openethereum/ethcore/sync/src/block_sync.rs

994 lines
33 KiB
Rust
Raw Normal View History

// Copyright 2015-2018 Parity Technologies (UK) Ltd.
// This file is part of Parity.
// Parity is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
// Parity is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
// You should have received a copy of the GNU General Public License
// along with Parity. If not, see <http://www.gnu.org/licenses/>.
///
/// Blockchain downloader
///
2017-07-29 17:12:07 +02:00
use std::collections::{HashSet, VecDeque};
use std::cmp;
use heapsize::HeapSizeOf;
use ethereum_types::H256;
use rlp::{self, Rlp};
use ethcore::header::BlockNumber;
use ethcore::client::{BlockStatus, BlockId};
use ethcore::error::{ImportErrorKind, QueueErrorKind, BlockError, Error as EthcoreError, ErrorKind as EthcoreErrorKind};
use sync_io::SyncIo;
use blocks::{BlockCollection, SyncBody, SyncHeader};
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
use chain::BlockSet;
const MAX_HEADERS_TO_REQUEST: usize = 128;
const MAX_BODIES_TO_REQUEST: usize = 32;
const MAX_RECEPITS_TO_REQUEST: usize = 128;
const SUBCHAIN_SIZE: u64 = 256;
const MAX_ROUND_PARENTS: usize = 16;
2016-11-16 19:34:12 +01:00
const MAX_PARALLEL_SUBCHAIN_DOWNLOAD: usize = 5;
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
const MAX_USELESS_HEADERS_PER_ROUND: usize = 3;
// logging macros prepend BlockSet context for log filtering
macro_rules! trace_sync {
($self:ident, $fmt:expr, $($arg:tt)+) => {
trace!(target: "sync", concat!("{:?}: ", $fmt), $self.block_set, $($arg)+);
};
($self:ident, $fmt:expr) => {
trace!(target: "sync", concat!("{:?}: ", $fmt), $self.block_set);
};
}
macro_rules! debug_sync {
($self:ident, $fmt:expr, $($arg:tt)+) => {
debug!(target: "sync", concat!("{:?}: ", $fmt), $self.block_set, $($arg)+);
};
($self:ident, $fmt:expr) => {
debug!(target: "sync", concat!("{:?}: ", $fmt), $self.block_set);
};
}
#[derive(Copy, Clone, Eq, PartialEq, Debug)]
/// Downloader state
pub enum State {
/// No active downloads.
Idle,
/// Downloading subchain heads
ChainHead,
/// Downloading blocks
Blocks,
/// Download is complete
Complete,
}
/// Data that needs to be requested from a peer.
pub enum BlockRequest {
Headers {
start: H256,
count: u64,
skip: u64,
},
Bodies {
hashes: Vec<H256>,
},
Receipts {
hashes: Vec<H256>,
},
}
2016-11-16 19:34:12 +01:00
/// Indicates sync action
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
#[derive(Eq, PartialEq, Debug)]
2016-11-16 19:34:12 +01:00
pub enum DownloadAction {
/// Do nothing
None,
/// Reset downloads for all peers
Reset
}
#[derive(Eq, PartialEq, Debug)]
pub enum BlockDownloaderImportError {
/// Imported data is rejected as invalid. Peer should be dropped.
Invalid,
/// Imported data is valid but rejected cause the downloader does not need it.
Useless,
}
impl From<rlp::DecoderError> for BlockDownloaderImportError {
fn from(_: rlp::DecoderError) -> BlockDownloaderImportError {
BlockDownloaderImportError::Invalid
}
}
/// Block downloader strategy.
/// Manages state and block data for a block download process.
pub struct BlockDownloader {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
/// Which set of blocks to download
block_set: BlockSet,
/// Downloader state
state: State,
/// Highest block number seen
highest_block: Option<BlockNumber>,
/// Downloaded blocks, holds `H`, `B` and `S`
blocks: BlockCollection,
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
/// Last imported block number
last_imported_block: BlockNumber,
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
/// Last imported block hash
last_imported_hash: H256,
/// Number of blocks imported this round
imported_this_round: Option<usize>,
/// Block number the last round started with.
last_round_start: BlockNumber,
last_round_start_hash: H256,
/// Block parents imported this round (hash, parent)
round_parents: VecDeque<(H256, H256)>,
/// Do we need to download block recetips.
download_receipts: bool,
/// Sync up to the block with this hash.
target_hash: Option<H256>,
/// Probing range for seeking common best block.
retract_step: u64,
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
/// consecutive useless headers this round
useless_headers_count: usize,
}
impl BlockDownloader {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
/// Create a new instance of syncing strategy.
