Follow-up to bitcoin/bitcoin#32497.
Clarify why the witness merkle test uses an odd leaf count (it exercises leaf duplication in `ComputeMerkleRoot()`), and make the coinbase witness hash initialization explicit.
Also simplify the leaf-copy loop in the MerkleRoot benchmark for readability.
No production code is changed in this follow-up, for simplicity and safety.
Co-authored-by: w0xlt <94266259+w0xlt@users.noreply.github.com>
Avoid calling `std::distance` on null pointers in `PoolResource::AllocateChunk`.
Compute remaining bytes with `m_available_memory_end - m_available_memory_it` instead, which is well-defined to be `0` when both are `nullptr`.
Co-authored-by: MarcoFalke <*~=`'#}+{/-|&$^_@721217.xyz>
`PartiallyDownloadedBlock::FillBlock()` computed the block header hash and summed missing transaction sizes for debug logging unconditionally, including when cmpctblock debug logging is disabled.
Guard the debug-only hash and size computations with `LogAcceptCategory`.
Since `txn_available` is invalidated after the first loop (needed for efficient moving), we compute `tx_missing_size` by iterating `vtx_missing` directly. This is safe because the later `tx_missing_offset` check guarantees `vtx_missing` was fully consumed during reconstruction.
Use `block.GetHash()` instead of `header.GetHash()`, since header is cleared before logging.
No behavior change when debug logging is enabled: the reported counts, hashes, and byte totals remain the same.
`PeerManagerImpl::SendBlockTransactions()` computed the total byte size of requested transactions for a debug log line by calling `ComputeTotalSize()` in a tight loop, triggering serialization even when debug logging is off.
Guard the size accumulation with `LogAcceptCategory` so the serialization work only happens when the log line can be emitted.
No behavior change when debug logging is enabled: the reported block hash, transaction count, and byte totals are the same.
The bounds checks still run unconditionally; the debug-only loop iterates the already-validated response contents.
Separating debug-only work from the critical path reduces risk and favors the performance-critical non-debug case.
This also narrows the racy scope of when logging is toggled from another thread.
Transaction hashes are cached, it may not be intuitive that their sizes are actually recalculated every time.
This is done before the other refactors to clarify why we want to avoid calling this method;
Co-authored-by: maflcko <6399679+maflcko@users.noreply.github.com>
Compute `uptime` from `SteadyClock` so it is unaffected by system time changes after startup.
Derive GUI startup time by subtracting the monotonic uptime from the wall clock time.
Add a functional test covering a large `setmocktime` jump.
Co-authored-by: MarcoFalke <*~=`'#}+{/-|&$^_@721217.xyz>
Log an info message when any `-debug` categories are enabled, noting they may contain privacy-sensitive information (e.g. transaction IDs) and should not be shared publicly.
Co-authored-by: Vasil Dimov <vd@FreeBSD.org>
Co-authored-by: MarcoFalke <*~=`'#}+{/-|&$^_@721217.xyz>
Private broadcast is a privacy feature, and users may share `debug.log` with support.
Unconditional log messages that mention private broadcast and/or include (w)txids can leak which transactions a user originated.
Move private broadcast event logging from `LogInfo()` to `LogDebug(BCLog::PRIVBROADCAST, ...)` so it is only emitted when debug logging is enabled, and drop the hardcoded "[privatebroadcast]" prefixes.
Keep warnings at the default log level without (w)txids, detailed context remains available under `-debug=privatebroadcast`.
Co-authored-by: Vasil Dimov <vd@FreeBSD.org>
Replace the last few instances of `.count() != 0` and `.count() == 0` and `.count()` patterns with the more expressive C++20 `.contains()` method:
* `std::set<std::string>` in `getblocktemplate` RPC;
* `std::map<std::string, ...>` in `transaction_tests`;
* other bare `std::unordered_set` and `std::map` count calls.
With no remaining violations, enable the `readability-container-contains`
clang-tidy check to prevent future regressions.
Update the integer fuzz test to move the vector into `ComputeMerkleRoot`, matching production usage patterns and avoiding unnecessary copies.
