go-ethereum/trie/sync.go
Jerry 7e2be02c26
Merge upstream_merge branch to develop (#1229)
* cmd, core, trie: verkle-capable `geth init` (#28270)

This change allows the creation of a genesis block for verkle testnets. This makes for a chunk of code that is easier to review and still touches many discussion points.

* eth/tracers/js: fix isPush for push0 (#28520)

Fixes so that `push0` opcode is correctly reported as `true` by the `IsPush` function

---------

Co-authored-by: Martin Holst Swende <martin@swende.se>

* trie: spelling - fix comments in hasher (#28507)

Co-authored-by: VM <arimas@foxmail.com>

* tests/fuzzers: move fuzzers into native packages (#28467)

This PR moves our fuzzers from tests/fuzzers into whatever their respective 'native' package is.

The historical reason why they were placed in an external location, is that when they were based on go-fuzz, they could not be "hidden" via the _test.go prefix. So in order to shove them away from the go-ethereum "production code", they were put aside.

But now we've rewritten them to be based on golang testing, and thus can be brought back. I've left (in tests/) the ones that are not production (bls128381), require non-standard imports (secp requires btcec, bn256 requires gnark/google/cloudflare deps).

This PR also adds a fuzzer for precompiled contracts, because why not.

This PR utilizes a newly rewritten replacement for go-118-fuzz-build, namely gofuzz-shim, which utilises the inputs from the fuzzing engine better.

* tests: skip tests on windows 32bit CI (#28521)

tests: skip half the blockchain- and state-tests on windows 32bit CI-tests

* cmd/geth: more special cases logging tests (#28527)

adds logging tests for errors and custom fmt.Stringer-types which output strings that needs to be quoted/escaped.

* accounts,cmd,console,les,metrics:  refactor some errors checked by (ST1005) go-staticcheck (#28532)

fix: fix some (ST1005)go-staticcheck

* miner: run tests in parallel (#28506)

Changes many of the tests in the miner package to run in parallel

* internal/jsre/deps: fix typo in jsdoc (#28511)

minor typo fix

* accounts/abi: improve readability of method-to-string conversion  (#28530)

refactor: improve readability of NewMethod print

* all: replace some cases of strings.SplitN with strings.Cut (#28446)

* ethdb/memorydb, trie: reduced allocations (#28473)

* trie: use pooling of iterator states in iterator

The node iterator burns through a lot of memory while iterating a trie, and a lot of
that can be avoided by using a fairly small pool (max 40 items).

name        old time/op    new time/op    delta
Iterator-8    6.22ms ± 3%    5.40ms ± 6%  -13.18%  (p=0.008 n=5+5)

name        old alloc/op   new alloc/op   delta
Iterator-8    2.36MB ± 0%    1.67MB ± 0%  -29.23%  (p=0.008 n=5+5)

name        old allocs/op  new allocs/op  delta
Iterator-8     37.0k ± 0%     29.8k ± 0%     ~     (p=0.079 n=4+5)

* ethdb/memorydb: avoid one copying of key

By making the transformation from []byte to string at an earlier point,
we save an allocation which otherwise happens later on.

name           old time/op    new time/op    delta
BatchAllocs-8     412µs ± 6%     382µs ± 2%   -7.18%  (p=0.016 n=5+4)

name           old alloc/op   new alloc/op   delta
BatchAllocs-8     480kB ± 0%     490kB ± 0%   +1.93%  (p=0.008 n=5+5)

name           old allocs/op  new allocs/op  delta
BatchAllocs-8     3.03k ± 0%     2.03k ± 0%  -32.98%  (p=0.008 n=5+5)

* Dockerfile: update Go to 1.21 (#28538)

* cmd/evm: validate blockchain tests poststate account storage (#28443)

This PR verifies the accounts' storage as specified in a blockchain test's postState field

The expect-section, it does really only check that the test works. It's meant for the test-author to verify that "If the test does what it's supposed to, then the nonce of X should be 2, and the slot Y at Z should be 0x123.

    This expect-section is not exhaustive (not full post-state)
    It is also not auto-generated, but put there manually by the author.

We can still check it, as a test-sanity-check, in geth

* signer: run tests in parallel (#28536)

marks tests as parallel-safe in package signer

* accounts, cmd: fix typos (#28526)

* core/txpool/legacypool: respect nolocals-setting (#28435)

This change adds a check to ensure that transactions added to the legacy pool are not treated as 'locals' if the global locals-management has been disabled. 

This change makes the pool enforce the --txpool.pricelimit setting.

* cmd: run tests in parallel (#28546)

* core/state/snapshot: print correct error from trie iterator (#28560)

* cmd/evm: capitalize evm commands (#28569)

* standard:fix for a unified standard

* standard:fix more as a complements

---------

Co-authored-by: haotian <haotian@haotiandeMacBook-Air.local>

* accounts/abi: context info on unpack-errors (#28529)

adds contextual information to errors returned by unpack

* core, trie, rpc: speed up tests (#28461)

* rpc: make subscription test faster

reduces time for TestClientSubscriptionChannelClose
from 25 sec to < 1 sec.

* trie: cache trie nodes for faster sanity check

This reduces the time spent on TestIncompleteSyncHash
from ~25s to ~16s.

* core/forkid: speed up validation test

This takes the validation test from > 5s to sub 1 sec

* core/state: improve snapshot test run
brings the time for TestSnapshotRandom from 13s down to 6s

* accounts/keystore: improve keyfile test

This removes some unnecessary waits and reduces the
runtime of TestUpdatedKeyfileContents from 5 to 3 seconds

* trie: remove resolver
* trie: only check ~5% of all trie nodes

* ethdb/pebble: don't double-close iterator inside pebbleIterator (#28566)

Adds 'released' flag to pebbleIterator to avoid double closing cockroachdb/pebble.Iterator as it is an invalid operation.

