62ed1f92ef txgraph: check that DoWork finds optimal if given high budget (tests) (Pieter Wuille)
f3c2fc867f txgraph: add work limit to DoWork(), try optimal (feature) (Pieter Wuille)
e96b00d99e txgraph: make number of acceptable iterations configurable (feature) (Pieter Wuille)
cfe9958852 txgraph: track amount of work done in linearization (preparation) (Pieter Wuille)
6ba316eaa0 txgraph: 1-or-2-tx split-off clusters are optimal (optimization) (Pieter Wuille)
fad0eb091e txgraph: reset quality when merging clusters (bugfix) (Pieter Wuille)
Pull request description:
Part of #30289. Builds on top of #31553.
So far, the `TxGraph::DoWork()` function took no parameters, and just made all clusters reach the "acceptable" internal quality level by performing a minimum number of improvement iterations on it, but:
* Did not attempt to go beyond that.
* Was broken, as the QualityLevel of optimal clusters that merge together was not being reset.
Fix this by adding an argument to `DoWork()` to control how much work it is allowed to do right now, which will first be used to get all clusters to the acceptable level, and if more budget remains, use it to try to get some or all clusters optimal. The function will now return `true` if all clusters are known to be optimal (and thus no further work remains). This is verified in the tests, by remembering whether the graph is optimal, and if it is at the end of the simulation run, verify that the overall linearization cannot be improved further.
ACKs for top commit:
instagibbs:
ACK 62ed1f92ef
ismaelsadeeq:
Code review ACK 62ed1f92ef
glozow:
ACK 62ed1f92ef
Tree-SHA512: 5f57d4052e369f3444e72e724f04c02004e0f66e365faa59c9f145323e606508380fc97bb038b68783a62ae9c10757f1b628b3b00b2ce9a46161fca2d4336d73
1cb2399703 doc: clarify the GetAddresses/GetAddressesUnsafe documentation (Daniela Brozzoni)
e5a7dfd79f p2p: rename GetAddresses -> GetAddressesUnsafe (Daniela Brozzoni)
Pull request description:
Rename GetAddresses to GetAddressesUnsafe to make it clearer that this function should only be used in trusted contexts. This helps avoid accidental privacy leaks by preventing the uncached version from being used in non-trusted scenarios, like P2P.
Additionally, better reflect in the documentation that the two methods should be used in different contexts.
Also update the outdated "call the function without a parameter" phrasing in the cached version. This wording was accurate when the cache was introduced in #18991, but became outdated after later commits (f26502e9fc, 81b00f8780) added parameters to each
function, and the previous commit changed the function naming completely.
ACKs for top commit:
stickies-v:
re-ACK 1cb2399703
l0rinc:
ACK 1cb2399703
luisschwab:
ACK 1cb2399703
brunoerg:
ACK 1cb2399703
theStack:
Code-review ACK 1cb2399703
mzumsande:
Code review ACK 1cb2399703
Tree-SHA512: 02c05d88436abcdfabad994f47ec5144e9ba47668667a2c1818f57bf8710727505faf8426fd0672c63de14fcf20b96f17cea2acc39fe3c1f56abbc2b1a9e9c23
fa1a14a13a fuzz: Reset chainman state in process_message(s) targets (MarcoFalke)
fa9a3de09b fuzz: DisableNextWrite (MarcoFalke)
aeeeeec9f7 fuzz: Reset dirty connman state in process_message(s) targets (MarcoFalke)
fa11eea405 fuzz: Avoid non-determinism in process_message(s) target (PeerMan) (MarcoFalke)
Pull request description:
`process_message(s)` are the least stable fuzz targets, according to OSS-Fuzz.
Tracking issue: https://github.com/bitcoin/bitcoin/issues/29018.
### Testing
Needs coverage compilation, as explained in `./contrib/devtools/README.md`. And then, using 32 threads:
```
cargo run --manifest-path ./contrib/devtools/deterministic-fuzz-coverage/Cargo.toml -- $PWD/bld-cmake/ $PWD/../b-c-qa-assets/fuzz_corpora/ process_messages 32
```
Each commit can be reverted to see more non-determinism re-appear.
