This pull request improves the condition to check if path state scheme is in use.
Originally, root node presence was used as the indicator if path scheme is used or not. However due to fact that root node will be deleted during the initial snap sync, this condition is no longer useful.
If PersistentStateID is present, it shows that we've already configured for path scheme.
Original problem was caused by #28595, where we made it so that as soon as we start to sync, the root of the disk layer is deleted. That is not wrong per se, but another part of the code uses the "presence of the root" as an init-check for the pathdb. And, since the init-check now failed, the code tried to re-initialize it which failed since a sync was already ongoing.
The total impact being: after a state-sync has begun, if the node for some reason is is shut down, it will refuse to start up again, with the error message: `Fatal: Failed to register the Ethereum service: waiting for sync.`.
This change also modifies how `geth removedb` works, so that the user is prompted for two things: `state data` and `ancient chain`. The former includes both the chaindb aswell as any state history stored in ancients.
---------
Co-authored-by: Martin HS <martin@swende.se>
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>
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.
This change addresses an issue in snap sync, specifically when the entire sync process can be halted due to an encountered empty storage range.
Currently, on the snap sync client side, the response to an empty (partial) storage range is discarded as a non-delivery. However, this response can be a valid response, when the particular range requested does not contain any slots.
For instance, consider a large contract where the entire key space is divided into 16 chunks, and there are no available slots in the last chunk [0xf] -> [end]. When the node receives a request for this particular range, the response includes:
The proof with origin [0xf]
A nil storage slot set
If we simply discard this response, the finalization of the last range will be skipped, halting the entire sync process indefinitely. The test case TestSyncWithUnevenStorage can reproduce the scenario described above.
In addition, this change also defines the common variables MaxAddress and MaxHash.
* cmd, core: resolve scheme from a read-write database
* cmd, core, eth: move the scheme check in the ethereum constructor
* cmd/geth: dump should in ro mode
* cmd: reverts
This changes implements faster post-selfdestruct iteration of storage slots for deletion, by using snapshot-storage+stacktrie to recover the trienodes to be deleted. This mechanism is only implemented for path-based schema.
For hash-based schema, the entire post-selfdestruct storage iteration is skipped, with this change, since hash-based does not actually perform deletion anyway.
---------
Co-authored-by: Martin Holst Swende <martin@swende.se>
* all: implement path-based state scheme
* all: edits from review
* core/rawdb, trie/triedb/pathdb: review changes
* core, light, trie, eth, tests: reimplement pbss history
* core, trie/triedb/pathdb: track block number in state history
* trie/triedb/pathdb: add history documentation
* core, trie/triedb/pathdb: address comments from Peter's review
Important changes to list:
- Cache trie nodes by path in clean cache
- Remove root->id mappings when history is truncated
* trie/triedb/pathdb: fallback to disk if unexpect node in clean cache
* core/rawdb: fix tests
* trie/triedb/pathdb: rename metrics, change clean cache key
* trie/triedb: manage the clean cache inside of disk layer
* trie/triedb/pathdb: move journal function
* trie/triedb/path: fix tests
* trie/triedb/pathdb: fix journal
* trie/triedb/pathdb: fix history
* trie/triedb/pathdb: try to fix tests on windows
* core, trie: address comments
* trie/triedb/pathdb: fix test issues
---------
Co-authored-by: Felix Lange <fjl@twurst.com>
Co-authored-by: Martin Holst Swende <martin@swende.se>
This change adds the ability to perform reads from freezer without size limitation. This can be useful in cases where callers are certain that out-of-memory will not happen (e.g. reading only a few elements).
The previous API was designed to behave both optimally and secure while servicing a request from a peer, whereas this change should _not_ be used when an untrusted peer can influence the query size.
This change makes the StateDB track the state key value diff of a block transition.
We already tracked current account and storage values for the purpose of updating
the state snapshot. With this PR, we now also track the original (pre-transition) values
of accounts and storage slots.
The clean trie cache is persisted periodically, therefore Geth can
quickly warmup the cache in next restart.
However it will reduce the robustness of system. The assumption is
held in Geth that if the parent trie node is present, then the entire
sub-trie associated with the parent are all prensent.
Imagine the scenario that Geth rewinds itself to a past block and
restart, but Geth finds the root node of "future state" in clean
cache then regard this state is present in disk, while is not in fact.
Another example is offline pruning tool. Whenever an offline pruning
is performed, the clean cache file has to be removed to aviod hitting
the root node of "deleted states" in clean cache.
All in all, compare with the minor performance gain, system robustness
is something we care more.
The state availability is checked during the creation of a state reader.
