Also known as Ephemeral Dust.
We try to ensure that dust is spent in blocks by requiring:
- ephemeral dust tx is 0-fee
- ephemeral dust tx only has one dust output
- If the ephemeral dust transaction has a child,
the dust is spent by by that child.
0-fee requirement means there is no incentive to mine
a transaction which doesn't have a child bringing its
own fees for the transaction package.
0de3e96e33 tracing: use bitcoind pid in bcc tracing examples (0xb10c)
411c6cfc6c tracing: only prepare tracepoint args if attached (0xb10c)
d524c1ec06 tracing: dedup TRACE macros & rename to TRACEPOINT (0xb10c)
Pull request description:
Currently, if the tracepoints are compiled (e.g. in depends and release builds), we always prepare the tracepoint arguments regardless of the tracepoints being used or not. We made sure that the argument preparation is as cheap as possible, but we can almost completely eliminate any overhead for users not interested in the tracepoints (the vast majority), by gating the tracepoint argument preparation with an `if(something is attached to this tracepoint)`. To achieve this, we use the optional semaphore feature provided by SystemTap.
The first commit simplifies and deduplicates our tracepoint macros from 13 TRACEx macros to a single TRACEPOINT macro. This makes them easier to use and also avoids more duplicate macro definitions in the second commit.
The Linux tracing tools I'm aware of (bcc, bpftrace, libbpf, and systemtap) all support the semaphore gating feature. Thus, all existing tracepoints and their argument preparation is gated in the second commit. For details, please refer to the commit messages and the updated documentation in `doc/tracing.md`.
Also adding unit tests that include all tracepoint macros to make sure there are no compiler problems with them (e.g. some varadiac extension not supported).
Reviewers might want to check:
- Do the tracepoints still work for you? Do the examples in `contrib/tracing/` run on your system (as bpftrace frequently breaks on every new version, please test master too if it should't work for you)? Do the CI interface tests still pass?
- Is the new documentation clear?
- The `TRACEPOINT_SEMAPHORE(event, context)` macros places global variables in our global namespace. Is this something we strictly want to avoid or maybe move to all `TRACEPOINT_SEMAPHORE`s to a separate .cpp file or even namespace? I like having the `TRACEPOINT_SEMAPHORE()` in same file as the `TRACEPOINT()`, but open for suggestion on alternative approaches.
- Are newly added tracepoints in the unit tests visible when using `readelf -n build/src/test/test_bitcoin`? You can run the new unit tests with `./build/src/test/test_bitcoin --run_test=util_trace_tests* --log_level=all`.
<details><summary>Two of the added unit tests demonstrate that we are only processing the tracepoint arguments when attached by having a test-failure condition in the tracepoint argument preparation. The following bpftrace script can be used to demonstrate that the tests do indeed fail when attached to the tracepoints.</summary>
`fail_tests.bt`:
```c
#!/usr/bin/env bpftrace
usdt:./build/src/test/test_bitcoin:test:check_if_attached {
printf("the 'check_if_attached' test should have failed\n");
}
usdt:./build/src/test/test_bitcoin:test:expensive_section {
printf("the 'expensive_section' test should have failed\n");
}
```
Run the unit tests with `./build/src/test/test_bitcoin` and start `bpftrace fail_tests.bt -p $(pidof test_bitcoin)` in a separate terminal. The unit tests should fail with:
```
Running 594 test cases...
test/util_trace_tests.cpp(31): error: in "util_trace_tests/test_tracepoint_check_if_attached": check false has failed
test/util_trace_tests.cpp(51): error: in "util_trace_tests/test_tracepoint_manual_tracepoint_active_check": check false has failed
*** 2 failures are detected in the test module "Bitcoin Core Test Suite"
```
</details>
These links might provide more contextual information for reviewers:
- [How SystemTap Userspace Probes Work by eklitzke](https://eklitzke.org/how-sytemtap-userspace-probes-work) (actually an example on Bitcoin Core; mentions that with semaphores "the overhead for an untraced process is effectively zero.")
- [libbpf comment on USDT semaphore handling](1596a09b5d/src/usdt.c (L83-L92)) (can recommend the whole comment for background on how the tracepoints and tracing tools work together)
- https://sourceware.org/systemtap/wiki/UserSpaceProbeImplementation#Semaphore_Handling
ACKs for top commit:
willcl-ark:
utACK 0de3e96e33
laanwj:
re-ACK 0de3e96e33
jb55:
utACK 0de3e96e33
vasild:
ACK 0de3e96e33
Tree-SHA512: 0e5e0dc5e0353beaf5c446e4be03d447e64228b1be71ee9972fde1d6fac3fac71a9d73c6ce4fa68975f87db2b2bf6eee2009921a2a145e24d83a475d007a559b
c189eec848 doc: release note for mempoolrullrbf removal (Greg Sanders)
d47297c6aa rpc: Mark fullrbf and bip125-replaceable as deprecated (Greg Sanders)
04a5dcee8a docs: remove requirement to signal bip125 (Greg Sanders)
111a23d9b3 Remove -mempoolfullrbf option (Greg Sanders)
Pull request description:
Given https://github.com/bitcoin/bitcoin/pull/30493 and the related discussion on network uptake it's probably not helpful to have an option for a feature that will not be respected by the network in any meaningful way.