ACKs for top commit:
marcofleon:
ReACK fa1a14a13a
dergoegge:
reACK fa1a14a13a
Tree-SHA512: 37b5b6dbdde6a39b4f83dc31e92cffb4a62a4b8f5befbf17029d943d0b2fd506f4a0833570dcdbf79a90b42af9caca44e98e838b03213d6bc1c3ecb70a6bb135
Rename GetAddresses to GetAddressesUnsafe to make it clearer that this
function should only be used in trusted contexts. This helps avoid
accidental privacy leaks by preventing the uncached version from being
used in non-trusted scenarios, like P2P.
31c4e77a25 test: fix ReadTopologicalSet unsigned integer overflow (ismaelsadeeq)
Pull request description:
This PR is a simple fix for a potential unsigned integer overflow in ReadTopologicalSet.
We obtain the value of `mask` from fuzz input, which can be the maximum representable value.
Adding 1 to it would then cause an overflow.
The fix skips the addition when the read value is already the maximum.
See https://github.com/bitcoin/bitcoin/pull/30605#discussion_r2215338569 for more context
ACKs for top commit:
maflcko:
lgtm ACK 31c4e77a25
Tree-SHA512: f58d7907f66a0de0ed8d4b1cad6a4971f65925a99f3c030537c21c4d84126b643257c65865242caf7d445b9cbb7a71a1816a9f870ab7520625c4c16cd41979cb
This is required in the process_message(s) fuzz targets to avoid leaking
the next write time from one run to the next. Also, disable it
completely because it is not needed and due to leveldb-internal
non-determinism.
The PeerManager has several members, such as the FastRandomContext,
which need to be reset before every run to avoid leaking state from one
run into the next.
Also, style fixups in p2p_handshake.cpp, where this code is copied from.
Since 31 byte xor-keys are not used in the codebase, using the common size (8 bytes) makes the benchmarks more realistic.
Co-authored-by: maflcko <6399679+maflcko@users.noreply.github.com>
This adds a large simulation fuzz test for all TxOrphanage public interface
functions, using a mix of comparison with expected behavior (in case it is
fully specified), and testing of properties exhibited otherwise.
For the default number of peers (125), allows each to relay a default
descendant package (up to 25-1=24 can be missing inputs) of small (9
inputs or fewer) transactions out of order.
This limit also gives acceptable bounds for worst case LimitOrphans iterations.
Functional tests aren't changed to check for larger cap because it would
make the runtime too long.
Also deletes the now-unused DEFAULT_MAX_ORPHAN_TRANSACTIONS.
Move towards a model where TxOrphanage is initialized with limits that
it remembers throughout its lifetime.
Remove the param. Limiting by number of unique orphans will be removed
in a later commit.
Now that -maxorphantx is gone, this does not change the node behavior.
The parameter is only used in tests.
This adds an `iters` parameter to DoWork(), which controls how much work it is
allowed to do right now.
Additionally, DoWork() won't stop at just getting everything ACCEPTABLE, but if
there is work budget left, will also attempt to get every cluster linearized
optimally.
a60f863d3e scripted-diff: Replace GenTxidVariant with GenTxid (marcofleon)
c8ba199598 Remove old GenTxid class (marcofleon)
072a198ea4 Convert remaining instances of GenTxid to GenTxidVariant (marcofleon)
1b528391c7 Convert `txrequest` to GenTxidVariant (marcofleon)
bde4579b07 Convert `txdownloadman_impl` to GenTxidVariant (marcofleon)
c876a892ec Replace GenTxid with Txid/Wtxid overloads in `txmempool` (marcofleon)
de858ce2be move-only: make GetInfo a private CTxMemPool member (stickies-v)
eee473d9f3 Convert `CompareInvMempoolOrder` to GenTxidVariant (marcofleon)
243553d590 refactor: replace get_iter_from_wtxid with GetIter(const Wtxid&) (stickies-v)
fcf92fd640 refactor: make CTxMemPool::GetIter strongly typed (marcofleon)
11d28f21bb Implement GenTxid as a variant (marcofleon)
Pull request description:
Part of the [type safety refactor](https://github.com/bitcoin/bitcoin/pull/32189).