- In hash-based database, if the specified root node does not exist on disk disk, then
the state reader won't be created and an error will be returned.
- In path-based database, if the specified state layer is not available, then the
state reader won't be created and an error will be returned.
This change also contains a stricter semantics regarding the `Commit` operation: once it has been performed, the trie is no longer usable, and certain operations will return an error.
This removes the feature where top nodes of the proof can be elided.
It was intended to be used by the LES server, to save bandwidth
when the client had already fetched parts of the state and only needed
some extra nodes to complete the proof. Alas, it never got implemented
in the client.
This changes the journal logic to mark the state object dirty immediately when it
is reset.
We're mostly adding this change to appease the fuzzer. Marking it dirty immediately
makes no difference in practice because accounts will always be modified by EVM
right after creation.
Drop the notions of uncles, and disables activities while syncing
- Disable activities (e.g. generate pending state) while node is syncing,
- Disable empty block submission (but empty block is still kept for payload building),
- Drop uncle notion since (ethash is already deprecated)
* trie: add node type common package
In trie/types package, a few node wrappers are defined, which will be used
in both trie package, trie/snap package, etc. Therefore, a standalone common
package is created to put these stuffs.
* trie: rename trie/types to trie/trienode
In this PR, all TryXXX(e.g. TryGet) APIs of trie are renamed to XXX(e.g. Get) with an error returned.
The original XXX(e.g. Get) APIs are renamed to MustXXX(e.g. MustGet) and does not return any error -- they print a log output. A future PR will change the behaviour to panic on errorrs.
This PR unifies the error handling in miner.
Whenever an error occur while applying a transaction, the transaction should be regarded as invalid and all following transactions from the same sender not executable because of the nonce restriction. The only exception is the `nonceTooLow` error which is handled separately.
Makes clear the distinction between Finalize and FinalizedAndAssemble:
- In Finalize function, a series of state operations are applied according to consensus rules. The statedb is mutated and the root hash can be checked and compared afterwards.
This function should be used in block processing(receive afrom network and apply it locally) but not block generation.
- In FinalizeAndAssemble function, after applying state mutations, the block is also to be assembled with the latest
state root computed, updating the header.
This function should be used in block generation only.
This change prints out more information about the problem, in the case where geth detects a gap between leveldb and ancients, so we can determine more exactly where the gap is (what the first missing is). Also prints out more metadata.
---------
Co-authored-by: Martin Holst Swende <martin@swende.se>
This change fixes a flaw where, in certain scenarios, the block sealer did not accurately reset the remaining gas after failing to include an invalid transaction. Fixes#26791
The EmptyRootHash and EmptyCodeHash are defined everywhere in the codebase, this PR replaces all of them with unified one defined in core/types package, and also defines constants for TxRoot, WithdrawalsRoot and UncleRoot
This change ports some changes from the main PBSS PR:
- get rid of callback function in `trie.Database.Commit` which is not required anymore
- rework the `nodeResolver` in `trie.Iterator` to make it compatible with multiple state scheme
- some other shallow changes in tests and typo-fixes
This change introduces a breaking change to miner.etherbase is configured.
Previously, users did not need to explicitly set the etherbase address via flag, since 'first' local account was used as etherbase automatically. This change removes the "default first account" feature.
In Proof-of-stake world, the fee recipient address is provided by CL, and not configured in Geth any more - meaning that miner.etherbase is mostly for legacy networks(pow, clique networks etc).
This PR fixes an issue which might result in data lost in freezer.
Whenever mutation happens in freezer, all data will be written into head data file
and it will be rotated with a new one in case the size of file reaches the threshold.
Theoretically, the rotated old data file should be fsync'd to prevent data loss.
In freezer.Sync function, we only fsync: (1) index file (2) meta file and (3) head
data file. So this PR forcibly fsync the head data file if mutation happens in the
boundary of data file.
This changes moves the tracking of "deleted in this block" out from snap-only domain, so that it happens regardless of whether the execution is snapshot-backed or trie-backed.
This changes the StorageTrie method to return an error when the trie
is not available. It used to return an 'empty trie' in this case, but that's
not possible anymore under PBSS.
This PR implements resettable freezer by adding a ResettableFreezer wrapper.
The resettable freezer wraps the original freezer in a way that makes it possible to ensure atomic resets. Implementation wise, it relies on the os.Rename and os.RemoveAll to atomically delete the original freezer data and re-create a new one from scratch.
This PR fixes an error in trie commit. If the trie.root is nil, it can be two possible scenarios:
- The trie was empty, and no change happens
- The trie was non-empty and all nodes are dropped
For the latter one, we should collect the deletions and apply them into database(e.g. in PBSS).
