
	offset := length - len(slice)
	if offset <= 0 {
		return slice
	}

	newSlice := make([]byte, length)
	copy(newSlice[offset:], slice)

	return newSlice
}

func splitMnemonicWords(mnemonic string) ([]string, bool) {
	// Create a list of all the words in the mnemonic sentence
	words := strings.Fields(mnemonic)

	// Get num of words
	numOfWords := len(words)

	// The number of words should be 12, 15, 18, 21 or 24
	if numOfWords%3 != 0 || numOfWords < 12 || numOfWords > 24 {
		return nil, false
	}

	return words, true
}