c5b36b1c1b Mempool Update Cut-Through Optimization (Jeremy Rubin)
c49daf9885 [TESTS] Increase limitancestorcount in tournament RPC test to showcase improved algorithm (Jeremy Rubin)
Pull request description:
Often when we're updating mempool entries we update entries that we ultimately end up removing the updated entries shortly thereafter. This patch makes it so that we filter for such entries a bit earlier in processing, which yields a mild improvement for these cases, and is negligible overhead otherwise.
There's potential for a better -- but more sophisticated -- algorithm that can be used taking advantage of epochs, but I figured it is better to do something that is simple and works first and upgrade it later as the other epoch mempool work proceeds as it makes the patches for the epoch algorithm simpler to understand, so you can consider this as preparatory work. It could either go in now if it is not controversial, or we could wait until the other patch is ready to go.
ACKs for top commit:
instagibbs:
reACK c5b36b1
sipa:
utACK c5b36b1c1b
mzumsande:
Code Review ACK c5b36b1c1b
Tree-SHA512: 78b16864f77a637d8a68a65e23c019a9757d8b2243486728ef601d212ae482f6084cf8e69d810958c356f1803178046e4697207ba40d6d10529ca57de647fae6
No behavior change.
This code was introduced in 5add7a7 before we required C++11, which is
why the struct was needed. As we are now using more modern C++ and this
is the only place where lockpoints are updated for mempool entries, it
is more idiomatic to call `modify` with a lambda.
Co-authored-by: Hennadii Stepanov <32963518+hebasto@users.noreply.github.com>
b4adc5ad67 [bugfix] update lockpoints correctly during reorg (glozow)
b6002b07a3 MOVEONLY: update_lock_points to txmempool.h (glozow)
Pull request description:
I introduced a bug in #22677 (sorry! 😅)
Mempool entries cache `LockPoints`, containing the first height/blockhash/`CBlockIndex*` at which the transaction becomes valid. During a reorg, we re-check timelocks on all mempool entries using `CheckSequenceLocks(useExistingLockPoints=false)` and remove any now-invalid entries. `CheckSequenceLocks()` also mutates the `LockPoints` passed in, and we update valid entries' `LockPoints` using `update_lock_points`. Thus, `update_lock_points(lp)` needs to be called right after `CheckSequenceLocks(lp)`, otherwise we lose the data in `lp`. I incorrectly assumed they could be called in separate loops.
The incorrect behavior introduced is: if we have a reorg in which a timelocked mempool transaction is still valid but becomes valid at a different block, the cached `LockPoints` will be incorrect.
This PR fixes the bug, adds a test, and adds an assertion at the end of `removeForReorg()` to check that all mempool entries' lockpoints are valid. You can reproduce the bug by running the test added in the [test] commit on the code before the [bugfix] commit.
ACKs for top commit:
jnewbery:
ACK b4adc5ad67
vasild:
ACK b4adc5ad67
mzumsande:
Code Review ACK b4adc5ad67
hebasto:
ACK b4adc5ad67
MarcoFalke:
re-ACK b4adc5ad67🏁
Tree-SHA512: 16b59f6ff8140d0229079ca1c6b04f2f4a00a2e49931275150e4f3fe5ac4ec109698b083fa6b223ba9511f328271cc1ab081263669d5da020af7fee83c13e401
Often when we're updating mempool entries we update entries that we
ultimately end up removing the updated entries shortly thereafter. This
patch makes it so that we filter for such entries a bit earlier in
processing, which yields a mild improvement for these cases, and is
negligible overhead otherwise.
No behavior change.
Parameterize removeForReorg using a CChain and callable that
encapsulates validation logic. The mempool shouldn't need to know a
bunch of details about coinbase maturity and lock finality. Instead,
just pass in a callable function that says true/false. Breaks circular
dependency by removing txmempool's dependency on validation.
