Merge pull request #7129

49fb8e8 Documentation updates for BIP 130 (Pieter Wuille)
50262d8 Allow block announcements with headers (Suhas Daftuar)

3 files changed, 549 insertions, 3 deletions

diff --git a/qa/pull-tester/rpc-tests.py b/qa/pull-tester/rpc-tests.py
index 3d156a2e7b..5004b09c18 100755
--- a/qa/pull-tester/rpc-tests.py
+++ b/qa/pull-tester/rpc-tests.py
@@ -91,6 +91,7 @@ testScripts = [
     'p2p-fullblocktest.py',
     'blockchain.py',
     'disablewallet.py',
+    'sendheaders.py',
 ]
 testScriptsExt = [
     'bip65-cltv.py',
diff --git a/qa/rpc-tests/sendheaders.py b/qa/rpc-tests/sendheaders.py
new file mode 100755
index 0000000000..d7f4292090
--- /dev/null
+++ b/qa/rpc-tests/sendheaders.py
@@ -0,0 +1,519 @@
+#!/usr/bin/env python2
+#
+# Distributed under the MIT/X11 software license, see the accompanying
+# file COPYING or http://www.opensource.org/licenses/mit-license.php.
+#
+
+from test_framework.mininode import *
+from test_framework.test_framework import BitcoinTestFramework
+from test_framework.util import *
+import time
+from test_framework.blocktools import create_block, create_coinbase
+
+'''
+SendHeadersTest -- test behavior of headers messages to announce blocks.
+
+Setup: 
+
+- Two nodes, two p2p connections to node0. One p2p connection should only ever
+  receive inv's (omitted from testing description below, this is our control).
+  Second node is used for creating reorgs.
+
+Part 1: No headers announcements before "sendheaders"
+a. node mines a block [expect: inv]
+   send getdata for the block [expect: block]
+b. node mines another block [expect: inv]
+   send getheaders and getdata [expect: headers, then block]
+c. node mines another block [expect: inv]
+   peer mines a block, announces with header [expect: getdata]
+d. node mines another block [expect: inv]
+
+Part 2: After "sendheaders", headers announcements should generally work.
+a. peer sends sendheaders [expect: no response]
+   peer sends getheaders with current tip [expect: no response]
+b. node mines a block [expect: tip header]
+c. for N in 1, ..., 10:
+   * for announce-type in {inv, header}
+     - peer mines N blocks, announces with announce-type
+       [ expect: getheaders/getdata or getdata, deliver block(s) ]
+     - node mines a block [ expect: 1 header ]
+
+Part 3: Headers announcements stop after large reorg and resume after getheaders or inv from peer.
+- For response-type in {inv, getheaders}
+  * node mines a 7 block reorg [ expect: headers announcement of 8 blocks ]
+  * node mines an 8-block reorg [ expect: inv at tip ]
+  * peer responds with getblocks/getdata [expect: inv, blocks ]
+  * node mines another block [ expect: inv at tip, peer sends getdata, expect: block ]
+  * node mines another block at tip [ expect: inv ]
+  * peer responds with getheaders with an old hashstop more than 8 blocks back [expect: headers]
+  * peer requests block [ expect: block ]
+  * node mines another block at tip [ expect: inv, peer sends getdata, expect: block ]
+  * peer sends response-type [expect headers if getheaders, getheaders/getdata if mining new block]
+  * node mines 1 block [expect: 1 header, peer responds with getdata]
+
+Part 4: Test direct fetch behavior
+a. Announce 2 old block headers.
+   Expect: no getdata requests.
+b. Announce 3 new blocks via 1 headers message.
+   Expect: one getdata request for all 3 blocks.
+   (Send blocks.)
+c. Announce 1 header that forks off the last two blocks.
+   Expect: no response.
+d. Announce 1 more header that builds on that fork.
+   Expect: one getdata request for two blocks.
+e. Announce 16 more headers that build on that fork.
+   Expect: getdata request for 14 more blocks.
+f. Announce 1 more header that builds on that fork.
+   Expect: no response.
