Also:
- Make DEFAULT_MAX_SIG_CACHE_SIZE into constexpr
DEFAULT_MAX_SIG_CACHE_BYTES to utilize the compile-time integer
arithmetic overflow checking available to constexpr.
- Fix comment (MiB instead of MB) for DEFAULT_MAX_SIG_CACHE_BYTES.
- Pass in max_size_bytes parameter to InitS*Cache(), modify log line to
no longer allude to maxsigcachesize being split evenly between the two
validation caches.
- Fix possible integer truncation and add a comment.
[META] I've kept the integer types as int64_t in order to not introduce
unintended behaviour changes, in the next commit we will make
them size_t.
Returning the approximate total size eliminates the need for
InitS*Cache() to do nElems*sizeof(uint256). The cuckoocache has a better
idea of this information.
dd065dae9fcebd6806ff67703ffa8128e80b97cc refactor: Make mapBlocksUnknownParent local, and rename it (Hennadii Stepanov)
Pull request description:
This PR is a second attempt at #19594. This PR has two motivations:
- Improve code hygiene by eliminating a global variable, `mapBlocksUnknownParent`
- Fix fuzz test OOM when running too long ([see #19594 comment](https://github.com/bitcoin/bitcoin/pull/19594#issuecomment-958801638))
A minor added advantage is to release `mapBlocksUnknownParent` memory when the reindexing phase is done. The current situation is somewhat similar to a memory leak because this map exists unused for the remaining lifetime of the process. It's true that this map should be empty of data elements after use, but its internal metadata (indexing structures, etc.) can have non-trivial size because there can be many thousands of simultaneous elements in this map.
This PR helps our efforts to reduce the use of global variables. This variable isn't just global, it's hidden inside a function (it looks like a local variable but has the `static` attribute).
This global variable exists because the `-reindex` processing code calls `LoadExternalBlockFile()` multiple times (once for each block file), but that function must preserve some state between calls (the `mapBlocksUnknownParent` map). This PR fixes this by allocating this map as a local variable in the caller's scope and passing it in on each call. When reindexing completes, the map goes out of scope and is deallocated.
I tested this manually by reindexing on mainnet and signet. Also, the existing `feature_reindex.py` functional test passes.
ACKs for top commit:
mzumsande:
re-ACK dd065dae9fcebd6806ff67703ffa8128e80b97cc
theStack:
re-ACK dd065dae9fcebd6806ff67703ffa8128e80b97cc
shaavan:
reACK dd065dae9fcebd6806ff67703ffa8128e80b97cc
Tree-SHA512: 9cd20e44d2fa1096dd405bc107bc065ea8f904f5b3f63080341b08d8cf57b790df565f58815c2f331377d044d5306708b4bf6bdfc5ef8d0ed85d8e97d744732c
facc2fa7b8a218a0df6a19772e1641ea68dda2e3 Use AutoFile where possible (MacroFake)
6666803c897e4ad27b45cb74e3a9aa74a335f1bf streams: Add AutoFile without ser-type and ser-version (MacroFake)
Pull request description:
This was done in the context of https://github.com/bitcoin/bitcoin/pull/25284 , but I think it also makes sense standalone.
The basic idea is that serialization type should not be initialized when it is not needed. Same for the serialization version.
So do this here for `AutoFile`. `CAutoFile` remains in places where it is not yet possible.
ACKs for top commit:
laanwj:
Code review ACK facc2fa7b8a218a0df6a19772e1641ea68dda2e3
fanquake:
ACK facc2fa7b8a218a0df6a19772e1641ea68dda2e3
Tree-SHA512: d82d024d55af57565ac53d9d1517afafc12b46964effba0332de62a6c77869356fa77f89e6d4834438fff44c45b64fccdf5a1358bfea03e28dfe55013b3c099d
cb3e9a1e3f8d72daaa361fc45dd853775e754b9d Move {Load,Dump}Mempool to kernel namespace (Carl Dong)
aa306765419f7dbea12b12e15553039835ba0e4d Move DEFAULT_PERSIST_MEMPOOL out of libbitcoinkernel (Carl Dong)
06b88ffb8ae7f2b2a93a32908cd80e77fafd270c LoadMempool: Pass in load_path, stop using gArgs (Carl Dong)
b857ac60d9a0433036519c26675378bbf56a1de1 test/fuzz: Invoke LoadMempool via CChainState (Carl Dong)
b3267258b052557fc136b9a4dcb754afb9219470 Move FopenFn to fsbridge namespace (Carl Dong)
ae1e8e37567fa603a5977d7d05105c682dd3f7db mempool: Use NodeClock+friends for LoadMempool (Carl Dong)
f9e8e5719f28d84f68f7d75e26c8e7fccac8e7d3 mempool: Improve comments for [GS]etLoadTried (Carl Dong)
813962da0b17b918941c6849996845e35d84a451 scripted-diff: Rename m_is_loaded -> m_load_tried (Carl Dong)
413f4bb52b72e082ad8716664ede48352b8e7e5a DumpMempool: Pass in dump_path, stop using gArgs (Carl Dong)
bd4407817e523e3c5b347bc6be25ed007cb27034 DumpMempool: Use std::chrono instead of weird int64_t arthmetics (Carl Dong)
c84390b741ab7b61c9f702d8b447c8cadc1257c8 test/mempool_persist: Test manual savemempool when -persistmempool=0 (Carl Dong)
Pull request description:
This is part of the `libbitcoinkernel` project: #24303, https://github.com/bitcoin/bitcoin/projects/18
-----
This PR moves `{Dump,Load}Mempool` into its own `kernel/mempool_persist` module and introduces `ArgsManager` `node::` helpers in `node/mempool_persist_args`to remove the scattered calls to `GetBoolArg("-persistmempool", DEFAULT_PERSIST_MEMPOOL)`.