65aaf9495d refactor: move `update_*` structs from txmempool.h to .cpp file (Sebastian Falbesoner)
9947ce6262 refactor: use const reference for parents in `CTxMemPool::UpdateAncestorsOf` (Sebastian Falbesoner)
Pull request description:
These helpers are exclusively used in txmempool.cpp, hence they should also be moved there. The PR also contains a commit which fixes const-correctness for parents in `CTxMemPool::UpdateAncestorsOf` and declares them as reference to avoid a copy.
ACKs for top commit:
promag:
Code review ACK 65aaf9495d. Verified move-only commit locally.
Tree-SHA512: 7ce29f3ba0e68b5355001f27725b00f6d54cc993015356eb40b61b8cdd17db49b980f4c3d798c8e0c940d245dc3a72c474bb9ff3c0ee971ead450786076812c2
082c5bf099 [refactor] pass coinsview and height to check() (glozow)
ed6115f1ea [mempool] simplify some check() logic (glozow)
9e8d7ad5d9 [validation/mempool] use Spend/AddCoin instead of UpdateCoins (glozow)
09d18916af MOVEONLY: remove single-use helper func CheckInputsAndUpdateCoins (glozow)
e8639ec26a [mempool] remove now-unnecessary code (glozow)
54c6f3c1da [mempool] speed up check() by using coins cache and iterating in topo order (glozow)
30e240f65e [bench] Benchmark CTxMemPool::check() (glozow)
cb1407196f [refactor/bench] make mempool_stress bench reusable and parameterizable (glozow)
Pull request description:
Remove the txmempool <-> validation circular dependency by removing txmempool's dependency on validation. There are two functions in txmempool that need validation right now: `check()` and `removeForReorg()`. This PR removes the dependencies in `check()`.
This PR also improves the performance of `CTxMemPool::check()` by walking through the entries exactly once, in ascending ancestorcount order, which guarantees that we see parents before children.
ACKs for top commit:
jnewbery:
reACK 082c5bf099
GeneFerneau:
tACK [082c5bf](082c5bf099)
rajarshimaitra:
tACK 082c5bf099
theStack:
Code-review ACK 082c5bf099
Tree-SHA512: 40ac622af1627b5c3e6abb4f0f035d833265a8c5e8dc88faf5354875dfb5137f137825e54bbd2a2668ed37b145c5d02285f776402629f58596e51853a9a79d29
This removes a bunch of boilerplate, makes the code easier to read.
Also, C++11 member initialization avoids accidental uninitialized
members.
Can be reviewed with the git option "--word-diff-regex=." or with "git
difftool --tool=meld".
When calculating ancestor/descendant counts for transactions in the
package, as a heuristic, count every transaction in the package as an
ancestor and descendant of every other transaction in the package.
This may overestimate, but will not underestimate, the
ancestor/descendant counts. This shortcut still produces an accurate
count for packages of 1 parent + 1 child.
This does not change existing behavior.
The ancestor/descendant limits are inclusive of the entries themselves,
but CalculateAncestorsAndCheckLimits() does not need access to them.
13650fe2e5 [policy] detect unsorted packages (glozow)
9ef643e21b [doc] add release note for package testmempoolaccept (glozow)
c4259f4b7e [test] functional test for packages in RPCs (glozow)
9ede34a6f2 [rpc] allow multiple txns in testmempoolaccept (glozow)
ae8e6df709 [policy] limit package sizes (glozow)
c9e1a26d1f [fuzz] add ProcessNewPackage call in tx_pool fuzzer (glozow)
363e3d916c [test] unit tests for ProcessNewPackage (glozow)
cd9a11ac96 [test] make submit optional in CreateValidMempoolTransaction (glozow)
2ef187941d [validation] package validation for test accepts (glozow)
578148ded6 [validation] explicit Success/Failure ctors for MempoolAcceptResult (glozow)
b88d77aec5 [policy] Define packages (glozow)
249f43f3cc [refactor] add option to disable RBF (glozow)
897e348f59 [coins/mempool] extend CCoinsViewMemPool to track temporary coins (glozow)
42cf8b25df [validation] make CheckSequenceLocks context-free (glozow)
Pull request description:
This PR enables validation dry-runs of packages through the `testmempoolaccept` RPC. The expectation is that the results returned from `testmempoolaccept` are what you'd get from test-then-submitting each transaction individually, in that order (this means the package is expected to be sorted in topological order, for now at least). The validation is also atomic: in the case of failure, it immediately halts and may return "unfinished" `MempoolAcceptResult`s for transactions that weren't fully validated. The API for 1 transaction stays the same.
