X-Git-Url: https://git.novaco.in/?a=blobdiff_plain;f=backends%2Fbitcoind%2Fblockchain_processor.py;h=7891f68262726c0b7a0af151be15bbca4a2a9d73;hb=fdc576873627546b8677b7acb560918fac9cafd0;hp=f92f6e4134a61259967d8a240f6ac51582b2960e;hpb=44c124628b5a24b4a3aea549f5dfd8b1bd23f3c4;p=electrum-server.git diff --git a/backends/bitcoind/blockchain_processor.py b/backends/bitcoind/blockchain_processor.py index f92f6e4..7891f68 100644 --- a/backends/bitcoind/blockchain_processor.py +++ b/backends/bitcoind/blockchain_processor.py @@ -3,50 +3,11 @@ import leveldb, urllib import deserialize import ast, time, threading, hashlib from Queue import Queue -import traceback, sys, os - - - -Hash = lambda x: hashlib.sha256(hashlib.sha256(x).digest()).digest() -hash_encode = lambda x: x[::-1].encode('hex') -hash_decode = lambda x: x.decode('hex')[::-1] - - - -def rev_hex(s): - return s.decode('hex')[::-1].encode('hex') - - -def int_to_hex(i, length=1): - s = hex(i)[2:].rstrip('L') - s = "0"*(2*length - len(s)) + s - return rev_hex(s) - -def header_to_string(res): - pbh = res.get('prev_block_hash') - if pbh is None: pbh = '0'*64 - s = int_to_hex(res.get('version'),4) \ - + rev_hex(pbh) \ - + rev_hex(res.get('merkle_root')) \ - + int_to_hex(int(res.get('timestamp')),4) \ - + int_to_hex(int(res.get('bits')),4) \ - + int_to_hex(int(res.get('nonce')),4) - return s - -def header_from_string( s): - hex_to_int = lambda s: eval('0x' + s[::-1].encode('hex')) - h = {} - h['version'] = hex_to_int(s[0:4]) - h['prev_block_hash'] = hash_encode(s[4:36]) - h['merkle_root'] = hash_encode(s[36:68]) - h['timestamp'] = hex_to_int(s[68:72]) - h['bits'] = hex_to_int(s[72:76]) - h['nonce'] = hex_to_int(s[76:80]) - return h - - +import traceback, sys, os, random +from util import Hash, hash_encode, hash_decode, rev_hex, int_to_hex +from util import bc_address_to_hash_160, hash_160_to_bc_address, header_to_string, header_from_string from processor import Processor, print_log class BlockchainProcessor(Processor): @@ -84,12 +45,13 @@ class BlockchainProcessor(Processor): config.get('bitcoind','port')) self.height = 0 + self.is_test = False self.sent_height = 0 self.sent_header = None try: - hist = self.deserialize(self.db.Get('0')) + hist = self.deserialize(self.db.Get('height')) self.last_hash, self.height, _ = hist[0] print_log( "hist", hist ) except: @@ -173,7 +135,7 @@ class BlockchainProcessor(Processor): print_log( "catching up missing headers:", height, db_height) try: - while height != db_height: + while height < db_height: height = height + 1 header = self.get_header(height) if height>1: @@ -257,7 +219,8 @@ class BlockchainProcessor(Processor): with self.dblock: try: - hist = self.deserialize(self.db.Get(addr)) + hash_160 = bc_address_to_hash_160(addr) + hist = self.deserialize(self.db.Get(hash_160)) is_known = True except: hist = [] @@ -320,37 +283,57 @@ class BlockchainProcessor(Processor): return {"block_height":height, "merkle":s, "pos":tx_pos} - def add_to_batch(self, addr, tx_hash, tx_pos, tx_height): - # we do it chronologically, so nothing wrong can happen... + + def add_to_history(self, addr, tx_hash, tx_pos, tx_height): + + # keep it sorted s = (tx_hash + int_to_hex(tx_pos, 4) + int_to_hex(tx_height, 4)).decode('hex') - self.batch_list[addr] += s + + serialized_hist = self.batch_list[addr] + + l = len(serialized_hist)/40 + for i in range(l-1, -1, -1): + item = serialized_hist[40*i:40*(i+1)] + item_height = int( rev_hex( item[36:40].encode('hex') ), 16 ) + if item_height < tx_height: + serialized_hist = serialized_hist[0:40*(i+1)] + s + serialized_hist[40*(i+1):] + break + else: + serialized_hist = s + serialized_hist + + self.batch_list[addr] = serialized_hist # backlink txo = (tx_hash + int_to_hex(tx_pos, 4)).decode('hex') self.batch_txio[txo] = addr - def remove_from_batch(self, tx_hash, tx_pos): + def remove_from_history(self, addr, tx_hash, tx_pos): txi = (tx_hash + int_to_hex(tx_pos, 4)).decode('hex') - try: - addr = self.batch_txio[txi] - except: - #raise BaseException(tx_hash, tx_pos) - print "WARNING: cannot find address for", (tx_hash, tx_pos) - return + if addr is None: + try: + addr = self.batch_txio[txi] + except: + raise BaseException(tx_hash, tx_pos) + serialized_hist = self.batch_list[addr] l = len(serialized_hist)/40 for i in range(l): - if serialized_hist[40*i:40*i+36] == txi: + item = serialized_hist[40*i:40*(i+1)] + if item[0:36] == txi: + height = int( rev_hex( item[36:40].encode('hex') ), 16 ) serialized_hist = serialized_hist[0:40*i] + serialized_hist[40*(i+1):] break else: + hist = self.deserialize(serialized_hist) raise BaseException("prevout not found", addr, hist, tx_hash, tx_pos) + self.batch_list[addr] = serialized_hist + return height, addr def deserialize_block(self, block): @@ -368,40 +351,59 @@ class BlockchainProcessor(Processor): is_coinbase = False return tx_hashes, txdict + def get_undo_info(self, height): + s = self.db.Get("undo%d"%(height%100)) + return eval(s) + + def write_undo_info(self, batch, height, undo_info): + if self.is_test or height > self.bitcoind_height - 100: + batch.Put("undo%d"%(height%100), repr(undo_info)) + def import_block(self, block, block_hash, block_height, sync, revert=False): self.batch_list = {} # address -> history self.batch_txio = {} # transaction i/o -> address - inputs_to_read = [] + block_inputs = [] + block_outputs = [] addr_to_read = [] # deserialize transactions t0 = time.time() tx_hashes, txdict = self.deserialize_block(block) - # read addresses of tx inputs t00 = time.time() - for tx in txdict.values(): - for x in tx.get('inputs'): - txi = (x.get('prevout_hash') + int_to_hex(x.get('prevout_n'), 4)).decode('hex') - inputs_to_read.append(txi) - inputs_to_read.sort() - for txi in inputs_to_read: - try: - addr = self.db.Get(txi) - except: - # the input could come from the same block - continue - self.batch_txio[txi] = addr - addr_to_read.append(addr) + if not revert: + # read addresses of tx inputs + for tx in txdict.values(): + for x in tx.get('inputs'): + txi = (x.get('prevout_hash') + int_to_hex(x.get('prevout_n'), 4)).decode('hex') + block_inputs.append(txi) + + block_inputs.sort() + for txi in block_inputs: + try: + addr = self.db.Get(txi) + except: + # the input could come from the same block + continue + self.batch_txio[txi] = addr + addr_to_read.append(addr) + + else: + for txid, tx in txdict.items(): + for x in tx.get('outputs'): + txo = (txid + int_to_hex(x.get('index'), 4)).decode('hex') + block_outputs.append(txo) + # read histories of addresses for txid, tx in txdict.items(): for x in tx.get('outputs'): - addr_to_read.append(x.get('address')) + hash_160 = bc_address_to_hash_160(x.get('address')) + addr_to_read.append(hash_160) addr_to_read.sort() for addr in addr_to_read: @@ -409,22 +411,48 @@ class BlockchainProcessor(Processor): self.batch_list[addr] = self.db.Get(addr) except: self.batch_list[addr] = '' - + + + if revert: + undo_info = self.get_undo_info(block_height) + # print "undo", block_height, undo_info + else: undo_info = {} + # process t1 = time.time() + if revert: tx_hashes = tx_hashes[::-1] for txid in tx_hashes: # must be ordered tx = txdict[txid] if not revert: + + undo = [] for x in tx.get('inputs'): - self.remove_from_batch( x.get('prevout_hash'), x.get('prevout_n')) + prevout_height, prevout_addr = self.remove_from_history( None, x.get('prevout_hash'), x.get('prevout_n')) + undo.append( (prevout_height, prevout_addr) ) + undo_info[txid] = undo + for x in tx.get('outputs'): - self.add_to_batch( x.get('address'), txid, x.get('index'), block_height) + hash_160 = bc_address_to_hash_160(x.get('address')) + self.add_to_history( hash_160, txid, x.get('index'), block_height) + else: for x in tx.get('outputs'): - self.remove_from_batch( x.get('prevout_hash'), x.get('prevout_n')) + hash_160 = bc_address_to_hash_160(x.get('address')) + self.remove_from_history( hash_160, txid, x.get('index')) + + i = 0 for x in tx.get('inputs'): - self.add_to_batch( x.get('address'), txid, x.get('index'), block_height) + prevout_height, prevout_addr = undo_info.get(txid)[i] + i += 1 + + # read the history into batch list + if self.batch_list.get(prevout_addr) is None: + self.batch_list[prevout_addr] = self.db.Get(prevout_addr) + + # re-add them to the history + self.add_to_history( prevout_addr, x.get('prevout_hash'), x.get('prevout_n'), prevout_height) + print_log( "new hist for", hash_160_to_bc_address(prevout_addr), self.deserialize(self.batch_list[prevout_addr]) ) # write max_len = 0 @@ -439,12 +467,26 @@ class BlockchainProcessor(Processor): max_len = l max_addr = addr - for txio, addr in self.batch_txio.items(): - batch.Put(txio, addr) - # delete spent inputs - for txi in inputs_to_read: - batch.Delete(txi) - batch.Put('0', self.serialize( [(block_hash, block_height, 0)] ) ) + if not revert: + # add new created outputs + for txio, addr in self.batch_txio.items(): + batch.Put(txio, addr) + # delete spent inputs + for txi in block_inputs: + batch.Delete(txi) + # add undo info + self.write_undo_info(batch, block_height, undo_info) + else: + # restore spent inputs + for txio, addr in self.batch_txio.items(): + batch.Put(txio, addr) + # delete spent outputs + for txo in block_outputs: + batch.Delete(txo) + + + # add the max + batch.Put('height', self.serialize( [(block_hash, block_height, 0)] ) ) # actual write self.db.Write(batch, sync = sync) @@ -578,14 +620,6 @@ class BlockchainProcessor(Processor): def catch_up(self, sync = True): - # - # -------> F ------> G -------> H - # / - # / - # A ------> B --------> C ------> E - # - # we always compare the hash in the headers file to the hash returned by bitcoind - t1 = time.time() @@ -593,8 +627,8 @@ class BlockchainProcessor(Processor): # are we done yet? info = self.bitcoind('getinfo') - bitcoind_height = info.get('blocks') - bitcoind_block_hash = self.bitcoind('getblockhash', [bitcoind_height]) + self.bitcoind_height = info.get('blocks') + bitcoind_block_hash = self.bitcoind('getblockhash', [self.bitcoind_height]) if self.last_hash == bitcoind_block_hash: self.up_to_date = True break @@ -602,16 +636,19 @@ class BlockchainProcessor(Processor): # not done.. self.up_to_date = False next_block_hash = self.bitcoind('getblockhash', [self.height+1]) - next_block = self.bitcoind('getblock', [block_hash, 1]) + next_block = self.bitcoind('getblock', [next_block_hash, 1]) + + # fixme: this is unsafe, if we revert when the undo info is not yet written + revert = (random.randint(1, 100)==1) if self.is_test else False - if next_block.get('previousblockhash') == self.last_hash: + if (next_block.get('previousblockhash') == self.last_hash) and not revert: self.import_block(next_block, next_block_hash, self.height+1, sync) self.height = self.height + 1 self.write_header(self.block2header(next_block), sync) self.last_hash = next_block_hash - if (self.height+1)%100 == 0 and not sync: + if (self.height)%100 == 0 and not sync: t2 = time.time() print_log( "catch_up: block %d (%.3fs)"%( self.height, t2 - t1 ) ) t1 = t2 @@ -619,14 +656,15 @@ class BlockchainProcessor(Processor): else: # revert current block block = self.bitcoind('getblock', [self.last_hash, 1]) - print_log( "bc2: reorg", self.height, block.get('previousblockhash'), self.last_hash ) - self.import_block(block, self.last_hash, self.height, revert=True) + print_log( "blockchain reorg", self.height, block.get('previousblockhash'), self.last_hash ) + self.import_block(block, self.last_hash, self.height, sync, revert=True) self.pop_header() + self.flush_headers() self.height = self.height -1 # read previous header from disk - self.header = self.read_header(self.height) + self.header = self.read_header(self.height) self.last_hash = self.hash_header(self.header)