/// For BlockSet::NewBlocks this won't reorganize to before the last kept state.
pub fn new(block_set: BlockSet, start_hash: &H256, start_number: BlockNumber) -> Self {
let sync_receipts = match block_set {
BlockSet::NewBlocks => false,
BlockSet::OldBlocks => true
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
};
BlockDownloader {
block_set,
state: State::Idle,
highest_block: None,
last_imported_block: start_number,
last_imported_hash: start_hash.clone(),
last_round_start: start_number,
last_round_start_hash: start_hash.clone(),
blocks: BlockCollection::new(sync_receipts),
imported_this_round: None,
round_parents: VecDeque::new(),
download_receipts: sync_receipts,
target_hash: None,
retract_step: 1,
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
useless_headers_count: 0,
}
}
/// Reset sync. Clear all local downloaded data.
pub fn reset(&mut self) {
self.blocks.clear();
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
self.useless_headers_count = 0;
self.state = State::Idle;
}
/// Mark a block as known in the chain
pub fn mark_as_known(&mut self, hash: &H256, number: BlockNumber) {
if number >= self.last_imported_block + 1 {
self.last_imported_block = number;
self.last_imported_hash = hash.clone();
self.imported_this_round = Some(self.imported_this_round.unwrap_or(0) + 1);
self.last_round_start = number;
self.last_round_start_hash = hash.clone();
}
}
/// Check if download is complete
pub fn is_complete(&self) -> bool {
self.state == State::Complete
}
/// Check if particular block hash is being downloaded
pub fn is_downloading(&self, hash: &H256) -> bool {
self.blocks.is_downloading(hash)
}
/// Set starting sync block
pub fn set_target(&mut self, hash: &H256) {
self.target_hash = Some(hash.clone());
}
/// Unmark header as being downloaded.
pub fn clear_header_download(&mut self, hash: &H256) {
self.blocks.clear_header_download(hash)
}
/// Unmark block body as being downloaded.
pub fn clear_body_download(&mut self, hashes: &[H256]) {
self.blocks.clear_body_download(hashes)
}
/// Unmark block receipt as being downloaded.
pub fn clear_receipt_download(&mut self, hashes: &[H256]) {
self.blocks.clear_receipt_download(hashes)
}
/// Reset collection for a new sync round with given subchain block hashes.
pub fn reset_to(&mut self, hashes: Vec<H256>) {
self.reset();
self.blocks.reset_to(hashes);
self.state = State::Blocks;
}
/// Returns used heap memory size.
pub fn heap_size(&self) -> usize {
self.blocks.heap_size() + self.round_parents.heap_size_of_children()
}
/// Returns best imported block number.
pub fn last_imported_block_number(&self) -> BlockNumber {
self.last_imported_block
}
/// Add new block headers.
pub fn import_headers(&mut self, io: &mut SyncIo, r: &Rlp, expected_hash: H256) -> Result<DownloadAction, BlockDownloaderImportError> {
let item_count = r.item_count().unwrap_or(0);
if self.state == State::Idle {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Ignored unexpected block headers");
2016-11-16 19:34:12 +01:00
return Ok(DownloadAction::None)
}
if item_count == 0 && (self.state == State::Blocks) {
return Err(BlockDownloaderImportError::Invalid);
}
// The request is generated in ::request_blocks.
let (max_count, skip) = if self.state == State::ChainHead {
(SUBCHAIN_SIZE as usize, (MAX_HEADERS_TO_REQUEST - 2) as u64)
} else {
(MAX_HEADERS_TO_REQUEST, 0)
};
if item_count > max_count {
debug!(target: "sync", "Headers response is larger than expected");
return Err(BlockDownloaderImportError::Invalid);
}
let mut headers = Vec::new();
let mut hashes = Vec::new();
let mut last_header = None;
for i in 0..item_count {
let info = SyncHeader::from_rlp(r.at(i)?.as_raw().to_vec())?;
let number = BlockNumber::from(info.header.number());
let hash = info.header.hash();
let valid_response = match last_header {
// First header must match expected hash.
None => expected_hash == hash,
Some((last_number, last_hash)) => {
// Subsequent headers must be spaced by skip interval.
let skip_valid = number == last_number + skip + 1;
// Consecutive headers must be linked by parent hash.
