1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967
//! The syncer downloads and verifies large numbers of blocks from peers to Zebra.
//!
//! It is used when Zebra is a long way behind the current chain tip.
use std::{collections::HashSet, pin::Pin, sync::Arc, task::Poll, time::Duration};
use color_eyre::eyre::{eyre, Report};
use futures::stream::{FuturesUnordered, StreamExt};
use indexmap::IndexSet;
use tokio::time::sleep;
use tower::{
builder::ServiceBuilder, hedge::Hedge, limit::ConcurrencyLimit, retry::Retry, timeout::Timeout,
Service, ServiceExt,
};
use zebra_chain::{
block::{self, Block},
chain_tip::ChainTip,
parameters::genesis_hash,
};
use zebra_consensus::{
chain::VerifyChainError, BlockError, VerifyBlockError, VerifyCheckpointError,
};
use zebra_network as zn;
use zebra_state as zs;
use crate::{
components::sync::downloads::BlockDownloadVerifyError, config::ZebradConfig, BoxError,
};
mod downloads;
mod gossip;
mod recent_sync_lengths;
mod status;
#[cfg(test)]
mod tests;
use downloads::{AlwaysHedge, Downloads};
pub use gossip::{gossip_best_tip_block_hashes, BlockGossipError};
pub use recent_sync_lengths::RecentSyncLengths;
pub use status::SyncStatus;
/// Controls the number of peers used for each ObtainTips and ExtendTips request.
const FANOUT: usize = 3;
/// Controls how many times we will retry each block download.
///
/// Failing block downloads is important because it defends against peers who
/// feed us bad hashes. But spurious failures of valid blocks cause the syncer to
/// restart from the previous checkpoint, potentially re-downloading blocks.
///
/// We also hedge requests, so we may retry up to twice this many times. Hedged
/// retries may be concurrent, inner retries are sequential.
const BLOCK_DOWNLOAD_RETRY_LIMIT: usize = 3;
/// A lower bound on the user-specified lookahead limit.
///
/// Set to the maximum checkpoint interval, so the pipeline holds around a checkpoint's
/// worth of blocks.
///
/// ## Security
///
/// If a malicious node is chosen for an ObtainTips or ExtendTips request, it can
/// provide up to 500 malicious block hashes. These block hashes will be
/// distributed across all available peers. Assuming there are around 50 connected
/// peers, the malicious node will receive approximately 10 of those block requests.
///
/// Malicious deserialized blocks can take up a large amount of RAM, see
/// [`super::inbound::downloads::MAX_INBOUND_CONCURRENCY`] and #1880 for details.
/// So we want to keep the lookahead limit reasonably small.
///
/// Once these malicious blocks start failing validation, the syncer will cancel all
/// the pending download and verify tasks, drop all the blocks, and start a new
/// ObtainTips with a new set of peers.
pub const MIN_LOOKAHEAD_LIMIT: usize = zebra_consensus::MAX_CHECKPOINT_HEIGHT_GAP;
/// The default for the user-specified lookahead limit.
///
/// See [`MIN_LOOKAHEAD_LIMIT`] for details.
///
/// TODO: increase to `MAX_CHECKPOINT_HEIGHT_GAP * 5`, after we implement orchard batching
pub const DEFAULT_LOOKAHEAD_LIMIT: usize = zebra_consensus::MAX_CHECKPOINT_HEIGHT_GAP * 3;
/// The expected maximum number of hashes in an ObtainTips or ExtendTips response.
///
/// This is used to allow block heights that are slightly beyond the lookahead limit,
/// but still limit the number of blocks in the pipeline between the downloader and
/// the state.
///
/// See [`MIN_LOOKAHEAD_LIMIT`] for details.
pub const MAX_TIPS_RESPONSE_HASH_COUNT: usize = 500;
/// Controls how long we wait for a tips response to return.
///
/// ## Correctness
///
/// If this timeout is removed (or set too high), the syncer will sometimes hang.