--syncTarget is a feature for development purpose in post-merge world. Previously
it's added into eth.Config. But it turns out that's a stupid idea.
- syncTarget is a block object, which is hard to be put in config file(large)
- syncTarget is just a dev feature, doesn't make too much sense to add it in config file
So I remove it from the eth config object. And it also fixes the #26328
This PR introduces a node scheme abstraction. The interface is only implemented by `hashScheme` at the moment, but will be extended by `pathScheme` very soon.
Apart from that, a few changes are also included which is worth mentioning:
- port the changes in the stacktrie, tracking the path prefix of nodes during commit
- use ethdb.Database for constructing trie.Database. This is not necessary right now, but it is required for path-based used to open reverse diff freezer
This PR adds a parameter to startup, --synctarget. The synctarget flag is a developer-flag, that can be useful in some scenarios as a replacement for a CL node. It defines a fixed block sync target:
geth --syncmode=full --synctarget=./block_15816882.hex_rlp
The --synctarget is only made available during syncmode=full
This PR ports a few changes from PBSS:
- Fix the snapshot generator waiter in case the generation is not even initialized
- Refactor db inspector for ancient store
This PR introduces a new mechanism in chain tracer for preventing creating too many trace states.
The workflow of chain tracer can be divided into several parts:
- state creator generates trace state in a thread
- state tracer retrieves the trace state and applies the tracing on top in another thread
- state collector gathers all result from state tracer and stream to users
It's basically a producer-consumer model here, while if we imagine that the state producer generates states too fast, then it will lead to accumulate lots of unused states in memory. Even worse, in path-based state scheme it will only keep the latest 128 states in memory, and the newly generated state will invalidate the oldest one by marking it as stale.
The solution for fixing it is to limit the speed of state generation. If there are over 128 states un-consumed in memory, then the creation will be paused until the states are be consumed properly.
This PR reworks tx indexer a bit. Compared to the original version, one scenario is no longer handled - upgrading from legacy geth without indexer support.
The tx indexer was introduced in 2020 and have been present through hardforks, so it can be assumed that all Geth nodes have tx indexer already. So we can simplify the tx indexer logic a bit:
- If the tail flag is not present, it means node is just initialized may or may not with an ancient store attached. In this case all blocks are regarded as unindexed
- If the tail flag is present, it means blocks below tail are unindexed, blocks above tail are indexed
This change also address some weird cornercases that could make the indexer not work after a crash.
This PR cleans up the configurations for pruner and snapshotter by passing a config struct.
And also, this PR disables the snapshot background generation if the chain is opened in "read-only" mode. The read-only mode is necessary in some cases. For example, we have a list of commands to open the etheruem node in "read-only" mode, like export-chain. In these cases, the snapshot background generation is non expected and should be banned explicitly.
This PR simplifies the logic of chain tracer and also adds the unit tests.
The most important change has been made in this PR is the state management. Whenever a tracing state is acquired there is a corresponding release function be returned as well. It must be called once the state is used up, otherwise resource leaking can happen.
And also the logic of state management has been simplified a lot. Specifically, the state provider(eth backend, les backend) should ensure the state is available and referenced. State customers can use the state according to their own needs, or build other states based on the given state. But once the release function is called, there is no guarantee of the availability of the state.
Co-authored-by: Sina Mahmoodi <1591639+s1na@users.noreply.github.com>
Co-authored-by: Péter Szilágyi <peterke@gmail.com>
It's a trivial PR to hide the error log when the trie node is not found in the database. The idea for this change is for all TryXXX functions, the error is already returned and we don't need to fire a log explicitly.
Recently there are a few tickets #25613#25589 reporting that the trie nodes are missing because of debug.SetHead. The root cause is after resetting, the chain rewinds to a historical point and re-imports the blocks on top.
Since the node is already synced and started to accept transactions previously, these transactions are still kept in the txpool and verified by txpool with a live state. This live state is constructed based on the live trie database, which is changed fast by node referencing and de-referencing.
Unfortunately, when we construct a live state(like the state in txpool), we don't reference the state we have. The blockchain will garbage collect the intermediate version nodes in another thread which leads the broken live state.
The best solution for this is to forcibly obtain a reference for all live states we create and call release function once it's used up. But it might end up with more junks persisted into disk. Will try to find an elegant solution later in the following PR.
This avoids copying the input []byte while decoding trie nodes. In most
cases, particularly when the input slice is provided by the underlying
database, this optimization is safe to use.
For cases where the origin of the input slice is unclear, the copying version
is retained. The new code performs better even when the input must be
copied, because it is now only copied once in decodeNode.
rjl493456442
