go-ethereum/trie/node.go

// Copyright 2014 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.

package trie

import (
	"bytes"
	"fmt"
	"io"
	"strings"

	"github.com/ethereum/go-ethereum/common"
	"github.com/ethereum/go-ethereum/crypto"
	"github.com/ethereum/go-ethereum/rlp"
)

var indices = []string{"0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "a", "b", "c", "d", "e", "f", "[17]"}

type node interface {
	cache() (hashNode, bool)
	encode(w rlp.EncoderBuffer)
	fstring(string) string
}

type (
	fullNode struct {
		Children [17]node // Actual trie node data to encode/decode (needs custom encoder)
		flags    nodeFlag
	}
	shortNode struct {
		Key   []byte
		Val   node
		flags nodeFlag
	}
	hashNode  []byte
	valueNode []byte

	// fullnodeEncoder is a type used exclusively for encoding fullNode.
	// Briefly instantiating a fullnodeEncoder and initializing with
	// existing slices is less memory intense than using the fullNode type.
	fullnodeEncoder struct {
		Children [17][]byte
	}

	// extNodeEncoder is a type used exclusively for encoding extension node.
	// Briefly instantiating a extNodeEncoder and initializing with existing
	// slices is less memory intense than using the shortNode type.
	extNodeEncoder struct {
		Key []byte
		Val []byte
	}

	// leafNodeEncoder is a type used exclusively for encoding leaf node.
	leafNodeEncoder struct {
		Key []byte
		Val []byte
	}
)

// EncodeRLP encodes a full node into the consensus RLP format.
func (n *fullNode) EncodeRLP(w io.Writer) error {
	eb := rlp.NewEncoderBuffer(w)
	n.encode(eb)
	return eb.Flush()
}

// nodeFlag contains caching-related metadata about a node.
type nodeFlag struct {
	hash  hashNode // cached hash of the node (may be nil)
	dirty bool     // whether the node has changes that must be written to the database
}

func (n nodeFlag) copy() nodeFlag {
	return nodeFlag{
		hash:  common.CopyBytes(n.hash),
		dirty: n.dirty,
	}
}

func (n *fullNode) cache() (hashNode, bool)  { return n.flags.hash, n.flags.dirty }
func (n *shortNode) cache() (hashNode, bool) { return n.flags.hash, n.flags.dirty }
func (n hashNode) cache() (hashNode, bool)   { return nil, true }
func (n valueNode) cache() (hashNode, bool)  { return nil, true }

// Pretty printing.
func (n *fullNode) String() string  { return n.fstring("") }
func (n *shortNode) String() string { return n.fstring("") }
func (n hashNode) String() string   { return n.fstring("") }
func (n valueNode) String() string  { return n.fstring("") }

func (n *fullNode) fstring(ind string) string {
	resp := fmt.Sprintf("[\n%s  ", ind)
	for i, node := range &n.Children {
		if node == nil {
			resp += fmt.Sprintf("%s: <nil> ", indices[i])
		} else {
			resp += fmt.Sprintf("%s: %v", indices[i], node.fstring(ind+"  "))
		}
	}
	return resp + fmt.Sprintf("\n%s] ", ind)
}

func (n *shortNode) fstring(ind string) string {
	return fmt.Sprintf("{%x: %v} ", n.Key, n.Val.fstring(ind+"  "))
}
func (n hashNode) fstring(ind string) string {
	return fmt.Sprintf("<%x> ", []byte(n))
}
func (n valueNode) fstring(ind string) string {
	return fmt.Sprintf("%x ", []byte(n))
}

// mustDecodeNode is a wrapper of decodeNode and panic if any error is encountered.
func mustDecodeNode(hash, buf []byte) node {
	n, err := decodeNode(hash, buf)
	if err != nil {
		panic(fmt.Sprintf("node %x: %v", hash, err))
	}
	return n
}

// mustDecodeNodeUnsafe is a wrapper of decodeNodeUnsafe and panic if any error is
// encountered.
func mustDecodeNodeUnsafe(hash, buf []byte) node {
	n, err := decodeNodeUnsafe(hash, buf)
	if err != nil {
		panic(fmt.Sprintf("node %x: %v", hash, err))
	}
	return n
}

// decodeNode parses the RLP encoding of a trie node. It will deep-copy the passed
// byte slice for decoding, so it's safe to modify the byte slice afterwards. The-
// decode performance of this function is not optimal, but it is suitable for most
// scenarios with low performance requirements and hard to determine whether the
// byte slice be modified or not.
func decodeNode(hash, buf []byte) (node, error) {
	return decodeNodeUnsafe(hash, common.CopyBytes(buf))
}