///
/// If this timeout is set too low, the syncer will sometimes get stuck in a
/// failure loop.
pub const TIPS_RESPONSE_TIMEOUT: Duration = Duration::from_secs(6);
/// Controls how long we wait for a block download request to complete.
///
/// This timeout makes sure that the syncer doesn't hang when:
/// - the lookahead queue is full, and
/// - we are waiting for a request that is stuck.
/// See [`BLOCK_VERIFY_TIMEOUT`] for details.
///
/// ## Correctness
///
/// If this timeout is removed (or set too high), the syncer will sometimes hang.
///
/// If this timeout is set too low, the syncer will sometimes get stuck in a
/// failure loop.
pub(super) const BLOCK_DOWNLOAD_TIMEOUT: Duration = Duration::from_secs(15);
/// Controls how long we wait for a block verify request to complete.
///
/// This timeout makes sure that the syncer doesn't hang when:
/// - the lookahead queue is full, and
/// - all pending verifications:
/// - are waiting on a missing download request,
/// - are waiting on a download or verify request that has failed, but we have
/// deliberately ignored the error,
/// - are for blocks a long way ahead of the current tip, or
/// - are for invalid blocks which will never verify, because they depend on
/// missing blocks or transactions.
/// These conditions can happen during normal operation - they are not bugs.
///
/// This timeout also mitigates or hides the following kinds of bugs:
/// - all pending verifications:
/// - are waiting on a download or verify request that has failed, but we have
/// accidentally dropped the error,
/// - are waiting on a download request that has hung inside Zebra,
/// - are on tokio threads that are waiting for blocked operations.
///
/// ## Correctness
///
/// If this timeout is removed (or set too high), the syncer will sometimes hang.
///
/// If this timeout is set too low, the syncer will sometimes get stuck in a
/// failure loop.
///
/// TODO: reduce to `6 * 60`, after we implement orchard batching?
pub(super) const BLOCK_VERIFY_TIMEOUT: Duration = Duration::from_secs(10 * 60);
/// Controls how long we wait to restart syncing after finishing a sync run.
///
/// This delay should be long enough to:
/// - allow zcashd peers to process pending requests. If the node only has a
/// few peers, we want to clear as much peer state as possible. In
/// particular, zcashd sends "next block range" hints, based on zcashd's
/// internal model of our sync progress. But we want to discard these hints,
/// so they don't get confused with ObtainTips and ExtendTips responses, and
/// - allow in-progress downloads to time out.
///
/// This delay is particularly important on instances with slow or unreliable
/// networks, and on testnet, which has a small number of slow peers.
///
/// Using a prime number makes sure that syncer fanouts don't synchronise with other crawls.
///
/// ## Correctness
///
/// If this delay is removed (or set too low), the syncer will
/// sometimes get stuck in a failure loop, due to leftover downloads from
/// previous sync runs.
const SYNC_RESTART_DELAY: Duration = Duration::from_secs(67);
/// Controls how long we wait to retry a failed attempt to download
/// and verify the genesis block.
///
/// This timeout gives the crawler time to find better peers.
///
/// ## Security
///
/// If this timeout is removed (or set too low), Zebra will immediately retry
/// to download and verify the genesis block from its peers. This can cause
/// a denial of service on those peers.
const GENESIS_TIMEOUT_RETRY: Duration = Duration::from_secs(5);
/// Helps work around defects in the bitcoin protocol by checking whether
/// the returned hashes actually extend a chain tip.