This PR changes the GenTxid class to a variant, which holds both Txids and Wtxids. This provides compile-time type safety and eliminates the manual type check (bool m_is_wtxid). Variables that can be either a Txid or a Wtxid are now using the new GenTxid variant, instead of uint256.
ACKs for top commit:
w0xlt:
ACK a60f863d3e
dergoegge:
Code review ACK a60f863d3e
maflcko:
review ACK a60f863d3e🎽
theStack:
Code-review ACK a60f863d3e
Tree-SHA512: da9b73b7bdffee2eb9281a409205519ac330d3336094d17681896703fbca8099608782c9c85801e388e4d90af5af8abf1f34931f57bbbe6e9674d802d6066047
fa8862723c fuzz: CheckGlobals in init (MarcoFalke)
fa26bfde98 test: Avoid resetting mocktime in testing setup (MarcoFalke)
fa6b45fa8e Add SetMockTime for time_point types (MarcoFalke)
Pull request description:
(Tracking issue https://github.com/bitcoin/bitcoin/issues/29018)
During fuzzing, `AppInitParameterInteraction` may actually disable a previously set mocktime. This is confusing and can also cause non-determinism.
Fix this issue, by
* fixing the erroneous `-mocktime` parsing in `AppInitParameterInteraction`.
* adding the missing `SetMockTime` calls to the affected fuzz init functions.
* adding a `CheckGlobals` to the fuzz init, to prevent this issue in the future.
This can be tested by
* Cherry-picking the `CheckGlobals`-commit onto current master and observing a fuzz failure in the touched fuzz targets.
* Reverting the touched fuzz fixups and observing a fuzz failure for each target.
ACKs for top commit:
w0xlt:
ACK fa8862723c
dergoegge:
utACK fa8862723c
Tree-SHA512: 5a9400f0467c82fa224713af4cc2b525afbefefc7c3f419077110925ad7af6c7fda3dcd2b50f7facf0ee7df2547c6ac20336906d707adcdfd1d652a9d9a735fe
d7fca5c171 clusterlin: add big comment explaning the relation between tests (Pieter Wuille)
b64e61d2de clusterlin: abstract try-permutations into ExhaustiveLinearize function (Pieter Wuille)
1fa55a64ed clusterlin tests: verify that chunks are minimal (Pieter Wuille)
da23ecef29 clusterlin tests: support non-empty ReadTopologicalSubset() (Pieter Wuille)
94f3e17c33 clusterlin tests: compare with fuzz-provided linearizations (Pieter Wuille)
5f92ebee0d clusterlin tests: compare with fuzz-provided topological sets (Pieter Wuille)
6e37824ac3 clusterlin tests: optimize clusterlin_simple_linearize (Pieter Wuille)
98c1c88b6f clusterlin tests: separate testing of SimpleLinearize and Linearize (Pieter Wuille)
10e90f7aef clusterlin tests: make SimpleCandidateFinder always find connected (Pieter Wuille)
a38c38951e clusterlin tests: separate testing of Search- and SimpleCandidateFinder (Pieter Wuille)
77a432ee70 clusterlin tests: count SimpleCandidateFinder iterations better (Pieter Wuille)
Pull request description:
Part of the cluster mempool project: #30289
The current cluster linearization fuzz tests contain two tests which combine testing of production code with testing of the test code itself:
* `clusterlin_search_finder`: establishes the correctness of `SearchCandidateFinder` by comparing against both `SimpleCandidateFinder` and `ExhaustiveCandidateFinder` (which is even more simple than `SimpleCandidateFinder`). If `SimpleCandidateFinder` works correctly, then this comparison with `ExhaustiveCandidateFinder` is redundant. If it isn't, we ought to find that in a test specific to `SimpleCandidateFinder` rather than as a side-effect of testing `SearchCandidateFinder`. Split this functionality out into a new `clusterlin_simple_finder`.