**Motivation:**
- This allows you to test validity for transaction chains (e.g. with multiple spending paths and where you don't want to broadcast yet); closes#18480.
- It's also a first step towards package validation in a minimally invasive way.
- The RPC commit happens to close#21074 by clarifying the "allowed" key.
There are a few added restrictions on the packages, mostly to simplify the logic for areas that aren't critical to main package use cases:
- No package can have conflicts, i.e. none of them can spend the same inputs, even if it would be a valid BIP125 replacement.
- The package cannot conflict with the mempool, i.e. RBF is disabled.
- The total count of the package cannot exceed 25 (the default descendant count limit), and total size cannot exceed 101KvB (the default descendant size limit).
If you're looking for review comments and github isn't loading them, I have a gist compiling some topics of discussion [here](https://gist.github.com/glozow/c3acaf161c95bba491fce31585b2aaf7)
ACKs for top commit:
laanwj:
Code review re-ACK 13650fe2e5
jnewbery:
Code review ACK 13650fe2e5
ariard:
ACK 13650fe
Tree-SHA512: 8c5cbfa91a6c714e1c8710bb281d5ff1c5af36741872a7c5df6b24874d6272b4a09f816cb8a4c7de33ef8e1c2a2c252c0df5105b7802f70bc6ff821ed7cc1a2f
e11b649650 validation: CVerifyDB::VerifyDB: Use locking annotation (Carl Dong)
03f75c42e1 validation: Use existing chain member in CChainState::LoadGenesisBlock (Carl Dong)
5e4af77380 validation: Use existing chain member in CChainState::AcceptBlock (Carl Dong)
fee73347c0 validation: Pass in chain to FindBlockPos+SaveBlockToDisk (Carl Dong)
a9d28bcd8d validation: Use *this in CChainState::ActivateBestChainStep (Carl Dong)
4744efc9ba validation: Pass in chainstate to CTxMemPool::check (Carl Dong)
1fb7b2c595 validation: Use *this in CChainState::InvalidateBlock (Carl Dong)
8cdb2f7e58 validation: Move LoadBlockIndexDB to CChainState (Carl Dong)
8b99efbcc0 validation: Move invalid block handling to CChainState (Carl Dong)
2bdf37fe18 validation: Pass in chainstate to CVerifyDB::VerifyDB (Carl Dong)
31eac50c72 validation: Remove global ::VersionBitsTip{State,SinceHeight,Statistics} (Carl Dong)
63e4c7316a validation: Pass in chainstate to ::PruneBlockFilesManual (Carl Dong)
4bada76237 validation: Pass in chainstate to UpdateTip (Carl Dong)
a3ba08ba7d validation: Remove global ::{{Precious,Invalidate}Block,ResetBlockFailureFlags} (Carl Dong)
4927c9e699 validation: Remove global ::LoadGenesisBlock (Carl Dong)
9da106be4d validation: Check chain tip is non-null in CheckFinalTx (Carl Dong)
Pull request description:
Overall PR: #20158 (tree-wide: De-globalize ChainstateManager)
Based on:
- [x] #20750 | [Bundle 2/n] Prune g_chainman usage in mempool-related validation functions
Note to reviewers:
1. This bundle may _apparently_ introduce usage of `g_chainman` or `::Chain(state|)Active()` globals, but these are resolved later on in the overall PR. [Commits of overall PR](https://github.com/bitcoin/bitcoin/pull/20158/commits)
2. There may be seemingly obvious local references to `ChainstateManager` or other validation objects which are not being used in callers of the current function in question, this is done intentionally to **_keep each commit centered around one function/method_** to ease review and to make the overall change systematic. We don't assume anything about our callers. Rest assured that once we are considering that particular caller in later commits, we will use the obvious local references. [Commits of overall PR](https://github.com/bitcoin/bitcoin/pull/20158/commits)