Fixes #28565

* eth/filters: reuse error msg for invalid block range (#28479)

* core/types: make 'v' optional for DynamicFeeTx and BlobTx (#28564)

This fixes an issue where transactions would not be accepted when they have only
'yParity' and not 'v'.

* rpc: improve performance of subscription notification encoding (#28328)

It turns out that encoding json.RawMessage is slow because
package json basically parses the message again to ensure it is valid.
We can avoid the slowdown by encoding the entire RPC notification once,
which yields a 30% speedup.

* cmd/utils: validate pre-existing genesis in --dev mode (#28468)

geth --dev can be used with an existing data directory and genesis block. Since
dev mode only works with PoS, we need to verify that the merge has happened.

Co-authored-by: Felix Lange <fjl@twurst.com>

* cmd/geth: add support for --dev flag in dumpgenesis (#28463)


Co-authored-by: Felix Lange <fjl@twurst.com>
Co-authored-by: lightclient <lightclient@protonmail.com>

* les/vflux: run tests in parallel (#28524)

* cmd/{geth,utils}: add cmd to export preimages in snap enumeration order (#28256)

Adds a subcommand: `geth snapshot export-preimages`, to export preimages of every hash found during a snapshot enumeration: that is, it exports _only the active state_, and not _all_ preimages that have been used but are no longer part of the state. 

This tool is needed for the verkle transition, in order to distribute the preimages needed for the conversion. Since only the 'active' preimages are exported, the output is shrunk from ~70GB to ~4GB.

The order of the output is the order used by the snapshot enumeration, which avoids database thrashing. However, it also means that storage-slot preimages are not deduplicated.

* cmd/geth: fix build error (#28585)

* cmd/devp2p/internal/ethtest: undo debug-hack (#28588)

cmd/devp2p/internal/ethtest: remove a debug-hack flaw which prevented certain tests from running

* params: update discV5 bootnodes (#28562)

update discV5 bootnodes from https://github.com/eth-clients/eth2-networks/blob/master/shared/mainnet/bootstrap_nodes.txt

* cmd, les, tests: remove light client code (#28586)

* cmd, les, tests: remove light client code

This commit removes the light client (LES) code.
Since the merge the light client has been broken and
it is hard to maintain it alongside the normal client.
We decided it would be best to remove it for now and
maybe rework and reintroduce it in the future.

* cmd, eth: remove some more mentions of light mode

* cmd: re-add flags and mark as deprecated

* cmd: warn the user about deprecated flags

* eth: better error message

* eth, internal/ethapi: drop some weird indirection (#28597)

* trie: fix random test generator early terminate (#28590)

This change fixes a minor bug in the `randTest.Generate` function, which caused the `quick.Check` to be a no-op.

* eth/gasestimator, internal/ethapi: move gas estimator out of rpc (#28600)

* go.mod: update uint256 to v1.2.4 (#28612)

* eth/catalyst, eth/downloader: expose more sync information (#28584)

This change exposes more information from sync module internally

* light: remove package light(#28614)

This changes removes the package 'light', which is currently unused.

* cmd/evm, core/state: fix post-exec dump of state (statetests, blockchaintests) (#28504)

There were several problems related to dumping state. 

- If a preimage was missing, even if we had set the `OnlyWithAddresses` to `false`, to export them anyway, the way the mapping was constructed (using `common.Address` as key) made the entries get lost anyway. Concerns both state- and blockchain tests. 
- Blockchain test execution was not configured to store preimages.

This changes makes it so that the block test executor takes a callback, just like the state test executor already does. This callback can be used to examine the post-execution state, e.g. to aid debugging of test failures.

* ethereum: remove TODO comment about subscription (#28609)

* eth/tracers/js: fix type inconsistencies (#28488)

This change fixes two type-inconsistencies in the JS tracer:

- In most places we return byte arrays as a `Uint8Array` to the tracer. However it seems we missed doing the conversion for `ctx` fields which are passed to the tracer during `result`. They are passed as simple arrays. I think Uint8Arrays are more suitable and we should change this inconsistency. Note: this will be a breaking-change. But I believe the effect is small. If we look at our tracers we see that these fields (`ctx.from`, `ctx.to`, etc.) are used in 2 ways. Passed to `toHex` which takes both array or buffer. Or the length was measured which is the same for both types.
- The `slice` taking in `int, int` params versus `memory.slice` taking `int64, int64` params. I suggest changing `slice` types to `int64`. This should have no effect almost in any case.

* crypto/secp256k1: fix 32-bit tests when CGO_ENABLED=0 (#28602)

* consensus: verify the nonexistence of shanghai- and cancun-specific header fields (#28605)

* eth/gasestimator: allow slight estimation error in favor of less iterations (#28618)

* eth/gasestimator: early exit for plain transfer and error allowance

* core, eth/gasestimator: hard guess at a possible required gas

* internal/ethapi: update estimation tests with the error ratio

* eth/gasestimator: I hate you linter

* graphql: fix gas estimation test

---------

Co-authored-by: Oren <orenyomtov@users.noreply.github.com>

* all: replace log15 with slog (#28187)

This PR replaces Geth's logger package (a fork of [log15](https://github.com/inconshreveable/log15)) with an implementation using slog, a logging library included as part of the Go standard library as of Go1.21.

Main changes are as follows:
* removes any log handlers that were unused in the Geth codebase.
* Json, logfmt, and terminal formatters are now slog handlers.
* Verbosity level constants are changed to match slog constant values.  Internal translation is done to make this opaque to the user and backwards compatible with existing `--verbosity` and `--vmodule` options.
* `--log.backtraceat` and `--log.debug` are removed.

The external-facing API is largely the same as the existing Geth logger.  Logger method signatures remain unchanged.

A small semantic difference is that a `Handler` can only be set once per `Logger` and not changed dynamically.  This just means that a new logger must be instantiated every time the handler of the root logger is changed.

