headerssync.h raw

   1  // Copyright (c) 2022 The Limenka developers
   2  // Distributed under the MIT software license, see the accompanying
   3  // file COPYING or http://www.opensource.org/licenses/mit-license.php.
   4  
   5  #ifndef LIMENKA_HEADERSSYNC_H
   6  #define LIMENKA_HEADERSSYNC_H
   7  
   8  #include <arith_uint256.h>
   9  #include <chain.h>
  10  #include <consensus/params.h>
  11  #include <net.h> // For NodeId
  12  #include <primitives/block.h>
  13  #include <uint256.h>
  14  #include <util/bitdeque.h>
  15  #include <util/hasher.h>
  16  
  17  #include <deque>
  18  #include <vector>
  19  
  20  // A compressed CBlockHeader, which leaves out the prevhash
  21  struct CompressedHeader {
  22      // header
  23      int32_t nVersion{0};
  24      uint256 hashMerkleRoot;
  25      uint32_t nTime{0};
  26      uint32_t nBits{0};
  27      uint32_t nNonce{0};
  28  
  29      CompressedHeader()
  30      {
  31          hashMerkleRoot.SetNull();
  32      }
  33  
  34      CompressedHeader(const CBlockHeader& header)
  35      {
  36          nVersion = header.nVersion;
  37          hashMerkleRoot = header.hashMerkleRoot;
  38          nTime = header.nTime;
  39          nBits = header.nBits;
  40          nNonce = header.nNonce;
  41      }
  42  
  43      CBlockHeader GetFullHeader(const uint256& hash_prev_block) {
  44          CBlockHeader ret;
  45          ret.nVersion = nVersion;
  46          ret.hashPrevBlock = hash_prev_block;
  47          ret.hashMerkleRoot = hashMerkleRoot;
  48          ret.nTime = nTime;
  49          ret.nBits = nBits;
  50          ret.nNonce = nNonce;
  51          return ret;
  52      };
  53  };
  54  
  55  /** HeadersSyncState:
  56   *
  57   * We wish to download a peer's headers chain in a DoS-resistant way.
  58   *
  59   * The Limenka protocol does not offer an easy way to determine the work on a
  60   * peer's chain. Currently, we can query a peer's headers by using a GETHEADERS
  61   * message, and our peer can return a set of up to 2000 headers that connect to
  62   * something we know. If a peer's chain has more than 2000 blocks, then we need
  63   * a way to verify that the chain actually has enough work on it to be useful to
  64   * us -- by being above our anti-DoS minimum-chain-work threshold -- before we
  65   * commit to storing those headers in memory. Otherwise, it would be cheap for
  66   * an attacker to waste all our memory by serving us low-work headers
  67   * (particularly for a new node coming online for the first time).
  68   *
  69   * To prevent memory-DoS with low-work headers, while still always being
  70   * able to reorg to whatever the most-work chain is, we require that a chain
  71   * meet a work threshold before committing it to memory. We can do this by
  72   * downloading a peer's headers twice, whenever we are not sure that the chain
  73   * has sufficient work:
  74   *
  75   * - In the first download phase, called pre-synchronization, we can calculate
  76   * the work on the chain as we go (just by checking the nBits value on each
  77   * header, and validating the proof-of-work).
  78   *
  79   * - Once we have reached a header where the cumulative chain work is
  80   * sufficient, we switch to downloading the headers a second time, this time
  81   * processing them fully, and possibly storing them in memory.
  82   *
  83   * To prevent an attacker from using (eg) the honest chain to convince us that
  84   * they have a high-work chain, but then feeding us an alternate set of
  85   * low-difficulty headers in the second phase, we store commitments to the
  86   * chain we see in the first download phase that we check in the second phase,
  87   * as follows:
  88   *
  89   * - In phase 1 (presync), store 1 bit (using a salted hash function) for every
  90   * N headers that we see. With a reasonable choice of N, this uses relatively
  91   * little memory even for a very long chain.