// decodeNodeUnsafe parses the RLP encoding of a trie node. The passed byte slice
// will be directly referenced by node without bytes deep copy, so the input MUST
// not be changed after.
func decodeNodeUnsafe(hash, buf []byte) (node, error) {
	if len(buf) == 0 {
		return nil, io.ErrUnexpectedEOF
	}
	elems, _, err := rlp.SplitList(buf)
	if err != nil {
		return nil, fmt.Errorf("decode error: %v", err)
	}
	c, err := rlp.CountValues(elems)
	switch {
	case err != nil:
		return nil, fmt.Errorf("invalid node list: %v", err)
	case c == 2:
		n, err := decodeShort(hash, elems)
		return n, wrapError(err, "short")
	case c == 17:
		n, err := decodeFull(hash, elems)
		return n, wrapError(err, "full")
	default:
		return nil, fmt.Errorf("invalid number of list elements: %v", c)
	}
}

func decodeShort(hash, elems []byte) (node, error) {
	kbuf, rest, err := rlp.SplitString(elems)
	if err != nil {
		return nil, err
	}
	flag := nodeFlag{hash: hash}
	key := compactToHex(kbuf)
	if hasTerm(key) {
		// value node
		val, _, err := rlp.SplitString(rest)
		if err != nil {
			return nil, fmt.Errorf("invalid value node: %v", err)
		}
		return &shortNode{key, valueNode(val), flag}, nil
	}
	r, _, err := decodeRef(rest)
	if err != nil {
		return nil, wrapError(err, "val")
	}
	return &shortNode{key, r, flag}, nil
}

func decodeFull(hash, elems []byte) (*fullNode, error) {
	n := &fullNode{flags: nodeFlag{hash: hash}}
	for i := 0; i < 16; i++ {
		cld, rest, err := decodeRef(elems)
		if err != nil {
			return n, wrapError(err, fmt.Sprintf("[%d]", i))
		}
		n.Children[i], elems = cld, rest
	}
	val, _, err := rlp.SplitString(elems)
	if err != nil {
		return n, err
	}
	if len(val) > 0 {
		n.Children[16] = valueNode(val)
	}
	return n, nil
}

const hashLen = len(common.Hash{})

func decodeRef(buf []byte) (node, []byte, error) {
	kind, val, rest, err := rlp.Split(buf)
	if err != nil {
		return nil, buf, err
	}
	switch {
	case kind == rlp.List:
		// 'embedded' node reference. The encoding must be smaller
		// than a hash in order to be valid.
		if size := len(buf) - len(rest); size >= hashLen {
			err := fmt.Errorf("oversized embedded node (size is %d bytes, want size < %d)", size, hashLen)
			return nil, buf, err
		}
		// The buffer content has already been copied or is safe to use;
		// no additional copy is required.
		n, err := decodeNodeUnsafe(nil, buf)
		return n, rest, err
	case kind == rlp.String && len(val) == 0:
		// empty node
		return nil, rest, nil
	case kind == rlp.String && len(val) == 32:
		return hashNode(val), rest, nil
	default:
		return nil, nil, fmt.Errorf("invalid RLP string size %d (want 0 or 32)", len(val))
	}
}

// decodeNodeElements parses the RLP encoding of a trie node and returns all the
// elements in raw byte format.
//
// For full node, it returns a slice of 17 elements;
// For short node, it returns a slice of 2 elements;
func decodeNodeElements(buf []byte) ([][]byte, error) {
	if len(buf) == 0 {
		return nil, io.ErrUnexpectedEOF
	}
	return rlp.SplitListValues(buf)
}

// encodeNodeElements encodes the provided node elements into a rlp list.
func encodeNodeElements(elements [][]byte) ([]byte, error) {
	if len(elements) != 2 && len(elements) != 17 {
		return nil, fmt.Errorf("invalid number of elements: %d", len(elements))
	}
	return rlp.MergeListValues(elements)
}

// NodeDifference accepts two RLP-encoding nodes and figures out the difference
// between them.
//
// An error is returned if any of the provided blob is nil, or the type of nodes
// are different.
func NodeDifference(oldvalue []byte, newvalue []byte) (int, []int, [][]byte, error) {
	oldElems, err := decodeNodeElements(oldvalue)
	if err != nil {
		return 0, nil, nil, err
	}
	newElems, err := decodeNodeElements(newvalue)
	if err != nil {
		return 0, nil, nil, err
	}
	if len(oldElems) != len(newElems) {
		return 0, nil, nil, fmt.Errorf("different node type, old elements: %d, new elements: %d", len(oldElems), len(newElems))
	}
	var (
		indices = make([]int, 0, len(oldElems))
		diff    = make([][]byte, 0, len(oldElems))
	)
	for i := 0; i < len(oldElems); i++ {
		if !bytes.Equal(oldElems[i], newElems[i]) {
			indices = append(indices, i)
			diff = append(diff, oldElems[i])
		}
	}
	return len(oldElems), indices, diff, nil
}