----
For users of the `go-ethereum/log` module. If you were using this module for your own project, you will need to change the initialization. If you previously did 
```golang
log.Root().SetHandler(log.LvlFilterHandler(log.LvlInfo, log.StreamHandler(os.Stderr, log.TerminalFormat(true))))
```
You now instead need to do 
```golang
log.SetDefault(log.NewLogger(log.NewTerminalHandlerWithLevel(os.Stderr, log.LevelInfo, true)))
```
See more about reasoning here: https://github.com/ethereum/go-ethereum/issues/28558#issuecomment-1820606613

* core/state: make stateobject.create selfcontain (#28459)

* trie/triedb/hashdb: take lock around access to dirties cache (#28542)

Add read locking of db lock around access to dirties cache in hashdb.Database to prevent
data race versus hashdb.Database.dereference which can modify the dirities map by deleting
an item.

Fixes #28541

---------

Co-authored-by: Gary Rong <garyrong0905@gmail.com>

* accounts/abi/bind: fix typo (#28630)

* slog: faster and less memory-consumption (#28621)

These changes improves the performance of the non-coloured terminal formatting, _quite a lot_. 

```
name               old time/op    new time/op    delta
TerminalHandler-8    10.2µs ±15%     5.4µs ± 9%  -47.02%  (p=0.008 n=5+5)

name               old alloc/op   new alloc/op   delta
TerminalHandler-8    2.17kB ± 0%    0.40kB ± 0%  -81.46%  (p=0.008 n=5+5)

name               old allocs/op  new allocs/op  delta
TerminalHandler-8      33.0 ± 0%       5.0 ± 0%  -84.85%  (p=0.008 n=5+5)
```

I tried to _somewhat_ organize the commits, but the it might still be a bit chaotic. Some core insights: 

- The function `terminalHandler.Handl` uses a mutex, and writes all output immediately to 'upstream'. Thus, it can reuse a scratch-buffer every time. 
- This buffer can be propagated internally, making all the internal formatters either write directly to it,
- OR, make  use of the `tmp := buf.AvailableBuffer()` in some cases, where a byte buffer "extra capacity" can be temporarily used. 
- The `slog` package  uses `Attr` by value. It makes sense to minimize operating on them, since iterating / collecting into a new slice, iterating again etc causes copy-on-heap. Better to operate on them only once. 
- If we want to do padding, it's better to copy from a constant `space`-buffer than to invoke `bytes.Repeat` every single time.

* eth/tracers: tx-level state in debug_traceCall (#28460)

* cmd/evm: fix Env struct json tag (#28635)

* accounts/abi/bind: fixed typos (#28634)

* Update auth.go

* Update backend.go

* Update bind.go

* Update bind_test.go

* eth/fetcher: fix invalid tracking of received at time for block (#28637)

eth/fetcher: fix invalid tracking of received at time

* accounts: run tests in parallel (#28544)

* eth/tracers/logger: make structlog/json-log stack hex again (#28628)

* common/hexutil: define hex wrappers for uint256.Int

* eth/tracers/logger: make structlog/json-log stack hex again

* common/hexutil: goimports

* log: remove lazy, remove unused interfaces, unexport methods (#28622)

This change 

- Removes interface `log.Format`, 
- Removes method `log.FormatFunc`, 
- unexports `TerminalHandler.TerminalFormat` formatting methods (renamed to `TerminalHandler.format`)
- removes the notion of `log.Lazy` values


The lazy handler was useful in the old log package, since it
could defer the evaluation of costly attributes until later in the
log pipeline: thus, if the logging was done at 'Trace', we could
skip evaluation if logging only was set to 'Info'.

With the move to slog, this way of deferring evaluation is no longer
needed, since slog introduced 'Enabled': the caller can thus do
the evaluate-or-not decision at the callsite, which is much more
straight-forward than dealing with lazy reflect-based evaluation.

Also, lazy evaluation would not work with 'native' slog, as in, these
two statements would be evaluated differently:

```golang
  log.Info("foo", "my lazy", lazyObj)
  slog.Info("foo", "my lazy", lazyObj)
```

* .github: use github actions to run 32-bit linux tests (#28549)

use github actions to run 32-bit linux tests

* ethdb/pebble: remove a dependency (#28627)

The dependency was not really used anyway, so we can get rid of it.

Co-authored-by: Felix Lange <fjl@twurst.com>

* tests/fuzzers/bls12381: deactivate BLS fuzzer when CGO_ENABLED=0 (#28653)

tests/fuzzers/bls12381: deactivate fuzzer when CGO_ENABLED=0

* build: upgrade -dlgo version to Go 1.21.5 (#28648)

* rpc: fix ns/µs mismatch in metrics (#28649)

The rpc/duration/all meter was in nanoseconds, the individual meter in microseconds.
This PR changes it so both of them use nanoseconds.

* cmd/evm: fix dump after state-test exec (#28650)

The dump after state-test didn't work, the problem was an error, "Already committed", which was silently ignored. 

This change re-initialises the state, so the dumping works again.

* beacon/light: add CommitteeChain (#27766)

This change implements CommitteeChain which is a key component of the beacon light client. It is a passive data structure that can validate, hold and update a chain of beacon light sync committees and updates, starting from a checkpoint that proves the starting committee through a beacon block hash, header and corresponding state. Once synced to the current sync period, CommitteeChain can also validate signed beacon headers.

* cmd/utils, eth: disallow invalid snap sync / snapshot flag combos (#28657)

* eth: prevent startup in snap mode without snapshots

* cmd/utils: try to fix bad flag combos wrt snap sync and snapshot generation

* trie: remove inconsistent trie nodes during sync in path mode (#28595)

This fixes a database corruption issue that could occur during state healing.
When sync is aborted while certain modifications were already committed, and a
reorg occurs, the database would contain incorrect trie nodes stored by path.
These nodes need to detected/deleted in order to obtain a complete and fully correct state
after state healing.