More context can be gleaned from the commit messages.
-----
One thing I was reflecting on as I wrote this was that in the long run, I think we should probably invert the validation <-> mempool relationship. Instead of mempool not depending on validation, it might make more sense to have validation not depend on mempool. Not super urgent since `libbitcoinkernel` will include both validation and mempool, but perhaps something for the future.
ACKs for top commit:
glozow:
re ACK cb3e9a1e3f8d72daaa361fc45dd853775e754b9d via `git range-diff 7ae032e...cb3e9a1`
MarcoFalke:
ACK cb3e9a1e3f 🔒
ryanofsky:
Code review ACK cb3e9a1e3f8d72daaa361fc45dd853775e754b9d
Tree-SHA512: 979d7237c3abb5a1dd9b5ad3dbf3b954f906a6d8320ed7b923557f41a4472deccae3e8a6bca0018c8e7a3c4a93afecc502acd1e26756f2054f157f1c0edd939d
Also:
1. Have CChainState::LoadMempool and ::ThreadImport take in paths and
pass it through untouched to LoadMempool.
2. Make LoadMempool exit early if the load_path is empty.
3. Adjust the call to ::ThreadImport in ::AppInitMain to correctly pass
in an empty path if mempool persistence is disabled.
Not only does this increase coverage, it is also more correct in that
when ::LoadMempool is called with a mempool and chainstate, it calls
AcceptToMemoryPool with just the chainstate.
AcceptToMemoryPool will then act on the chainstate's mempool via
CChainState::GetMempool, which may be different from the mempool
originally passed to ::LoadMempool. (In this fuzz test's case, it
definitely is different)
Also, move DummyChainstate to its own file since it's now used by the
validation_load_mempool fuzz test to replace CChainState's m_mempool.
Also introduce node::{ShouldPersistMempool,MempoolPath} helper functions
in node/mempool_persist_args.{h,cpp} which are used by non-kernel
DumpMempool callers to determine whether or not to automatically dump
the mempool and where to dump it to.
- Store the mempool expiry (-mempoolexpiry) in CTxMemPool as a
std::chrono::seconds member.
- Remove the requirement to explicitly specify a mempool expiry for
LimitMempoolSize(...), just use the newly-introduced member.
- Remove all now-unnecessary instances of:
std::chrono::hours{gArgs.GetIntArg("-mempoolexpiry", DEFAULT_MEMPOOL_EXPIRY)}
ce893c0497fc9b8ab9752153dfcc77c9f427545e doc: Update developer notes (Anthony Towns)
d2852917eecad6ab422a7b2c9892d351a7f0cc96 sync.h: Imply negative assertions when calling LOCK (Anthony Towns)
bba87c0553780eacf0317fbfec7330ea27aa02f8 scripted-diff: Convert global Mutexes to GlobalMutexes (Anthony Towns)
a559509a0b8cade27199740212d7b589f71a0e3b sync.h: Add GlobalMutex type (Anthony Towns)
be6aa72f9f8d50b6b5b19b319a74abe7ab4099ff qt/clientmodel: thread safety annotation for m_cached_tip_mutex (Anthony Towns)
f24bd45b37e1b2d19e5a053dbfefa30306c1d41a net_processing: thread safety annotation for m_tx_relay_mutex (Anthony Towns)
Pull request description:
This changes `LOCK(mutex)` for non-global, non-recursive mutexes to be annotated with the negative capability for the mutex it refers to, to prevent . clang applies negative capabilities recursively, so this helps avoid forgetting to annotate functions.