Update `merkle_test_BlockWitness` to use an odd number of transactions to ensure the test covers the scenario where leaf duplication occurs. Also switch to `GetWitnessHash` to match `BlockWitnessMerkleRoot` semantics.
The manual vector setup retains the exact-size `resize` to explicitly verify the behavior against the calculated root.
Split the repeated `SipHash` v[0..3] initialization into a small `SipHashState` helper that is used by both `CSipHasher` and `PresaltedSipHasher`.
Added explanatory comments to clarify behavior, documenting the equivalence of `PresaltedSipHasher` `operator()` overloads to `CSipHasher` usage.
Co-authored-by: Ryan Ofsky <ryan@ofsky.org>
Replaces separate `shorttxidk0`/`shorttxidk1` members with a cached `PresaltedSipHasher`, so `GetShortID()` reuses the precomputed `SipHash` state instead of rebuilding it on every call.
`CBlockHeaderAndShortTxIDs` was never intended to be used before `FillShortTxIDSelector()` runs; doing so already relied on indeterminate salt values.
The new `Assert(m_hasher)` just makes this invariant explicit and fails fast if the object is used in an uninitialized state.
Replaces standalone `SipHashUint256` with an `operator()` overload in `PresaltedSipHasher`.
Updates all hasher classes (`SaltedUint256Hasher`, `SaltedTxidHasher`, `SaltedWtxidHasher`) to use `PresaltedSipHasher` internally, enabling the same constant-state caching optimization while keeping behavior unchanged.
Benchmark was also adjusted to cache the salting part.
Aligns test variable naming with the `k0`/`k1` convention used consistently throughout the codebase for `SipHash` keys.
Also splits the single-param `SipHash` test from the one with extra, for clarity.
The changes made here were:
| From | To |
|-------------------|------------------|
| `m.count(k) == 1` | `m.contains(k)` |
| `m.count(k) == 0` | `!m.contains(k)` |
| `m.count(k) != 1` | `!m.contains(k)` |
| `m.count(k) < 1` | `!m.contains(k)` |
* `mapInfo` is instance of `std::unordered_map` and can only contain 0 or 1 value for a given key;
* similarly, `g_enabled_filter_types` and `setClientRules` are both `std::set` instances;
* lastly, while `mapTxSpends` is `std::unordered_multimap` that could potentially hold multiple values, having a size less than 1 means that the value is missing.
`QMap<WalletModel*, WalletView*> mapWalletViews` values were also migrated manually.
Co-authored-by: pablomartin4btc <pablomartin4btc@gmail.com>
Co-authored-by: fanquake <fanquake@gmail.com>
The changes made here were:
| From | To |
|-------------------|------------------|
| `m.count(k)` | `m.contains(k)` |
| `!m.count(k)` | `!m.contains(k)` |
| `m.count(k) == 0` | `!m.contains(k)` |
| `m.count(k) != 0` | `m.contains(k)` |
| `m.count(k) > 0` | `m.contains(k)` |
The commit contains the trivial, mechanical refactors where it doesn't matter if the container can have multiple elements or not
Co-authored-by: Jan B <608446+janb84@users.noreply.github.com>
The changes made here were:
| From | To |
|------------------------|------------------|
| `m.find(k) == m.end()` | `!m.contains(k)` |
| `m.find(k) != m.end()` | `m.contains(k)` |
The `mutated` parameter is never used at any call site - all callers pass `nullptr`.
The explicit comment in `validation.cpp` explains the reason:
// The malleation check is ignored; as the transaction tree itself
// already does not permit it, it is impossible to trigger in the
// witness tree.
There's a single call to the methods from `ComputeMerklePath` where the last parameter is always provided.
This simplifies the implementation by not having to check for missing parameter.
A few temporary `CCoinsViewCache`'s are destructed right after the `Flush()`, therefore it is not necessary to call `ReallocateCache` to recreate them right before they're killed anyway.
* `Flush()` - retains existing functionality;
* `Flush(/*will_reuse_cache=*/false)` - skips destruction and reallocation of the parent cache since it will soon go out of scope anyway;
For the `will_reuse_cache` parameter we want to see exactly which ones will reallocate memory and which won't - since both can be valid usages.