#[derive(Copy, Clone, Debug, Hash, PartialEq, Eq)]
struct CheckedTip {
tip: block::Hash,
expected_next: block::Hash,
}
pub struct ChainSync<ZN, ZS, ZV, ZSTip>
where
ZN: Service<zn::Request, Response = zn::Response, Error = BoxError>
+ Send
+ Sync
+ Clone
+ 'static,
ZN::Future: Send,
ZS: Service<zs::Request, Response = zs::Response, Error = BoxError>
+ Send
+ Sync
+ Clone
+ 'static,
ZS::Future: Send,
ZV: Service<Arc<Block>, Response = block::Hash, Error = BoxError>
+ Send
+ Sync
+ Clone
+ 'static,
ZV::Future: Send,
ZSTip: ChainTip + Clone + Send + 'static,
{
// Configuration
/// The genesis hash for the configured network
genesis_hash: block::Hash,
/// The configured lookahead limit, after applying the minimum limit.
lookahead_limit: usize,
// Services
/// A network service which is used to perform ObtainTips and ExtendTips
/// requests.
///
/// Has no retry logic, because failover is handled using fanout.
tip_network: Timeout<ZN>,
/// A service which downloads and verifies blocks, using the provided
/// network and verifier services.
downloads: Pin<
Box<
Downloads<
Hedge<ConcurrencyLimit<Retry<zn::RetryLimit, Timeout<ZN>>>, AlwaysHedge>,
Timeout<ZV>,
ZSTip,
>,
>,
>,
/// The cached block chain state.
state: ZS,
/// Allows efficient access to the best tip of the blockchain.
latest_chain_tip: ZSTip,
// Internal sync state
/// The tips that the syncer is currently following.
prospective_tips: HashSet<CheckedTip>,
/// The lengths of recent sync responses.
recent_syncs: RecentSyncLengths,
}
/// Polls the network to determine whether further blocks are available and
/// downloads them.
///
/// This component is used for initial block sync, but the `Inbound` service is
/// responsible for participating in the gossip protocols used for block
/// diffusion.
impl<ZN, ZS, ZV, ZSTip> ChainSync<ZN, ZS, ZV, ZSTip>
where
ZN: Service<zn::Request, Response = zn::Response, Error = BoxError>
+ Send
+ Sync
+ Clone
+ 'static,
ZN::Future: Send,
ZS: Service<zs::Request, Response = zs::Response, Error = BoxError>
+ Send
+ Sync
+ Clone
+ 'static,
ZS::Future: Send,
ZV: Service<Arc<Block>, Response = block::Hash, Error = BoxError>
+ Send
+ Sync
+ Clone
+ 'static,
ZV::Future: Send,
ZSTip: ChainTip + Clone + Send + 'static,
{
/// Returns a new syncer instance, using:
/// - chain: the zebra-chain `Network` to download (Mainnet or Testnet)
/// - peers: the zebra-network peers to contact for downloads
/// - verifier: the zebra-consensus verifier that checks the chain
/// - state: the zebra-state that stores the chain
/// - latest_chain_tip: the latest chain tip from `state`
///
/// Also returns a [`SyncStatus`] to check if the syncer has likely reached the chain tip.
pub fn new(
config: &ZebradConfig,
peers: ZN,
verifier: ZV,
state: ZS,
latest_chain_tip: ZSTip,
) -> (Self, SyncStatus) {
let tip_network = Timeout::new(peers.clone(), TIPS_RESPONSE_TIMEOUT);
// The Hedge middleware is the outermost layer, hedging requests
// between two retry-wrapped networks. The innermost timeout
// layer is relatively unimportant, because slow requests will
// probably be pre-emptively hedged.
//
// The Hedge goes outside the Retry, because the Retry layer
// abstracts away spurious failures from individual peers
// making a less-fallible network service, and the Hedge layer
// tries to reduce latency of that less-fallible service.