let parent_valid = (number != last_number + 1) || *info.header.parent_hash() == last_hash;
skip_valid && parent_valid
}
};
// Disable the peer for this syncing round if it gives invalid chain
if !valid_response {
debug!(target: "sync", "Invalid headers response");
return Err(BlockDownloaderImportError::Invalid);
}
last_header = Some((number, hash));
if self.blocks.contains(&hash) {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Skipping existing block header {} ({:?})", number, hash);
continue;
}
match io.chain().block_status(BlockId::Hash(hash.clone())) {
BlockStatus::InChain | BlockStatus::Queued => {
match self.state {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
State::Blocks => trace_sync!(self, "Header already in chain {} ({})", number, hash),
_ => trace_sync!(self, "Header already in chain {} ({}), state = {:?}", number, hash, self.state),
}
headers.push(info);
hashes.push(hash);
},
BlockStatus::Bad => {
return Err(BlockDownloaderImportError::Invalid);
},
BlockStatus::Unknown => {
headers.push(info);
hashes.push(hash);
}
}
}
if let Some((number, _)) = last_header {
if self.highest_block.as_ref().map_or(true, |n| number > *n) {
self.highest_block = Some(number);
}
}
match self.state {
State::ChainHead => {
if !headers.is_empty() {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Received {} subchain heads, proceeding to download", headers.len());
self.blocks.reset_to(hashes);
self.state = State::Blocks;
2016-11-16 19:34:12 +01:00
return Ok(DownloadAction::Reset);
2016-11-28 16:30:36 +01:00
} else {
trace_sync!(self, "No useful subchain heads received, expected hash {:?}", expected_hash);
2016-11-28 16:30:36 +01:00
let best = io.chain().chain_info().best_block_number;
let oldest_reorg = io.chain().pruning_info().earliest_state;
let last = self.last_imported_block;
match self.block_set {
BlockSet::NewBlocks if best > last && (last == 0 || last < oldest_reorg) => {
trace_sync!(self, "No common block, disabling peer");
return Err(BlockDownloaderImportError::Invalid)
},
BlockSet::OldBlocks => {
trace_sync!(self, "Expected some useful headers for downloading OldBlocks. Try a different peer");
return Err(BlockDownloaderImportError::Useless)
},
_ => (),
2016-11-28 16:30:36 +01:00
}
}
},
State::Blocks => {
let count = headers.len();
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
// At least one of the headers must advance the subchain. Otherwise they are all useless.
if count == 0 {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
self.useless_headers_count += 1;
trace_sync!(self, "No useful headers ({:?} this round), expected hash {:?}", self.useless_headers_count, expected_hash);
// only reset download if we have multiple subchain heads, to avoid unnecessary resets
// when we are at the head of the chain when we may legitimately receive no useful headers
if self.blocks.heads_len() > 1 && self.useless_headers_count >= MAX_USELESS_HEADERS_PER_ROUND {
trace_sync!(self, "Received {:?} useless responses this round. Resetting sync", MAX_USELESS_HEADERS_PER_ROUND);
self.reset();
}
2016-11-16 19:34:12 +01:00
return Err(BlockDownloaderImportError::Useless);
}
self.blocks.insert_headers(headers);
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Inserted {} headers", count);
},
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
_ => trace_sync!(self, "Unexpected headers({})", headers.len()),
}
2016-11-16 19:34:12 +01:00
Ok(DownloadAction::None)
}
/// Called by peer once it has new block bodies
pub fn import_bodies(&mut self, r: &Rlp, expected_hashes: &[H256]) -> Result<(), BlockDownloaderImportError> {
let item_count = r.item_count().unwrap_or(0);
if item_count == 0 {
return Err(BlockDownloaderImportError::Useless);
} else if self.state != State::Blocks {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Ignored unexpected block bodies");
} else {
let mut bodies = Vec::with_capacity(item_count);
for i in 0..item_count {
let body = SyncBody::from_rlp(r.at(i)?.as_raw())?;
bodies.push(body);
}
let hashes = self.blocks.insert_bodies(bodies);
if hashes.len() != item_count {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Deactivating peer for giving invalid block bodies");
return Err(BlockDownloaderImportError::Invalid);
}
if !all_expected(hashes.as_slice(), expected_hashes, |&a, &b| a == b) {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Deactivating peer for giving unexpected block bodies");
return Err(BlockDownloaderImportError::Invalid);
}
}
Ok(())
}
/// Called by peer once it has new block bodies
pub fn import_receipts(&mut self, r: &Rlp, expected_hashes: &[H256]) -> Result<(), BlockDownloaderImportError> {
let item_count = r.item_count().unwrap_or(0);
if item_count == 0 {
return Err(BlockDownloaderImportError::Useless);
}
else if self.state != State::Blocks {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Ignored unexpected block receipts");
}
else {
let mut receipts = Vec::with_capacity(item_count);
for i in 0..item_count {
let receipt = r.at(i).map_err(|e| {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Error decoding block receipts RLP: {:?}", e);
BlockDownloaderImportError::Invalid
})?;
receipts.push(receipt.as_raw().to_vec());
}
let hashes = self.blocks.insert_receipts(receipts);
if hashes.len() != item_count {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Deactivating peer for giving invalid block receipts");
return Err(BlockDownloaderImportError::Invalid);
}
if !all_expected(hashes.as_slice(), expected_hashes, |a, b| a.contains(b)) {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Deactivating peer for giving unexpected block receipts");
return Err(BlockDownloaderImportError::Invalid);
}
}
Ok(())
}
fn start_sync_round(&mut self, io: &mut SyncIo) {
self.state = State::ChainHead;
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Starting round (last imported count = {:?}, last started = {}, block = {:?}", self.imported_this_round, self.last_round_start, self.last_imported_block);
// Check if need to retract to find the common block. The problem is that the peers still return headers by hash even
// from the non-canonical part of the tree. So we also retract if nothing has been imported last round.
