Merge bitcoin/bitcoin#22387: Rate limit the processing of rumoured addresses

a4bcd687c9 Improve tests using statistics (John Newbery)
f424d601e1 Add logging and addr rate limiting statistics (Pieter Wuille)
b4ece8a1cd Functional tests for addr rate limiting (Pieter Wuille)
5648138f59 Randomize the order of addr processing (Pieter Wuille)
0d64b8f709 Rate limit the processing of incoming addr messages (Pieter Wuille)

Pull request description:

  The rate at which IP addresses are rumoured (through ADDR and ADDRV2 messages) on the network seems to vary from 0 for some non-participating nodes, to 0.005-0.025 addr/s for recent Bitcoin Core nodes. However, the current codebase will happily accept and process an effectively unbounded rate from attackers. There are measures to limit the influence attackers can have on the addrman database (bucket restrictions based on source IPs), but still - there is no need to permit them to feed us addresses at a rate that's orders of magnitude larger than what is common on the network today, especially as it will cause us to spam our peers too.

  This PR implements a [token bucket](https://en.wikipedia.org/wiki/Token_bucket) based rate limiter, allowing an average of 0.1 addr/s per connection, with bursts up to 1000 addresses at once. Whitelisted peers as well as responses to GETADDR requests are exempt from the limit. New connections start with 1 token, so as to not interfere with the common practice of peers' self-announcement.

ACKs for top commit:
  laanwj:
    ACK a4bcd687c9
  vasild:
    ACK a4bcd687c9
  jnewbery:
    ACK a4bcd687c9
  jonatack:
    ACK a4bcd687c9

Tree-SHA512: b757de76ad78a53035b622944c4213b29b3b55d3d98bf23585afa84bfba10808299d858649f92269a16abfa75eb4366ea047eae3216f7e2f6d3c455782a16bea

src/net_permissions.h

@@ -31,7 +31,8 @@ enum class NetPermissionFlags : uint32_t {
     NoBan = (1U << 4) | Download,
     // Can query the mempool
     Mempool = (1U << 5),
-    // Can request addrs without hitting a privacy-preserving cache
+    // Can request addrs without hitting a privacy-preserving cache, and send us
+    // unlimited amounts of addrs.
     Addr = (1U << 7),
 
     // True if the user did not specifically set fine grained permissions

src/net_processing.cpp

@@ -155,6 +155,13 @@ static constexpr uint32_t MAX_GETCFHEADERS_SIZE = 2000;
 static constexpr size_t MAX_PCT_ADDR_TO_SEND = 23;
 /** The maximum number of address records permitted in an ADDR message. */
 static constexpr size_t MAX_ADDR_TO_SEND{1000};
+/** The maximum rate of address records we're willing to process on average. Can be bypassed using
+ *  the NetPermissionFlags::Addr permission. */
+static constexpr double MAX_ADDR_RATE_PER_SECOND{0.1};
+/** The soft limit of the address processing token bucket (the regular MAX_ADDR_RATE_PER_SECOND
+ *  based increments won't go above this, but the MAX_ADDR_TO_SEND increment following GETADDR
+ *  is exempt from this limit. */
+static constexpr size_t MAX_ADDR_PROCESSING_TOKEN_BUCKET{MAX_ADDR_TO_SEND};
 
 // Internal stuff
 namespace {
@@ -233,6 +240,15 @@ struct Peer {
     std::atomic_bool m_wants_addrv2{false};
     /** Whether this peer has already sent us a getaddr message. */
     bool m_getaddr_recvd{false};
+    /** Number of addr messages that can be processed from this peer. Start at 1 to
+     *  permit self-announcement. */
+    double m_addr_token_bucket{1.0};
+    /** When m_addr_token_bucket was last updated */
+    std::chrono::microseconds m_addr_token_timestamp{GetTime<std::chrono::microseconds>()};
+    /** Total number of addresses that were dropped due to rate limiting. */
+    std::atomic<uint64_t> m_addr_rate_limited{0};
+    /** Total number of addresses that were processed (excludes rate limited ones). */
+    std::atomic<uint64_t> m_addr_processed{0};
 