This change was based on a subset of https://github.com/bitcoin/bitcoin/pull/28945.
Co-authored-by: Martin Ankerl <martin.ankerl@gmail.com>
Extends 7703884 to guard environ declaration on all Windows builds, not just MSVC.
In the mingw-w64 headers (used by llvm-mingw), environ is defined as a macro which expands through _environ to (* __p__environ()), a call to a dllimport function, causing the same inconsistent linkage warning as MSVC.
Use WIN32 instead of _MSC_VER to match the platform-specific guards already used throughout the file.
The warning occurs with llvm-mingw (both UCRT and MSVCRT variants as tested by Hebasto), but not with the mingw-w64 toolchain currently used in CI (as mentioned by fanquake).
Co-authored-by: Hennadii Stepanov <32963518+hebasto@users.noreply.github.com>
`EmplaceCoinInternalDANGER()` incremented `cachedCoinsUsage` even when `try_emplace` did not insert (duplicate key), inflating the counter.
This is mostly reachable in tests today since `AssumeUTXO` does not overwrite.
Increment only on successful insert, and capture `coin.DynamicMemoryUsage()` before the move so accounting uses the correct value.
Fuzz: add an `EmplaceCoinInternalDANGER` path to exercise insert-only accounting.
Unit test: emplace two different coins at the same outpoint (with different `DynamicMemoryUsage()`), verify `SelfTest()` passes and `AccessCoin(outpoint)` returns the first coin.
Co-authored-by: Andrew Toth <andrewstoth@gmail.com>
Co-authored-by: w0xlt <woltx@protonmail.com>
Move the `cachedCoinsUsage` subtract in `AddCoin()` to after the `possible_overwrite` check.
Previously a throw before assignment decremented the counter without changing the entry, which corrupted accounting and later underflowed.
In `Flush()`, reset `cachedCoinsUsage` to `0` only when `BatchWrite()` succeeds and `cacheCoins` is actually cleared. In production `BatchWrite()` returns `true`, so this mostly affects tests. On failure, leave the counter unchanged to keep it in sync with the cache.
The existing `Flush()` workaround in fuzzing was also removed now that the source of the problem was fixed, so the fuzzer no longer needs `coins_view_cache.Flush()` to realign `cachedCoinsUsage` after an exception.
Replace the prior `expected_code_path` tracking with direct assertions. The role of the variable was to verify that code execution follows only expected paths, either successful addition, or if it's an exception, the message is verified and checked that overwrite was disallowed.
With these changes the counter stays consistent across success and exception paths, so we can finally remove the `UBSan` suppressions for `CCoinsViewCache` that were masking the issue.
Included a unit test as well, attempting to add a different coin to the same outpoint without allowing overwrites and make sure it throws.
We use `SelfTest()` to validates accounting, and check that the cache remains usable.
Co-authored-by: Ryan Ofsky <ryan@ofsky.org>
Co-authored-by: w0xlt <woltx@protonmail.com>
When a coin is spent via `SpendCoin()`, `cachedCoinsUsage` is already decremented and the coin's `scriptPubKey` is cleared, so `DynamicMemoryUsage()` is `0`.
`CoinsViewCacheCursor::NextAndMaybeErase()` was subtracting usage again when erasing spent entries.
Replace it with an assert that documents spent coins have zero dynamic memory usage by the time the cursor encounters them.
Remove the now-unnecessary `usage` reference from the cursor's constructor and member variables.
During `BatchWrite`, the parent entry is created under a guard that guarantees insertion, so the new `Coin` is default-constructed and empty.
Assert this invariant to document why there is no `cachedCoinsUsage` decrement before the assignment at this site.
Co-authored-by: Andrew Toth <andrewstoth@gmail.com>
When the assumevalid ancestry check fails, log a precise reason:
- "block height above assumevalid height" if the block is above the assumevalid block (the default reason)
- "block not in of assumevalid chain" otherwise
The new split was added under the existing condition to simplify conceptually that the two cases are related.