* `clusterlin_linearize`: establishes the correctness of `Linearize` by comparing against both `SimpleLinearize` and literally every valid linearization for the cluster. Again, if `SimpleLinearize` works correctly, then this comparison with all valid linearizations is redundant, and if it isn't we should find it in a test for `SimpleLinearize`. Do so by splitting off that functionality into `clusterlin_simple_linearize`.
After that, a few general improvements to the affected tests are made (comparing with linearizations and subsets read from the fuzz input, plus a performance improvement).
ACKs for top commit:
marcofleon:
Re ACK d7fca5c171
ismaelsadeeq:
re-ACK d7fca5c171
monlovesmango:
ACK d7fca5c171
Tree-SHA512: 33cb76bd9b9547a5f3ee231fa452e928f064ad03af98e3d9e64246eb972f2b026c13e7367257ccdac1ae57982ee8ef98c907684588ecbb4bc4c82cbec160b3e8
8cc3ac6c23 validation: Don't use IsValid() to filter for invalid blocks (Martin Zumsande)
86d98b94e5 test: verify that ancestors of a reconsidered block can become the chain tip (stratospher)
3c39a55e64 validation: Add ancestors of reconsiderblock to setBlockIndexCandidates (Martin Zumsande)
Pull request description:
When we call `reconsiderblock` for some block, `Chainstate::ResetBlockFailureFlags` puts the descendants of that block into `setBlockIndexCandidates` (if they meet the criteria, i.e. have more work than the tip etc.), but never put any ancestors into the set even though we do clear their failure flags.
I think that this is wrong, because `setBlockIndexCandidates` should always contain all eligible indexes that have at least as much work as the current tip, which can include ancestors of the reconsidered block. This is being checked by `CheckBlockIndex()`, which could fail if it was invoked after `ActivateBestChain` connects a block and releases `cs_main`:
``` diff
diff --git a/src/validation.cpp b/src/validation.cpp
index 7b04bd9a5b..ff0c3c9f58 100644
--- a/src/validation.cpp
+++ b/src/validation.cpp
@@ -3551,6 +3551,7 @@ bool Chainstate::ActivateBestChain(BlockValidationState& state, std::shared_ptr<
}
}
// When we reach this point, we switched to a new tip (stored in pindexNewTip).
+ m_chainman.CheckBlockIndex();
if (exited_ibd) {
// If a background chainstate is in use, we may need to rebalance our
```
makes `rpc_invalidateblock.py` fail on master.
Even though we don't currently have a `CheckBlockIndex()` in that place, after `cs_main` is released other threads could invoke it, which is happening in the rare failures of #16444 where an invalid header received from another peer could trigger a `CheckBlockIndex()` call that would fail.
Fix this by adding eligible ancestors to `setBlockIndexCandidates` in `Chainstate::ResetBlockFailureFlags` (also simplifying that function a bit).
Fixes#16444
ACKs for top commit:
achow101:
ACK 8cc3ac6c23
TheCharlatan:
Re-ACK 8cc3ac6c23
stratospher:
reACK 8cc3ac6.
Tree-SHA512: 53f27591916246be4093d64b86a0494e55094abd8c586026b1247e4a36747bc3d6dbe46dc26ee4a22f47b8eb0d9699d13e577dee0e7198145f3c9b11ab2a30b7
1632fc104b txgraph: Track multiple potential would-be clusters in Trim (improvement) (Pieter Wuille)
4608df37e0 txgraph: add Trim benchmark (benchmark) (Pieter Wuille)
9c436ff01c txgraph: add fuzz test scenario that avoids cycles inside Trim() (tests) (Pieter Wuille)
938e86f8fe txgraph: add unit test for TxGraph::Trim (tests) (glozow)
a04e205ab0 txgraph: Add ability to trim oversized clusters (feature) (Pieter Wuille)
eabcd0eb6f txgraph: remove unnecessary m_group_oversized (simplification) (Greg Sanders)
19b14e61ea txgraph: Permit transactions that exceed cluster size limit (feature) (Pieter Wuille)
c4287b9b71 txgraph: Add ability to configure maximum cluster size/weight (feature) (Pieter Wuille)
Pull request description:
Part of cluster mempool (#30289).
