This PR adds `rawdb.SafeDeleteRange` and uses it for range deletion in
`core/filtermaps`. This includes deleting the old bloombits database,
resetting the log index database and removing index data for unindexed
tail epochs (which previously weren't properly implemented for the
fallback case).
`SafeDeleteRange` either calls `ethdb.DeleteRange` if the node uses the
new path based state scheme or uses an iterator based fallback method
that safely skips trie nodes in the range if the old hash based state
scheme is used. Note that `ethdb.DeleteRange` also has its own iterator
based fallback implementation in `ethdb/leveldb`. If a path based state
scheme is used and the backing db is pebble (as it is on the majority of
new nodes) then `rawdb.SafeDeleteRange` uses the fast native range
delete.
Also note that `rawdb.SafeDeleteRange` has different semantics from
`ethdb.DeleteRange`, it does not automatically return if the operation
takes a long time. Instead it receives a `stopCallback` that can
interrupt the process if necessary. This is because in the safe mode
potentially a lot of entries are iterated without being deleted (this is
definitely the case when deleting the old bloombits database which has a
single byte prefix) and therefore restarting the process every time a
fixed number of entries have been iterated would result in a quadratic
run time in the number of skipped entries.
When running in safe mode, unindexing an epoch takes about a second,
removing bloombits takes around 10s while resetting a full log index
might take a few minutes. If a range delete operation takes a
significant amount of time then log messages are printed. Also, any
range delete operation can be interrupted by shutdown (tail uinindexing
can also be interrupted by head indexing, similarly to how tail indexing
works). If the last unindexed epoch might have "dirty" index data left
then the indexed map range points to the first valid epoch and
`cleanedEpochsBefore` points to the previous, potentially dirty one. At
startup it is always assumed that the epoch before the first fully
indexed one might be dirty. New tail maps are never rendered and also no
further maps are unindexed before the previous unindexing is properly
cleaned up.
---------
Co-authored-by: Gary Rong <garyrong0905@gmail.com>
Co-authored-by: Felix Lange <fjl@twurst.com>
This PR changes log indexer error handling so that if an indexing error
happens then it disables the indexer and reverts to unindexed more
without resetting the database (except in case of a failed database
init).
Resetting the database on the first error would probably be overkill as
a client update might fix this without having to reindex the entire
history. It would also make debugging very hard. On the other hand,
these errors do not resolve themselves automatically so constantly
retrying makes no sense either. With these changes a new attempt to
resume indexing is made every time the client is restarted.
The PR also fixes https://github.com/ethereum/go-ethereum/issues/31491
which originated from the tail indexer trying to resume processing a
failed map renderer.
---------
Co-authored-by: Felix Lange <fjl@twurst.com>
Add support for state overrides in eth_createAccessList. This will make the method consistent
with other execution methods.
---------
Co-authored-by: Sina Mahmoodi <itz.s1na@gmail.com>
This pull request removes the node copy operation to reduce memory
allocation. Key Changes as below:
**(a) Use `decodeNodeUnsafe` for decoding nodes retrieved from the trie
node reader**
In the current implementation of the MPT, once a trie node blob is
retrieved, it is passed to `decodeNode` for decoding. However,
`decodeNode` assumes the supplied byte slice might be mutated later, so
it performs a deep copy internally before parsing the node.
Given that the node reader is implemented by the path database and the
hash database, both of which guarantee the immutability of the returned
byte slice. By restricting the node reader interface to explicitly
guarantee that the returned byte slice will not be modified, we can
safely replace `decodeNode` with `decodeNodeUnsafe`. This eliminates the
need for a redundant byte copy during each node resolution.
**(b) Modify the trie in place**
In the current implementation of the MPT, a copy of a trie node is
created before any modifications are made. These modifications include:
- Node resolution: Converting the value from a hash to the actual node.
- Node hashing: Tagging the hash into its cache.
- Node commit: Replacing the children with its hash.
- Structural changes: For example, adding a new child to a fullNode or
replacing a child of a shortNode.
This mechanism ensures that modifications only affect the live tree,
leaving all previously created copies unaffected.
Unfortunately, this property leads to a huge memory allocation
requirement. For example, if we want to modify the fullNode for n times,
the node will be copied for n times.
In this pull request, all the trie modifications are made in place. In
order to make sure all previously created copies are unaffected, the
`Copy` function now will deep-copy all the live nodes rather than the
root node itself.
With this change, while the `Copy` function becomes more expensive, it's
totally acceptable as it's not a frequently used one. For the normal
trie operations (Get, GetNode, Hash, Commit, Insert, Delete), the node
copy is not required anymore.
This PR adds an extra condition to the log indexer initialization in
order to avoid initializing with block 0 as target head. Previously this
caused the indexer to initialize without a checkpoint. Later, when the
real chain head was set, it indexed the entire history, then unindexed
most of it if only the recent history was supposed to be indexed. Now
the init only happens when there is an actual synced chain head and
therefore the index is initialized at the most recent checkpoint and
only the last year is indexed according to the default parameters.