It could still be useful to know when the block is just above the assumevalid block or when it's not even on the same chain.
Update the functional test to assert the new reason strings. No behavior change.
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
Flatten nested conditionals into a linear gating sequence for readability and precise logging. No functional change, TODOs are addressed in next commit
Replaced `atomic<bool>` with `std::optional<bool>` (logged once on first observation). Safe because `ConnectBlock` holds `cs_main`.\
After this change, the state is logged before the very first `UpdateTip` line.
Co-authored-by: Eunovo <eunovo9@gmail.com>
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
Co-authored-by: w0xlt <woltx@protonmail.com>
Increase the test to 6 nodes and add flows for baseline, deep anchor, and too-recent cases, plus scaffolding for off-best-header, not-in-assumevalid,
and reindex gates.
Assertions are minimal here; follow-ups add reason checks.
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
After an incomplete reindex the blocks will need to be replayed.
This results in excessive `Rolling back` and `Rolling forward` messages which quickly triggers the recently introduced log rate limiter.
Change the logging strategy to:
- Add single `LogInfo` messages showing the full range being replayed for both rollback and roll forward;
- Log progress at `LogInfo` level only every 10,000 blocks to track the long operations.
Reproducer:
* Start a normal IBD, stop after some progress
* Do a reindex, stop before it finishes
* Restart the node normally without specifying the reindex parameter
It should start rolling the blocks forward.
Before this change the excessive logging would show:
```
[*] Rolling forward 000000002f4f55aecfccc911076dc3f73ac0288c83dc1d79db0a026441031d40 (46245)
[*] Rolling forward 0000000017ffcf34c8eac010c529670ba6745ea59cf1edf7b820928e3b40acf6 (46246)
```
After the change it shows:
```
Replaying blocks
Rolling forward to 00000000000000001034012d7e4facaf16ca747ea94b8ea66743086cfe298ef8 (326223 to 340991)
Rolling forward 00000000000000000faabab19f17c0178c754dbed023e6c871dcaf74159c5f02 (330000)
Rolling forward 00000000000000000d9b2508615d569e18f00c034d71474fc44a43af8d4a5003 (340000)
...
Rolled forward to 00000000000000001034012d7e4facaf16ca747ea94b8ea66743086cfe298ef8
```
(similarly to rolling back)
Co-authored-by: Anthony Towns <aj@erisian.com.au>
Co-authored-by: Vasil Dimov <vd@freebsd.org>
Even though not all script verification is turned off currently (e.g. we're still doing the cheaper sigop counts), this naming is more consistent with other usages.
Oversized allocations can cause out-of-memory errors or [heavy swapping](https://github.com/getumbrel/umbrel-os/issues/64#issuecomment-663637321), [grinding the system to a halt](https://x.com/murchandamus/status/1964432335849607224).
`LogOversizedDbCache()` now emits a startup warning if the configured `-dbcache` exceeds a cap derived from system RAM, using the same parsing/clamping as cache sizing via CalculateDbCacheBytes(). This isn't meant as a recommended setting, rather a likely upper limit.
Note that we're not modifying the set value, just issuing a warning.
Also note that the 75% calculation is rounded for the last two numbers since we have to divide first before multiplying, otherwise we wouldn't stay inside size_t on 32-bit systems - and this was simpler than casting back and forth.
We could have chosen the remaining free memory for the warning (e.g. warn if free memory is less than 1 GiB), but this is just a heuristic, we assumed that on systems with a lot of memory, other processes are also running, while memory constrained ones run only Core.
If total RAM < 2 GiB, cap is `DEFAULT_DB_CACHE` (`450 MiB`), otherwise it's 75% of total RAM.
The threshold is chosen to be close to values commonly used in [raspiblitz](https://github.com/raspiblitz/raspiblitz/blob/dev/home.admin/_provision.setup.sh#L98-L115) for common setups:
| Total RAM | `dbcache` (MiB) | raspiblitz % | proposed cap (MiB) |
|----------:|----------------:|-------------:|-------------------:|
| 1 GiB | 512 | 50.0% | 450* |
| 2 GiB | 1536 | 75.0% | 1536 |
| 4 GiB | 2560 | 62.5% | 3072 |
| 8 GiB | 4096 | 50.0% | 6144 |
| 16 GiB | 4096 | 25.0% | 12288 |
| 32 GiB | 4096 | 12.5% | 24576 |
[Umbrel issues](https://github.com/getumbrel/umbrel-os/issues/64#issuecomment-663816367) also mention 75% being the upper limit.