//
// XXX add ServiceBuilder::hedge() so this becomes
// ServiceBuilder::new().hedge(...).retry(...)...
let block_network = Hedge::new(
ServiceBuilder::new()
.concurrency_limit(config.sync.max_concurrent_block_requests)
.retry(zn::RetryLimit::new(BLOCK_DOWNLOAD_RETRY_LIMIT))
.timeout(BLOCK_DOWNLOAD_TIMEOUT)
.service(peers),
AlwaysHedge,
20,
0.95,
2 * SYNC_RESTART_DELAY,
);
// We apply a timeout to the verifier to avoid hangs due to missing earlier blocks.
let verifier = Timeout::new(verifier, BLOCK_VERIFY_TIMEOUT);
let mut lookahead_limit = config.sync.lookahead_limit;
if lookahead_limit < MIN_LOOKAHEAD_LIMIT {
warn!(
"configured lookahead limit {} too low, increasing to {}",
config.sync.lookahead_limit, MIN_LOOKAHEAD_LIMIT,
);
lookahead_limit = MIN_LOOKAHEAD_LIMIT;
}
let (sync_status, recent_syncs) = SyncStatus::new();
let new_syncer = Self {
genesis_hash: genesis_hash(config.network.network),
lookahead_limit,
tip_network,
downloads: Box::pin(Downloads::new(
block_network,
verifier,
latest_chain_tip.clone(),
lookahead_limit,
)),
state,
latest_chain_tip,
prospective_tips: HashSet::new(),
recent_syncs,
};
(new_syncer, sync_status)
}
/// Runs the syncer to synchronize the chain and keep it synchronized.
#[instrument(skip(self))]
pub async fn sync(mut self) -> Result<(), Report> {
// We can't download the genesis block using our normal algorithm,
// due to protocol limitations
self.request_genesis().await?;
loop {
if self.try_to_sync().await.is_err() {
self.downloads.cancel_all();
}
self.update_metrics();
info!(
timeout = ?SYNC_RESTART_DELAY,
state_tip = ?self.latest_chain_tip.best_tip_height(),
"waiting to restart sync"
);
sleep(SYNC_RESTART_DELAY).await;
}
}
/// Tries to synchronize the chain as far as it can.
///
/// Obtains some prospective tips and iteratively tries to extend them and download the missing
/// blocks.
///
/// Returns `Ok` if it was able to synchronize as much of the chain as it could, and then ran
/// out of prospective tips. This happens when synchronization finishes or if Zebra ended up
/// following a fork. Either way, Zebra should attempt to obtain some more tips.
///
/// Returns `Err` if there was an unrecoverable error and restarting the synchronization is
/// necessary.
#[instrument(skip(self))]
async fn try_to_sync(&mut self) -> Result<(), Report> {
self.prospective_tips = HashSet::new();
info!(
state_tip = ?self.latest_chain_tip.best_tip_height(),
"starting sync, obtaining new tips"
);
if let Err(e) = self.obtain_tips().await {
info!("temporary error obtaining tips: {:#}", e);
return Err(e);
}
self.update_metrics();
while !self.prospective_tips.is_empty() {
// Check whether any block tasks are currently ready:
while let Poll::Ready(Some(rsp)) = futures::poll!(self.downloads.next()) {
Self::handle_block_response(rsp)?;
}
self.update_metrics();
// If we have too many pending tasks, wait for some to finish.
//
// Starting to wait is interesting, but logging each wait can be
// very verbose.
if self.downloads.in_flight() > self.lookahead_limit {
tracing::info!(
tips.len = self.prospective_tips.len(),
in_flight = self.downloads.in_flight(),
lookahead_limit = self.lookahead_limit,
"waiting for pending blocks",
);
}
while self.downloads.in_flight() > self.lookahead_limit {
trace!(
tips.len = self.prospective_tips.len(),
in_flight = self.downloads.in_flight(),
lookahead_limit = self.lookahead_limit,
state_tip = ?self.latest_chain_tip.best_tip_height(),
"waiting for pending blocks",
);
let response = self.downloads.next().await.expect("downloads is nonempty");
Self::handle_block_response(response)?;
self.update_metrics();
}
// Once we're below the lookahead limit, we can keep extending the tips.
info!(
tips.len = self.prospective_tips.len(),
in_flight = self.downloads.in_flight(),
lookahead_limit = self.lookahead_limit,
state_tip = ?self.latest_chain_tip.best_tip_height(),
"extending tips",
);
if let Err(e) = self.extend_tips().await {
info!("temporary error extending tips: {:#}", e);
return Err(e);
}
self.update_metrics();
}
info!("exhausted prospective tip set");
Ok(())
}
/// Given a block_locator list fan out request for subsequent hashes to
/// multiple peers
#[instrument(skip(self))]
async fn obtain_tips(&mut self) -> Result<(), Report> {
let block_locator = self
.state
.ready()
.await
.map_err(|e| eyre!(e))?
