|
| 1 | +#include "merkle.h" |
| 2 | +#include "hash.h" |
| 3 | +#include "utilstrencodings.h" |
| 4 | + |
| 5 | +/* WARNING! If you're reading this because you're learning about crypto |
| 6 | + and/or designing a new system that will use merkle trees, keep in mind |
| 7 | + that the following merkle tree algorithm has a serious flaw related to |
| 8 | + duplicate txids, resulting in a vulnerability (CVE-2012-2459). |
| 9 | +
|
| 10 | + The reason is that if the number of hashes in the list at a given time |
| 11 | + is odd, the last one is duplicated before computing the next level (which |
| 12 | + is unusual in Merkle trees). This results in certain sequences of |
| 13 | + transactions leading to the same merkle root. For example, these two |
| 14 | + trees: |
| 15 | +
|
| 16 | + A A |
| 17 | + / \ / \ |
| 18 | + B C B C |
| 19 | + / \ | / \ / \ |
| 20 | + D E F D E F F |
| 21 | + / \ / \ / \ / \ / \ / \ / \ |
| 22 | + 1 2 3 4 5 6 1 2 3 4 5 6 5 6 |
| 23 | +
|
| 24 | + for transaction lists [1,2,3,4,5,6] and [1,2,3,4,5,6,5,6] (where 5 and |
| 25 | + 6 are repeated) result in the same root hash A (because the hash of both |
| 26 | + of (F) and (F,F) is C). |
| 27 | +
|
| 28 | + The vulnerability results from being able to send a block with such a |
| 29 | + transaction list, with the same merkle root, and the same block hash as |
| 30 | + the original without duplication, resulting in failed validation. If the |
| 31 | + receiving node proceeds to mark that block as permanently invalid |
| 32 | + however, it will fail to accept further unmodified (and thus potentially |
| 33 | + valid) versions of the same block. We defend against this by detecting |
| 34 | + the case where we would hash two identical hashes at the end of the list |
| 35 | + together, and treating that identically to the block having an invalid |
| 36 | + merkle root. Assuming no double-SHA256 collisions, this will detect all |
| 37 | + known ways of changing the transactions without affecting the merkle |
| 38 | + root. |
| 39 | +*/ |
| 40 | + |
| 41 | +/* This implements a constant-space merkle root/path calculator, limited to 2^32 leaves. */ |
| 42 | +static void MerkleComputation(const std::vector<uint256>& leaves, uint256* proot, bool* pmutated, uint32_t branchpos, std::vector<uint256>* pbranch) { |
| 43 | + if (pbranch) pbranch->clear(); |
| 44 | + if (leaves.size() == 0) { |
| 45 | + if (pmutated) *pmutated = false; |
| 46 | + if (proot) *proot = uint256(); |
| 47 | + return; |
| 48 | + } |
| 49 | + bool mutated = false; |
| 50 | + // count is the number of leaves processed so far. |
| 51 | + uint32_t count = 0; |
| 52 | + // inner is an array of eagerly computed subtree hashes, indexed by tree |
| 53 | + // level (0 being the leaves). |
| 54 | + // For example, when count is 25 (11001 in binary), inner[4] is the hash of |
| 55 | + // the first 16 leaves, inner[3] of the next 8 leaves, and inner[0] equal to |
| 56 | + // the last leaf. The other inner entries are undefined. |
| 57 | + uint256 inner[32]; |
| 58 | + // Which position in inner is a hash that depends on the matching leaf. |
| 59 | + int matchlevel = -1; |
| 60 | + // First process all leaves into 'inner' values. |
| 61 | + while (count < leaves.size()) { |
| 62 | + uint256 h = leaves[count]; |
| 63 | + bool matchh = count == branchpos; |
| 64 | + count++; |
| 65 | + int level; |
| 66 | + // For each of the lower bits in count that are 0, do 1 step. Each |
| 67 | + // corresponds to an inner value that existed before processing the |
| 68 | + // current leaf, and each needs a hash to combine it. |
| 69 | + for (level = 0; !(count & (((uint32_t)1) << level)); level++) { |
| 70 | + if (pbranch) { |
| 71 | + if (matchh) { |
| 72 | + pbranch->push_back(inner[level]); |
| 73 | + } else if (matchlevel == level) { |
| 74 | + pbranch->push_back(h); |
| 75 | + matchh = true; |
| 76 | + } |
| 77 | + } |
| 78 | + mutated |= (inner[level] == h); |