Wallet changes can be done in another PR on its own cadence to account for possible fingerprinting, waiting for fullrbf logic to permeate the network, etc.
ACKs for top commit:
stickies-v:
re-ACK c189eec848
achow101:
ACK c189eec848
murchandamus:
ACK c189eec848
rkrux:
reACK c189eec848
Tree-SHA512: 9447f88f8f291c56c5bde70af0a91b0a4f5163aaaf173370fbfdaa3c3fd0b44120b14d3a1977f7ee10e27ffe9453f8a70dd38aad0ffb8c39cf145049d2550730
3a4a788ee0 init: Correct coins db cache size setting (TheCharlatan)
Pull request description:
The chainstate caches are currently re-balanced on startup even in the non-assumeutxo case, leading to the database being needlessly re-opened and its cache re-allocated.
Similar to `InitCoinsCache` and `m_coinstip_cache_size_bytes`, the `m_coinsdb_cache_size_bytes` should be set in `InitCoinsDB`.
Together with only conservatively setting the cache values when a assumeutxo chainstate is present, this allows for skipping the cache re-balance during initialization in the normal non-assumeutxo case.
Before:
```
2024-10-09T21:22:17Z Checking all blk files are present...
2024-10-09T21:22:17Z Initializing chainstate Chainstate [ibd] @ height -1 (null)
2024-10-09T21:22:17Z Opening LevelDB in /home/drgrid/.bitcoin/signet/chainstate
2024-10-09T21:22:17Z Opened LevelDB successfully
2024-10-09T21:22:17Z Using obfuscation key for /home/drgrid/.bitcoin/signet/chainstate: b0a6f4e95fd05c92
2024-10-09T21:22:17Z Loaded best chain: hashBestChain=0000000e119967d4937dad58456885ae43fb1761db686947e2f8e168c9a39a4f height=216852 date=2024-10-09T21:06:16Z progress=0.999989
2024-10-09T21:22:17Z Opening LevelDB in /home/drgrid/.bitcoin/signet/chainstate
2024-10-09T21:22:17Z Opened LevelDB successfully
2024-10-09T21:22:17Z Using obfuscation key for /home/drgrid/.bitcoin/signet/chainstate: b0a6f4e95fd05c92
2024-10-09T21:22:17Z [Chainstate [ibd] @ height 216852 (0000000e119967d4937dad58456885ae43fb1761db686947e2f8e168c9a39a4f)] resized coinsdb cache to 8.0 MiB
2024-10-09T21:22:17Z [Chainstate [ibd] @ height 216852 (0000000e119967d4937dad58456885ae43fb1761db686947e2f8e168c9a39a4f)] resized coinstip cache to 440.0 MiB
2024-10-09T21:22:17Z init message: Verifying blocks…
```
After:
```
2024-10-09T21:21:37Z Checking all blk files are present...
2024-10-09T21:21:37Z Initializing chainstate Chainstate [ibd] @ height -1 (null)
2024-10-09T21:21:37Z Opening LevelDB in /home/drgrid/.bitcoin/signet/chainstate
2024-10-09T21:21:37Z Opened LevelDB successfully
2024-10-09T21:21:37Z Using obfuscation key for /home/drgrid/.bitcoin/signet/chainstate: b0a6f4e95fd05c92
2024-10-09T21:21:37Z Loaded best chain: hashBestChain=0000012c12b48011a7d9150ce96ed6a44bbf32b09eeecaff4a667789dda2a566 height=216850 date=2024-10-09T20:37:05Z progress=0.999971
2024-10-09T21:21:37Z init message: Verifying blocks…
```
The change may also be verified by looking at the `feature_assumeutxo.py` functional test debug logs.
ACKs for top commit:
fjahr:
utACK 3a4a788ee0
achow101:
ACK 3a4a788ee0
laanwj:
Code review ACK 3a4a788ee0
BrandonOdiwuor:
Code Review ACK 3a4a788ee0
Tree-SHA512: 87878d0d196bb426370d4b4bd180ca52a34017a0799ecea651c2532461fd2927b0f7cc8182276a7d9bb1fe0ede7d0ad677e3714ca22f321917d711c643acc578
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.