let start = self.last_round_start;
let start_hash = self.last_round_start_hash;
match self.imported_this_round {
Some(n) if n == 0 && start > 0 => {
// nothing was imported last round, step back to a previous block
// search parent in last round known parents first
if let Some(&(_, p)) = self.round_parents.iter().find(|&&(h, _)| h == start_hash) {
self.last_imported_block = start - 1;
self.last_imported_hash = p.clone();
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Searching common header from the last round {} ({})", self.last_imported_block, self.last_imported_hash);
} else {
let best = io.chain().chain_info().best_block_number;
let oldest_reorg = io.chain().pruning_info().earliest_state;
if self.block_set == BlockSet::NewBlocks && best > start && start < oldest_reorg {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
debug_sync!(self, "Could not revert to previous ancient block, last: {} ({})", start, start_hash);
self.reset();
} else {
2017-07-29 17:12:07 +02:00
let n = start - cmp::min(self.retract_step, start);
self.retract_step *= 2;
match io.chain().block_hash(BlockId::Number(n)) {
Some(h) => {
self.last_imported_block = n;
self.last_imported_hash = h;
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Searching common header in the blockchain {} ({})", start, self.last_imported_hash);
}
None => {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
debug_sync!(self, "Could not revert to previous block, last: {} ({})", start, self.last_imported_hash);
self.reset();
}
}
}
}
},
_ => {
self.retract_step = 1;
},
}
self.last_round_start = self.last_imported_block;
self.last_round_start_hash = self.last_imported_hash;
self.imported_this_round = None;
}
/// Find some headers or blocks to download for a peer.
2016-11-16 19:34:12 +01:00
pub fn request_blocks(&mut self, io: &mut SyncIo, num_active_peers: usize) -> Option<BlockRequest> {
match self.state {
State::Idle => {
self.start_sync_round(io);
if self.state == State::ChainHead {
return self.request_blocks(io, num_active_peers);
}
},
State::ChainHead => {
2016-11-16 19:34:12 +01:00
if num_active_peers < MAX_PARALLEL_SUBCHAIN_DOWNLOAD {
// Request subchain headers
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Starting sync with better chain");
2016-11-16 19:34:12 +01:00
// Request MAX_HEADERS_TO_REQUEST - 2 headers apart so that
// MAX_HEADERS_TO_REQUEST would include headers for neighbouring subchains
return Some(BlockRequest::Headers {
start: self.last_imported_hash.clone(),
count: SUBCHAIN_SIZE,
skip: (MAX_HEADERS_TO_REQUEST - 2) as u64,
});
}
},
State::Blocks => {
// check to see if we need to download any block bodies first
let needed_bodies = self.blocks.needed_bodies(MAX_BODIES_TO_REQUEST, false);
if !needed_bodies.is_empty() {
return Some(BlockRequest::Bodies {
hashes: needed_bodies,
});
}
if self.download_receipts {
let needed_receipts = self.blocks.needed_receipts(MAX_RECEPITS_TO_REQUEST, false);
if !needed_receipts.is_empty() {
return Some(BlockRequest::Receipts {
hashes: needed_receipts,
});
}
}
// find subchain to download
if let Some((h, count)) = self.blocks.needed_headers(MAX_HEADERS_TO_REQUEST, false) {
return Some(BlockRequest::Headers {
start: h,
count: count as u64,
skip: 0,
});
}
},
State::Complete => (),
}
None
}
/// Checks if there are blocks fully downloaded that can be imported into the blockchain and does the import.