This can't reasonably be used for globals, because clang would require every function to be annotated with `EXCLUSIVE_LOCKS_REQUIRED(!g_mutex)` for each global mutex; so this introduces a trivial `GlobalMutex` subclass of `Mutex`, and reduces the annotations for both `GlobalMutex` to `LOCKS_EXCLUDED` which only catches trivial errors (eg (`LOCK(x); LOCK(x);`).
ACKs for top commit:
MarcoFalke:
review ACK ce893c0497fc9b8ab9752153dfcc77c9f427545e 🐦
hebasto:
ACK ce893c0497fc9b8ab9752153dfcc77c9f427545e
Tree-SHA512: 5c35e8c7677ce3d994a7e3774f4344adad496223a51b3a1d1d3b5f20684b2e1d5cff688eb3fbc8d33e1b9940dfa76e515f9434e21de6f3ce3c935e29a319f529
We want m_chainparams to be alive for the duration of
ChainstateManager's lifetime since ChainstateManager's behaviour depends
on m_chainparams.
We could allow for a std::shared_ptr to be passed in as m_chainparams,
but that complicates things further. Given that CChainParams is not an
entity class or struct, we can just copy it and have ChainstateManager
own it.
This decouples validation.cpp from netaddress.cpp (transitively,
timedata.cpp, and asmap.cpp).
This is important for libbitcoinkernel as:
- There is no reason for the consensus engine to be coupled with
netaddress, timedata, and asmap
- Users of libbitcoinkernel can now easily supply their own
std::function that provides the adjusted time.
See the src/Makefile.am changes for some satisfying removals.
[META] Although it seems like we don't need it for just one option,
we're going to introduce another member to this struct *in the
next commit*. In future patchsets for libbitcoinkernel decoupling
it from ArgsManager, even more members will be added here.
f64aa9c411ad78259756a28756ec1eb8069b5ab4 Disallow more unsafe string->path conversions allowed by path append operators (Ryan Ofsky)
Pull request description:
Add more `fs::path` `operator/` and `operator+` overloads to prevent unsafe string->path conversions on Windows that would cause strings to be decoded according to the current Windows locale & code page instead of the correct string encoding.
Update application code to deal with loss of implicit string->path conversions by calling `fs::u8path` or `fs::PathFromString` explicitly, or by just changing variable types from `std::string` to `fs::path` to avoid conversions altogether, or make them happen earlier.
In all cases, there's no change in behavior either (1) because strings only contained ASCII characters and would be decoded the same regardless of what encoding was used, or (2) because of the 1:1 mapping between paths and strings using the `PathToString` and `PathFromString` functions.
Motivation for this PR was just that I was experimenting with #24469 and noticed that operations like `fs::path / std::string` were allowed, and I thought it would be better not to allow them.
ACKs for top commit:
hebasto:
ACK f64aa9c411ad78259756a28756ec1eb8069b5ab4
Tree-SHA512: 944cce49ed51537ee7a35ea4ea7f5feaf0c8fff2fa67ee81ec5adebfd3dcbaf41b73eb35e49973d5f852620367f13506fd12a7a9b5ae3a7a0007414d5c9df50f
In previous commits in this patchset, we've made sure that every
Unload/UnloadBlockIndex member function resets its own members, and does
not reach out to globals.
This means that their corresponding classes' default destructors can now
replace them, and do an even more thorough job without the need to be
updated for every new member variable.
Therefore, we can remove them, and also remove UnloadBlockIndex since
that's not used anymore.
Unfortunately, chainstatemanager_loadblockindex relies on
CChainState::UnloadBlockIndex, so that needs to stay for now.
fa1970f075292d7312654730a994a68c2ca8bc06 Make BlockManager::LoadBlockIndex private (MarcoFalke)
Pull request description:
* After commit fa27f03b4943540aa2eab283d4cf50ad4a1a01f8 `BlockManager::LoadBlockIndex` is only called by `BlockManager::LoadBlockIndexDB`. Thus, it can be made `private`.
* After commit c600ee38168a460d3026eae0e289c976194aad14 `m_best_invalid` is no longer accessed by `BlockManager::LoadBlockIndex`. Thus, the unused `friend` can be removed.
ACKs for top commit:
mruddy:
ACK fa1970f075292d7312654730a994a68c2ca8bc06 I verified by double checking references, then applying the patch, and running `make check`. LGTM.
Tree-SHA512: 9b36b4c59bf7ad01171764ce61b1be9750fc92d105c4fe939b1a6a70027ab6300d5d2a2fc3e82f981e22c3987f2ca84e092d2e1f8463fa320af9f05048580c0a
Add more fs::path operator/ and operator+ overloads to prevent unsafe
string->path conversions on Windows that would cause strings to be
decoded according to the current Windows locale & code page instead of
the correct string encoding.
