//! The scanner task and scanning APIs. use std::{collections::HashMap, sync::Arc, time::Duration}; use color_eyre::{eyre::eyre, Report}; use itertools::Itertools; use tower::{buffer::Buffer, util::BoxService, Service, ServiceExt}; use tracing::info; use zcash_client_backend::{ data_api::ScannedBlock, encoding::decode_extended_full_viewing_key, proto::compact_formats::{ ChainMetadata, CompactBlock, CompactSaplingOutput, CompactSaplingSpend, CompactTx, }, scanning::{ScanError, ScanningKey}, }; use zcash_primitives::{ constants::*, sapling::SaplingIvk, zip32::{AccountId, DiversifiableFullViewingKey, Scope}, }; use zebra_chain::{ block::Block, chain_tip::ChainTip, diagnostic::task::WaitForPanics, parameters::Network, serialization::ZcashSerialize, transaction::{self, Transaction}, }; use zebra_state::{ChainTipChange, SaplingScannedResult}; use crate::storage::{SaplingScanningKey, Storage}; /// The generic state type used by the scanner. pub type State = Buffer< BoxService, zebra_state::Request, >; /// Wait a few seconds at startup for some blocks to get verified. /// /// But sometimes the state might be empty if the network is slow. const INITIAL_WAIT: Duration = Duration::from_secs(15); /// The amount of time between checking for new blocks and starting new scans. /// /// This is just under half the target block interval. const CHECK_INTERVAL: Duration = Duration::from_secs(30); /// We log an info log with progress after this many blocks. const INFO_LOG_INTERVAL: u32 = 100_000; /// Start a scan task that reads blocks from `state`, scans them with the configured keys in /// `storage`, and then writes the results to `storage`. pub async fn start( mut state: State, chain_tip_change: ChainTipChange, storage: Storage, ) -> Result<(), Report> { let network = storage.network(); let mut height = storage.min_sapling_birthday_height(); // Read keys from the storage on disk, which can block async execution. let key_storage = storage.clone(); let key_birthdays = tokio::task::spawn_blocking(move || key_storage.sapling_keys()) .wait_for_panics() .await; // Parse and convert keys once, then use them to scan all blocks. // There is some cryptography here, but it should be fast even with thousands of keys. let parsed_keys: HashMap< SaplingScanningKey, (Vec, Vec), > = key_birthdays .keys() .map(|key| { let parsed_keys = sapling_key_to_scan_block_keys(key, network)?; Ok::<_, Report>((key.clone(), parsed_keys)) }) .try_collect()?; // Give empty states time to verify some blocks before we start scanning. tokio::time::sleep(INITIAL_WAIT).await; loop { // Get a block from the state. // We can't use ServiceExt::oneshot() here, because it causes lifetime errors in init(). let block = state .ready() .await .map_err(|e| eyre!(e))? .call(zebra_state::Request::Block(height.into())) .await .map_err(|e| eyre!(e))?; let block = match block { zebra_state::Response::Block(Some(block)) => block, zebra_state::Response::Block(None) => { // If we've reached the tip, sleep for a while then try and get the same block. tokio::time::sleep(CHECK_INTERVAL).await; continue; } _ => unreachable!("unmatched response to a state::Tip request"), }; // Only log at info level every 100,000 blocks let is_info_log = height == storage.min_sapling_birthday_height() || height.0 % INFO_LOG_INTERVAL == 0; // TODO: add debug logs? if is_info_log { info!( "Scanning the blockchain: now at block {:?}, current tip {:?}", height, chain_tip_change .latest_chain_tip() .best_tip_height_and_hash(), ); } for (key_num, (sapling_key, birthday_height)) in key_birthdays.iter().enumerate() { // # Security // // We can't log `sapling_key` here because it is a private viewing key. Anyone who reads // the logs could use the key to view those transactions. if is_info_log { info!( "Scanning the blockchain for key {}, started at block {:?}", key_num, birthday_height, ); } // Get the pre-parsed keys for this configured key. let (dfvks, ivks) = parsed_keys.get(sapling_key).cloned().unwrap_or_default(); // Scan the block, which blocks async execution until the scan is complete. // // TODO: skip scanning before birthday height (#8022) // TODO: scan each key in parallel (after MVP?) let sapling_key = sapling_key.clone(); let block = block.clone(); let mut storage = storage.clone(); // We use a dummy size of the Sapling note commitment tree. // // We can't set the size to zero, because the underlying scanning function would return // `zcash_client_backeng::scanning::ScanError::TreeSizeUnknown`. // // And we can't set them close to 0, because the scanner subtracts the number of notes // in the block, and panics with "attempt to subtract with overflow". The number of // notes in a block must be less than this value, this is a consensus rule. let sapling_tree_size = 1 << 16; tokio::task::spawn_blocking(move || { let dfvk_res = scan_block(network, &block, sapling_tree_size, &dfvks).map_err(|e| eyre!(e))?; let ivk_res = scan_block(network, &block, sapling_tree_size, &ivks).map_err(|e| eyre!