+'''
+
+class BaseNode(NodeConnCB):
+    def __init__(self):
+        NodeConnCB.__init__(self)
+        self.create_callback_map()
+        self.connection = None
+        self.last_inv = None
+        self.last_headers = None
+        self.last_block = None
+        self.ping_counter = 1
+        self.last_pong = msg_pong(0)
+        self.last_getdata = None
+        self.sleep_time = 0.05
+        self.block_announced = False
+
+    def clear_last_announcement(self):
+        with mininode_lock:
+            self.block_announced = False
+            self.last_inv = None
+            self.last_headers = None
+
+    def add_connection(self, conn):
+        self.connection = conn
+
+    # Request data for a list of block hashes
+    def get_data(self, block_hashes):
+        msg = msg_getdata()
+        for x in block_hashes:
+            msg.inv.append(CInv(2, x))
+        self.connection.send_message(msg)
+
+    def get_headers(self, locator, hashstop):
+        msg = msg_getheaders()
+        msg.locator.vHave = locator
+        msg.hashstop = hashstop
+        self.connection.send_message(msg)
+
+    def send_block_inv(self, blockhash):
+        msg = msg_inv()
+        msg.inv = [CInv(2, blockhash)]
+        self.connection.send_message(msg)
+
+    # Wrapper for the NodeConn's send_message function
+    def send_message(self, message):
+        self.connection.send_message(message)
+
+    def on_inv(self, conn, message):
+        self.last_inv = message
+        self.block_announced = True
+
+    def on_headers(self, conn, message):
+        self.last_headers = message
+        self.block_announced = True
+
+    def on_block(self, conn, message):
+        self.last_block = message.block
+        self.last_block.calc_sha256()
+
+    def on_getdata(self, conn, message):
+        self.last_getdata = message
+
+    def on_pong(self, conn, message):
+        self.last_pong = message
+
+    # Test whether the last announcement we received had the
+    # right header or the right inv
+    # inv and headers should be lists of block hashes
+    def check_last_announcement(self, headers=None, inv=None):
+        expect_headers = headers if headers != None else []
+        expect_inv = inv if inv != None else []
+        test_function = lambda: self.block_announced
+        self.sync(test_function)
+        with mininode_lock:
+            self.block_announced = False
+
+            success = True
+            compare_inv = []
+            if self.last_inv != None:
+                compare_inv = [x.hash for x in self.last_inv.inv]
+            if compare_inv != expect_inv:
+                success = False
+
+            hash_headers = []
+            if self.last_headers != None:
+                # treat headers as a list of block hashes
+                hash_headers = [ x.sha256 for x in self.last_headers.headers ]
+            if hash_headers != expect_headers:
+                success = False
+
+            self.last_inv = None
+            self.last_headers = None
+        return success
+
+    # Syncing helpers
+    def sync(self, test_function, timeout=60):
+        while timeout > 0:
+            with mininode_lock:
+                if test_function():
+                    return
+            time.sleep(self.sleep_time)
+            timeout -= self.sleep_time
+        raise AssertionError("Sync failed to complete")
+        
+    def sync_with_ping(self, timeout=60):
+        self.send_message(msg_ping(nonce=self.ping_counter))
+        test_function = lambda: self.last_pong.nonce == self.ping_counter
+        self.sync(test_function, timeout)
+        self.ping_counter += 1
+        return
+
+    def wait_for_block(self, blockhash, timeout=60):
+        test_function = lambda: self.last_block != None and self.last_block.sha256 == blockhash
+        self.sync(test_function, timeout)
+        return
+
+    def wait_for_getdata(self, hash_list, timeout=60):
+        if hash_list == []:
+            return
+
+        test_function = lambda: self.last_getdata != None and [x.hash for x in self.last_getdata.inv] == hash_list
+        self.sync(test_function, timeout)
+        return
+
+    def send_header_for_blocks(self, new_blocks):
+        headers_message = msg_headers()
+        headers_message.headers = [ CBlockHeader(b) for b in new_blocks ]
+        self.send_message(headers_message)
+
+    def send_getblocks(self, locator):
+        getblocks_message = msg_getblocks()
+        getblocks_message.locator.vHave = locator
+        self.send_message(getblocks_message)
+
+# InvNode: This peer should only ever receive inv's, because it doesn't ever send a
+# "sendheaders" message.