Starting `bitcoind` on an 8 GiB rpi4b with a dbcache of 7 GiB:
> ./build/bin/bitcoind -dbcache=7000
warns now as follows:
```
2025-09-07T17:24:29Z [warning] A 7000 MiB dbcache may be too large for a system memory of only 7800 MiB.
2025-09-07T17:24:29Z Cache configuration:
2025-09-07T17:24:29Z * Using 2.0 MiB for block index database
2025-09-07T17:24:29Z * Using 8.0 MiB for chain state database
2025-09-07T17:24:29Z * Using 6990.0 MiB for in-memory UTXO set (plus up to 286.1 MiB of unused mempool space)
```
Besides the [godbolt](https://godbolt.org/z/EPsaE3xTj) reproducers for the new total memory method, we also tested the warnings manually on:
- [x] Apple M4 Max, macOS 15.6.1
- [x] Intel Core i9-9900K, Ubuntu 24.04.2 LTS
- [x] Raspberry Pi 4 Model B, Armbian Linux 6.12.22-current-bcm2711
- [x] Intel Xeon x64, Windows 11 Home Version 24H2, OS Build 26100.4351
Co-authored-by: stickies-v <stickies-v@protonmail.com>
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
Co-authored-by: w0xlt <woltx@protonmail.com>
Added a minimal system helper to query total physical RAM on [Linux/macOS/Windows](https://stackoverflow.com/a/2513561) (on other platforms we just return an empty optional).
The added test checks if the value is roughly correct by checking if the CI platforms are returning any value and if the value is at least 1 GiB and not more than 10 TiB.
The max value is only validated on 64 bits, since it's not unreasonable for 32 bits to have max memory, but on 64 bits it's likely an error.
https://learn.microsoft.com/en-us/windows/win32/api/sysinfoapi/ns-sysinfoapi-memorystatusex
> ullTotalPhys The amount of actual physical memory, in bytes.
https://man7.org/linux/man-pages/man3/sysconf.3.html:
> _SC_PHYS_PAGES The number of pages of physical memory. Note that it is possible for the product of this value and the value of _SC_PAGESIZE to overflow.
> _SC_PAGESIZE Size of a page in bytes. Must not be less than 1.
See https://godbolt.org/z/ec81Tjvrj for further details
### Summary
When triggering a reindex, users had no indication of how many files remained or how far along the process was.
### Fix
This patch prefetches the target file block file count to be able to show progress information.
Instead of just displaying which block file is being processed, it now indicates how many files remain.
### Reproducer + expected results
Running
```bash
cmake -B build && make -C build -DCMAKE_BUILD_TYPE=Release && ./build/bin/bitcoind -datadir=demo -reindex
```
Shows the block files one-by-one currently, there's no way to see how much work is left:
```
Reindexing block file blk00000.dat...
Loaded 119920 blocks from external file in 1228ms
Reindexing block file blk00001.dat...
Loaded 10671 blocks from external file in 284ms
Reindexing block file blk00002.dat...
Loaded 5459 blocks from external file in 263ms
Reindexing block file blk00003.dat...
Loaded 5595 blocks from external file in 267ms
```
After the change:
```
Reindexing block file blk00000.dat (0% complete)...
Loaded 119920 blocks from external file in 1255ms
Reindexing block file blk00001.dat (1% complete)...
Loaded 10671 blocks from external file in 303ms
Reindexing block file blk00002.dat (2% complete)...
Loaded 5459 blocks from external file in 278ms
Reindexing block file blk00003.dat (3% complete)...
Loaded 5595 blocks from external file in 285ms
```
Regenerated `.clang-format` from current configs to replace deprecated keys with up-to-date equivalents.
Also added all current formatter default values to guard against version differences.