     /** Set of txids to reconsider once their parent transactions have been accepted **/
     std::set<uint256> m_orphan_work_set GUARDED_BY(g_cs_orphans);
@@ -1239,6 +1255,8 @@ bool PeerManagerImpl::GetNodeStateStats(NodeId nodeid, CNodeStateStats& stats) c
     }
 
     stats.m_ping_wait = ping_wait;
+    stats.m_addr_processed = peer->m_addr_processed.load();
+    stats.m_addr_rate_limited = peer->m_addr_rate_limited.load();
 
     return true;
 }
@@ -2583,6 +2601,9 @@ void PeerManagerImpl::ProcessMessage(CNode& pfrom, const std::string& msg_type,
             // Get recent addresses
             m_connman.PushMessage(&pfrom, CNetMsgMaker(greatest_common_version).Make(NetMsgType::GETADDR));
             peer->m_getaddr_sent = true;
+            // When requesting a getaddr, accept an additional MAX_ADDR_TO_SEND addresses in response
+            // (bypassing the MAX_ADDR_PROCESSING_TOKEN_BUCKET limit).
+            peer->m_addr_token_bucket += MAX_ADDR_TO_SEND;
         }
 
         if (!pfrom.IsInboundConn()) {
@@ -2777,11 +2798,34 @@ void PeerManagerImpl::ProcessMessage(CNode& pfrom, const std::string& msg_type,
         std::vector<CAddress> vAddrOk;
         int64_t nNow = GetAdjustedTime();
         int64_t nSince = nNow - 10 * 60;
+
+        // Update/increment addr rate limiting bucket.
+        const auto current_time = GetTime<std::chrono::microseconds>();
+        if (peer->m_addr_token_bucket < MAX_ADDR_PROCESSING_TOKEN_BUCKET) {
+            // Don't increment bucket if it's already full
+            const auto time_diff = std::max(current_time - peer->m_addr_token_timestamp, 0us);
+            const double increment = CountSecondsDouble(time_diff) * MAX_ADDR_RATE_PER_SECOND;
+            peer->m_addr_token_bucket = std::min<double>(peer->m_addr_token_bucket + increment, MAX_ADDR_PROCESSING_TOKEN_BUCKET);
+        }
+        peer->m_addr_token_timestamp = current_time;
+
+        const bool rate_limited = !pfrom.HasPermission(NetPermissionFlags::Addr);
+        uint64_t num_proc = 0;
+        uint64_t num_rate_limit = 0;
+        Shuffle(vAddr.begin(), vAddr.end(), FastRandomContext());
         for (CAddress& addr : vAddr)
         {
             if (interruptMsgProc)
                 return;
 
+            // Apply rate limiting.
+            if (rate_limited) {
+                if (peer->m_addr_token_bucket < 1.0) {
+                    ++num_rate_limit;
+                    continue;
+                }
+                peer->m_addr_token_bucket -= 1.0;
+            }
             // We only bother storing full nodes, though this may include
             // things which we would not make an outbound connection to, in
             // part because we may make feeler connections to them.
@@ -2795,6 +2839,7 @@ void PeerManagerImpl::ProcessMessage(CNode& pfrom, const std::string& msg_type,
                 // Do not process banned/discouraged addresses beyond remembering we received them
                 continue;
             }
+            ++num_proc;
             bool fReachable = IsReachable(addr);
             if (addr.nTime > nSince && !peer->m_getaddr_sent && vAddr.size() <= 10 && addr.IsRoutable()) {
                 // Relay to a limited number of other nodes
@@ -2804,6 +2849,15 @@ void PeerManagerImpl::ProcessMessage(CNode& pfrom, const std::string& msg_type,
             if (fReachable)
                 vAddrOk.push_back(addr);
         }
+        peer->m_addr_processed += num_proc;
+        peer->m_addr_rate_limited += num_rate_limit;
+        LogPrint(BCLog::NET, "Received addr: %u addresses (%u processed, %u rate-limited) from peer=%d%s\n",
+                 vAddr.size(),
+                 num_proc,
+                 num_rate_limit,
+                 pfrom.GetId(),
+                 fLogIPs ? ", peeraddr=" + pfrom.addr.ToString() : "");
+
         m_addrman.Add(vAddrOk, pfrom.addr, 2 * 60 * 60);
         if (vAddr.size() < 1000) peer->m_getaddr_sent = false;
         if (pfrom.IsAddrFetchConn()) {