+class InvNode(BaseNode):
+    def __init__(self):
+        BaseNode.__init__(self)
+
+# TestNode: This peer is the one we use for most of the testing.
+class TestNode(BaseNode):
+    def __init__(self):
+        BaseNode.__init__(self)
+
+class SendHeadersTest(BitcoinTestFramework):
+    def setup_chain(self):
+        initialize_chain_clean(self.options.tmpdir, 2)
+
+    def setup_network(self):
+        self.nodes = []
+        self.nodes = start_nodes(2, self.options.tmpdir, [["-debug", "-logtimemicros=1"]]*2)
+        connect_nodes(self.nodes[0], 1)
+
+    # mine count blocks and return the new tip
+    def mine_blocks(self, count):
+        self.nodes[0].generate(count)
+        return int(self.nodes[0].getbestblockhash(), 16)
+
+    # mine a reorg that invalidates length blocks (replacing them with
+    # length+1 blocks).
+    # peers is the p2p nodes we're using; we clear their state after the
+    # to-be-reorged-out blocks are mined, so that we don't break later tests.
+    # return the list of block hashes newly mined
+    def mine_reorg(self, length, peers):
+        self.nodes[0].generate(length) # make sure all invalidated blocks are node0's
+        sync_blocks(self.nodes, wait=0.1)
+        [x.clear_last_announcement() for x in peers]
+
+        tip_height = self.nodes[1].getblockcount()
+        hash_to_invalidate = self.nodes[1].getblockhash(tip_height-(length-1))
+        self.nodes[1].invalidateblock(hash_to_invalidate)
+        all_hashes = self.nodes[1].generate(length+1) # Must be longer than the orig chain
+        sync_blocks(self.nodes, wait=0.1)
+        return [int(x, 16) for x in all_hashes]
+
+    def run_test(self):
+        # Setup the p2p connections and start up the network thread.
+        inv_node = InvNode()
+        test_node = TestNode()
+
+        connections = []
+        connections.append(NodeConn('127.0.0.1', p2p_port(0), self.nodes[0], inv_node))
+        # Set nServices to 0 for test_node, so no block download will occur outside of
+        # direct fetching
+        connections.append(NodeConn('127.0.0.1', p2p_port(0), self.nodes[0], test_node, services=0))
+        inv_node.add_connection(connections[0])
+        test_node.add_connection(connections[1])
+
+        NetworkThread().start() # Start up network handling in another thread
+
+        # Test logic begins here
+        inv_node.wait_for_verack()
+        test_node.wait_for_verack()
+
+        tip = int(self.nodes[0].getbestblockhash(), 16)
+
+        # PART 1
+        # 1. Mine a block; expect inv announcements each time
+        print "Part 1: headers don't start before sendheaders message..."
+        for i in xrange(4):
+            old_tip = tip
+            tip = self.mine_blocks(1)
+            assert_equal(inv_node.check_last_announcement(inv=[tip]), True)
+            assert_equal(test_node.check_last_announcement(inv=[tip]), True)
+            # Try a few different responses; none should affect next announcement
+            if i == 0:
+                # first request the block
+                test_node.get_data([tip])
+                test_node.wait_for_block(tip, timeout=5)
+            elif i == 1:
+                # next try requesting header and block
+                test_node.get_headers(locator=[old_tip], hashstop=tip)
+                test_node.get_data([tip])
+                test_node.wait_for_block(tip)
+                test_node.clear_last_announcement() # since we requested headers...
+            elif i == 2:
+                # this time announce own block via headers
+                height = self.nodes[0].getblockcount()
+                last_time = self.nodes[0].getblock(self.nodes[0].getbestblockhash())['time']
+                block_time = last_time + 1
+                new_block = create_block(tip, create_coinbase(height+1), block_time)
+                new_block.solve()
+                test_node.send_header_for_blocks([new_block])
+                test_node.wait_for_getdata([new_block.sha256], timeout=5)
+                test_node.send_message(msg_block(new_block))
+                test_node.sync_with_ping() # make sure this block is processed
+                inv_node.clear_last_announcement()
+                test_node.clear_last_announcement()
+
+        print "Part 1: success!"