---------

Co-authored-by: Felix Lange <fjl@twurst.com>

* cmd/utils: fix HTTPHost, WSHost flag priority (#28669)


Co-authored-by: Felix Lange <fjl@twurst.com>

* eth/protocols/eth: fix typos in comments (#28652)

* core/txpool : small cleanup refactors (#28654)

* eth/fetcher, eth/gasestimator: fix typos in comments (#28675)

* all: fix typos in comments (#28662)


Co-authored-by: Felix Lange <fjl@twurst.com>

* miner: eliminate the dead loop possibility for `newWorkLoop` and `mainLoop` (#28677)

discard the intervalAdjust message if the channel is full

* all: fix typos in comments (#28682)

chore(core,eth):fix a couple of typos

* p2p/discover: add liveness check in collectTableNodes (#28686)

* p2p/discover: add liveness check in collectTableNodes

* p2p/discover: fix test

* p2p/discover: rename to appendLiveNodes

* p2p/discover: add dedup logic back

* p2p/discover: simplify

* p2p/discover: fix issue found by test

* internal/flags: add missing flag types for auto-env-var generation (#28692)

Certain flags, such as `--rpc.txfeecap` currently do not have an env-var auto-generated for them. This change adds three missing cli flag types to the auto env-var helper function to fix this.

* cmd/evm:  default to mirror mainnet forks enabled (#28691)

cmd/evm:  default to using dev chain config (all mainnet HFs activated at block/timestamp 0

* cmd/evm, cmd/clef, cmd/bootnode: fix / unify logging (#28696)

This change fixes a problem with our non-core binaries: evm, clef, bootnode.

First of all, they failed to convert from legacy loglevels 1 to 5, to the new slog loglevels -4 to 4.

Secondly, the logging was actually setup in the init phase, and then overridden in the main. This is not needed for evm, since it used the same flag name as the main geth verbosity. Better to let the flags/internal handle the logging init.

* cmd/evm: t8n support custom tracers (#28557)

This change implements ability for the `evm t8n` tool to use custom tracers; either 'native' golang tracers or javascript tracers.

* params: release go-ethereum v1.13.6 stable

* Fix build errors

* Fix test-integration

---------

Co-authored-by: Guillaume Ballet <3272758+gballet@users.noreply.github.com>
Co-authored-by: Sina Mahmoodi <1591639+s1na@users.noreply.github.com>
Co-authored-by: Martin Holst Swende <martin@swende.se>
Co-authored-by: VM <112189277+sysvm@users.noreply.github.com>
Co-authored-by: VM <arimas@foxmail.com>
Co-authored-by: jwasinger <j-wasinger@hotmail.com>
Co-authored-by: Zoro <40222601+BabyHalimao@users.noreply.github.com>
Co-authored-by: Håvard Anda Estensen <haavard.ae@gmail.com>
Co-authored-by: aliening <128203330+aliening@users.noreply.github.com>
Co-authored-by: Halimao <1065621723@qq.com>
Co-authored-by: danceratopz <danceratopz@gmail.com>
Co-authored-by: levisyin <150114626+levisyin@users.noreply.github.com>
Co-authored-by: jp-imx <109574657+jp-imx@users.noreply.github.com>
Co-authored-by: rjl493456442 <garyrong0905@gmail.com>
Co-authored-by: Haotian <51777534+tmelhao@users.noreply.github.com>
Co-authored-by: haotian <haotian@haotiandeMacBook-Air.local>
Co-authored-by: Marius van der Wijden <m.vanderwijden@live.de>
Co-authored-by: Maciej Kulawik <10907694+magicxyyz@users.noreply.github.com>
Co-authored-by: ucwong <ucwong@126.com>
Co-authored-by: Mario Vega <marioevz@gmail.com>
Co-authored-by: Delweng <delweng@gmail.com>
Co-authored-by: Felix Lange <fjl@twurst.com>
Co-authored-by: lightclient <lightclient@protonmail.com>
Co-authored-by: Mikel Cortes <45786396+cortze@users.noreply.github.com>
Co-authored-by: Péter Szilágyi <peterke@gmail.com>
Co-authored-by: Ng Wei Han <47109095+weiihann@users.noreply.github.com>
Co-authored-by: lightclient <14004106+lightclient@users.noreply.github.com>
Co-authored-by: Shivam Sandbhor <shivam.sandbhor@gmail.com>
Co-authored-by: Jakub Freebit <49676311+jakub-freebit@users.noreply.github.com>
Co-authored-by: Oren <orenyomtov@users.noreply.github.com>
Co-authored-by: BorkBorked <107079055+BorkBorked@users.noreply.github.com>
Co-authored-by: ddl <dengdiliang@gmail.com>
Co-authored-by: Manav Darji <manavdarji.india@gmail.com>
Co-authored-by: Marius Kjærstad <sandakersmann@users.noreply.github.com>
Co-authored-by: Felföldi Zsolt <zsfelfoldi@gmail.com>
Co-authored-by: Ford <153042616+guerrierindien@users.noreply.github.com>
Co-authored-by: Ursulafe <152976968+Ursulafe@users.noreply.github.com>
Co-authored-by: Elias Rad <146735585+nnsW3@users.noreply.github.com>
Co-authored-by: FletcherMan <fanciture@163.com>
Co-authored-by: alex <152680487+bodhi-crypo@users.noreply.github.com>
Co-authored-by: Sebastian Stammler <seb@oplabs.co>
2024-04-29 12:21:11 +02:00