src/net_processing.h

@@ -29,6 +29,8 @@ struct CNodeStateStats {
     int m_starting_height = -1;
     std::chrono::microseconds m_ping_wait;
     std::vector<int> vHeightInFlight;
+    uint64_t m_addr_processed = 0;
+    uint64_t m_addr_rate_limited = 0;
 };
 
 class PeerManager : public CValidationInterface, public NetEventsInterface

src/rpc/net.cpp

@@ -242,6 +242,8 @@ static RPCHelpMan getpeerinfo()
                 heights.push_back(height);
             }
             obj.pushKV("inflight", heights);
+            obj.pushKV("addr_processed", statestats.m_addr_processed);
+            obj.pushKV("addr_rate_limited", statestats.m_addr_rate_limited);
         }
         UniValue permissions(UniValue::VARR);
         for (const auto& permission : NetPermissions::ToStrings(stats.m_permissionFlags)) {

test/functional/p2p_addr_relay.py

@@ -13,17 +13,20 @@ from test_framework.messages import (
     msg_addr,
     msg_getaddr
 )
-from test_framework.p2p import P2PInterface
-from test_framework.test_framework import BitcoinTestFramework
-from test_framework.util import (
-    assert_equal,
+from test_framework.p2p import (
+    P2PInterface,
+    p2p_lock,
 )
+from test_framework.test_framework import BitcoinTestFramework
+from test_framework.util import assert_equal
+import random
 import time
 
 
 class AddrReceiver(P2PInterface):
     num_ipv4_received = 0
     test_addr_contents = False
+    _tokens = 1
 
     def __init__(self, test_addr_contents=False):
         super().__init__()
@@ -40,6 +43,20 @@ class AddrReceiver(P2PInterface):
                     raise AssertionError("Invalid addr.port of {} (8333-8342 expected)".format(addr.port))
                 assert addr.ip.startswith('123.123.123.')
 
+    def on_getaddr(self, message):
+        # When the node sends us a getaddr, it increments the addr relay tokens for the connection by 1000
+        self._tokens += 1000
+
+    @property
+    def tokens(self):
+        with p2p_lock:
+            return self._tokens
+
+    def increment_tokens(self, n):
+        # When we move mocktime forward, the node increments the addr relay tokens for its peers
+        with p2p_lock:
+            self._tokens += n
+
     def addr_received(self):
         return self.num_ipv4_received != 0
 
@@ -53,12 +70,14 @@ class AddrTest(BitcoinTestFramework):
 
     def set_test_params(self):
         self.num_nodes = 1
+        self.extra_args = [["-whitelist=addr@127.0.0.1"]]
 
     def run_test(self):
         self.oversized_addr_test()
         self.relay_tests()
         self.getaddr_tests()
         self.blocksonly_mode_tests()
+        self.rate_limit_tests()
 
     def setup_addr_msg(self, num):
         addrs = []
@@ -75,6 +94,19 @@ class AddrTest(BitcoinTestFramework):
         msg.addrs = addrs
         return msg
 
+    def setup_rand_addr_msg(self, num):
+        addrs = []
+        for i in range(num):
+            addr = CAddress()
+            addr.time = self.mocktime + i
+            addr.nServices = NODE_NETWORK | NODE_WITNESS
+            addr.ip = f"{random.randrange(128,169)}.{random.randrange(1,255)}.{random.randrange(1,255)}.{random.randrange(1,255)}"
+            addr.port = 8333
+            addrs.append(addr)
+        msg = msg_addr()
+        msg.addrs = addrs
+        return msg
+
     def send_addr_msg(self, source, msg, receivers):
         source.send_and_ping(msg)
         # pop m_next_addr_send timer
@@ -191,7 +223,7 @@ class AddrTest(BitcoinTestFramework):
 
     def blocksonly_mode_tests(self):
         self.log.info('Test addr relay in -blocksonly mode')
-        self.restart_node(0, ["-blocksonly"])
+        self.restart_node(0, ["-blocksonly", "-whitelist=addr@127.0.0.1"])
         self.mocktime = int(time.time())
 
         self.log.info('Check that we send getaddr messages')
@@ -207,6 +239,63 @@ class AddrTest(BitcoinTestFramework):
 
         self.nodes[0].disconnect_p2ps()
 