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
/// Returns DownloadAction::Reset if it is imported all the the blocks it can and all downloading peers should be reset
pub fn collect_blocks(&mut self, io: &mut SyncIo, allow_out_of_order: bool) -> DownloadAction {
let mut download_action = DownloadAction::None;
let mut imported = HashSet::new();
let blocks = self.blocks.drain();
let count = blocks.len();
for block_and_receipts in blocks {
let block = block_and_receipts.block;
let receipts = block_and_receipts.receipts;
let h = block.header.hash();
let number = block.header.number();
let parent = *block.header.parent_hash();
if self.target_hash.as_ref().map_or(false, |t| t == &h) {
self.state = State::Complete;
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Sync target reached");
return download_action;
}
let result = if let Some(receipts) = receipts {
io.chain().queue_ancient_block(block, receipts)
} else {
io.chain().import_block(block)
};
match result {
Err(EthcoreError(EthcoreErrorKind::Import(ImportErrorKind::AlreadyInChain), _)) => {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Block already in chain {:?}", h);
self.block_imported(&h, number, &parent);
},
Err(EthcoreError(EthcoreErrorKind::Import(ImportErrorKind::AlreadyQueued), _)) => {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Block already queued {:?}", h);
self.block_imported(&h, number, &parent);
},
Ok(_) => {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Block queued {:?}", h);
imported.insert(h.clone());
self.block_imported(&h, number, &parent);
},
Err(EthcoreError(EthcoreErrorKind::Block(BlockError::UnknownParent(_)), _)) if allow_out_of_order => {
break;
},
Err(EthcoreError(EthcoreErrorKind::Block(BlockError::UnknownParent(_)), _)) => {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Unknown new block parent, restarting sync");
break;
},
Err(EthcoreError(EthcoreErrorKind::Block(BlockError::TemporarilyInvalid(_)), _)) => {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
debug_sync!(self, "Block temporarily invalid: {:?}, restarting sync", h);
break;
},
Err(EthcoreError(EthcoreErrorKind::Queue(QueueErrorKind::Full(limit)), _)) => {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
debug_sync!(self, "Block import queue full ({}), restarting sync", limit);
download_action = DownloadAction::Reset;
break;
},
Err(e) => {
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
debug_sync!(self, "Bad block {:?} : {:?}", h, e);
download_action = DownloadAction::Reset;
break;
}
}
}
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Imported {} of {}", imported.len(), count);
self.imported_this_round = Some(self.imported_this_round.unwrap_or(0) + imported.len());
if self.blocks.is_empty() {
// complete sync round
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
trace_sync!(self, "Sync round complete");
download_action = DownloadAction::Reset;
}
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
download_action
}
fn block_imported(&mut self, hash: &H256, number: BlockNumber, parent: &H256) {
self.last_imported_block = number;
self.last_imported_hash = hash.clone();
self.round_parents.push_back((hash.clone(), parent.clone()));
if self.round_parents.len() > MAX_ROUND_PARENTS {
self.round_parents.pop_front();
}
}
}
// Determines if the first argument matches an ordered subset of the second, according to some predicate.
fn all_expected<A, B, F>(values: &[A], expected_values: &[B], is_expected: F) -> bool
where F: Fn(&A, &B) -> bool
{
let mut expected_iter = expected_values.iter();
values.iter().all(|val1| {
while let Some(val2) = expected_iter.next() {
if is_expected(val1, val2) {
return true;
}
}
false
})
}
#[cfg(test)]
mod tests {
use super::*;
use ethcore::client::TestBlockChainClient;
use ethcore::header::Header as BlockHeader;
use ethcore::spec::Spec;
use ethkey::{Generator,Random};
use hash::keccak;
use parking_lot::RwLock;
use rlp::{encode_list,RlpStream};
use tests::helpers::TestIo;
use tests::snapshot::TestSnapshotService;
use transaction::{Transaction,SignedTransaction};
use triehash_ethereum::ordered_trie_root;
fn dummy_header(number: u64, parent_hash: H256) -> BlockHeader {
let mut header = BlockHeader::new();
header.set_gas_limit(0.into());
header.set_difficulty((number * 100).into());
header.set_timestamp(number * 10);
header.set_number(number);
header.set_parent_hash(parent_hash);
header.set_state_root(H256::zero());
header
}
fn dummy_signed_tx() -> SignedTransaction {
let keypair = Random.generate().unwrap();
Transaction::default().sign(keypair.secret(), None)
}
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
fn import_headers(headers: &[BlockHeader], downloader: &mut BlockDownloader, io: &mut SyncIo) -> Result<DownloadAction, BlockDownloaderImportError> {
let mut stream = RlpStream::new();
stream.append_list(headers);
let bytes = stream.out();
let rlp = Rlp::new(&bytes);
let expected_hash = headers.first().unwrap().hash();
downloader.import_headers(io, &rlp, expected_hash)
}
fn import_headers_ok(headers: &[BlockHeader], downloader: &mut BlockDownloader, io: &mut SyncIo) {
let res = import_headers(headers, downloader, io);
assert!(res.is_ok());
}
#[test]
fn import_headers_in_chain_head_state() {
::env_logger::try_init().ok();
let spec = Spec::new_test();
let genesis_hash = spec.genesis_header().hash();
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
let mut downloader = BlockDownloader::new(BlockSet::NewBlocks, &genesis_hash, 0);
downloader.state = State::ChainHead;
let mut chain = TestBlockChainClient::new();
let snapshot_service = TestSnapshotService::new();
let queue = RwLock::new(VecDeque::new());
let mut io = TestIo::new(&mut chain, &snapshot_service, &queue, None);
// Valid headers sequence.
let valid_headers = [
spec.genesis_header(),
dummy_header(127, H256::random()),
dummy_header(254, H256::random()),
];
let rlp_data = encode_list(&valid_headers);
let valid_rlp = Rlp::new(&rlp_data);
match downloader.import_headers(&mut io, &valid_rlp, genesis_hash) {
Ok(DownloadAction::Reset) => assert_eq!(downloader.state, State::Blocks),
_ => panic!("expected transition to Blocks state"),
};
// Headers are rejected because the expected hash does not match.