747 lines
26 KiB
Go

// Copyright 2015 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library 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 Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
package trie
import (
"errors"
"fmt"
"sync"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/common/prque"
"github.com/ethereum/go-ethereum/core/rawdb"
"github.com/ethereum/go-ethereum/core/types"
"github.com/ethereum/go-ethereum/ethdb"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum/go-ethereum/metrics"
)
// ErrNotRequested is returned by the trie sync when it's requested to process a
// node it did not request.
var ErrNotRequested = errors.New("not requested")
// ErrAlreadyProcessed is returned by the trie sync when it's requested to process a
// node it already processed previously.
var ErrAlreadyProcessed = errors.New("already processed")
// maxFetchesPerDepth is the maximum number of pending trie nodes per depth. The
// role of this value is to limit the number of trie nodes that get expanded in
// memory if the node was configured with a significant number of peers.
const maxFetchesPerDepth = 16384
var (
// deletionGauge is the metric to track how many trie node deletions
// are performed in total during the sync process.
deletionGauge = metrics.NewRegisteredGauge("trie/sync/delete", nil)
// lookupGauge is the metric to track how many trie node lookups are
// performed to determine if node needs to be deleted.
lookupGauge = metrics.NewRegisteredGauge("trie/sync/lookup", nil)
// accountNodeSyncedGauge is the metric to track how many account trie
// node are written during the sync.
accountNodeSyncedGauge = metrics.NewRegisteredGauge("trie/sync/nodes/account", nil)
// storageNodeSyncedGauge is the metric to track how many account trie
// node are written during the sync.
storageNodeSyncedGauge = metrics.NewRegisteredGauge("trie/sync/nodes/storage", nil)
// codeSyncedGauge is the metric to track how many contract codes are
// written during the sync.
codeSyncedGauge = metrics.NewRegisteredGauge("trie/sync/codes", nil)
)
// SyncPath is a path tuple identifying a particular trie node either in a single
// trie (account) or a layered trie (account -> storage).
//
// Content wise the tuple either has 1 element if it addresses a node in a single
// trie or 2 elements if it addresses a node in a stacked trie.
//
// To support aiming arbitrary trie nodes, the path needs to support odd nibble
// lengths. To avoid transferring expanded hex form over the network, the last
// part of the tuple (which needs to index into the middle of a trie) is compact
// encoded. In case of a 2-tuple, the first item is always 32 bytes so that is
// simple binary encoded.
//
// Examples:
// - Path 0x9 -> {0x19}
// - Path 0x99 -> {0x0099}
// - Path 0x01234567890123456789012345678901012345678901234567890123456789019 -> {0x0123456789012345678901234567890101234567890123456789012345678901, 0x19}
// - Path 0x012345678901234567890123456789010123456789012345678901234567890199 -> {0x0123456789012345678901234567890101234567890123456789012345678901, 0x0099}
type SyncPath [][]byte
// NewSyncPath converts an expanded trie path from nibble form into a compact
// version that can be sent over the network.
func NewSyncPath(path []byte) SyncPath {
// If the hash is from the account trie, append a single item, if it
// is from a storage trie, append a tuple. Note, the length 64 is
// clashing between account leaf and storage root. It's fine though
// because having a trie node at 64 depth means a hash collision was
// found and we're long dead.
if len(path) < 64 {
return SyncPath{hexToCompact(path)}
}
return SyncPath{hexToKeybytes(path[:64]), hexToCompact(path[64:])}
}
// LeafCallback is a callback type invoked when a trie operation reaches a leaf
// node.
//
// The keys is a path tuple identifying a particular trie node either in a single
// trie (account) or a layered trie (account -> storage). Each key in the tuple
// is in the raw format(32 bytes).
//
// The path is a composite hexary path identifying the trie node. All the key
// bytes are converted to the hexary nibbles and composited with the parent path
// if the trie node is in a layered trie.
//
// It's used by state sync and commit to allow handling external references
// between account and storage tries. And also it's used in the state healing
// for extracting the raw states(leaf nodes) with corresponding paths.
type LeafCallback func(keys [][]byte, path []byte, leaf []byte, parent common.Hash, parentPath []byte) error
// nodeRequest represents a scheduled or already in-flight trie node retrieval request.
type nodeRequest struct {
hash common.Hash // Hash of the trie node to retrieve
path []byte // Merkle path leading to this node for prioritization
data []byte // Data content of the node, cached until all subtrees complete
parent *nodeRequest // Parent state node referencing this entry
deps int // Number of dependencies before allowed to commit this node
callback LeafCallback // Callback to invoke if a leaf node it reached on this branch
}
// codeRequest represents a scheduled or already in-flight bytecode retrieval request.
type codeRequest struct {
hash common.Hash // Hash of the contract bytecode to retrieve
path []byte // Merkle path leading to this node for prioritization
data []byte // Data content of the node, cached until all subtrees complete
parents []*nodeRequest // Parent state nodes referencing this entry (notify all upon completion)
}
// NodeSyncResult is a response with requested trie node along with its node path.
type NodeSyncResult struct {
Path string // Path of the originally unknown trie node
Data []byte // Data content of the retrieved trie node
}
// CodeSyncResult is a response with requested bytecode along with its hash.
type CodeSyncResult struct {
Hash common.Hash // Hash the originally unknown bytecode
Data []byte // Data content of the retrieved bytecode
}
// nodeOp represents an operation upon the trie node. It can either represent a
// deletion to the specific node or a node write for persisting retrieved node.
type nodeOp struct {
