Before this commit, we would always prepare tracepoint arguments
regardless of the tracepoint being used or not. While we already made
sure not to include expensive arguments in our tracepoints, this
commit introduces gating to make sure the arguments are only prepared
if the tracepoints are actually used. This is a win-win improvement
to our tracing framework. For users not interested in tracing, the
overhead is reduced to a cheap 'greater than 0' compare. As the
semaphore-gating technique used here is available in bpftrace, bcc,
and libbpf, users interested in tracing don't have to change their
tracing scripts while profiting from potential future tracepoints
passing slightly more expensive arguments. An example are mempool
tracepoints that pass serialized transactions. We've avoided the
serialization in the past as it was too expensive.
Under the hood, the semaphore-gating works by placing a 2-byte
semaphore in the '.probes' ELF section. The address of the semaphore
is contained in the ELF note providing the tracepoint information
(`readelf -n ./src/bitcoind | grep NT_STAPSDT`). Tracing toolkits
like bpftrace, bcc, and libbpf increase the semaphore at the address
upon attaching to the tracepoint. We only prepare the arguments and
reach the tracepoint if the semaphore is greater than zero. The
semaphore is decreased when detaching from the tracepoint.
This also extends the "Adding a new tracepoint" documentation to
include information about the semaphores and updated step-by-step
instructions on how to add a new tracepoint.
Enhanced efficiency and readability of CCoinsViewCache::FetchCoin by replacing separate find() and emplace() calls with a single try_emplace(), reducing map lookups and potential insertions.
Erase spent cache entries and clear flags of unspent
entries inside the BatchWrite loop, instead of an
additional loop after BatchWrite.
Co-Authored-By: Pieter Wuille <pieter@wuille.net>
No behavior change because any entries that are added in EmplaceCoinInternalDANGER
have DIRTY assigned to them after, and if they
are not inserted then they will not be
modified as before.
This prepares moving the cache entry
flags field to private access.
Co-Authored-By: Martin Leitner-Ankerl <martin.ankerl@gmail.com>
As per doc/developer-notes#logging, LogError should be used for
severe problems that require the node to shut down.
Co-Authored-By: stickies-v <stickies-v@protonmail.com>
ed52e71176 Periodically check disk space to avoid corruption (Aurèle Oulès)
7fe537f7a4 Implement CCoinsViewErrorCatcher::HaveCoin (Aurèle Oulès)
Pull request description:
Attempt to fix#26112.
As suggested by sipa in https://github.com/bitcoin/bitcoin/issues/26112#issuecomment-1249683401:
> CCoinsViewErrorCatcher, the wrapper class used around CCoinsViewDB that's supposed to detect these problems and forcefully exit the application, has an override for GetCoins. But in CheckTxInputs, HaveInputs is first invoked, which on its turn calls HaveCoin. HaveCoin is implemented in CCoinsViewDB, but not in CCoinsViewErrorCatcher, and thus the disk read exception escapes.
> A solution may be to just add an override for HaveCoin in CCoinsViewErrorCatcher.
I implemented `CCoinsViewErrorCatcher::HaveCoin` and also added a periodic disk space check that shutdowns the node if there is not enough space left on disk, the minimum here is 50MB.
For reviewers, it's possible to saturate disk space to test the PR by creating large files with `fallocate -l 50G test.bin`
ACKs for top commit:
achow101:
ACK ed52e71176
w0xlt:
Code Review ACK ed52e71176
sipa:
utACK ed52e71176
Tree-SHA512: 456aa7b996023df42b4fbb5158ee429d9abf7374b7b1ec129b21aea1188ad19be8da4ae8e0edd90b85b7a3042b8e44e17d3742e33808a4234d5ddbe9bcef1b78
In my benchmarks, using this pool allocator for CCoinsMap gives about
20% faster `-reindex-chainstate` with -dbcache=5000 with practically the
same memory usage. The change in max RSS changed was 0.3%.
The `validation_flush_tests` tests need to be updated because
memory allocation is now done in large pools instead of one node at a
time, so the limits need to be updated accordingly.
This frees up all associated memory with the map, not only the nodes.
This is necessary in preparation for using the PoolAllocator for
CCoinsMap, which does not actually free any memory on clear().
In certain circumstances, we may want to flush to disk without
emptying `cacheCoins`, which affects performance. UTXO snapshot
activation is one such case.
This method is currently unused and this commit does not
change any behavior.
Incorporates feedback from John Newbery.
Co-authored-by: Suhas Daftuar <sdaftuar@gmail.com>
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
Also adds CCoinsViewCache::ReallocateCache() to attempt to free
memory that the cacheCoins's allocator may be hanging onto when
downsizing the cache.
Adds `CChainState::m_coins{tip,db}_cache_size_bytes` data members
so that we can reference cache size on a per-chainstate basis for
flushing.
and into coins.cpp. This move is necessary so that we can later include a
CCoinsViewErrorCatcher instance under CChainState.
Co-authored-by: MarcoFalke <falke.marco@gmail.com>
std::unordered_map::erase( const_iterator pos ) returns an iterator to the element following the removed one. Use that to optimize (probably minor-performance-wise, and definitely code-structure-wise) the implementation of CCoinsViewCache::BatchWrite().