During checkpoint initialization the best available checkpoint is also
checked against the history cutoff point and fails if the indexing would
have to start from a block older than the cutoff. If initialization
fails then the indexer reverts to unindexed mode instead of retrying
because the the failure conditions cannot be expected to recover later.
Instead of reporting all filtermaps stuff in one line, I'm breaking it
down into the three separate kinds of entries here.
```
+-----------------------+-----------------------------+------------+------------+
| DATABASE | CATEGORY | SIZE | ITEMS |
+-----------------------+-----------------------------+------------+------------+
| Key-Value store | Log index filter-map rows | 59.21 GiB | 616077345 |
| Key-Value store | Log index last-block-of-map | 12.35 MiB | 269755 |
| Key-Value store | Log index block-lv | 421.70 MiB | 22109169 |
```
Also added some other changes to make it easier to debug:
- restored bloombits into the inspect output, so we notice if it doesn't
get deleted for some reason
- tracking of unaccounted key examples
This adds a new subcommand 'geth prune-history' that removes the pre-merge history
on supported networks. Geth is not fully ready to work in this mode, please do not run
this command on your production node.
---------
Co-authored-by: Felix Lange <fjl@twurst.com>
In #31384 we unindex TXes prior to the merge block. However when the
node starts up it will try to re-index those back if the config is to index the
whole chain. This change makes the indexer aware of the history cutoff block,
avoiding reindexing in that segment.
---------
Co-authored-by: Gary Rong <garyrong0905@gmail.com>
Co-authored-by: Felix Lange <fjl@twurst.com>
This pull request improves the protection mechanism in the txpool for
senders with delegation. A sender with either delegation or pending
delegation is now limited to a maximum of one in-flight executable
transaction, while gapped transactions will be rejected.
Reason:
If nonce-gapped transaction from delegated/pending-delegated senders
can be acceptable, then it's no-longer possible to send another
"executable" transaction with correct nonce due to the policy of at most
one inflight tx. The gapped transaction will be stuck in the txpool, with no
meaningful way to unlock the sender.
---------
Co-authored-by: lightclient <lightclient@protonmail.com>
This fixes the handshake in a scenario where the remote end sends two unknown
packets in a row. When this happens, we would previously respond to both with
a WHOAREYOU challenge, but keep only the latest sent challenge. Transmission is
assumed to be unreliable, so any client that sends two request packets simultaneously
has to be prepared to follow up on whichever request leads to a handshake. With
this fix, we force them to do the handshake that we can actually complete.
Fixes#30581
This PR changes the matcher syncing conditions so that it is possible to
run a search while head indexing is in progress. Previously it was a
requirement to have the head indexed in order to perform matcher sync
before and after a search. This was unnecessarily strict as the purpose
was just to avoid syncing the valid range with the temporary shortened
indexed range applied while updating existing head maps. Now the sync
condition explicitly checks whether the indexer has a temporary indexed
range with some head maps being partially updated.
It also fixes a deadlock that happened when matcher synchronization was
attempted in the event handler called from the `writeFinishedMaps`
periodical callback.
This pull request fixes a broken unit test
```
=== CONT TestTracingWithOverrides
api_test.go:1012: result: {"gas":21167,"failed":false,"returnValue":"0x0000000000000000000000000000000000000000000000000000000000000002","structLogs":[{"pc":0,"op":"PUSH1","gas":24978860,"gasCost":3,"depth":1,"stack":[]},{"pc":2,"op":"CALLDATALOAD","gas":24978857,"gasCost":3,"depth":1,"stack":["0x0"]},{"pc":3,"op":"PUSH1","gas":24978854,"gasCost":3,"depth":1,"stack":["0x1"]},{"pc":5,"op":"ADD","gas":24978851,"gasCost":3,"depth":1,"stack":["0x1","0x1"]},{"pc":6,"op":"PUSH1","gas":24978848,"gasCost":3,"depth":1,"stack":["0x2"]},{"pc":8,"op":"MSTORE","gas":24978845,"gasCost":6,"depth":1,"stack":["0x2","0x0"]},{"pc":9,"op":"PUSH1","gas":24978839,"gasCost":3,"depth":1,"stack":[]},{"pc":11,"op":"PUSH1","gas":24978836,"gasCost":3,"depth":1,"stack":["0x20"]},{"pc":13,"op":"RETURN","gas":24978833,"gasCost":0,"depth":1,"stack":["0x20","0x0"]}]}
api_test.go:1013: test 10, result mismatch, have
{21167 false 0x0000000000000000000000000000000000000000000000000000000000000002}
, want
{21167 false 0000000000000000000000000000000000000000000000000000000000000002}
api_test.go:1012: result: {"gas":25664,"failed":false,"returnValue":"0x000000000000000000000000c6e93f4c1920eaeaa1e699f76a7a8c18e3056074","structLogs":[]}
api_test.go:1013: test 11, result mismatch, have
{25664 false 0x000000000000000000000000c6e93f4c1920eaeaa1e699f76a7a8c18e3056074}
, want
{25664 false 000000000000000000000000c6e93f4c1920eaeaa1e699f76a7a8c18e3056074}
```
This is a **breaking change** to the opcode tracer. The top-level
`returnValue` field of a trace will be now hex-encoded. If the return
data is empty, this field will contain "0x".
