ipld-eth-statedb/trie_by_cid/trie/proof.go

// Copyright 2015 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"
	"errors"
	"fmt"

	"github.com/ethereum/go-ethereum/common"
	"github.com/ethereum/go-ethereum/ethdb"
	"github.com/ethereum/go-ethereum/log"
)

// Prove constructs a merkle proof for key. The result contains all encoded nodes
// on the path to the value at key. The value itself is also included in the last
// node and can be retrieved by verifying the proof.
//
// If the trie does not contain a value for key, the returned proof contains all
// nodes of the longest existing prefix of the key (at least the root node), ending
// with the node that proves the absence of the key.
func (t *Trie) Prove(key []byte, fromLevel uint, proofDb ethdb.KeyValueWriter) error {
	// Collect all nodes on the path to key.
	var (
		prefix []byte
		nodes  []node
		tn     = t.root
	)
	key = keybytesToHex(key)
	for len(key) > 0 && tn != nil {
		switch n := tn.(type) {
		case *shortNode:
			if len(key) < len(n.Key) || !bytes.Equal(n.Key, key[:len(n.Key)]) {
				// The trie doesn't contain the key.
				tn = nil
			} else {
				tn = n.Val
				prefix = append(prefix, n.Key...)
				key = key[len(n.Key):]
			}
			nodes = append(nodes, n)
		case *fullNode:
			tn = n.Children[key[0]]
			prefix = append(prefix, key[0])
			key = key[1:]
			nodes = append(nodes, n)
		case hashNode:
			var err error
			tn, err = t.resolveHash(n, prefix)
			if err != nil {
				log.Error(fmt.Sprintf("Unhandled trie error: %v", err))
				return err
			}
		default:
			panic(fmt.Sprintf("%T: invalid node: %v", tn, tn))
		}
	}
	hasher := newHasher(false)
	defer returnHasherToPool(hasher)

	for i, n := range nodes {
		if fromLevel > 0 {
			fromLevel--
			continue
		}
		var hn node
		n, hn = hasher.proofHash(n)
		if hash, ok := hn.(hashNode); ok || i == 0 {
			// If the node's database encoding is a hash (or is the
			// root node), it becomes a proof element.
			enc := nodeToBytes(n)
			if !ok {
				hash = hasher.hashData(enc)
			}
			proofDb.Put(hash, enc)
		}
	}
	return nil
}

// Prove constructs a merkle proof for key. The result contains all encoded nodes
// on the path to the value at key. The value itself is also included in the last
// node and can be retrieved by verifying the proof.
//
// If the trie does not contain a value for key, the returned proof contains all
// nodes of the longest existing prefix of the key (at least the root node), ending
// with the node that proves the absence of the key.
func (t *StateTrie) Prove(key []byte, fromLevel uint, proofDb ethdb.KeyValueWriter) error {
	return t.trie.Prove(key, fromLevel, proofDb)
}

// // VerifyProof checks merkle proofs. The given proof must contain the value for
// // key in a trie with the given root hash. VerifyProof returns an error if the
// // proof contains invalid trie nodes or the wrong value.
// func VerifyProof(rootHash common.Hash, key []byte, proofDb ethdb.KeyValueReader) (value []byte, err error) {
// 	key = keybytesToHex(key)
// 	wantHash := rootHash
// 	for i := 0; ; i++ {
// 		buf, _ := proofDb.Get(wantHash[:])
// 		if buf == nil {
// 			return nil, fmt.Errorf("proof node %d (hash %064x) missing", i, wantHash)
// 		}
// 		n, err := decodeNode(wantHash[:], buf)
// 		if err != nil {
// 			return nil, fmt.Errorf("bad proof node %d: %v", i, err)
// 		}
// 		keyrest, cld := get(n, key, true)
// 		switch cld := cld.(type) {
// 		case nil:
// 			// The trie doesn't contain the key.
// 			return nil, nil
// 		case hashNode:
// 			key = keyrest
// 			copy(wantHash[:], cld)
// 		case valueNode:
// 			return cld, nil
// 		}
// 	}
// }