owner common.Hash // identifier of the trie (empty for account trie)
path []byte // path from the root to the specified node.
blob []byte // the content of the node (nil for deletion)
hash common.Hash // hash of the node content (empty for node deletion)
}
// isDelete indicates if the operation is a database deletion.
func (op *nodeOp) isDelete() bool {
return len(op.blob) == 0
}
// syncMemBatch is an in-memory buffer of successfully downloaded but not yet
// persisted data items.
type syncMemBatch struct {
scheme string // State scheme identifier
codes map[common.Hash][]byte // In-memory batch of recently completed codes
nodes []nodeOp // In-memory batch of recently completed/deleted nodes
size uint64 // Estimated batch-size of in-memory data.
}
// newSyncMemBatch allocates a new memory-buffer for not-yet persisted trie nodes.
func newSyncMemBatch(scheme string) *syncMemBatch {
return &syncMemBatch{
scheme: scheme,
codes: make(map[common.Hash][]byte),
}
}
// hasCode reports the contract code with specific hash is already cached.
func (batch *syncMemBatch) hasCode(hash common.Hash) bool {
_, ok := batch.codes[hash]
return ok
}
// addCode caches a contract code database write operation.
func (batch *syncMemBatch) addCode(hash common.Hash, code []byte) {
batch.codes[hash] = code
batch.size += common.HashLength + uint64(len(code))
}
// addNode caches a node database write operation.
func (batch *syncMemBatch) addNode(owner common.Hash, path []byte, blob []byte, hash common.Hash) {
if batch.scheme == rawdb.PathScheme {
if owner == (common.Hash{}) {
batch.size += uint64(len(path) + len(blob))
} else {
batch.size += common.HashLength + uint64(len(path)+len(blob))
}
} else {
batch.size += common.HashLength + uint64(len(blob))
}
batch.nodes = append(batch.nodes, nodeOp{
owner: owner,
path: path,
blob: blob,
hash: hash,
})
}
// delNode caches a node database delete operation.
func (batch *syncMemBatch) delNode(owner common.Hash, path []byte) {
if batch.scheme != rawdb.PathScheme {
log.Error("Unexpected node deletion", "owner", owner, "path", path, "scheme", batch.scheme)
return // deletion is not supported in hash mode.
}
if owner == (common.Hash{}) {
batch.size += uint64(len(path))
} else {
batch.size += common.HashLength + uint64(len(path))
}
batch.nodes = append(batch.nodes, nodeOp{
owner: owner,
path: path,
})
}
// Sync is the main state trie synchronisation scheduler, which provides yet
// unknown trie hashes to retrieve, accepts node data associated with said hashes
// and reconstructs the trie step by step until all is done.
type Sync struct {
scheme string // Node scheme descriptor used in database.
database ethdb.KeyValueReader // Persistent database to check for existing entries
membatch *syncMemBatch // Memory buffer to avoid frequent database writes
nodeReqs map[string]*nodeRequest // Pending requests pertaining to a trie node path
codeReqs map[common.Hash]*codeRequest // Pending requests pertaining to a code hash
queue *prque.Prque[int64, any] // Priority queue with the pending requests
fetches map[int]int // Number of active fetches per trie node depth
}
// NewSync creates a new trie data download scheduler.
func NewSync(root common.Hash, database ethdb.KeyValueReader, callback LeafCallback, scheme string) *Sync {
ts := &Sync{
scheme: scheme,
database: database,
membatch: newSyncMemBatch(scheme),
nodeReqs: make(map[string]*nodeRequest),
codeReqs: make(map[common.Hash]*codeRequest),
queue: prque.New[int64, any](nil), // Ugh, can contain both string and hash, whyyy
fetches: make(map[int]int),
}
ts.AddSubTrie(root, nil, common.Hash{}, nil, callback)
return ts
}
// AddSubTrie registers a new trie to the sync code, rooted at the designated
// parent for completion tracking. The given path is a unique node path in
// hex format and contain all the parent path if it's layered trie node.
func (s *Sync) AddSubTrie(root common.Hash, path []byte, parent common.Hash, parentPath []byte, callback LeafCallback) {
if root == types.EmptyRootHash {
return
}
owner, inner := ResolvePath(path)
exist, inconsistent := s.hasNode(owner, inner, root)
if exist {
// The entire subtrie is already present in the database.
return
} else if inconsistent {
// There is a pre-existing node with the wrong hash in DB, remove it.
s.membatch.delNode(owner, inner)
}
// Assemble the new sub-trie sync request
req := &nodeRequest{
hash: root,
path: path,
callback: callback,
}
// If this sub-trie has a designated parent, link them together
if parent != (common.Hash{}) {
ancestor := s.nodeReqs[string(parentPath)]
if ancestor == nil {
panic(fmt.Sprintf("sub-trie ancestor not found: %x", parent))
}
ancestor.deps++
req.parent = ancestor
}
s.scheduleNodeRequest(req)
}
// AddCodeEntry schedules the direct retrieval of a contract code that should not
// be interpreted as a trie node, but rather accepted and stored into the database
// as is.
func (s *Sync) AddCodeEntry(hash common.Hash, path []byte, parent common.Hash, parentPath []byte) {
// Short circuit if the entry is empty or already known
if hash == types.EmptyCodeHash {
return
}
if s.membatch.hasCode(hash) {
return
}
// If database says duplicate, the blob is present for sure.
// Note we only check the existence with new code scheme, snap
// sync is expected to run with a fresh new node. Even there
// exists the code with legacy format, fetch and store with
// new scheme anyway.
if rawdb.HasCodeWithPrefix(s.database, hash) {
return
}
// Assemble the new sub-trie sync request
req := &codeRequest{
path: path,
hash: hash,
}
// If this sub-trie has a designated parent, link them together
if parent != (common.Hash{}) {
ancestor := s.nodeReqs[string(parentPath)] // the parent of codereq can ONLY be nodereq
if ancestor == nil {
panic(fmt.Sprintf("raw-entry ancestor not found: %x", parent))
}
ancestor.deps++
req.parents = append(req.parents, ancestor)
}
s.scheduleCodeRequest(req)
}
// Missing retrieves the known missing nodes from the trie for retrieval. To aid
// both eth/6x style fast sync and snap/1x style state sync, the paths of trie
// nodes are returned too, as well as separate hash list for codes.