3. When changing a function/method that has many callers (e.g. `LookupBlockIndex` with 55 callers), it is sometimes easier (and less error-prone) to use a scripted-diff. When doing so, there will be 3 commits in sequence so that every commit compiles like so:
1. Add `new_function`, make `old_function` a wrapper of `new_function`, divert all calls to `old_function` to `new_function` **in the local module only**
2. Scripted-diff to divert all calls to `old_function` to `new_function` **in the rest of the codebase**
3. Remove `old_function`
Note to self:
- [x] Address: https://github.com/bitcoin/bitcoin/pull/20750#discussion_r579400663
ACKs for top commit:
laanwj:
Code review ACK e11b649650
Tree-SHA512: 205a451a741e32f17d5966de289f2f5a3f0817738c0087b70ff4755ddd217b53d01050ed396669bda2b1d216a88d927b9778777f9ff95ab1fe20e59c5f341776
fd6580e405 [refactor] txmempool: split epoch logic into class (Anthony Towns)
Pull request description:
Splits the epoch logic introduced in #17925 into a separate class.
Uses clang's thread safety annotations and encapsulates the data more strongly to reduce chances of bugs from API misuse.
ACKs for top commit:
jonatack:
ACK fd6580e405 using clang thread safety annotations looks like a very good idea, and the encapsulation this change adds should improve robustness (and possible unit test-ability) of the code. Verified that changing some of the locking duly provoked build-time warnings with Clang 9 on Debian and that small changes in the new `Epoch` class were covered by failing functional test assertions in `mempool_updatefromblock.py`, `mempool_resurrect.py`, and `mempool_reorg.py`
hebasto:
re-ACK fd6580e405, since my [previous](https://github.com/bitcoin/bitcoin/pull/18017#pullrequestreview-569619362) review:
Tree-SHA512: 7004623faa02b56639aa05ab7a078320a6d8d54ec62d8022876221e33f350f47df51ddff056c0de5be798f8eb39b5c03c2d3f035698555d70abc218e950f2f8c
Several other parameters are now redundant since they can be safely
obtained from the chainstate given that ::cs_main is locked. These are
now removed.
Currently, CTxMemPool::check locks CTxMemPool's own cs member, then
calls GetSpendHeight which locks cs_main. This can potentially cause an
undesirable lock invesion since CTxMemPool's cs is supposed to be locked
after cs_main.
This does not cause us any problems right now because all callers of
CTxMemPool already lock cs_main before calling CTxMemPool::check, which
means that the LOCK(cs_main) in GetSpendHeight becomes benign.
However, it is currently possible for new code to be added which calls
CTxMemPool::check without locking cs_main (which would be dangerous).
Therefore we should make it explicit that cs_main needs to be held
before calling CTxMemPool::check.
NOTE: After all review-only assertions are removed in "#20158 |
tree-wide: De-globalize ChainstateManager", and assuming that we
keep the changes in "validation: Pass in spendheight to
CTxMemPool::check", we can re-evaluate to see if this annotation
is still necessary.
281fd1a4a0 Replace KeyIDHasher with SaltedSipHasher (Andrew Chow)
210b693db6 Add generic SaltedSipHasher (Andrew Chow)
95e61c1cf2 Move Hashers to util/hasher.{cpp/h} (Andrew Chow)
Pull request description:
There are existing `SaltedOutPointHasher` and `SaltedTxidHasher` classes used for `std::unordered_map` and `std::unordered_set` that could be useful in other places in the codebase. So we these to their own `saltedhash.{cpp/h}` file. An existing `KeyIDHasher` is moved there too. Additionally, `ScriptIDHasher`, `SaltedPubkeyHasher`, and `SaltedScriptHasher` are added so that they can be used in future work.