.call(zebra_state::Request::BlockLocator)
.await
.map(|response| match response {
zebra_state::Response::BlockLocator(block_locator) => block_locator,
_ => unreachable!(
"GetBlockLocator request can only result in Response::BlockLocator"
),
})
.map_err(|e| eyre!(e))?;
debug!(
tip = ?block_locator.first().expect("we have at least one block locator object"),
?block_locator,
"got block locator and trying to obtain new chain tips"
);
let mut requests = FuturesUnordered::new();
for attempt in 0..FANOUT {
if attempt > 0 {
// Let other tasks run, so we're more likely to choose a different peer.
//
// TODO: move fanouts into the PeerSet, so we always choose different peers (#2214)
tokio::task::yield_now().await;
}
let ready_tip_network = self.tip_network.ready().await;
requests.push(tokio::spawn(ready_tip_network.map_err(|e| eyre!(e))?.call(
zn::Request::FindBlocks {
known_blocks: block_locator.clone(),
stop: None,
},
)));
}
let mut download_set = IndexSet::new();
while let Some(res) = requests.next().await {
match res
.expect("panic in spawned obtain tips request")
.map_err::<Report, _>(|e| eyre!(e))
{
Ok(zn::Response::BlockHashes(hashes)) => {
trace!(?hashes);
// zcashd sometimes appends an unrelated hash at the start
// or end of its response.
//
// We can't discard the first hash, because it might be a
// block we want to download. So we just accept any
// out-of-order first hashes.
// We use the last hash for the tip, and we want to avoid bad
// tips. So we discard the last hash. (We don't need to worry
// about missed downloads, because we will pick them up again
// in ExtendTips.)
let hashes = match hashes.as_slice() {
[] => continue,
[rest @ .., _last] => rest,
};
let mut first_unknown = None;
for (i, &hash) in hashes.iter().enumerate() {
if !self.state_contains(hash).await? {
first_unknown = Some(i);
break;
}
}
debug!(hashes.len = ?hashes.len(), ?first_unknown);
let unknown_hashes = if let Some(index) = first_unknown {
&hashes[index..]
} else {
continue;
};
trace!(?unknown_hashes);
let new_tip = if let Some(end) = unknown_hashes.rchunks_exact(2).next() {
CheckedTip {
tip: end[0],
expected_next: end[1],
}
} else {
debug!("discarding response that extends only one block");
continue;
};
// Make sure we get the same tips, regardless of the
// order of peer responses
if !download_set.contains(&new_tip.expected_next) {
debug!(?new_tip,
"adding new prospective tip, and removing existing tips in the new block hash list");
self.prospective_tips
.retain(|t| !unknown_hashes.contains(&t.expected_next));
self.prospective_tips.insert(new_tip);
} else {
debug!(
?new_tip,
"discarding prospective tip: already in download set"
);
}
// security: the first response determines our download order
//
// TODO: can we make the download order independent of response order?
let prev_download_len = download_set.len();
download_set.extend(unknown_hashes);
let new_download_len = download_set.len();
let new_hashes = new_download_len - prev_download_len;
debug!(new_hashes, "added hashes to download set");
metrics::histogram!("sync.obtain.response.hash.count", new_hashes as f64);
}
Ok(_) => unreachable!("network returned wrong response"),
// We ignore this error because we made multiple fanout requests.