// proofToPath converts a merkle proof to trie node path. The main purpose of
// this function is recovering a node path from the merkle proof stream. All
// necessary nodes will be resolved and leave the remaining as hashnode.
//
// The given edge proof is allowed to be an existent or non-existent proof.
func proofToPath(rootHash common.Hash, root node, key []byte, proofDb ethdb.KeyValueReader, allowNonExistent bool) (node, []byte, error) {
	// resolveNode retrieves and resolves trie node from merkle proof stream
	resolveNode := func(hash common.Hash) (node, error) {
		buf, _ := proofDb.Get(hash[:])
		if buf == nil {
			return nil, fmt.Errorf("proof node (hash %064x) missing", hash)
		}
		n, err := decodeNode(hash[:], buf)
		if err != nil {
			return nil, fmt.Errorf("bad proof node %v", err)
		}
		return n, err
	}
	// If the root node is empty, resolve it first.
	// Root node must be included in the proof.
	if root == nil {
		n, err := resolveNode(rootHash)
		if err != nil {
			return nil, nil, err
		}
		root = n
	}
	var (
		err           error
		child, parent node
		keyrest       []byte
		valnode       []byte
	)
	key, parent = keybytesToHex(key), root
	for {
		keyrest, child = get(parent, key, false)
		switch cld := child.(type) {
		case nil:
			// The trie doesn't contain the key. It's possible
			// the proof is a non-existing proof, but at least
			// we can prove all resolved nodes are correct, it's
			// enough for us to prove range.
			if allowNonExistent {
				return root, nil, nil
			}
			return nil, nil, errors.New("the node is not contained in trie")
		case *shortNode:
			key, parent = keyrest, child // Already resolved
			continue
		case *fullNode:
			key, parent = keyrest, child // Already resolved
			continue
		case hashNode:
			child, err = resolveNode(common.BytesToHash(cld))
			if err != nil {
				return nil, nil, err
			}
		case valueNode:
			valnode = cld
		}
		// Link the parent and child.
		switch pnode := parent.(type) {
		case *shortNode:
			pnode.Val = child
		case *fullNode:
			pnode.Children[key[0]] = child
		default:
			panic(fmt.Sprintf("%T: invalid node: %v", pnode, pnode))
		}
		if len(valnode) > 0 {
			return root, valnode, nil // The whole path is resolved
		}
		key, parent = keyrest, child
	}
}

// unsetInternal removes all internal node references(hashnode, embedded node).
// It should be called after a trie is constructed with two edge paths. Also
// the given boundary keys must be the one used to construct the edge paths.
//
// It's the key step for range proof. All visited nodes should be marked dirty
// since the node content might be modified. Besides it can happen that some
// fullnodes only have one child which is disallowed. But if the proof is valid,
// the missing children will be filled, otherwise it will be thrown anyway.
//
// Note we have the assumption here the given boundary keys are different
// and right is larger than left.
func unsetInternal(n node, left []byte, right []byte) (bool, error) {
	left, right = keybytesToHex(left), keybytesToHex(right)

	// Step down to the fork point. There are two scenarios can happen:
	// - the fork point is a shortnode: either the key of left proof or
	//   right proof doesn't match with shortnode's key.
	// - the fork point is a fullnode: both two edge proofs are allowed
	//   to point to a non-existent key.
	var (
		pos    = 0
		parent node

		// fork indicator, 0 means no fork, -1 means proof is less, 1 means proof is greater
		shortForkLeft, shortForkRight int
	)
findFork:
	for {
		switch rn := (n).(type) {
		case *shortNode:
			rn.flags = nodeFlag{dirty: true}