func (s *Sync) Missing(max int) ([]string, []common.Hash, []common.Hash) {
var (
nodePaths []string
nodeHashes []common.Hash
codeHashes []common.Hash
)
for !s.queue.Empty() && (max == 0 || len(nodeHashes)+len(codeHashes) < max) {
// Retrieve the next item in line
item, prio := s.queue.Peek()
// If we have too many already-pending tasks for this depth, throttle
depth := int(prio >> 56)
if s.fetches[depth] > maxFetchesPerDepth {
break
}
// Item is allowed to be scheduled, add it to the task list
s.queue.Pop()
s.fetches[depth]++
switch item := item.(type) {
case common.Hash:
codeHashes = append(codeHashes, item)
case string:
req, ok := s.nodeReqs[item]
if !ok {
log.Error("Missing node request", "path", item)
continue // System very wrong, shouldn't happen
}
nodePaths = append(nodePaths, item)
nodeHashes = append(nodeHashes, req.hash)
}
}
return nodePaths, nodeHashes, codeHashes
}
// ProcessCode injects the received data for requested item. Note it can
// happen that the single response commits two pending requests(e.g.
// there are two requests one for code and one for node but the hash
// is same). In this case the second response for the same hash will
// be treated as "non-requested" item or "already-processed" item but
// there is no downside.
func (s *Sync) ProcessCode(result CodeSyncResult) error {
// If the code was not requested or it's already processed, bail out
req := s.codeReqs[result.Hash]
if req == nil {
return ErrNotRequested
}
if req.data != nil {
return ErrAlreadyProcessed
}
req.data = result.Data
return s.commitCodeRequest(req)
}
// ProcessNode injects the received data for requested item. Note it can
// happen that the single response commits two pending requests(e.g.
// there are two requests one for code and one for node but the hash
// is same). In this case the second response for the same hash will
// be treated as "non-requested" item or "already-processed" item but
// there is no downside.
func (s *Sync) ProcessNode(result NodeSyncResult) error {
// If the trie node was not requested or it's already processed, bail out
req := s.nodeReqs[result.Path]
if req == nil {
return ErrNotRequested
}
if req.data != nil {
return ErrAlreadyProcessed
}
// Decode the node data content and update the request
node, err := decodeNode(req.hash.Bytes(), result.Data)
if err != nil {
return err
}
req.data = result.Data
// Create and schedule a request for all the children nodes
requests, err := s.children(req, node)
if err != nil {
return err
}
if len(requests) == 0 && req.deps == 0 {
_ = s.commitNodeRequest(req)
} else {
req.deps += len(requests)
for _, child := range requests {
s.scheduleNodeRequest(child)
}
}
return nil
}
// Commit flushes the data stored in the internal membatch out to persistent
// storage, returning any occurred error. The whole data set will be flushed
// in an atomic database batch.
func (s *Sync) Commit(dbw ethdb.Batch) error {
// Flush the pending node writes into database batch.
var (
account int
storage int
)
for _, op := range s.membatch.nodes {
if op.isDelete() {
// node deletion is only supported in path mode.
if op.owner == (common.Hash{}) {
rawdb.DeleteAccountTrieNode(dbw, op.path)
} else {
rawdb.DeleteStorageTrieNode(dbw, op.owner, op.path)
}
deletionGauge.Inc(1)
} else {
if op.owner == (common.Hash{}) {
account += 1
} else {
storage += 1
}
rawdb.WriteTrieNode(dbw, op.owner, op.path, op.hash, op.blob, s.scheme)
}
}
accountNodeSyncedGauge.Inc(int64(account))
storageNodeSyncedGauge.Inc(int64(storage))
// Flush the pending code writes into database batch.
for hash, value := range s.membatch.codes {
rawdb.WriteCode(dbw, hash, value)
}
codeSyncedGauge.Inc(int64(len(s.membatch.codes)))
s.membatch = newSyncMemBatch(s.scheme) // reset the batch
return nil
}
// MemSize returns an estimated size (in bytes) of the data held in the membatch.
func (s *Sync) MemSize() uint64 {
return s.membatch.size
}
// Pending returns the number of state entries currently pending for download.
func (s *Sync) Pending() int {
return len(s.nodeReqs) + len(s.codeReqs)
}
// scheduleNodeRequest inserts a new state retrieval request into the fetch queue. If there
// is already a pending request for this node, the new request will be discarded
// and only a parent reference added to the old one.
func (s *Sync) scheduleNodeRequest(req *nodeRequest) {
s.nodeReqs[string(req.path)] = req
// Schedule the request for future retrieval. This queue is shared
// by both node requests and code requests.
prio := int64(len(req.path)) << 56 // depth >= 128 will never happen, storage leaves will be included in their parents
for i := 0; i < 14 && i < len(req.path); i++ {
prio |= int64(15-req.path[i]) << (52 - i*4) // 15-nibble => lexicographic order
}
s.queue.Push(string(req.path), prio)
}
// scheduleCodeRequest inserts a new state retrieval request into the fetch queue. If there
// is already a pending request for this node, the new request will be discarded
// and only a parent reference added to the old one.
func (s *Sync) scheduleCodeRequest(req *codeRequest) {
// If we're already requesting this node, add a new reference and stop
if old, ok := s.codeReqs[req.hash]; ok {
old.parents = append(old.parents, req.parents...)
return
}
s.codeReqs[req.hash] = req
// Schedule the request for future retrieval. This queue is shared
// by both node requests and code requests.
prio := int64(len(req.path)) << 56 // depth >= 128 will never happen, storage leaves will be included in their parents
for i := 0; i < 14 && i < len(req.path); i++ {
prio |= int64(15-req.path[i]) << (52 - i*4) // 15-nibble => lexicographic order
}
s.queue.Push(req.hash, prio)
}
// children retrieves all the missing children of a state trie entry for future
// retrieval scheduling.
// nolint:gocognit
func (s *Sync) children(req *nodeRequest, object node) ([]*nodeRequest, error) {
// Gather all the children of the node, irrelevant whether known or not
type childNode struct {
path []byte
node node
}
var children []childNode
switch node := (object).(type) {
case *shortNode:
key := node.Key
if hasTerm(key) {
key = key[:len(key)-1]
}
children = []childNode{{