Err(e) => debug!(?e),
}
}
debug!(?self.prospective_tips);
// Check that the new tips we got are actually unknown.
for hash in &download_set {
debug!(?hash, "checking if state contains hash");
if self.state_contains(*hash).await? {
return Err(eyre!("queued download of hash behind our chain tip"));
}
}
let new_downloads = download_set.len();
debug!(new_downloads, "queueing new downloads");
metrics::gauge!("sync.obtain.queued.hash.count", new_downloads as f64);
// security: use the actual number of new downloads from all peers,
// so the last peer to respond can't toggle our mempool
self.recent_syncs.push_obtain_tips_length(new_downloads);
let response = self.request_blocks(download_set).await;
Self::handle_response(response)?;
Ok(())
}
#[instrument(skip(self))]
async fn extend_tips(&mut self) -> Result<(), Report> {
let tips = std::mem::take(&mut self.prospective_tips);
let mut download_set = IndexSet::new();
debug!(tips = ?tips.len(), "trying to extend chain tips");
for tip in tips {
debug!(?tip, "asking peers to extend chain tip");
let mut responses = FuturesUnordered::new();
for attempt in 0..FANOUT {
if attempt > 0 {
// Let other tasks run, so we're more likely to choose a different peer.
//
// TODO: move fanouts into the PeerSet, so we always choose different peers (#2214)
tokio::task::yield_now().await;
}
let ready_tip_network = self.tip_network.ready().await;
responses.push(tokio::spawn(ready_tip_network.map_err(|e| eyre!(e))?.call(
zn::Request::FindBlocks {
known_blocks: vec![tip.tip],
stop: None,
},
)));
}
while let Some(res) = responses.next().await {
match res
.expect("panic in spawned extend tips request")
.map_err::<Report, _>(|e| eyre!(e))
{
Ok(zn::Response::BlockHashes(hashes)) => {
debug!(first = ?hashes.first(), len = ?hashes.len());
trace!(?hashes);
// zcashd sometimes appends an unrelated hash at the
// start or end of its response. Check the first hash
// against the previous response, and discard mismatches.
let unknown_hashes = match hashes.as_slice() {
[expected_hash, rest @ ..] if expected_hash == &tip.expected_next => {
rest
}
// If the first hash doesn't match, retry with the second.
[first_hash, expected_hash, rest @ ..]
if expected_hash == &tip.expected_next =>
{
debug!(?first_hash,
?tip.expected_next,
?tip.tip,
"unexpected first hash, but the second matches: using the hashes after the match");
rest
}
// We ignore these responses
[] => continue,
[single_hash] => {
debug!(?single_hash,
?tip.expected_next,
?tip.tip,
"discarding response containing a single unexpected hash");
continue;
}
[first_hash, second_hash, rest @ ..] => {
debug!(?first_hash,
?second_hash,
rest_len = ?rest.len(),
?tip.expected_next,
?tip.tip,
"discarding response that starts with two unexpected hashes");
continue;
}
};
// We use the last hash for the tip, and we want to avoid
// bad tips. So we discard the last hash. (We don't need
// to worry about missed downloads, because we will pick
// them up again in the next ExtendTips.)
let unknown_hashes = match unknown_hashes {
[] => continue,
[rest @ .., _last] => rest,
};
let new_tip = if let Some(end) = unknown_hashes.rchunks_exact(2).next() {
CheckedTip {
tip: end[0],
expected_next: end[1],
}
} else {
debug!("discarding response that extends only one block");
continue;
};
trace!(?unknown_hashes);
// Make sure we get the same tips, regardless of the
// order of peer responses
if !download_set.contains(&new_tip.expected_next) {
debug!(?new_tip,
"adding new prospective tip, and removing any existing tips in the new block hash list");
self.prospective_tips
.retain(|t| !unknown_hashes.contains(&t.expected_next));
self.prospective_tips.insert(new_tip);
} else {
debug!(
?new_tip,
"discarding prospective tip: already in download set"
);
}
// security: the first response determines our download order
//
// TODO: can we make the download order independent of response order?
let prev_download_len = download_set.len();
download_set.extend(unknown_hashes);
let new_download_len = download_set.len();
let new_hashes = new_download_len - prev_download_len;
debug!(new_hashes, "added hashes to download set");
metrics::histogram!("sync.extend.response.hash.count", new_hashes as f64);
}
Ok(_) => unreachable!("network returned wrong response"),
// We ignore this error because we made multiple fanout requests.