During reorganisations, it is possible that dependencies get added which would result in clusters that violate policy limits (cluster count, cluster weight), when linking the new from-block transactions to the old from-mempool transactions. Unlike RBF scenarios, we cannot simply reject the changes when they are due to received blocks. To accommodate this, add a `TxGraph::Trim()`, which removes some subset of transactions (including descendants) in order to make all resulting clusters satisfy the limits.
Conceptually, the way this is done is by defining a rudimentary linearization for the entire would-be too-large cluster, iterating it from beginning to end, and reasoning about the counts and weights of the clusters that would be reached using transactions up to that point. If a transaction is encountered whose addition would violate the limit, it is removed, together with all its descendants.
This rudimentary linearization is like a merge sort of the chunks of the clusters being combined, but respecting topology. More specifically, it is continuously picking the highest-chunk-feerate remaining transaction among those which have no unmet dependencies left. For efficiency, this rudimentary linearization is computed lazily, by putting all viable transactions in a heap, sorted by chunk feerate, and adding new transactions to it as they become viable.
The `Trim()` function is rather unusual compared to the `TxGraph` functionality added in previous PRs, in that `Trim()` makes it own decisions about what the resulting graph contents will be, without good specification of how it makes that decision - it is just a best-effort attempt (which is improved in the last commit). All other `TxGraph` mutators are simply to inform the graph about changes the calling mempool code decided on; this one lets the decision be made by txgraph.
As part of this, the "oversized" property is expanded to also encompass a configurable cluster weight limit (in addition to cluster count limit).
ACKs for top commit:
instagibbs:
reACK 1632fc104b
glozow:
reACK 1632fc104b via range-diff
ismaelsadeeq:
reACK 1632fc104b🛰️
Tree-SHA512: ccacb54be8ad622bd2717905fc9b7e42aea4b07f824de1924da9237027a97a9a2f1b862bc6a791cbd2e1a01897ad2c7c73c398a2d5ccbce90bfbeac0bcebc9ce
c10e382d2a flatfile: check whether the file has been closed successfully (Vasil Dimov)
4bb5dd78ea util: check that a file has been closed before ~AutoFile() is called (Vasil Dimov)
8bb34f07df Explicitly close all AutoFiles that have been written (Vasil Dimov)
a69c4098b2 rpc: take ownership of the file by WriteUTXOSnapshot() (Hodlinator)
Pull request description:
`fclose(3)` may fail to flush the previously written data to disk, thus a failing `fclose(3)` is as serious as a failing `fwrite(3)`.
Previously the code ignored `fclose(3)` failures. This PR improves that by changing all users of `AutoFile` that use it to write data to explicitly close the file and handle a possible error.
---
Other alternatives are:
1. `fflush(3)` after each write to the file (and throw if it fails from the `AutoFile::write()` method) and hope that `fclose(3)` will then always succeed. Assert that it succeeds from the destructor 🙄. Will hurt performance.
2. Throw nevertheless from the destructor. Exception within the exception in C++ I think results in terminating the program without a useful message.
3. (this is implemented in the latest incarnation of this PR) Redesign `AutoFile` so that its destructor cannot fail. Adjust _all_ its users 😭. For example, if the file has been written to, then require the callers to explicitly call the `AutoFile::fclose()` method before the object goes out of scope. In the destructor, as a sanity check, assume/assert that this is indeed the case. Defeats the purpose of a RAII wrapper for `FILE*` which automatically closes the file when it goes out of scope and there are a lot of users of `AutoFile`.
4. Pass a new callback function to the `AutoFile` constructor which will be called from the destructor to handle `fclose()` errors, as described in https://github.com/bitcoin/bitcoin/pull/29307#issuecomment-2243842400. My thinking is that if that callback is going to only log a message, then we can log the message directly from the destructor without needing a callback. If the callback is going to do more complicated error handling then it is easier to do that at the call site by directly calling `AutoFile::fclose()` instead of getting the `AutoFile` object out of scope (so that its destructor is called) and inspecting for side effects done by the callback (e.g. set a variable to indicate a failed `fclose()`).