  92   *
  93   * - In phase 2 (redownload), keep a lookahead buffer and only accept headers
  94   * from that buffer into the block index (permanent memory usage) once they
  95   * have some target number of verified commitments on top of them. With this
  96   * parametrization, we can achieve a given security target for potential
  97   * permanent memory usage, while choosing N to minimize memory use during the
  98   * sync (temporary, per-peer storage).
  99   */
 100  
 101  class HeadersSyncState {
 102  public:
 103      ~HeadersSyncState() = default;
 104  
 105      enum class State {
 106          /** PRESYNC means the peer has not yet demonstrated their chain has
 107           * sufficient work and we're only building commitments to the chain they
 108           * serve us. */
 109          PRESYNC,
 110          /** REDOWNLOAD means the peer has given us a high-enough-work chain,
 111           * and now we're redownloading the headers we saw before and trying to
 112           * accept them */
 113          REDOWNLOAD,
 114          /** We're done syncing with this peer and can discard any remaining state */
 115          FINAL
 116      };
 117  
 118      /** Return the current state of our download */
 119      State GetState() const { return m_download_state; }
 120  
 121      /** Return the height reached during the PRESYNC phase */
 122      int64_t GetPresyncHeight() const { return m_current_height; }
 123  
 124      /** Return the block timestamp of the last header received during the PRESYNC phase. */
 125      uint32_t GetPresyncTime() const { return m_last_header_received.nTime; }
 126  
 127      /** Return the amount of work in the chain received during the PRESYNC phase. */
 128      arith_uint256 GetPresyncWork() const { return m_current_chain_work; }
 129  
 130      /** Construct a HeadersSyncState object representing a headers sync via this
 131       *  download-twice mechanism).
 132       *
 133       * id: node id (for logging)
 134       * consensus_params: parameters needed for difficulty adjustment validation
 135       * chain_start: best known fork point that the peer's headers branch from
 136       * minimum_required_work: amount of chain work required to accept the chain
 137       */
 138      HeadersSyncState(NodeId id, const Consensus::Params& consensus_params,
 139              const CBlockIndex* chain_start, const arith_uint256& minimum_required_work);
 140  
 141      /** Result data structure for ProcessNextHeaders. */
 142      struct ProcessingResult {
 143          std::vector<CBlockHeader> pow_validated_headers;
 144          bool success{false};
 145          bool request_more{false};
 146      };
 147  
 148      /** Process a batch of headers, once a sync via this mechanism has started
 149       *
 150       * received_headers: headers that were received over the network for processing.
 151       *                   Assumes the caller has already verified the headers
 152       *                   are continuous, and has checked that each header
 153       *                   satisfies the proof-of-work target included in the
 154       *                   header (but not necessarily verified that the
 155       *                   proof-of-work target is correct and passes consensus
 156       *                   rules).
 157       * full_headers_message: true if the message was at max capacity,
 158       *                       indicating more headers may be available
 159       * ProcessingResult.pow_validated_headers: will be filled in with any
 160       *                       headers that the caller can fully process and
 161       *                       validate now (because these returned headers are
 162       *                       on a chain with sufficient work)
 163       * ProcessingResult.success: set to false if an error is detected and the sync is
 164       *                       aborted; true otherwise.
 165       * ProcessingResult.request_more: if true, the caller is suggested to call
 166       *                       NextHeadersRequestLocator and send a getheaders message using it.
 167       */
 168      ProcessingResult ProcessNextHeaders(const std::vector<CBlockHeader>&
 169              received_headers, bool full_headers_message);
 170  
 171      /** Issue the next GETHEADERS message to our peer.
 172       *
 173       * This will return a locator appropriate for the current sync object, to continue the
 174       * synchronization phase it is in.
 175       */
 176      CBlockLocator NextHeadersRequestLocator() const;
 177  
 178  protected:
 179      /** The (secret) offset on the heights for which to create commitments.
 180       *
 181       * m_header_commitments entries are created at any height h for which
 182       * (h % HEADER_COMMITMENT_PERIOD) == m_commit_offset. */
 183      const unsigned m_commit_offset;
 184  
 185  private:
 186      /** Clear out all download state that might be in progress (freeing any used
 187       * memory), and mark this object as no longer usable.