Fixes#31196
This PR fixes the broken request error handling of the workload filter
tests. Until now `validateHistoryPruneErr` was invoked with `fq.Err` as
an input which was always nil and a timeout or http error was reported
as a result content mismatch.
Also, in case of `errPrunedHistory` it is wrong to return here without
setting an error because then it will look like a valid empty result and
the check will later fail. So instead `errPrunedHistory` is always
returned now (without printing an error message) and the callers of
`run` should handle this special case (typically ignore silently).
This removes the signer type-train in favor of defining a single object
that can handle all tx types. Supported types are enabled via a map.
Notably, the new signer also supports disabling legacy transactions.
This PR roughly halves the number of allocations needed to compute the
sigHash for a transaction.
This sigHash is used whenever we recover a signature of a transaction,
so quite often. During a recent benchmark full syncing on Holesky,
roughly 2.8% of all allocations were happening here because the fields
from the transaction would be copied multiple times.
```
66168733 153175654 (flat, cum) 2.80% of Total
. . 368:func (s londonSigner) Hash(tx *Transaction) common.Hash {
. . 369: if tx.Type() != DynamicFeeTxType {
. . 370: return s.eip2930Signer.Hash(tx)
. . 371: }
. 19169966 372: return prefixedRlpHash(
. . 373: tx.Type(),
26442187 26442187 374: []interface{}{
. . 375: s.chainId,
6848616 6848616 376: tx.Nonce(),
. 19694077 377: tx.GasTipCap(),
. 18956774 378: tx.GasFeeCap(),
6357089 6357089 379: tx.Gas(),
. 12321050 380: tx.To(),
. 16865054 381: tx.Value(),
13435187 13435187 382: tx.Data(),
13085654 13085654 383: tx.AccessList(),
. . 384: })
. . 385:}
```
This PR reduces the allocations and speeds up the computation of the
sigHash by ~22%, which is quite significantly given that this operation
involves a call to Keccak
```
// BenchmarkHash-8 440082 2639 ns/op 384 B/op 13 allocs/op
// BenchmarkHash-8 493566 2033 ns/op 240 B/op 6 allocs/op
```
```
Hash-8 2.691µ ± 8% 2.097µ ± 9% -22.07% (p=0.000 n=10)
```
It also kinda cleans up stuff in my opinion, since the transaction
should itself know best how to compute the sighash
---------
Co-authored-by: Gary Rong <garyrong0905@gmail.com>
Currently, when answering GetPooledTransaction request, txpool.Get() is
used. When the requested hash is blob transaction, blobpool.Get() is
called. This function loads the RLP-encoded transaction from limbo then
decodes and returns. Later, in answerGetPooledTransactions, we need to
RLP encode again. This decode then encode is wasteful. This commit adds
GetRLP to transaction pool interface so that answerGetPooledTransactions
can use the RLP-encoded from limbo directly.
---------
Co-authored-by: Gary Rong <garyrong0905@gmail.com>
The main use case I see of this is that it allows users to estimate gas
against the same state that they query for their nonce, and the same
state they base the data of their transaction against. This helps ensure
that gas estimation won't fail and the transaction won't revert on-chain
because of a mismatch between the state used for gas estimation and the
state used to generate the inputs to gas estimation or the transaction's
nonce when submitted to the mempool.
This PR also updates the EstimateGas comment based on the new geth
`eth_estimateGas` default of using latest state as of v1.12.0:
https://github.com/ethereum/go-ethereum/pull/24363
---------
Co-authored-by: Felix Lange <fjl@twurst.com>
When I'm running `geth import --metrics`, the metrics is different to
normal `geth --metrics`, so the grafana dashboard needs to be updated,
eg: `eth_db_chaindata_disk_read` vs `disk_read`.
So I think we should always set the name to `eth/db/chaindata` for more
convenient.
---------
Signed-off-by: jsvisa <delweng@gmail.com>
This PR fixes a bug in the `lastMapBoundaryBefore` logic that resulted
in incorrect checkpoint initialization (started rendering from the
previous epoch boundary which caused the `needTailEpoch` check to fail).
Apparently the bug was present before but went unnoticed because
`needTailEpoch` behaved differently.
Fixes https://github.com/ethereum/go-ethereum/issues/31413