+        print "Part 2: announce blocks with headers after sendheaders message..."
+        # PART 2
+        # 2. Send a sendheaders message and test that headers announcements
+        # commence and keep working.
+        test_node.send_message(msg_sendheaders())
+        prev_tip = int(self.nodes[0].getbestblockhash(), 16)
+        test_node.get_headers(locator=[prev_tip], hashstop=0L)
+        test_node.sync_with_ping()
+        test_node.clear_last_announcement() # Clear out empty headers response
+
+        # Now that we've synced headers, headers announcements should work
+        tip = self.mine_blocks(1)
+        assert_equal(inv_node.check_last_announcement(inv=[tip]), True)
+        assert_equal(test_node.check_last_announcement(headers=[tip]), True)
+
+        height = self.nodes[0].getblockcount()+1
+        block_time += 10  # Advance far enough ahead
+        for i in xrange(10):
+            # Mine i blocks, and alternate announcing either via
+            # inv (of tip) or via headers. After each, new blocks
+            # mined by the node should successfully be announced
+            # with block header, even though the blocks are never requested
+            for j in xrange(2):
+                blocks = []
+                for b in xrange(i+1):
+                    blocks.append(create_block(tip, create_coinbase(height), block_time))
+                    blocks[-1].solve()
+                    tip = blocks[-1].sha256
+                    block_time += 1
+                    height += 1
+                if j == 0:
+                    # Announce via inv
+                    test_node.send_block_inv(tip)
+                    test_node.wait_for_getdata([tip], timeout=5)
+                    # Test that duplicate inv's won't result in duplicate
+                    # getdata requests, or duplicate headers announcements
+                    inv_node.send_block_inv(tip)
+                    # Should have received a getheaders as well!
+                    test_node.send_header_for_blocks(blocks)
+                    test_node.wait_for_getdata([x.sha256 for x in blocks[0:-1]], timeout=5)
+                    [ inv_node.send_block_inv(x.sha256) for x in blocks[0:-1] ]
+                    inv_node.sync_with_ping()
+                else:
+                    # Announce via headers
+                    test_node.send_header_for_blocks(blocks)
+                    test_node.wait_for_getdata([x.sha256 for x in blocks], timeout=5)
+                    # Test that duplicate headers won't result in duplicate
+                    # getdata requests (the check is further down)
+                    inv_node.send_header_for_blocks(blocks)
+                    inv_node.sync_with_ping()
+                [ test_node.send_message(msg_block(x)) for x in blocks ]
+                test_node.sync_with_ping()
+                inv_node.sync_with_ping()
+                # This block should not be announced to the inv node (since it also
+                # broadcast it)
+                assert_equal(inv_node.last_inv, None)
+                assert_equal(inv_node.last_headers, None)
+                inv_node.clear_last_announcement()
+                test_node.clear_last_announcement()
+                tip = self.mine_blocks(1)
+                assert_equal(inv_node.check_last_announcement(inv=[tip]), True)
+                assert_equal(test_node.check_last_announcement(headers=[tip]), True)
+                height += 1
+                block_time += 1
+
+        print "Part 2: success!"
+
+        print "Part 3: headers announcements can stop after large reorg, and resume after headers/inv from peer..."
+
+        # PART 3.  Headers announcements can stop after large reorg, and resume after
+        # getheaders or inv from peer.