let invalid_start_block_headers = [
dummy_header(0, H256::random()),
dummy_header(127, H256::random()),
dummy_header(254, H256::random()),
];
let rlp_data = encode_list(&invalid_start_block_headers);
let invalid_start_block_rlp = Rlp::new(&rlp_data);
match downloader.import_headers(&mut io, &invalid_start_block_rlp, genesis_hash) {
Err(BlockDownloaderImportError::Invalid) => (),
_ => panic!("expected BlockDownloaderImportError"),
};
// Headers are rejected because they are not spaced as expected.
let invalid_skip_headers = [
spec.genesis_header(),
dummy_header(128, H256::random()),
dummy_header(256, H256::random()),
];
let rlp_data = encode_list(&invalid_skip_headers);
let invalid_skip_rlp = Rlp::new(&rlp_data);
match downloader.import_headers(&mut io, &invalid_skip_rlp, genesis_hash) {
Err(BlockDownloaderImportError::Invalid) => (),
_ => panic!("expected BlockDownloaderImportError"),
};
// Invalid because the packet size is too large.
let mut too_many_headers = Vec::with_capacity((SUBCHAIN_SIZE + 1) as usize);
too_many_headers.push(spec.genesis_header());
for i in 1..(SUBCHAIN_SIZE + 1) {
too_many_headers.push(dummy_header((MAX_HEADERS_TO_REQUEST as u64 - 1) * i, H256::random()));
}
let rlp_data = encode_list(&too_many_headers);
let too_many_rlp = Rlp::new(&rlp_data);
match downloader.import_headers(&mut io, &too_many_rlp, genesis_hash) {
Err(BlockDownloaderImportError::Invalid) => (),
_ => panic!("expected BlockDownloaderImportError"),
};
}
#[test]
fn import_headers_in_blocks_state() {
::env_logger::try_init().ok();
let mut chain = TestBlockChainClient::new();
let snapshot_service = TestSnapshotService::new();
let queue = RwLock::new(VecDeque::new());
let mut io = TestIo::new(&mut chain, &snapshot_service, &queue, None);
let mut headers = Vec::with_capacity(3);
let parent_hash = H256::random();
headers.push(dummy_header(127, parent_hash));
let parent_hash = headers[0].hash();
headers.push(dummy_header(128, parent_hash));
let parent_hash = headers[1].hash();
headers.push(dummy_header(129, parent_hash));
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
let mut downloader = BlockDownloader::new(BlockSet::NewBlocks, &H256::random(), 0);
downloader.state = State::Blocks;
downloader.blocks.reset_to(vec![headers[0].hash()]);
let rlp_data = encode_list(&headers);
let headers_rlp = Rlp::new(&rlp_data);
match downloader.import_headers(&mut io, &headers_rlp, headers[0].hash()) {
Ok(DownloadAction::None) => (),
_ => panic!("expected successful import"),
};
// Invalidate parent_hash link.
headers[2] = dummy_header(129, H256::random());
let rlp_data = encode_list(&headers);
let headers_rlp = Rlp::new(&rlp_data);
match downloader.import_headers(&mut io, &headers_rlp, headers[0].hash()) {
Err(BlockDownloaderImportError::Invalid) => (),
_ => panic!("expected BlockDownloaderImportError"),
};
// Invalidate header sequence by skipping a header.
headers[2] = dummy_header(130, headers[1].hash());
let rlp_data = encode_list(&headers);
let headers_rlp = Rlp::new(&rlp_data);
match downloader.import_headers(&mut io, &headers_rlp, headers[0].hash()) {
Err(BlockDownloaderImportError::Invalid) => (),
_ => panic!("expected BlockDownloaderImportError"),
};
}
#[test]
fn import_bodies() {
::env_logger::try_init().ok();
let mut chain = TestBlockChainClient::new();
let snapshot_service = TestSnapshotService::new();
let queue = RwLock::new(VecDeque::new());
let mut io = TestIo::new(&mut chain, &snapshot_service, &queue, None);
// Import block headers.
let mut headers = Vec::with_capacity(4);
let mut bodies = Vec::with_capacity(4);
let mut parent_hash = H256::zero();
for i in 0..4 {
// Construct the block body
let mut uncles = if i > 0 {
encode_list(&[dummy_header(i - 1, H256::random())])
} else {
::rlp::EMPTY_LIST_RLP.to_vec()
};
let mut txs = encode_list(&[dummy_signed_tx()]);
let tx_root = ordered_trie_root(Rlp::new(&txs).iter().map(|r| r.as_raw()));
let mut rlp = RlpStream::new_list(2);
rlp.append_raw(&txs, 1);
rlp.append_raw(&uncles, 1);
bodies.push(rlp.out());
// Construct the block header
let mut header = dummy_header(i, parent_hash);
header.set_transactions_root(tx_root);
header.set_uncles_hash(keccak(&uncles));
parent_hash = header.hash();
headers.push(header);
}
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
let mut downloader = BlockDownloader::new(BlockSet::NewBlocks, &headers[0].hash(), 0);
downloader.state = State::Blocks;
downloader.blocks.reset_to(vec![headers[0].hash()]);
// Only import the first three block headers.
