* core/state: trie prefetcher change: calling trie() doesn't stop the associated subfetcher
Co-authored-by: Martin HS <martin@swende.se>
Co-authored-by: Péter Szilágyi <peterke@gmail.com>
* core/state: improve prefetcher
* core/state: restore async prefetcher stask scheduling
* core/state: finish prefetching async and process storage updates async
* core/state: don't use the prefetcher for missing snapshot items
* core/state: remove update concurrency for Verkle tries
* core/state: add some termination checks to prefetcher async shutdowns
* core/state: differentiate db tries and prefetched tries
* core/state: teh teh teh
---------
Co-authored-by: Jared Wasinger <j-wasinger@hotmail.com>
Co-authored-by: Martin HS <martin@swende.se>
Co-authored-by: Gary Rong <garyrong0905@gmail.com>
* core/state, internal/workerpool: parallelize parts of state commit
* core, internal: move workerpool into syncx
* core/state: use errgroups, commit accounts concurrently
* core: resurrect detailed commit timers to almost-accuracy
This PR fixes some flaws with the existing tests.
The randomized testing (TestSnapshotRandom) executes a series of steps which modify the state and create journal-events. Later on, we compare the forward-going-states against the backwards-unrolling-journal-states, and check that they are identical.
The "identical" check is performed using various accessors. It turned out that we failed to check some things:
- the accesslist contents
- the transient storage contents
- the 'newContract' flag
- the dirty storage map
This change adds these new checks
Currently our state journal tracks each storage update to a contract, having the ability to revert those changes to the previously set value.
For the very first modification however, it behaves a bit wonky. Reverting the update doesn't actually remove the dirty-ness of the slot, rather leaves it as "change this slot to it's original value". This can cause issues down the line with for example write witnesses needing to gather an unneeded proof.
This PR modifies the storageChange journal entry to not only track the previous value of a slot, but also whether there was any previous value at all set in the current execution context. In essence, the PR changes the semantic of storageChange so it does not simply track storage changes, rather it tracks dirty storage changes, an important distinction for being able to cleanly revert the journal item.
This pull request defines a gentrie for snap sync purpose.
The stackTrie is used to generate the merkle tree nodes upon receiving a state batch. Several additional options have been added into stackTrie to handle incomplete states (either missing states before or after).
In this pull request, these options have been relocated from stackTrie to genTrie, which serves as a wrapper for stackTrie specifically for snap sync purposes.
Further, the logic for managing incomplete state has been enhanced in this change. Originally, there are two cases handled:
- boundary node filtering
- internal (covered by extension node) node clearing
This changes adds one more:
- Clearing leftover nodes on the boundaries.
This feature is necessary if there are leftover trie nodes in database, otherwise node inconsistency may break the state healing.
This addresses an edge-case (detailed in the code comment) where the computation of the intermediate trie root would force the unnecessary resolution of a hash node. The change makes it so that when we process changes from a block, we first process trie-updates and afterwards process trie-deletions.
Here we add a Go API for running tracing plugins within the main block import process.
As an advanced user of geth, you can now create a Go file in eth/tracers/live/, and within
that file register your custom tracer implementation. Then recompile geth and select your tracer
on the command line. Hooks defined in the tracer will run whenever a block is processed.
The hook system is defined in package core/tracing. It uses a struct with callbacks, instead of
requiring an interface, for several reasons:
- We plan to keep this API stable long-term. The core/tracing hook API does not depend on
on deep geth internals.
- There are a lot of hooks, and tracers will only need some of them. Using a struct allows you
to implement only the hooks you want to actually use.
All existing tracers in eth/tracers/native have been rewritten to use the new hook system.
This change breaks compatibility with the vm.EVMLogger interface that we used to have.
If you are a user of vm.EVMLogger, please migrate to core/tracing, and sorry for breaking
your stuff. But we just couldn't have both the old and new tracing APIs coexist in the EVM.
---------
Co-authored-by: Matthieu Vachon <matthieu.o.vachon@gmail.com>
Co-authored-by: Delweng <delweng@gmail.com>
Co-authored-by: Martin HS <martin@swende.se>
As SELF-DESTRUCT opcode is disabled in the cancun fork(unless the
account is created within the same transaction, nothing to delete
in this case). The account will only be deleted in the following
cases:
- The account is created within the same transaction. In this case
the original storage was empty.
- The account is empty(zero nonce, zero balance, zero code) and
is touched within the transaction. Fortunately this kind of accounts
are not-existent on ethereum-mainnet.
All in all, after cancun, we are pretty sure there is no large contract
deletion and we don't need this mechanism for oom protection.
This change switches from using the `Hasher` interface to add/query the bloomfilter to implementing it as methods.
This significantly reduces the allocations for Search and Rebloom.
This change makes use of uin256 to represent balance in state. It touches primarily upon statedb, stateobject and state processing, trying to avoid changes in transaction pools, core types, rpc and tracers.
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.
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
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.
* 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
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.
This PR removes panics from stacktrie (mostly), and makes the Update return errors instead. While adding tests for this, I also found that one case of possible corruption was not caught, which is now fixed.
This change enhances the stacktrie constructor by introducing an option struct. It also simplifies the `Hash` and `Commit` operations, getting rid of the special handling round root node.