			// If either the key of left proof or right proof doesn't match with
			// shortnode, stop here and the forkpoint is the shortnode.
			if len(left)-pos < len(rn.Key) {
				shortForkLeft = bytes.Compare(left[pos:], rn.Key)
			} else {
				shortForkLeft = bytes.Compare(left[pos:pos+len(rn.Key)], rn.Key)
			}
			if len(right)-pos < len(rn.Key) {
				shortForkRight = bytes.Compare(right[pos:], rn.Key)
			} else {
				shortForkRight = bytes.Compare(right[pos:pos+len(rn.Key)], rn.Key)
			}
			if shortForkLeft != 0 || shortForkRight != 0 {
				break findFork
			}
			parent = n
			n, pos = rn.Val, pos+len(rn.Key)
		case *fullNode:
			rn.flags = nodeFlag{dirty: true}

			// If either the node pointed by left proof or right proof is nil,
			// stop here and the forkpoint is the fullnode.
			leftnode, rightnode := rn.Children[left[pos]], rn.Children[right[pos]]
			if leftnode == nil || rightnode == nil || leftnode != rightnode {
				break findFork
			}
			parent = n
			n, pos = rn.Children[left[pos]], pos+1
		default:
			panic(fmt.Sprintf("%T: invalid node: %v", n, n))
		}
	}
	switch rn := n.(type) {
	case *shortNode:
		// There can have these five scenarios:
		// - both proofs are less than the trie path => no valid range
		// - both proofs are greater than the trie path => no valid range
		// - left proof is less and right proof is greater => valid range, unset the shortnode entirely
		// - left proof points to the shortnode, but right proof is greater
		// - right proof points to the shortnode, but left proof is less
		if shortForkLeft == -1 && shortForkRight == -1 {
			return false, errors.New("empty range")
		}
		if shortForkLeft == 1 && shortForkRight == 1 {
			return false, errors.New("empty range")
		}
		if shortForkLeft != 0 && shortForkRight != 0 {
			// The fork point is root node, unset the entire trie
			if parent == nil {
				return true, nil
			}
			parent.(*fullNode).Children[left[pos-1]] = nil
			return false, nil
		}
		// Only one proof points to non-existent key.
		if shortForkRight != 0 {
			if _, ok := rn.Val.(valueNode); ok {
				// The fork point is root node, unset the entire trie
				if parent == nil {
					return true, nil
				}
				parent.(*fullNode).Children[left[pos-1]] = nil
				return false, nil
			}
			return false, unset(rn, rn.Val, left[pos:], len(rn.Key), false)
		}
		if shortForkLeft != 0 {
			if _, ok := rn.Val.(valueNode); ok {
				// The fork point is root node, unset the entire trie
				if parent == nil {
					return true, nil
				}
				parent.(*fullNode).Children[right[pos-1]] = nil
				return false, nil
			}
			return false, unset(rn, rn.Val, right[pos:], len(rn.Key), true)
		}
		return false, nil
	case *fullNode:
		// unset all internal nodes in the forkpoint
		for i := left[pos] + 1; i < right[pos]; i++ {
			rn.Children[i] = nil
		}
		if err := unset(rn, rn.Children[left[pos]], left[pos:], 1, false); err != nil {
			return false, err
		}
		if err := unset(rn, rn.Children[right[pos]], right[pos:], 1, true); err != nil {
			return false, err
		}
		return false, nil
	default:
		panic(fmt.Sprintf("%T: invalid node: %v", n, n))
	}
}