+        for j in xrange(2):
+            # First try mining a reorg that can propagate with header announcement
+            new_block_hashes = self.mine_reorg(length=7, peers=[test_node, inv_node])
+            tip = new_block_hashes[-1]
+            assert_equal(inv_node.check_last_announcement(inv=[tip]), True)
+            assert_equal(test_node.check_last_announcement(headers=new_block_hashes), True)
+
+            block_time += 8 
+
+            # Mine a too-large reorg, which should be announced with a single inv
+            new_block_hashes = self.mine_reorg(length=8, peers=[test_node, inv_node])
+            tip = new_block_hashes[-1]
+            assert_equal(inv_node.check_last_announcement(inv=[tip]), True)
+            assert_equal(test_node.check_last_announcement(inv=[tip]), True)
+
+            block_time += 9
+
+            fork_point = self.nodes[0].getblock("%02x" % new_block_hashes[0])["previousblockhash"]
+            fork_point = int(fork_point, 16)
+
+            # Use getblocks/getdata
+            test_node.send_getblocks(locator = [fork_point])
+            assert_equal(test_node.check_last_announcement(inv=new_block_hashes[0:-1]), True)
+            test_node.get_data(new_block_hashes)
+            test_node.wait_for_block(new_block_hashes[-1])
+
+            for i in xrange(3):
+                # Mine another block, still should get only an inv
+                tip = self.mine_blocks(1)
+                assert_equal(inv_node.check_last_announcement(inv=[tip]), True)
+                assert_equal(test_node.check_last_announcement(inv=[tip]), True)
+                if i == 0:
+                    # Just get the data -- shouldn't cause headers announcements to resume
+                    test_node.get_data([tip])
+                    test_node.wait_for_block(tip)
+                elif i == 1:
+                    # Send a getheaders message that shouldn't trigger headers announcements
+                    # to resume (best header sent will be too old)
+                    test_node.get_headers(locator=[fork_point], hashstop=new_block_hashes[1])
+                    test_node.get_data([tip])
+                    test_node.wait_for_block(tip)
+                    test_node.clear_last_announcement()
+                elif i == 2:
+                    test_node.get_data([tip])
+                    test_node.wait_for_block(tip)
+                    # This time, try sending either a getheaders to trigger resumption
+                    # of headers announcements, or mine a new block and inv it, also 
+                    # triggering resumption of headers announcements.
+                    if j == 0:
+                        test_node.get_headers(locator=[tip], hashstop=0L)
+                        test_node.sync_with_ping()
+                    else:
+                        test_node.send_block_inv(tip)
+                        test_node.sync_with_ping()
+            # New blocks should now be announced with header
+            tip = self.mine_blocks(1)
+            assert_equal(inv_node.check_last_announcement(inv=[tip]), True)
+            assert_equal(test_node.check_last_announcement(headers=[tip]), True)
+
+        print "Part 3: success!"
+
+        print "Part 4: Testing direct fetch behavior..."
+        tip = self.mine_blocks(1)
+        height = self.nodes[0].getblockcount() + 1
+        last_time = self.nodes[0].getblock(self.nodes[0].getbestblockhash())['time']
+        block_time = last_time + 1
+
+        # Create 2 blocks.  Send the blocks, then send the headers.
+        blocks = []
+        for b in xrange(2):
+            blocks.append(create_block(tip, create_coinbase(height), block_time))
+            blocks[-1].solve()
+            tip = blocks[-1].sha256
+            block_time += 1
+            height += 1
+            inv_node.send_message(msg_block(blocks[-1]))
+
+        inv_node.sync_with_ping() # Make sure blocks are processed
+        test_node.last_getdata = None
+        test_node.send_header_for_blocks(blocks);
+        test_node.sync_with_ping()
+        # should not have received any getdata messages
+        with mininode_lock:
+            assert_equal(test_node.last_getdata, None)
+
+        # This time, direct fetch should work
+        blocks = []
+        for b in xrange(3):
+            blocks.append(create_block(tip, create_coinbase(height), block_time))
+            blocks[-1].solve()
+            tip = blocks[-1].sha256
+            block_time += 1
+            height += 1
+