During snap-sync, we request ranges of values: either a range of accounts or a range of storage values. For any large trie, e.g. the main account trie or a large storage trie, we cannot fetch everything at once.
Short version; we split it up and request in multiple stages. To do so, we use an origin field, to say "Give me all storage key/values where key > 0x20000000000000000". When the server fulfils this, the server provides the first key after origin, let's say 0x2e030000000000000 -- never providing the exact origin. However, the client-side needs to be able to verify that the 0x2e03.. indeed is the first one after 0x2000.., and therefore the attached proof concerns the origin, not the first key.
So, short-short version: the left-hand side of the proof relates to the origin, and is free-standing from the first leaf.
On the other hand, (pun intended), the right-hand side, there's no such 'gap' between "along what path does the proof walk" and the last provided leaf. The proof must prove the last element (unless there are no elements).
Therefore, we can simplify the semantics for trie.VerifyRangeProof by removing an argument. This doesn't make much difference in practice, but makes it so that we can remove some tests. The reason I am raising this is that the upcoming stacktrie-based verifier does not support such fancy features as standalone right-hand borders.
This change
- Removes the owner-notion from a stacktrie; the owner is only ever needed for comitting to the database, but the commit-function, the `writeFn` is provided by the caller, so the caller can just set the owner into the `writeFn` instead of having it passed through the stacktrie.
- Removes the `encoding.BinaryMarshaler`/`encoding.BinaryUnmarshaler` interface from stacktrie. We're not using it, and it is doubtful whether anyone downstream is either.
This change includes a lot of things, listed below.
### Split up interfaces, write vs read
The interfaces have been split up into one write-interface and one read-interface, with `Snapshot` being the gateway from write to read. This simplifies the semantics _a lot_.
Example of splitting up an interface into one readonly 'snapshot' part, and one updatable writeonly part:
```golang
type MeterSnapshot interface {
Count() int64
Rate1() float64
Rate5() float64
Rate15() float64
RateMean() float64
}
// Meters count events to produce exponentially-weighted moving average rates
// at one-, five-, and fifteen-minutes and a mean rate.
type Meter interface {
Mark(int64)
Snapshot() MeterSnapshot
Stop()
}
```
### A note about concurrency
This PR makes the concurrency model clearer. We have actual meters and snapshot of meters. The `meter` is the thing which can be accessed from the registry, and updates can be made to it.
- For all `meters`, (`Gauge`, `Timer` etc), it is assumed that they are accessed by different threads, making updates. Therefore, all `meters` update-methods (`Inc`, `Add`, `Update`, `Clear` etc) need to be concurrency-safe.
- All `meters` have a `Snapshot()` method. This method is _usually_ called from one thread, a backend-exporter. But it's fully possible to have several exporters simultaneously: therefore this method should also be concurrency-safe.
TLDR: `meter`s are accessible via registry, all their methods must be concurrency-safe.
For all `Snapshot`s, it is assumed that an individual exporter-thread has obtained a `meter` from the registry, and called the `Snapshot` method to obtain a readonly snapshot. This snapshot is _not_ guaranteed to be concurrency-safe. There's no need for a snapshot to be concurrency-safe, since exporters should not share snapshots.
Note, though: that by happenstance a lot of the snapshots _are_ concurrency-safe, being unmutable minimal representations of a value. Only the more complex ones are _not_ threadsafe, those that lazily calculate things like `Variance()`, `Mean()`.
Example of how a background exporter typically works, obtaining the snapshot and sequentially accessing the non-threadsafe methods in it:
```golang
ms := metric.Snapshot()
...
fields := map[string]interface{}{
"count": ms.Count(),
"max": ms.Max(),
"mean": ms.Mean(),
"min": ms.Min(),
"stddev": ms.StdDev(),
"variance": ms.Variance(),
```
TLDR: `snapshots` are not guaranteed to be concurrency-safe (but often are).
### Sample changes
I also changed the `Sample` type: previously, it iterated the samples fully every time `Mean()`,`Sum()`, `Min()` or `Max()` was invoked. Since we now have readonly base data, we can just iterate it once, in the constructor, and set all four values at once.
The same thing has been done for runtimehistogram.
### ResettingTimer API
Back when ResettingTImer was implemented, as part of https://github.com/ethereum/go-ethereum/pull/15910, Anton implemented a `Percentiles` on the new type. However, the method did not conform to the other existing types which also had a `Percentiles`.
1. The existing ones, on input, took `0.5` to mean `50%`. Anton used `50` to mean `50%`.
2. The existing ones returned `float64` outputs, thus interpolating between values. A value-set of `0, 10`, at `50%` would return `5`, whereas Anton's would return either `0` or `10`.
This PR removes the 'new' version, and uses only the 'legacy' percentiles, also for the ResettingTimer type.
The resetting timer snapshot was also defined so that it would expose the internal values. This has been removed, and getters for `Max, Min, Mean` have been added instead.
### Unexport types
A lot of types were exported, but do not need to be. This PR unexports quite a lot of them.
rjl493456442
Péter Szilágyi
Guillaume Ballet
Hteev Oli
Martin HS
haoran
Aaron Chen
Abirdcfly
cui
jwasinger
Sina M
buddho
lmittmann
Dimitris Apostolou
Marius van der Wijden
Darioush Jalali
cygaar
aaronbuchwald