Err(e) => debug!(?e),
}
}
}
let new_downloads = download_set.len();
debug!(new_downloads, "queueing new downloads");
metrics::gauge!("sync.extend.queued.hash.count", new_downloads as f64);
// security: use the actual number of new downloads from all peers,
// so the last peer to respond can't toggle our mempool
self.recent_syncs.push_extend_tips_length(new_downloads);
let response = self.request_blocks(download_set).await;
Self::handle_response(response)?;
Ok(())
}
/// Download and verify the genesis block, if it isn't currently known to
/// our node.
async fn request_genesis(&mut self) -> Result<(), Report> {
// Due to Bitcoin protocol limitations, we can't request the genesis
// block using our standard tip-following algorithm:
// - getblocks requires at least one hash
// - responses start with the block *after* the requested block, and
// - the genesis hash is used as a placeholder for "no matches".
//
// So we just download and verify the genesis block here.
while !self.state_contains(self.genesis_hash).await? {
info!("starting genesis block download and verify");
let response = self.downloads.download_and_verify(self.genesis_hash).await;
Self::handle_response(response).map_err(|e| eyre!(e))?;
let response = self.downloads.next().await.expect("downloads is nonempty");
match response {
Ok(hash) => trace!(?hash, "verified and committed block to state"),
Err(error) => {
// TODO: exit syncer on permanent service errors (NetworkError, VerifierError)
if Self::should_restart_sync(&error) {
warn!(
?error,
"could not download or verify genesis block, retrying"
);
} else {
info!(
?error,
"temporary error downloading or verifying genesis block, retrying"
);
}
tokio::time::sleep(GENESIS_TIMEOUT_RETRY).await;
}
}
}
Ok(())
}
/// Queue download and verify tasks for each block that isn't currently known to our node
async fn request_blocks(
&mut self,
hashes: IndexSet<block::Hash>,
) -> Result<(), BlockDownloadVerifyError> {
debug!(hashes.len = hashes.len(), "requesting blocks");
for hash in hashes.into_iter() {
self.downloads.download_and_verify(hash).await?;
}
Ok(())
}
/// Handles a response for a requested block.
///
/// Returns `Ok` if the block was successfully verified and committed to the state, or if an
/// expected error occurred, so that the synchronization can continue normally.
///
/// Returns `Err` if an unexpected error occurred, to force the synchronizer to restart.
fn handle_block_response(
response: Result<block::Hash, BlockDownloadVerifyError>,
) -> Result<(), BlockDownloadVerifyError> {
match response {
Ok(hash) => trace!(?hash, "verified and committed block to state"),
Err(_) => return Self::handle_response(response.map(|_| ())),
}
Ok(())
}
/// Handles a response to a syncer request.
///
/// Returns `Ok` if the request was successful, or if an expected error occurred,
/// so that the synchronization can continue normally.
///
/// Returns `Err` if an unexpected error occurred, to force the synchronizer to restart.
fn handle_response(
response: Result<(), BlockDownloadVerifyError>,
) -> Result<(), BlockDownloadVerifyError> {
if let Err(error) = response {
// TODO: exit syncer on permanent service errors (NetworkError, VerifierError)
if Self::should_restart_sync(&error) {
return Err(error);
}
}
Ok(())
}
/// Returns `true` if the hash is present in the state, and `false`
/// if the hash is not present in the state.