+        test_node.send_header_for_blocks(blocks)
+        test_node.sync_with_ping()
+        test_node.wait_for_getdata([x.sha256 for x in blocks], timeout=test_node.sleep_time)
+
+        [ test_node.send_message(msg_block(x)) for x in blocks ]
+
+        test_node.sync_with_ping()
+
+        # Now announce a header that forks the last two blocks
+        tip = blocks[0].sha256
+        height -= 1
+        blocks = []
+
+        # Create extra blocks for later
+        for b in xrange(20):
+            blocks.append(create_block(tip, create_coinbase(height), block_time))
+            blocks[-1].solve()
+            tip = blocks[-1].sha256
+            block_time += 1
+            height += 1
+
+        # Announcing one block on fork should not trigger direct fetch
+        # (less work than tip)
+        test_node.last_getdata = None
+        test_node.send_header_for_blocks(blocks[0:1])
+        test_node.sync_with_ping()
+        with mininode_lock:
+            assert_equal(test_node.last_getdata, None)
+
+        # Announcing one more block on fork should trigger direct fetch for
+        # both blocks (same work as tip)
+        test_node.send_header_for_blocks(blocks[1:2])
+        test_node.sync_with_ping()
+        test_node.wait_for_getdata([x.sha256 for x in blocks[0:2]], timeout=test_node.sleep_time)
+
+        # Announcing 16 more headers should trigger direct fetch for 14 more
+        # blocks
+        test_node.send_header_for_blocks(blocks[2:18])
+        test_node.sync_with_ping()
+        test_node.wait_for_getdata([x.sha256 for x in blocks[2:16]], timeout=test_node.sleep_time)
+
+        # Announcing 1 more header should not trigger any response
+        test_node.last_getdata = None
+        test_node.send_header_for_blocks(blocks[18:19])
+        test_node.sync_with_ping()
+        with mininode_lock:
+            assert_equal(test_node.last_getdata, None)
+
+        print "Part 4: success!"
+
+        # Finally, check that the inv node never received a getdata request,
+        # throughout the test
+        assert_equal(inv_node.last_getdata, None)
+
+if __name__ == '__main__':
+    SendHeadersTest().main()
diff --git a/qa/rpc-tests/test_framework/mininode.py b/qa/rpc-tests/test_framework/mininode.py
index b7d78e74fa..64985d58e2 100755
--- a/qa/rpc-tests/test_framework/mininode.py
+++ b/qa/rpc-tests/test_framework/mininode.py
@@ -751,8 +751,8 @@ class msg_inv(object):
 class msg_getdata(object):
     command = "getdata"
 
-    def __init__(self):
-        self.inv = []
+    def __init__(self, inv=None):
+        self.inv = inv if inv != None else []
 
     def deserialize(self, f):
         self.inv = deser_vector(f, CInv)
@@ -905,6 +905,20 @@ class msg_mempool(object):
     def __repr__(self):
         return "msg_mempool()"
 
+class msg_sendheaders(object):
+    command = "sendheaders"
+
+    def __init__(self):
+        pass
+
+    def deserialize(self, f):
+        pass
+
+    def serialize(self):
+        return ""
+
+    def __repr__(self):
+        return "msg_sendheaders()"
 
 # getheaders message has
 # number of entries
@@ -990,6 +1004,17 @@ class NodeConnCB(object):
     def __init__(self):
         self.verack_received = False
 
+    # Spin until verack message is received from the node.
+    # Tests may want to use this as a signal that the test can begin.
+    # This can be called from the testing thread, so it needs to acquire the
+    # global lock.
+    def wait_for_verack(self):
+        while True:
+            with mininode_lock:
+                if self.verack_received:
+                    return
+            time.sleep(0.05)
+
     # Derived classes should call this function once to set the message map
     # which associates the derived classes' functions to incoming messages
     def create_callback_map(self):
@@ -1084,7 +1109,7 @@ class NodeConn(asyncore.dispatcher):
         "regtest": "\xfa\xbf\xb5\xda"    # regtest
     }
 
-    def __init__(self, dstaddr, dstport, rpc, callback, net="regtest"):
+    def __init__(self, dstaddr, dstport, rpc, callback, net="regtest", services=1):
         asyncore.dispatcher.__init__(self, map=mininode_socket_map)
         self.log = logging.getLogger("NodeConn(%s:%d)" % (dstaddr, dstport))
         self.dstaddr = dstaddr
@@ -1102,6 +1127,7 @@ class NodeConn(asyncore.dispatcher):
 
         # stuff version msg into sendbuf
         vt = msg_version()
+        vt.nServices = services
         vt.addrTo.ip = self.dstaddr
         vt.addrTo.port = self.dstport
         vt.addrFrom.ip = "0.0.0.0"