Update application code to deal with loss of implicit string->path
conversions by calling fs::u8path or fs::PathFromString explicitly, or
by just changing variable types from std::string to fs::path to avoid
conversions altoghther, or make them happen earlier.
In all cases, there's no change in behavior either (1) because strings
only contained ASCII characters and would be decoded the same regardless
of what encoding was used, or (2) because of the 1:1 mapping between
paths and strings using the PathToString and PathFromString functions.
Co-authored-by: Hennadii Stepanov <32963518+hebasto@users.noreply.github.com>
f865cf8ded2b2fbc82a6fbc41226d991909a6880 Add and use BlockManager::GetAllBlockIndices (Carl Dong)
28ba0313eac37e4a900b7e97af7169ce999c4024 Add and use CBlockIndexHeightOnlyComparator (Carl Dong)
12eb05df63f930969115af6dc66e2e5d02f2a517 move-only: Move CBlockIndexWorkComparator to blockstorage (Carl Dong)
c600ee38168a460d3026eae0e289c976194aad14 Only load BlockMan in BlockMan member functions (Carl Dong)
42e56d9b188f97c077ed2269e24acc0be35ece17 style-only: No need for std::pair for vSortedByHeight (Carl Dong)
3bbb6fea051f4e19bd2448e401a6c4e9b4cc7a41 style-only: Various blockstorage.cpp cleanups (Carl Dong)
5be9ee3c54dcb396ff52fc8c8b7e4e6e39ec4a3b refactor: more const annotations for uses of CBlockIndex* (Anthony Towns)
Pull request description:
The only important commit is "Only load BlockMan in BlockMan member functions", everything else is all just small style changes.
Here's the commit message, reproduced:
```
This commit effectively splits the "load block index itself" logic from
"derive Chainstate variables from loaded block index" logic.
This means that BlockManager::LoadBlockIndex{,DB} will only load what's
relevant to the BlockManager.
```
ACKs for top commit:
ajtowns:
ACK f865cf8ded2b2fbc82a6fbc41226d991909a6880 ; code review only
MarcoFalke:
review ACK f865cf8ded2b2fbc82a6fbc41226d991909a6880 🗂
Tree-SHA512: 7b204d782834e06fd7329d022e2ae860181b4e8105c33bfb928539a4ec24161dc7438a9c4d4ee279dcad77de310c160b997bb8aa18923243d0fd55ccf4ad7c3a
This commit effectively splits the "load block index itself" logic from
"derive Chainstate variables from loaded block index" logic.
This means that BlockManager::LoadBlockIndex{,DB} will only load what's
relevant to the BlockManager.
I strongly recommend reviewing with the following git-diff flags:
--color-moved=dimmed_zebra --color-moved-ws=allow-indentation-change
fa8d4d9128c35de0fe715f2e2b99269d23c09cc1 scripted-diff: Clarify CheckFinalTxAtTip name (MarcoFalke)
fa4e30b0f36f2e7a09db7d30dca9008ed9dbcb35 policy: Remove unused locktime flags (MarcoFalke)
Pull request description:
The locktime flags have many issues:
* They are passed in by a default argument, which is fragile. It has already lead to bugs like the one fixed in commit e30b6ea194fee3bb95a45e7b732a99566b88f1f5.
* They are negative (signed), which doesn't make sense for flags (unsigned in general). According to the review comments when the code was added: "The max on the flags is a fairly weird operation." (https://github.com/bitcoin/bitcoin/pull/6566#issuecomment-150310861)
* No call site relies on the default argument and they all pass in a single compile-time constant, rendering most of the code dead and untested.
* The dead code calls `GetAdjustedTime` (network adjusted time), which has its own issues. See https://github.com/bitcoin/bitcoin/issues/4521
Fix all issues by removing them
ACKs for top commit:
ajtowns:
ACK fa8d4d9128c35de0fe715f2e2b99269d23c09cc1
theStack:
Code-review ACK fa8d4d9128c35de0fe715f2e2b99269d23c09cc1
glozow:
ACK fa8d4d9128c35de0fe715f2e2b99269d23c09cc1, agree the default arg `flags` is a massive footgun and just setting max flags is weird. Adding `AtTip` to the names makes sense to me, since they're both testing for *next* block and only ever used for {,re}addition to mempool.
Tree-SHA512: 79f4a52f34909eb598d88bbae7afe8abe5f85f45c128483d16aa83dacd0e5579e561b725d01b1e9a931d1821012a51ad2bc6fb2867f8d09ee541f9d234d696f8