The configs were updated with the following command (using v16 for maximal compatibility):
$(brew --prefix llvm@16)/bin/clang-format -dump-config -style=file:src/.clang-format
The new config was tested with:
$(brew --prefix llvm@16)/bin/clang-format -i src/deploymentinfo.h
Co-authored-by: MarcoFalke <*~=`'#}+{/-|&$^_@721217.xyz>
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
The default database write batch size is increased from 16 MiB to 32 MiB to improve I/O efficiency and performance during UTXO flushes, particularly during Initial Block Download and `assumeutxo` loads.
On systems with slower I/O, a larger batch size reduces overhead from numerous small writes. Measurements show this change provides a modest performance improvement on most hardware during a critical section, with a minimal peak memory increase (approx. 75 MiB on default settings).
Did both in this commit, since the return value of `WriteReindexing` was ignored anyway - which existed only because of the constant `Erase` being called
`WriteBatch` can only ever return `true` - its errors are handled by throwing a `throw dbwrapper_error` instead.
The boolean return value is quite confusing, especially since it's symmetric with `CDBWrapper::Read`, which catches the exceptions and returns a boolean instead.
We're removing the constant return value and inlining `true` for its usages.
The `-assumevalid` option skips script verification for a specified block and all its ancestors during Initial Block Download.
Many new users are surprised when this suddenly slows their node to a halt.
This commit adds a log message to clearly indicate when this optimization ends and full validation begins (and vice versa).
When using `-assumeutxo`, logging is suppressed for the active assumed-valid chainstate and for the background validation chainstate to avoid the confusing toggles.
-------
> cmake -B build && cmake --build build && mkdir -p demo && build/bin/bitcoind -datadir=demo -stopatheight=500 | grep 'signature validation'
```
2025-08-08T20:59:21Z Disabling signature validations at block #1 (00000000839a8e6886ab5951d76f411475428afc90947ee320161bbf18eb6048).
2025-08-08T20:59:21Z Enabling signature validations at block #100 (000000007bc154e0fa7ea32218a72fe2c1bb9f86cf8c9ebf9a715ed27fdb229a).
2025-08-08T20:59:21Z Disabling signature validations at block #200 (000000008f1a7008320c16b8402b7f11e82951f44ca2663caf6860ab2eeef320).
2025-08-08T20:59:21Z Enabling signature validations at block #300 (0000000062b69e4a2c3312a5782d7798b0711e9ebac065cd5d19f946439f8609).
```
* parameter name uses underscores
* commit message typo fixed and compacted
* used `10_MiB` to avoid having to comment
* swapped order of operands in (9 * x / 10) to make it obvious that we're calculating 90%
* inlined return value
Move-only commit, enabled reusing the large cache size calculation logic later. The only difference is the removal of the `static` keyword (since in a constexpr function it's a C++23 extension)
The current `prevector` size of 28 bytes (chosen to fill the `sizeof(CScript)` aligned size) was introduced in 2015 (https://github.com/bitcoin/bitcoin/pull/6914) before SegWit and TapRoot.
However, the increasingly common `P2WSH` and `P2TR` scripts are both 34 bytes, and are forced to use heap (re)allocation rather than efficient inline storage.
The core trade-off of this change is to eliminate heap allocations for common 34-36 byte scripts at the cost of increasing the base memory footprint of all `CScript` objects by 8 bytes (while still respecting peak memory usage defined by `-dbcache`).
Increasing the `prevector` size allows these scripts to be stored inline, avoiding extra heap allocations, reducing potential memory fragmentation, and improving performance during cache flushes. Massif analysis confirms a lower stable memory usage after flushing, suggesting the elimination of heap allocations outweighs the larger base size for common workloads.
Due to memory alignment, increasing the `prevector` size to 36 bytes doesn't change the overall `sizeof(CScript)` compared to an increase to 34 bytes, allowing us to include `P2PK` scripts as well at no additional memory cost.