///
/// BUG: check if the hash is in any chain (#862)
/// Depth only checks the main chain.
async fn state_contains(&mut self, hash: block::Hash) -> Result<bool, Report> {
match self
.state
.ready()
.await
.map_err(|e| eyre!(e))?
.call(zebra_state::Request::Depth(hash))
.await
.map_err(|e| eyre!(e))?
{
zs::Response::Depth(Some(_)) => Ok(true),
zs::Response::Depth(None) => Ok(false),
_ => unreachable!("wrong response to depth request"),
}
}
fn update_metrics(&mut self) {
metrics::gauge!(
"sync.prospective_tips.len",
self.prospective_tips.len() as f64,
);
metrics::gauge!(
"sync.downloads.in_flight",
self.downloads.in_flight() as f64,
);
}
/// Return if the sync should be restarted based on the given error
/// from the block downloader and verifier stream.
fn should_restart_sync(e: &BlockDownloadVerifyError) -> bool {
match e {
// Structural matches: downcasts
BlockDownloadVerifyError::Invalid {
error: VerifyChainError::Checkpoint(VerifyCheckpointError::AlreadyVerified { .. }),
..
} => {
debug!(error = ?e, "block was already verified, possibly from a previous sync run, continuing");
false
}
BlockDownloadVerifyError::Invalid {
error:
VerifyChainError::Block(VerifyBlockError::Block {
source: BlockError::AlreadyInChain(_, _),
}),
..
} => {
debug!(error = ?e, "block is already in chain, possibly from a previous sync run, continuing");
false
}
// Structural matches: direct
BlockDownloadVerifyError::CancelledDuringDownload { .. }
| BlockDownloadVerifyError::CancelledDuringVerification { .. } => {
debug!(error = ?e, "block verification was cancelled, continuing");
false
}
BlockDownloadVerifyError::BehindTipHeightLimit { .. } => {
debug!(
error = ?e,
"block height is behind the current state tip, \
assuming the syncer will eventually catch up to the state, continuing"
);
false
}
BlockDownloadVerifyError::DuplicateBlockQueuedForDownload { .. } => {
debug!(
error = ?e,
"queued duplicate block hash for download, \
assuming the syncer will eventually resolve duplicates, continuing"
);
false
}
// String matches
BlockDownloadVerifyError::Invalid {
error: VerifyChainError::Block(VerifyBlockError::Commit(ref source)),
..
} if format!("{:?}", source).contains("block is already committed to the state") => {
// TODO: improve this by checking the type (#2908)
debug!(error = ?e, "block is already committed, possibly from a previous sync run, continuing");
false
}
BlockDownloadVerifyError::DownloadFailed { ref error, .. }
if format!("{:?}", error).contains("NotFound") =>
{
// Covers these errors:
// - NotFoundResponse
// - NotFoundRegistry
//
// TODO: improve this by checking the type (#2908)
// restart after a certain number of NotFound errors?
debug!(error = ?e, "block was not found, possibly from a peer that doesn't have the block yet, continuing");
false
}
_ => {
// download_and_verify downcasts errors from the block verifier
// into VerifyChainError, and puts the result inside one of the
// BlockDownloadVerifyError enumerations. This downcast could
// become incorrect e.g. after some refactoring, and it is difficult
// to write a test to check it. The test below is a best-effort
// attempt to catch if that happens and log it.
//
// TODO: add a proper test and remove this
// https://github.com/ZcashFoundation/zebra/issues/2909
let err_str = format!("{:?}", e);
if err_str.contains("AlreadyVerified")
|| err_str.contains("AlreadyInChain")
|| err_str.contains("block is already committed to the state")
|| err_str.contains("NotFound")
{
error!(?e,
"a BlockDownloadVerifyError that should have been filtered out was detected, \
which possibly indicates a programming error in the downcast inside \
zebrad::components::sync::downloads::Downloads::download_and_verify"
)
}
warn!(?e, "error downloading and verifying block");
true
}
}
}
}