 188       */
 189      void Finalize();
 190  
 191      /**
 192       *  Only called in PRESYNC.
 193       *  Validate the work on the headers we received from the network, and
 194       *  store commitments for later. Update overall state with successfully
 195       *  processed headers.
 196       *  On failure, this invokes Finalize() and returns false.
 197       */
 198      bool ValidateAndStoreHeadersCommitments(const std::vector<CBlockHeader>& headers);
 199  
 200      /** In PRESYNC, process and update state for a single header */
 201      bool ValidateAndProcessSingleHeader(const CBlockHeader& current);
 202  
 203      /** In REDOWNLOAD, check a header's commitment (if applicable) and add to
 204       * buffer for later processing */
 205      bool ValidateAndStoreRedownloadedHeader(const CBlockHeader& header);
 206  
 207      /** Return a set of headers that satisfy our proof-of-work threshold */
 208      std::vector<CBlockHeader> PopHeadersReadyForAcceptance();
 209  
 210  private:
 211      /** NodeId of the peer (used for log messages) **/
 212      const NodeId m_id;
 213  
 214      /** We use the consensus params in our anti-DoS calculations */
 215      const Consensus::Params& m_consensus_params;
 216  
 217      /** Store the last block in our block index that the peer's chain builds from */
 218      const CBlockIndex* m_chain_start{nullptr};
 219  
 220      /** Minimum work that we're looking for on this chain. */
 221      const arith_uint256 m_minimum_required_work;
 222  
 223      /** Work that we've seen so far on the peer's chain */
 224      arith_uint256 m_current_chain_work;
 225  
 226      /** m_hasher is a salted hasher for making our 1-bit commitments to headers we've seen. */
 227      const SaltedTxidHasher m_hasher;
 228  
 229      /** A queue of commitment bits, created during the 1st phase, and verified during the 2nd. */
 230      bitdeque<> m_header_commitments;
 231  
 232      /** m_max_commitments is a bound we calculate on how long an honest peer's chain could be,
 233       * given the MTP rule.
 234       *
 235       * Any peer giving us more headers than this will have its sync aborted. This serves as a
 236       * memory bound on m_header_commitments. */
 237      uint64_t m_max_commitments{0};
 238  
 239      /** Store the latest header received while in PRESYNC (initialized to m_chain_start) */
 240      CBlockHeader m_last_header_received;
 241  
 242      /** Height of m_last_header_received */
 243      int64_t m_current_height{0};
 244  
 245      /** During phase 2 (REDOWNLOAD), we buffer redownloaded headers in memory
 246       *  until enough commitments have been verified; those are stored in
 247       *  m_redownloaded_headers */
 248      std::deque<CompressedHeader> m_redownloaded_headers;
 249  
 250      /** Height of last header in m_redownloaded_headers */
 251      int64_t m_redownload_buffer_last_height{0};
 252  
 253      /** Hash of last header in m_redownloaded_headers (initialized to
 254       * m_chain_start). We have to cache it because we don't have hashPrevBlock
 255       * available in a CompressedHeader.
 256       */
 257      uint256 m_redownload_buffer_last_hash;
 258  
 259      /** The hashPrevBlock entry for the first header in m_redownloaded_headers
 260       * We need this to reconstruct the full header when it's time for
 261       * processing.
 262       */
 263      uint256 m_redownload_buffer_first_prev_hash;
 264  
 265      /** The accumulated work on the redownloaded chain. */
 266      arith_uint256 m_redownload_chain_work;
 267  
 268      /** Set this to true once we encounter the target blockheader during phase
 269       * 2 (REDOWNLOAD). At this point, we can process and store all remaining
 270       * headers still in m_redownloaded_headers.
 271       */
 272      bool m_process_all_remaining_headers{false};
 273  
 274      /** Current state of our headers sync. */
 275      State m_download_state{State::PRESYNC};
 276  };
 277  
 278  #endif // LIMENKA_HEADERSSYNC_H
 279