Performance benchmarks for AssumeUTXO load and flush show:
* Small dbcache (450MB): ~1-3% performance improvement (despite more frequent flushes)
* Large dbcache (4500MB): ~6-8% performance improvement due to fewer heap allocations (and basically the number of flushes)
* Very large dbcache (4500MB): ~5-6% performance improvement due to fewer heap allocations (and memory limit not being reached, so there's no memory penalty)
Full IBD and reindex-chainstate with larger `dbcache` values also show an overall ~3-4% speedup.
Co-authored-by: Ava Chow <github@achow101.com>
Co-authored-by: Andrew Toth <andrewstoth@gmail.com>
Co-authored-by: maflcko <6399679+maflcko@users.noreply.github.com>
Verifies that script types are correctly allocated using prevector's direct or indirect storage based on their size:
Direct allocated script types (size ≤ 28 bytes):
* OP_RETURN (small)
* P2WPKH
* P2SH
* P2PKH
Indirect allocated script types (size > 28 bytes):
* P2WSH
* P2TR
* P2PK
* MULTISIG (small)
This test provides a baseline for verifying changes to prevector's inline capacity.
The `CHECK_SCRIPT_STATIC_SIZE` and `CHECK_SCRIPT_DYNAMIC_SIZE` macros were added to differentiate the two cases - while preserving the correct source code line in case of failure.
During Initial Block Download, the mempool is usually empty, but `CTxMemPool::removeForBlock` is still called for every connected block where we:
* iterate over every transaction in the block even though none will be found in the empty `mapTx`, always leaving `txs_removed_for_block` empty...
* which is pre-allocated regardless with 40 bytes * vtx.size(), even though it will always remain empty.
This change introduces a minor performance optimization by only executing the loop if any of the core mempool maps have any contents.
The call to `MempoolTransactionsRemovedForBlock` and the updates to the rolling fee logic remain unchanged.
The `removeForBlock` was also updated stylistically to match the surrounding methods and a clarification was added to clarify that it affects fee estimation as well.
Benchmarks indicated that obfuscating multiple bytes already gives an order of magnitude speed-up, but:
* GCC still emitted scalar code;
* Clang’s auto-vectorized loop ran on the slow unaligned-load path.
Fix contains:
* peeling the misaligned head enabled the hot loop starting at an 8-byte address;
* `std::assume_aligned<8>` tells the optimizer the promise holds - required to keep Apple Clang happy;
* manually unrolling the body to 64 bytes enabled GCC to auto-vectorize.
Note that `target.size() > KEY_SIZE` condition is just an optimization, the aligned and unaligned loops work without it as well - it's why the alignment calculation still contains `std::min`.
> C++ compiler .......................... GNU 14.2.0
| ns/byte | byte/s | err% | ins/byte | cyc/byte | IPC | bra/byte | miss% | total | benchmark
|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
| 0.03 | 32,464,658,919.11 | 0.0% | 0.50 | 0.11 | 4.474 | 0.08 | 0.0% | 5.29 | `ObfuscationBench`
> C++ compiler .......................... Clang 20.1.7
| ns/byte | byte/s | err% | ins/byte | cyc/byte | IPC | bra/byte | miss% | total | benchmark
|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
| 0.02 | 41,231,547,045.17 | 0.0% | 0.30 | 0.09 | 3.463 | 0.02 | 0.0% | 5.47 | `ObfuscationBench`
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
All former `std::vector<std::byte>` keys were replaced with `uint64_t` (we still serialize them as vectors but convert immediately to `uint64_t` on load).
This is why some tests still generate vector keys and convert them to `uint64_t` later instead of generating them directly.
In `Obfuscation::Unserialize` we can safely throw an `std::ios_base::failure` since during mempool fuzzing `mempool_persist.cpp#L141` catches and ignored these errors.
> C++ compiler .......................... GNU 14.2.0
| ns/byte | byte/s | err% | ins/byte | cyc/byte | IPC | bra/byte | miss% | total | benchmark
|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
| 0.04 | 28,365,698,819.44 | 0.0% | 0.34 | 0.13 | 2.714 | 0.07 | 0.0% | 5.33 | `ObfuscationBench`
> C++ compiler .......................... Clang 20.1.7
| ns/byte | byte/s | err% | ins/byte | cyc/byte | IPC | bra/byte | miss% | total | benchmark
|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
| 0.08 | 13,012,464,203.00 | 0.0% | 0.65 | 0.28 | 2.338 | 0.13 | 0.8% | 5.50 | `ObfuscationBench`
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
Co-authored-by: Ryan Ofsky <ryan@ofsky.org>
This is meant to focus the usages to narrow the scope of the obfuscation optimization.