let rlp_data = encode_list(&headers[0..3]);
let headers_rlp = Rlp::new(&rlp_data);
assert!(downloader.import_headers(&mut io, &headers_rlp, headers[0].hash()).is_ok());
// Import first body successfully.
let mut rlp_data = RlpStream::new_list(1);
rlp_data.append_raw(&bodies[0], 1);
let bodies_rlp = Rlp::new(rlp_data.as_raw());
assert!(downloader.import_bodies(&bodies_rlp, &[headers[0].hash(), headers[1].hash()]).is_ok());
// Import second body successfully.
let mut rlp_data = RlpStream::new_list(1);
rlp_data.append_raw(&bodies[1], 1);
let bodies_rlp = Rlp::new(rlp_data.as_raw());
assert!(downloader.import_bodies(&bodies_rlp, &[headers[0].hash(), headers[1].hash()]).is_ok());
// Import unexpected third body.
let mut rlp_data = RlpStream::new_list(1);
rlp_data.append_raw(&bodies[2], 1);
let bodies_rlp = Rlp::new(rlp_data.as_raw());
match downloader.import_bodies(&bodies_rlp, &[headers[0].hash(), headers[1].hash()]) {
Err(BlockDownloaderImportError::Invalid) => (),
_ => panic!("expected BlockDownloaderImportError"),
};
}
#[test]
fn import_receipts() {
::env_logger::try_init().ok();
let mut chain = TestBlockChainClient::new();
let snapshot_service = TestSnapshotService::new();
let queue = RwLock::new(VecDeque::new());
let mut io = TestIo::new(&mut chain, &snapshot_service, &queue, None);
// Import block headers.
let mut headers = Vec::with_capacity(4);
let mut receipts = Vec::with_capacity(4);
let mut parent_hash = H256::zero();
for i in 0..4 {
// Construct the receipts. Receipt root for the first two blocks is the same.
//
// The RLP-encoded integers are clearly not receipts, but the BlockDownloader treats
// all receipts as byte blobs, so it does not matter.
let mut receipts_rlp = if i < 2 {
encode_list(&[0u32])
} else {
encode_list(&[i as u32])
};
let receipts_root = ordered_trie_root(Rlp::new(&receipts_rlp).iter().map(|r| r.as_raw()));
receipts.push(receipts_rlp);
// Construct the block header.
let mut header = dummy_header(i, parent_hash);
header.set_receipts_root(receipts_root);
parent_hash = header.hash();
headers.push(header);
}
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
let mut downloader = BlockDownloader::new(BlockSet::OldBlocks, &headers[0].hash(), 0);
downloader.state = State::Blocks;
downloader.blocks.reset_to(vec![headers[0].hash()]);
// Only import the first three block headers.
let rlp_data = encode_list(&headers[0..3]);
let headers_rlp = Rlp::new(&rlp_data);
assert!(downloader.import_headers(&mut io, &headers_rlp, headers[0].hash()).is_ok());
// Import second and third receipts successfully.
let mut rlp_data = RlpStream::new_list(2);
rlp_data.append_raw(&receipts[1], 1);
rlp_data.append_raw(&receipts[2], 1);
let receipts_rlp = Rlp::new(rlp_data.as_raw());
assert!(downloader.import_receipts(&receipts_rlp, &[headers[1].hash(), headers[2].hash()]).is_ok());
// Import unexpected fourth receipt.