node: node.Val,
path: append(append([]byte(nil), req.path...), key...),
}}
// Mark all internal nodes between shortNode and its **in disk**
// child as invalid. This is essential in the case of path mode
// scheme; otherwise, state healing might overwrite existing child
// nodes silently while leaving a dangling parent node within the
// range of this internal path on disk and the persistent state
// ends up with a very weird situation that nodes on the same path
// are not inconsistent while they all present in disk. This property
// would break the guarantee for state healing.
//
// While it's possible for this shortNode to overwrite a previously
// existing full node, the other branches of the fullNode can be
// retained as they are not accessible with the new shortNode, and
// also the whole sub-trie is still untouched and complete.
//
// This step is only necessary for path mode, as there is no deletion
// in hash mode at all.
if _, ok := node.Val.(hashNode); ok && s.scheme == rawdb.PathScheme {
owner, inner := ResolvePath(req.path)
for i := 1; i < len(key); i++ {
// While checking for a non-existent item in Pebble can be less efficient
// without a bloom filter, the relatively low frequency of lookups makes
// the performance impact negligible.
var exists bool
if owner == (common.Hash{}) {
exists = rawdb.ExistsAccountTrieNode(s.database, append(inner, key[:i]...))
} else {
exists = rawdb.ExistsStorageTrieNode(s.database, owner, append(inner, key[:i]...))
}
if exists {
s.membatch.delNode(owner, append(inner, key[:i]...))
log.Debug("Detected dangling node", "owner", owner, "path", append(inner, key[:i]...))
}
}
lookupGauge.Inc(int64(len(key) - 1))
}
case *fullNode:
for i := 0; i < 17; i++ {
if node.Children[i] != nil {
children = append(children, childNode{
node: node.Children[i],
path: append(append([]byte(nil), req.path...), byte(i)),
})
}
}
default:
panic(fmt.Sprintf("unknown node: %+v", node))
}
// Iterate over the children, and request all unknown ones
var (
missing = make(chan *nodeRequest, len(children))
pending sync.WaitGroup
batchMu sync.Mutex
)
for _, child := range children {
// Notify any external watcher of a new key/value node
if req.callback != nil {
if node, ok := (child.node).(valueNode); ok {
var paths [][]byte
if len(child.path) == 2*common.HashLength {
paths = append(paths, hexToKeybytes(child.path))
} else if len(child.path) == 4*common.HashLength {
paths = append(paths, hexToKeybytes(child.path[:2*common.HashLength]))
paths = append(paths, hexToKeybytes(child.path[2*common.HashLength:]))
}
if err := req.callback(paths, child.path, node, req.hash, req.path); err != nil {
return nil, err
}
}
}
// If the child references another node, resolve or schedule.
// We check all children concurrently.
if node, ok := (child.node).(hashNode); ok {
path := child.path
hash := common.BytesToHash(node)
pending.Add(1)
go func() {
defer pending.Done()
owner, inner := ResolvePath(path)
exist, inconsistent := s.hasNode(owner, inner, hash)
if exist {
return
} else if inconsistent {
// There is a pre-existing node with the wrong hash in DB, remove it.
batchMu.Lock()
s.membatch.delNode(owner, inner)
batchMu.Unlock()
}
// Locally unknown node, schedule for retrieval
missing <- &nodeRequest{
path: path,
hash: hash,
parent: req,
callback: req.callback,
}
}()
}
}
pending.Wait()
requests := make([]*nodeRequest, 0, len(children))
for done := false; !done; {
select {
case miss := <-missing:
requests = append(requests, miss)
default:
done = true
}
}
return requests, nil
}
// commitNodeRequest finalizes a retrieval request and stores it into the membatch. If any
// of the referencing parent requests complete due to this commit, they are also
// committed themselves.
func (s *Sync) commitNodeRequest(req *nodeRequest) error {
// Write the node content to the membatch
owner, path := ResolvePath(req.path)
s.membatch.addNode(owner, path, req.data, req.hash)
// Removed the completed node request
delete(s.nodeReqs, string(req.path))
s.fetches[len(req.path)]--
// Check parent for completion
if req.parent != nil {
req.parent.deps--
if req.parent.deps == 0 {
if err := s.commitNodeRequest(req.parent); err != nil {
return err
}
}
}
return nil
}
// commitCodeRequest finalizes a retrieval request and stores it into the membatch. If any
// of the referencing parent requests complete due to this commit, they are also
// committed themselves.
func (s *Sync) commitCodeRequest(req *codeRequest) error {
// Write the node content to the membatch
s.membatch.addCode(req.hash, req.data)
// Removed the completed code request
delete(s.codeReqs, req.hash)
s.fetches[len(req.path)]--
// Check all parents for completion
for _, parent := range req.parents {
parent.deps--
if parent.deps == 0 {
if err := s.commitNodeRequest(parent); err != nil {
return err
}
}
}
return nil
}
// hasNode reports whether the specified trie node is present in the database.
// 'exists' is true when the node exists in the database and matches the given root
// hash. The 'inconsistent' return value is true when the node exists but does not
// match the expected hash.
func (s *Sync) hasNode(owner common.Hash, path []byte, hash common.Hash) (exists bool, inconsistent bool) {
// If node is running with hash scheme, check the presence with node hash.
if s.scheme == rawdb.HashScheme {
return rawdb.HasLegacyTrieNode(s.database, hash), false
}
// If node is running with path scheme, check the presence with node path.
var blob []byte
var dbHash common.Hash
if owner == (common.Hash{}) {
blob, dbHash = rawdb.ReadAccountTrieNode(s.database, path)
} else {
blob, dbHash = rawdb.ReadStorageTrieNode(s.database, owner, path)
}
exists = hash == dbHash
inconsistent = !exists && len(blob) != 0
return exists, inconsistent
}
// ResolvePath resolves the provided composite node path by separating the
// path in account trie if it's existent.
func ResolvePath(path []byte) (common.Hash, []byte) {
var owner common.Hash
if len(path) >= 2*common.HashLength {
owner = common.BytesToHash(hexToKeybytes(path[:2*common.HashLength]))
path = path[2*common.HashLength:]
}
return owner, path
}