`Obfuscation::Xor` is mostly a move.
Co-authored-by: maflcko <6399679+maflcko@users.noreply.github.com>
Since `FastRandomContext` delegates to `GetRandBytes` anyway, we can simplify new key generation to a Write/Read combo, unifying the flow of enabling obfuscation via `Read`.
The comments were also adjusted to clarify that the `m_obfuscation` field affects the behavior of `Read` and `Write` methods.
These changes are meant to simplify the diffs for the riskier optimization commits later.
Mechanical refactor of the low-level "xor" wording to signal the intent instead of the implementation used.
The renames are ordered by heaviest-hitting substitutions first, and were constructed such that after each replacement the code is still compilable.
-BEGIN VERIFY SCRIPT-
sed -i \
-e 's/\bGetObfuscateKey\b/GetObfuscation/g' \
-e 's/\bxor_key\b/obfuscation/g' \
-e 's/\bxor_pat\b/obfuscation/g' \
-e 's/\bm_xor_key\b/m_obfuscation/g' \
-e 's/\bm_xor\b/m_obfuscation/g' \
-e 's/\bobfuscate_key\b/m_obfuscation/g' \
-e 's/\bOBFUSCATE_KEY_KEY\b/OBFUSCATION_KEY_KEY/g' \
-e 's/\bSetXor(/SetObfuscation(/g' \
-e 's/\bdata_xor\b/obfuscation/g' \
-e 's/\bCreateObfuscateKey\b/CreateObfuscation/g' \
-e 's/\bobfuscate key\b/obfuscation key/g' \
$(git ls-files '*.cpp' '*.h')
-END VERIFY SCRIPT-
The two tests are doing different things - `xor_roundtrip_random_chunks` does black-box style property-based testing to validate that certain invariants hold - that deobfuscating an obfuscation results in the original message (higher level, it doesn't have to know about the implementation details).
The `xor_bytes_reference` test makes sure the optimized xor implementation behaves in every imaginable scenario exactly as the simplest possible obfuscation - with random chunks, random alignment, random data, random key.
Since we're touching the file, other related small refactors were also applied:
* `nullpt` typo fixed;
* manual byte-by-byte xor key creations were replaced with `_hex` factories;
* since we're only using 64 bit keys in production, smaller keys were changed to reflect real-world usage;
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
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>
Dropped the default expected_hash parameter from `ReadBlock()`.
In `blockmanager_flush_block_file` tests, we pass {} since the tests would already fail at PoW validation for corrupted blocks.
In `ChainstateManager::LoadExternalBlockFile`, we pass {} when processing child blocks because their hashes aren't known beforehand.
Switch to the index-aware `ReadBlock()` overload in `ComputeFilter` so that filter creation will abort if the stored block header hash doesn't match the expected one.
In the `readwriteblock` benchmark, pass the expected hash to `ReadBlock()` to match the new signature without affecting benchmark performance.
The non-recent-block code path in `ProcessGetBlockData` already has `inv.hash` available (equaling `pindex->GetBlockHash()`).
Pass it to `ReadBlock()` and assert that the on-disk header matches the requested hash.
The `GETBLOCKTXN` message handler in `ProcessMessage` receives `req.blockhash` from the peer (equaling `pindex->GetBlockHash()`).
Pass this hash to `ReadBlock()` for verification and assert that the index lookup matches.
Co-authored-by: TheCharlatan <seb.kung@gmail.com>
Co-authored-by: Hodlinator <172445034+hodlinator@users.noreply.github.com>
Eliminate one SHA‑256 double‑hash computation of the header per block read by reusing the hash for:
* proof‑of‑work verification;
* (optional) integrity check against the supplied hash.
Lőrinc