`KeyIDHasher` and `ScriptIDHasher` are not salted so that equality comparisons of maps and sets keyed by `CKeyID` and `CScriptID` will actually work.
Split from #19602 (and a few other PRs/branches I have).
ACKs for top commit:
laanwj:
Code review ACK 281fd1a4a0
jonatack:
ACK 281fd1a4a0, code review, debug build and ran bitcoind after rebasing to master @ dff0f6f753
fjahr:
utACK 281fd1a4a0
Tree-SHA512: bb03b231ccf3c9ecefc997b8da9c3770af4819f9be5b0a72997a103864e84046a2ac39b8eadf0dc9247bdccd53f86f433642e3a098882e6748341a9e7736271b
Move the hashers that we use for hash tables to a common place.
Moved hashers:
- SaltedTxidHasher
- SaltedOutpointHasher
- FilterHeaderHasher
- SignatureCacheHasher
- BlockHasher
Since m_check_ratio is only set once and since the CTxMemPool object is
no longer a global variable, m_check_ratio can be passed into the
constructor of CTxMemPool. Since it is only read from after
initialization, m_check_ratio can also be made a const and hence no
longer needs to be guarded by the cs mutex.
Using the zmq notifications to avoid excessive mempool polling can be difficult
given the current notifications available. It announces all transactions
being added to mempool or included in blocks, but announces no evictions
and gives no indication if the transaction is in the mempool or a block.
Block notifications for zmq are also substandard, in that it only announces
block tips, while all block transactions are still announced.
This commit adds a unified stream which can be used to closely track mempool:
1) getrawmempool to fill out mempool knowledge
2) if txhash is announced, add or remove from set
based on add/remove flag
3) if blockhash is announced, get block txn list,
remove from those transactions local view of mempool
4) if we drop a sequence number, go to (1)
The mempool sequence number starts at the value 1, and
increments each time a transaction enters the mempool,
or is evicted from the mempool for any reason, including
block inclusion. The mempool sequence number is published
via ZMQ for any transaction-related notification.
These features allow for ZMQ/RPC consumer to track mempool
state in a more exacting way, without unnecesarily polling
getrawmempool. See interface_zmq.py::test_mempool_sync for
example usage.
a8a64acaf3 [BroadcastTransaction] Remove unsafe move operator (Amiti Uttarwar)
125c038126 [p2p] Remove dead code (Amiti Uttarwar)
fc66d0a65c [p2p] Check for nullptr before dereferencing pointer (Adam Jonas)
cb79b9dbf4 [mempool] Revert unbroadcast set to tracking just txid (Amiti Uttarwar)
Pull request description:
Addresses some outstanding review comments from #18044
- reverts unbroadcast txids to a set instead of a map (simpler, communicates intent better, takes less space, no efficiency advantages of map)
- adds safety around two touchpoints (check for nullptr before dereferencing pointer, remove an inaccurate std::move operator)
- removes some dead code
Links to comments on wtxid PR: [1](https://github.com/bitcoin/bitcoin/pull/18044#discussion_r460495254) [2](https://github.com/bitcoin/bitcoin/pull/18044#discussion_r460496023) [3](https://github.com/bitcoin/bitcoin/pull/18044#discussion_r463532611)
thanks to jnewbery & adamjonas for flagging these ! !
ACKs for top commit:
sdaftuar:
utACK a8a64acaf3
naumenkogs:
utACK a8a64acaf3
jnewbery:
utACK a8a64acaf3
Tree-SHA512: 7be669cb30cc17fb9e06b50e636ef7887c6a27354697987e4e4d38dba4b8f50e175647587430cd9bc3295bec01ce8b1e6639a50a4249d8fff9b1ca1b9ead3277
When I originally implemented the unbroadcast set in 18038, it just tracked
txids. After 18038 was merged, I offered a patch to 18044 to make the
unbroadcast changes compatible with wtxid relay. In this patch, I updated
`unbroadcast_txids` to a map of txid -> wtxid. Post merge review comments shed
light on the fact that this update was unnecessary, and distracting. So, this
commit updates the unbroadcast ids back to a set.