let mut rlp_data = RlpStream::new_list(1);
rlp_data.append_raw(&receipts[3], 1);
let bodies_rlp = Rlp::new(rlp_data.as_raw());
match downloader.import_bodies(&bodies_rlp, &[headers[1].hash(), headers[2].hash()]) {
Err(BlockDownloaderImportError::Invalid) => (),
_ => panic!("expected BlockDownloaderImportError"),
};
}
Fix ancient blocks sync (#9531) * Log block set in block_sync for easier debugging * logging macros * Match no args in sync logging macros * Add QueueFull error * Only allow importing headers if the first matches requested * WIP * Test for chain head gaps and log * Calc distance even with 2 heads * Revert previous commits, preparing simple fix This reverts commit 5f38aa885b22ebb0e3a1d60120cea69f9f322628. * Reject headers with no gaps when ChainHead * Reset block sync download when queue full * Simplify check for subchain heads * Add comment to explain subchain heads filter * Fix is_subchain_heads check and comment * Prevent premature round completion after restart This is a problem on mainnet where multiple stale peer requests will force many rounds to complete quickly, forcing the retraction. * Reset stale old blocks request after queue full * Revert "Reject headers with no gaps when ChainHead" This reverts commit 0eb865539e5dee37ab34f168f5fb643300de5ace. * Add BlockSet to BlockDownloader logging Currently it is difficult to debug this because there are two instances, one for OldBlocks and one for NewBlocks. This adds the BlockSet to all log messages for easy log filtering. * Reset OldBlocks download from last enqueued Previously when the ancient block queue was full it would restart the download from the last imported block, so the ones still in the queue would be redownloaded. Keeping the existing downloader instance and just resetting it will start again from the last enqueued block.:wq * Ignore expired Body and Receipt requests * Log when ancient block download being restarted * Only request old blocks from peers with >= difficulty https://github.com/paritytech/parity-ethereum/pull/9226 might be too permissive and causing the behaviour of the retraction soon after the fork block. With this change the peer difficulty has to be greater than or euqal to our syncing difficulty, so should still fix https://github.com/paritytech/parity-ethereum/issues/9225 * Some logging and clear stalled blocks head * Revert "Some logging and clear stalled blocks head" This reverts commit 757641d9b817ae8b63fec684759b0815af9c4d0e. * Reset stalled header if useless more than once * Store useless headers in HashSet * Add sync target to logging macro * Don't disable useless peer and fix log macro * Clear useless headers on reset and comments * Use custom error for collecting blocks Previously we resued BlockImportError, however only the Invalid case and this made little sense with the QueueFull error. * Remove blank line * Test for reset sync after consecutive useless headers * Don't reset after consecutive headers when chain head * Delete commented out imports * Return DownloadAction from collect_blocks instead of error * Don't reset after round complete, was causing test hangs * Add comment explaining reset after useless * Replace HashSet with counter for useless headers * Refactor sync reset on bad block/queue full * Add missing target for log message * Fix compiler errors and test after merge * ethcore: revert ethereum tests submodule update
2018-10-09 15:31:40 +02:00
#[test]
fn reset_after_multiple_sets_of_useless_headers() {
::env_logger::try_init().ok();
let spec = Spec::new_test();
let genesis_hash = spec.genesis_header().hash();
let mut downloader = BlockDownloader::new(BlockSet::NewBlocks, &genesis_hash, 0);
downloader.state = State::ChainHead;
let mut chain = TestBlockChainClient::new();
let snapshot_service = TestSnapshotService::new();
let queue = RwLock::new(VecDeque::new());
let mut io = TestIo::new(&mut chain, &snapshot_service, &queue, None);
let heads = [
spec.genesis_header(),
dummy_header(127, H256::random()),
dummy_header(254, H256::random()),
];
let short_subchain = [dummy_header(1, genesis_hash)];
import_headers_ok(&heads, &mut downloader, &mut io);
import_headers_ok(&short_subchain, &mut downloader, &mut io);
assert_eq!(downloader.state, State::Blocks);
assert!(!downloader.blocks.is_empty());
// simulate receiving useless headers
let head = vec![short_subchain.last().unwrap().clone()];
for _ in 0..MAX_USELESS_HEADERS_PER_ROUND {
let res = import_headers(&head, &mut downloader, &mut io);
assert!(res.is_err());
}
assert_eq!(downloader.state, State::Idle);
assert!(downloader.blocks.is_empty());
}
#[test]
fn dont_reset_after_multiple_sets_of_useless_headers_for_chain_head() {
::env_logger::try_init().ok();
let spec = Spec::new_test();
let genesis_hash = spec.genesis_header().hash();
let mut downloader = BlockDownloader::new(BlockSet::NewBlocks, &genesis_hash, 0);
downloader.state = State::ChainHead;
let mut chain = TestBlockChainClient::new();
let snapshot_service = TestSnapshotService::new();
let queue = RwLock::new(VecDeque::new());
let mut io = TestIo::new(&mut chain, &snapshot_service, &queue, None);
let heads = [
spec.genesis_header()
];
let short_subchain = [dummy_header(1, genesis_hash)];
import_headers_ok(&heads, &mut downloader, &mut io);
import_headers_ok(&short_subchain, &mut downloader, &mut io);
assert_eq!(downloader.state, State::Blocks);
assert!(!downloader.blocks.is_empty());
// simulate receiving useless headers
let head = vec![short_subchain.last().unwrap().clone()];
for _ in 0..MAX_USELESS_HEADERS_PER_ROUND {
let res = import_headers(&head, &mut downloader, &mut io);
assert!(res.is_err());
}
// download shouldn't be reset since this is the chain head for a single subchain.
// this state usually occurs for NewBlocks when it has reached the chain head.
assert_eq!(downloader.state, State::Blocks);
assert!(!downloader.blocks.is_empty());
}
}