ACKs for top commit:
l0rinc:
ACK c10e382d2a
achow101:
ACK c10e382d2a
hodlinator:
re-ACK c10e382d2a
Tree-SHA512: 3994ca57e5b2b649fc84f24dad144173b7500fc0e914e06291d5c32fbbf8d2b1f8eae0040abd7a5f16095ddf4e11fe1636c6092f49058cda34f3eb2ee536d7ba
Trim internally builds an approximate dependency graph of the merged cluster,
replacing all existing dependencies within existing clusters with a simple
linear chain of dependencies. This helps keep the complexity of the merging
operation down, but may result in cycles to appear in the general case, even
though in real scenarios (where Trim is called for stitching re-added mempool
transactions after a reorg back to the existing mempool transactions) such
cycles are not possible.
Add a test that specifically targets Trim() but in scenarios where it is
guaranteed not to have any cycles. It is a special case, is much more a
whitebox test than a blackbox test, and relies on randomness rather than
fuzz input. The upside is that somewhat stronger properties can be tested.
Co-authored-by: Greg Sanders <gsanders87@gmail.com>
During reorganisations, it is possible that dependencies get add which
result in clusters that violate limits (count, size), when linking the
new from-block transactions to the old from-mempool transactions.
Unlike RBF scenarios, we cannot simply reject these policy violations
when they are due to received blocks. To accomodate this, add a Trim()
function to TxGraph, which removes transactions (including descendants)
in order to make all resulting clusters satisfy the limits.
In the initial version of the function added here, the following approach
is used:
- Lazily compute a naive linearization for the to-be-merged cluster (using
an O(n log n) algorithm, optimized for far larger groups of transactions
than the normal linearization code).
- Initialize a set of accepted transactions to {}
- Iterate over the transactions in this cluster one by one:
- If adding the transaction to the set makes it exceed the max cluster size
or count limit, stop.
- Add the transaction to the set.
- Remove all transactions from the cluster that were not included in the set
(note that this necessarily includes all descendants too, because they
appear later in the naive linearization).
Co-authored-by: Greg Sanders <gsanders87@gmail.com>
This removes the restriction added in the previous commit that individual
transactions do not exceed the max cluster size limit.
With this change, the responsibility for enforcing cluster size limits can
be localized purely in TxGraph, without callers (and in particular, tests)
needing to duplicate the enforcement for individual transactions.
This is integrated with the oversized property: the graph is oversized when
any connected component within it contains more than the cluster count limit
many transactions, or when their combined size/weight exceeds the cluster size
limit.
It becomes disallowed to call AddTransaction with a size larger than this limit,
though this limit will be lifted in the next commit.
In addition, SetTransactionFeeRate becomes SetTransactionFee, so that we do not
need to deal with the case that a call to this function might affect the
oversizedness.
28299ce776 p2p: remove vestigial READ_STATUS_CHECKBLOCK_FAILED (Greg Sanders)
bac9ee4830 p2p: Add witness mutation check inside FillBlock (Greg Sanders)
Pull request description:
Since #29412, we have not allowed mutated blocks to continue being processed immediately the block is received, but this is only done for the legacy BLOCK message.
Extend these checks as belt-and-suspenders to not allow similar mutation strategies to affect relay by honest peers by applying the check inside `PartiallyDownloadedBlock::FillBlock`, immediately before returning `READ_STATUS_OK`.
ACKs for top commit:
Crypt-iQ:
ACK 28299ce776
achow101:
ACK 28299ce776
stratospher:
ACK 28299ce7.
dergoegge:
Code review ACK 28299ce776
Tree-SHA512: 883d7c12ca096234b425e6fe12e46b0611607600916e6ac8d1c8112224aa76924b7b074754910163ac2ec15379075d618a9ece3642649ac7629cddb0d4e432ea
There is no way to report a close error from `AutoFile` destructor.
Such an error could be serious if the file has been written to because
it may mean the file is now corrupted (same as if write fails).
So, change all users of `AutoFile` that use it to write data to
explicitly close the file and handle a possible error.