// unset removes all internal node references either the left most or right most.
// It can meet these scenarios:
//
//   - The given path is existent in the trie, unset the associated nodes with the
//     specific direction
//   - The given path is non-existent in the trie
//   - the fork point is a fullnode, the corresponding child pointed by path
//     is nil, return
//   - the fork point is a shortnode, the shortnode is included in the range,
//     keep the entire branch and return.
//   - the fork point is a shortnode, the shortnode is excluded in the range,
//     unset the entire branch.
func unset(parent node, child node, key []byte, pos int, removeLeft bool) error {
	switch cld := child.(type) {
	case *fullNode:
		if removeLeft {
			for i := 0; i < int(key[pos]); i++ {
				cld.Children[i] = nil
			}
			cld.flags = nodeFlag{dirty: true}
		} else {
			for i := key[pos] + 1; i < 16; i++ {
				cld.Children[i] = nil
			}
			cld.flags = nodeFlag{dirty: true}
		}
		return unset(cld, cld.Children[key[pos]], key, pos+1, removeLeft)
	case *shortNode:
		if len(key[pos:]) < len(cld.Key) || !bytes.Equal(cld.Key, key[pos:pos+len(cld.Key)]) {
			// Find the fork point, it's an non-existent branch.
			if removeLeft {
				if bytes.Compare(cld.Key, key[pos:]) < 0 {
					// The key of fork shortnode is less than the path
					// (it belongs to the range), unset the entrie
					// branch. The parent must be a fullnode.
					fn := parent.(*fullNode)
					fn.Children[key[pos-1]] = nil
				}
				//else {
				// The key of fork shortnode is greater than the
				// path(it doesn't belong to the range), keep
				// it with the cached hash available.
				//}
			} else {
				if bytes.Compare(cld.Key, key[pos:]) > 0 {
					// The key of fork shortnode is greater than the
					// path(it belongs to the range), unset the entrie
					// branch. The parent must be a fullnode.
					fn := parent.(*fullNode)
					fn.Children[key[pos-1]] = nil
				}
				//else {
				// The key of fork shortnode is less than the
				// path(it doesn't belong to the range), keep
				// it with the cached hash available.
				//}
			}
			return nil
		}
		if _, ok := cld.Val.(valueNode); ok {
			fn := parent.(*fullNode)
			fn.Children[key[pos-1]] = nil
			return nil
		}
		cld.flags = nodeFlag{dirty: true}
		return unset(cld, cld.Val, key, pos+len(cld.Key), removeLeft)
	case nil:
		// If the node is nil, then it's a child of the fork point
		// fullnode(it's a non-existent branch).
		return nil
	default:
		panic("it shouldn't happen") // hashNode, valueNode
	}
}

// hasRightElement returns the indicator whether there exists more elements
// on the right side of the given path. The given path can point to an existent
// key or a non-existent one. This function has the assumption that the whole
// path should already be resolved.
func hasRightElement(node node, key []byte) bool {
	pos, key := 0, keybytesToHex(key)
	for node != nil {
		switch rn := node.(type) {
		case *fullNode:
			for i := key[pos] + 1; i < 16; i++ {
				if rn.Children[i] != nil {
					return true
				}
			}
			node, pos = rn.Children[key[pos]], pos+1
		case *shortNode:
			if len(key)-pos < len(rn.Key) || !bytes.Equal(rn.Key, key[pos:pos+len(rn.Key)]) {
				return bytes.Compare(rn.Key, key[pos:]) > 0
			}
			node, pos = rn.Val, pos+len(rn.Key)
		case valueNode:
			return false // We have resolved the whole path
		default:
			panic(fmt.Sprintf("%T: invalid node: %v", node, node)) // hashnode
		}
	}
	return false
}

// get returns the child of the given node. Return nil if the
// node with specified key doesn't exist at all.
//
// There is an additional flag `skipResolved`. If it's set then
// all resolved nodes won't be returned.
func get(tn node, key []byte, skipResolved bool) ([]byte, node) {
	for {
		switch n := tn.(type) {
		case *shortNode:
			if len(key) < len(n.Key) || !bytes.Equal(n.Key, key[:len(n.Key)]) {
				return nil, nil
			}
			tn = n.Val
			key = key[len(n.Key):]
			if !skipResolved {
				return key, tn
			}
		case *fullNode:
			tn = n.Children[key[0]]
			key = key[1:]
			if !skipResolved {
				return key, tn
			}
		case hashNode:
			return key, n
		case nil:
			return key, nil
		case valueNode:
			return nil, n
		default:
			panic(fmt.Sprintf("%T: invalid node: %v", tn, tn))
		}
	}
}