+    def send_addrs_and_test_rate_limiting(self, peer, no_relay, new_addrs, total_addrs):
+        """Send an addr message and check that the number of addresses processed and rate-limited is as expected"""
+
+        peer.send_and_ping(self.setup_rand_addr_msg(new_addrs))
+
+        peerinfo = self.nodes[0].getpeerinfo()[0]
+        addrs_processed = peerinfo['addr_processed']
+        addrs_rate_limited = peerinfo['addr_rate_limited']
+        self.log.debug(f"addrs_processed = {addrs_processed}, addrs_rate_limited = {addrs_rate_limited}")
+
+        if no_relay:
+            assert_equal(addrs_processed, 0)
+            assert_equal(addrs_rate_limited, 0)
+        else:
+            assert_equal(addrs_processed, min(total_addrs, peer.tokens))
+            assert_equal(addrs_rate_limited, max(0, total_addrs - peer.tokens))
+
+    def rate_limit_tests(self):
+
+        self.mocktime = int(time.time())
+        self.restart_node(0, [])
+        self.nodes[0].setmocktime(self.mocktime)
+
+        for contype, no_relay in [("outbound-full-relay", False), ("block-relay-only", True), ("inbound", False)]:
+            self.log.info(f'Test rate limiting of addr processing for {contype} peers')
+            if contype == "inbound":
+                peer = self.nodes[0].add_p2p_connection(AddrReceiver())
+            else:
+                peer = self.nodes[0].add_outbound_p2p_connection(AddrReceiver(), p2p_idx=0, connection_type=contype)
+
+            # Send 600 addresses. For all but the block-relay-only peer this should result in addresses being processed.
+            self.send_addrs_and_test_rate_limiting(peer, no_relay, 600, 600)
+
+            # Send 600 more addresses. For the outbound-full-relay peer (which we send a GETADDR, and thus will
+            # process up to 1001 incoming addresses), this means more addresses will be processed.
+            self.send_addrs_and_test_rate_limiting(peer, no_relay, 600, 1200)
+
+            # Send 10 more. As we reached the processing limit for all nodes, no more addresses should be procesesd.
+            self.send_addrs_and_test_rate_limiting(peer, no_relay, 10, 1210)
+
+            # Advance the time by 100 seconds, permitting the processing of 10 more addresses.
+            # Send 200 and verify that 10 are processed.
+            self.mocktime += 100
+            self.nodes[0].setmocktime(self.mocktime)
+            peer.increment_tokens(10)
+
+            self.send_addrs_and_test_rate_limiting(peer, no_relay, 200, 1410)
+
+            # Advance the time by 1000 seconds, permitting the processing of 100 more addresses.
+            # Send 200 and verify that 100 are processed.
+            self.mocktime += 1000
+            self.nodes[0].setmocktime(self.mocktime)
+            peer.increment_tokens(100)
+
+            self.send_addrs_and_test_rate_limiting(peer, no_relay, 200, 1610)
+
+            self.nodes[0].disconnect_p2ps()
 
 if __name__ == '__main__':
     AddrTest().main()

test/functional/p2p_addrv2_relay.py

@@ -42,7 +42,9 @@ class AddrReceiver(P2PInterface):
         super().__init__(support_addrv2 = True)
 
     def on_addrv2(self, message):
-        if ADDRS == message.addrs:
+        expected_set = set((addr.ip, addr.port) for addr in ADDRS)
+        received_set = set((addr.ip, addr.port) for addr in message.addrs)
+        if expected_set == received_set:
             self.addrv2_received_and_checked = True
 
     def wait_for_addrv2(self):
@@ -53,6 +55,7 @@ class AddrTest(BitcoinTestFramework):
     def set_test_params(self):
         self.setup_clean_chain = True
         self.num_nodes = 1
+        self.extra_args = [["-whitelist=addr@127.0.0.1"]]
 
     def run_test(self):
         self.log.info('Create connection that sends addrv2 messages')

test/functional/p2p_invalid_messages.py

@@ -58,6 +58,7 @@ class InvalidMessagesTest(BitcoinTestFramework):
     def set_test_params(self):
         self.num_nodes = 1
         self.setup_clean_chain = True
+        self.extra_args = [["-whitelist=addr@127.0.0.1"]]
 
     def run_test(self):
         self.test_buffer()