// ReassembleNode accepts a RLP-encoding node along with a set of mutations,
// applying the modification diffs according to the indices and re-assemble.
func ReassembleNode(blob []byte, mutations [][][]byte, indices [][]int) ([]byte, error) {
	if len(mutations) == 0 && len(indices) == 0 {
		return blob, nil
	}
	elements, err := decodeNodeElements(blob)
	if err != nil {
		return nil, err
	}
	for i := 0; i < len(mutations); i++ {
		for j, pos := range indices[i] {
			elements[pos] = mutations[i][j]
		}
	}
	return encodeNodeElements(elements)
}

// wraps a decoding error with information about the path to the
// invalid child node (for debugging encoding issues).
type decodeError struct {
	what  error
	stack []string
}

func wrapError(err error, ctx string) error {
	if err == nil {
		return nil
	}
	if decErr, ok := err.(*decodeError); ok {
		decErr.stack = append(decErr.stack, ctx)
		return decErr
	}
	return &decodeError{err, []string{ctx}}
}

func (err *decodeError) Error() string {
	return fmt.Sprintf("%v (decode path: %s)", err.what, strings.Join(err.stack, "<-"))
}

// StripPartitionRoot strips the leading nibble n from a partition subtree
// root blob produced by a StackTrie built over keys that all share that
// nibble. Returns the hash the canonical top-level branch should mount in
// slot n and, if a new node had to be constructed during stripping, the
// blob the caller must persist.
func StripPartitionRoot(blob []byte, n byte) (hash common.Hash, writeBlob []byte, err error) {
	elems, err := decodeNodeElements(blob)
	if err != nil {
		return common.Hash{}, nil, fmt.Errorf("decode partition root: %w", err)
	}
	if len(elems) != 2 {
		return common.Hash{}, nil, fmt.Errorf("expected shortNode (2 elements), got %d", len(elems))
	}

	// Elements from SplitListValues come with their RLP tag. Strip the
	// tag off the compact-key element to get the raw compact bytes.
	compactKey, _, err := rlp.SplitString(elems[0])
	if err != nil {
		return common.Hash{}, nil, fmt.Errorf("parse compact key: %w", err)
	}
	hex := compactToHex(compactKey)
	if len(hex) == 0 {
		return common.Hash{}, nil, fmt.Errorf("partition root has empty key")
	}
	if hex[0] != n {
		return common.Hash{}, nil, fmt.Errorf("partition root key starts with nibble %d, want %d", hex[0], n)
	}
	childOrValue := elems[1]

	// Case 1: extension of exactly [n] -> reuse the existing child. This is the
	// common case, the partition has many accounts, they all share exactly the
	// leading N, and diverge at the second nibble.
	if !hasTerm(hex) && len(hex) == 1 {
		content, _, err := rlp.SplitString(childOrValue)
		if err != nil {
			return common.Hash{}, nil, fmt.Errorf("parse child ref: %w", err)
		}
		if len(content) != common.HashLength {
			return common.Hash{}, nil, fmt.Errorf("child ref is %d bytes, expected 32", len(content))
		}
		return common.BytesToHash(content), nil, nil
	}

	// Case 2: extension with path [n, more...] -> The new node is an extension
	// with path [more...], same child. All accounts in the partition happen to
	// share the second nibble (or more) too. The extension "ate" more than just
	// the leading N.
	strippedCompact := hexToCompact(hex[1:])
	newKeyRLP, err := rlp.EncodeToBytes(strippedCompact)
	if err != nil {
		return common.Hash{}, nil, fmt.Errorf("encode stripped key: %w", err)
	}

	// Case 3: leaf with path [n, more..., term] -> The new node is a leaf with
	// path [more..., term], same value. The partition's single account produces
	// a leaf whose path is the full 64-nibble account hash plus terminator.
	writeBlob, err = encodeNodeElements([][]byte{newKeyRLP, childOrValue})
	if err != nil {
		return common.Hash{}, nil, fmt.Errorf("encode stripped node: %w", err)
	}
	return crypto.Keccak256Hash(writeBlob), writeBlob, nil
}

// AssembleBranch constructs a fullNode (17-slot branch) from the given
// children and returns its RLP encoding and 32-byte hash.
func AssembleBranch(children [17][]byte) ([]byte, common.Hash, error) {
	fn := &fullnodeEncoder{Children: children}
	w := rlp.NewEncoderBuffer(nil)
	fn.encode(w)
	blob := w.ToBytes()
	w.Flush()
	return blob, crypto.Keccak256Hash(blob), nil
}