(e))?; let dfvk_res = scanned_block_to_db_result(dfvk_res); let ivk_res = scanned_block_to_db_result(ivk_res); storage.add_sapling_result(sapling_key.clone(), height, dfvk_res); storage.add_sapling_result(sapling_key, height, ivk_res); Ok::<_, Report>(()) }) .wait_for_panics() .await?; } height = height .next() .expect("a valid blockchain never reaches the max height"); } } /// Returns transactions belonging to the given `ScanningKey`. This list of keys should come from /// a single configured `SaplingScanningKey`. /// /// # Performance / Hangs /// /// This method can block while reading database files, so it must be inside spawn_blocking() /// in async code. /// /// TODO: /// - Remove the `sapling_tree_size` parameter or turn it into an `Option` once we have access to /// Zebra's state, and we can retrieve the tree size ourselves. /// - Add prior block metadata once we have access to Zebra's state. pub fn scan_block( network: Network, block: &Arc, sapling_tree_size: u32, scanning_keys: &[K], ) -> Result, ScanError> { // TODO: Implement a check that returns early when the block height is below the Sapling // activation height. let network: zcash_primitives::consensus::Network = network.into(); let chain_metadata = ChainMetadata { sapling_commitment_tree_size: sapling_tree_size, // Orchard is not supported at the moment so the tree size can be 0. orchard_commitment_tree_size: 0, }; // Use a dummy `AccountId` as we don't use accounts yet. let dummy_account = AccountId::from(0); let scanning_keys: Vec<_> = scanning_keys .iter() .map(|key| (&dummy_account, key)) .collect(); zcash_client_backend::scanning::scan_block( &network, block_to_compact(block, chain_metadata), scanning_keys.as_slice(), // Ignore whether notes are change from a viewer's own spends for now. &[], // Ignore previous blocks for now. None, ) } /// Converts a Zebra-format scanning key into some `scan_block()` keys. /// /// Currently only accepts extended full viewing keys, and returns both their diversifiable full /// viewing key and their individual viewing key, for testing purposes. /// /// TODO: work out what string format is used for SaplingIvk, if any, and support it here /// performance: stop returning both the dfvk and ivk for the same key pub fn sapling_key_to_scan_block_keys( sapling_key: &SaplingScanningKey, network: Network, ) -> Result<(Vec, Vec), Report> { let hrp = if network.is_a_test_network() { // Assume custom testnets have the same HRP // // TODO: add the regtest HRP here testnet::HRP_SAPLING_EXTENDED_FULL_VIEWING_KEY } else { mainnet::HRP_SAPLING_EXTENDED_FULL_VIEWING_KEY }; let efvk = decode_extended_full_viewing_key(hrp, sapling_key).map_err(|e| eyre!(e))?; // Just return all the keys for now, so we can be sure our code supports them. let dfvk = efvk.to_diversifiable_full_viewing_key(); let eivk = dfvk.to_ivk(Scope::External); let iivk = dfvk.to_ivk(Scope::Internal); Ok((vec![dfvk], vec![eivk, iivk])) } /// Converts a zebra block and meta data into a compact block. pub fn block_to_compact(block: &Arc, chain_metadata: ChainMetadata) -> CompactBlock { CompactBlock { height: block .coinbase_height() .expect("verified block should have a valid height") .0 .into(), // TODO: performance: look up the block hash from the state rather than recalculating it hash: block.hash().bytes_in_display_order().to_vec(), prev_hash: block .header .previous_block_hash .bytes_in_display_order() .to_vec(), time: block .header .time .timestamp() .try_into() .expect("unsigned 32-bit times should work until 2105"), header: block .header .zcash_serialize_to_vec() .expect("verified block should serialize"), vtx: block .transactions .iter() .cloned() .enumerate() .map(transaction_to_compact) .collect(), chain_metadata: Some(chain_metadata), // The protocol version is used for the gRPC wire format, so it isn't needed here. proto_version: 0, } } /// Converts a zebra transaction into a compact transaction. fn transaction_to_compact((index, tx): (usize, Arc)) -> CompactTx { CompactTx { index: index .try_into() .expect("tx index in block should fit in u64"), // TODO: performance: look up the tx hash from the state rather than recalculating it hash: tx.hash().bytes_in_display_order().to_vec(), // `fee` is not checked by the `scan_block` function. It is allowed to be unset. // fee: 0, spends: tx .sapling_nullifiers() .map(|nf| CompactSaplingSpend { nf: <[u8; 32]>::from(*nf).to_vec(), }) .collect(), // > output encodes the cmu field, ephemeralKey field, and a 52-byte prefix of the encCiphertext field of a Sapling Output // // outputs: tx .sapling_outputs() .map(|output| CompactSaplingOutput { cmu: output.cm_u.to_bytes().to_vec(), ephemeral_key: output .ephemeral_key .zcash_serialize_to_vec() .expect("verified output should serialize successfully"), ciphertext: output .enc_ciphertext .zcash_serialize_to_vec() .expect("verified output should serialize successfully") .into_iter() .take(52) .collect(), }) .collect(), // `actions` is not checked by the `scan_block` function. actions: vec![], } } /// Convert a scanned block to a list of scanner database results. fn scanned_block_to_db_result(scanned_block: ScannedBlock) -> Vec { scanned_block .transactions() .iter() .map(|tx| transaction::Hash::from_bytes_in_display_order(tx.txid.as_ref())) .collect() }