| 79 | + CHash256().Write(inner[level].begin(), 32).Write(h.begin(), 32).Finalize(h.begin()); |
| 80 | + } |
| 81 | + // Store the resulting hash at inner position level. |
| 82 | + inner[level] = h; |
| 83 | + if (matchh) { |
| 84 | + matchlevel = level; |
| 85 | + } |
| 86 | + } |
| 87 | + // Do a final 'sweep' over the rightmost branch of the tree to process |
| 88 | + // odd levels, and reduce everything to a single top value. |
| 89 | + // Level is the level (counted from the bottom) up to which we've sweeped. |
| 90 | + int level = 0; |
| 91 | + // As long as bit number level in count is zero, skip it. It means there |
| 92 | + // is nothing left at this level. |
| 93 | + while (!(count & (((uint32_t)1) << level))) { |
| 94 | + level++; |
| 95 | + } |
| 96 | + uint256 h = inner[level]; |
| 97 | + bool matchh = matchlevel == level; |
| 98 | + while (count != (((uint32_t)1) << level)) { |
| 99 | + // If we reach this point, h is an inner value that is not the top. |
| 100 | + // We combine it with itself (Bitcoin's special rule for odd levels in |
| 101 | + // the tree) to produce a higher level one. |
| 102 | + if (pbranch && matchh) { |
| 103 | + pbranch->push_back(h); |
| 104 | + } |
| 105 | + CHash256().Write(h.begin(), 32).Write(h.begin(), 32).Finalize(h.begin()); |
| 106 | + // Increment count to the value it would have if two entries at this |
| 107 | + // level had existed. |
| 108 | + count += (((uint32_t)1) << level); |
| 109 | + level++; |
| 110 | + // And propagate the result upwards accordingly. |
| 111 | + while (!(count & (((uint32_t)1) << level))) { |
| 112 | + if (pbranch) { |
| 113 | + if (matchh) { |
| 114 | + pbranch->push_back(inner[level]); |
| 115 | + } else if (matchlevel == level) { |
| 116 | + pbranch->push_back(h); |
| 117 | + matchh = true; |
| 118 | + } |
| 119 | + } |
| 120 | + CHash256().Write(inner[level].begin(), 32).Write(h.begin(), 32).Finalize(h.begin()); |
| 121 | + level++; |
| 122 | + } |
| 123 | + } |
| 124 | + // Return result. |
| 125 | + if (pmutated) *pmutated = mutated; |
| 126 | + if (proot) *proot = h; |
| 127 | +} |
| 128 | + |
| 129 | +uint256 ComputeMerkleRoot(const std::vector<uint256>& leaves, bool* mutated) { |
| 130 | + uint256 hash; |
| 131 | + MerkleComputation(leaves, &hash, mutated, -1, NULL); |
| 132 | + return hash; |
| 133 | +} |
| 134 | + |
| 135 | +std::vector<uint256> ComputeMerkleBranch(const std::vector<uint256>& leaves, uint32_t position) { |
| 136 | + std::vector<uint256> ret; |
| 137 | + MerkleComputation(leaves, NULL, NULL, position, &ret); |
| 138 | + return ret; |
| 139 | +} |
| 140 | + |
| 141 | +uint256 ComputeMerkleRootFromBranch(const uint256& leaf, const std::vector<uint256>& vMerkleBranch, uint32_t nIndex) { |
| 142 | + uint256 hash = leaf; |
| 143 | + for (std::vector<uint256>::const_iterator it = vMerkleBranch.begin(); it != vMerkleBranch.end(); ++it) { |
| 144 | + if (nIndex & 1) { |
| 145 | + hash = Hash(BEGIN(*it), END(*it), BEGIN(hash), END(hash)); |
| 146 | + } else { |
| 147 | + hash = Hash(BEGIN(hash), END(hash), BEGIN(*it), END(*it)); |
| 148 | + } |
| 149 | + nIndex >>= 1; |
| 150 | + } |
| 151 | + return hash; |
| 152 | +} |
| 153 | + |
| 154 | +uint256 BlockMerkleRoot(const CBlock& block, bool* mutated) |
| 155 | +{ |
| 156 | + std::vector<uint256> leaves; |
| 157 | + leaves.resize(block.vtx.size()); |
| 158 | + for (size_t s = 0; s < block.vtx.size(); s++) { |
| 159 | + leaves[s] = block.vtx[s].GetHash(); |
| 160 | + } |
| 161 | + return ComputeMerkleRoot(leaves, mutated); |
| 162 | +} |
| 163 | + |
| 164 | +std::vector<uint256> BlockMerkleBranch(const CBlock& block, uint32_t position) |
| 165 | +{ |
| 166 | + std::vector<uint256> leaves; |
| 167 | + leaves.resize(block.vtx.size()); |
| 168 | + for (size_t s = 0; s < block.vtx.size(); s++) { |
| 169 | + leaves[s] = block.vtx[s].GetHash(); |
| 170 | + } |
| 171 | + return ComputeMerkleBranch(leaves, position); |
| 172 | +} |
0 commit comments