4feaa28728 refactor: Rely on returned value of GetCoin instead of parameter (Lőrinc)
46dfbf169b refactor: Return optional of Coin in GetCoin (Lőrinc)
e31bfb26c2 refactor: Remove unrealistic simulation state (Lőrinc)
Pull request description:
While reviewing [the removal of the unreachable combinations from the Coin cache logic](https://github.com/bitcoin/bitcoin/pull/30673#discussion_r1721727681), we've noticed that the related tests often [reflect impossible states](https://github.com/bitcoin/bitcoin/pull/30673/files#r1740154464).
Browsing the Coin cache refactoring history revealed that migrating `bool GetCoin` to `optional<Coin> GetCoin` was [already proposed a few times before](https://github.com/bitcoin/bitcoin/pull/18746#issuecomment-842393167).
This refactor makes certain invalid states impossible, reducing the possibility of errors and making the code easier to understand. This will let us remove test code that exercises the impossible states as well.
The PR is done in multiple small steps, first swapping the new `optional` return value, slowly strangling out the usages of the return parameter, followed by the removal of the parameter.
Most of the invalid test states were still kept, except for https://github.com/bitcoin/bitcoin/pull/30673/files#r1748087322, where the new design prohibits invalid usage and https://github.com/bitcoin/bitcoin/pull/30673/files#r1749350258 was just marked with a TODO, will be removed in a follow-up PR.
ACKs for top commit:
andrewtoth:
re-ACK 4feaa28728
achow101:
ACK 4feaa28728
laanwj:
Code review ACK 4feaa28728
theStack:
Code-review ACK 4feaa28728
Tree-SHA512: 818d60b2e97f58c489a61120fe761fb67a08dffbefe7a3fce712d362fc9eb8c2cced23074f1bec55fe71c616a3561b5a8737919ad6ffb2635467ec4711683df7
86e2a6b749 [test] A non-standard transaction which is also consensus-invalid should return the consensus error (Antoine Poinsot)
f859ff8a4e [validation] Improve script check error reporting (dergoegge)
Pull request description:
An input script might be invalid for multiple reasons. For example, it might fail both a standardness check and a consensus check, which can lead to a `mandatory-script-verify-flag-failed` error being reported that includes the script error string from the standardness failure (e.g. `mandatory-script-verify-flag-failed (Using OP_CODESEPARATOR in non-witness script)`), which is confusing.
ACKs for top commit:
darosior:
re-ACK 86e2a6b749
ariard:
Re-Code Review ACK 86e2a6b7
instagibbs:
ACK 86e2a6b749
Tree-SHA512: 053939107c0bcd6643e9006b2518ddc3a6de47d2c6c66af71a04e8af5cf9ec207f19e54583b7a056efd77571edf5fd4f36c31ebe80d1f0777219c756c055eb42
cd0edf26c0 tracing: cast block_connected duration to nanoseconds (0xb10c)
Pull request description:
When the `validation:block_connected` tracepoint was introduced in 8f37f5c2a5, the connect block duration was passed in microseconds `µs`. By starting to use steady clock in fabf1cdb20 this changed to nanoseconds `ns`. As the test only checked if the duration value is `> 0` as a plausibility check, this went unnoticed. This was detected this when setting up monitoring for block validation time as part of the Great Consensus Cleanup Revival discussion.
This change casts the duration explicitly to nanoseconds, updates the documentation, and adds a check for an upper bound to the tracepoint interface tests. The upper bound is quite lax as mining the block takes much longer than connecting the empty test block. It's however able to detect a duration passed in an incorrect unit (1000x off).
A previous version of this PR casted the duration to microseconds `µs` - however, as the last three major releases have had the duration as nanoseconds (and this went unnoticed), we assume that this is the API now and changeing it back to microseconds would break the API again. See also https://github.com/bitcoin/bitcoin/pull/29877#issuecomment-2067867597
ACKs for top commit:
maflcko:
re-lgtm ACK cd0edf26c0
laanwj:
re-ACK cd0edf26c0
Tree-SHA512: 54a1eea0297e01c07c2d071ffafbf97dbd080f763e1dc0014ff086a913b739637c1634b1cf87c90b94a3c2f66006acfaada0414a15769cac761e03bc4aab2a77
The chainstate caches are currently re-balanced on startup
even in the non-assumeutxo case, leading to the database being
needlessly re-opened and its cache re-allocated.
Similar to `InitCoinsCache` and `m_coinstip_cache_size_bytes` the
`m_coinsdb_cache_size_bytes` should be set in `InitCoinsDB`.
Together with only conservatively setting the cache values when a
assumeutxo chainstate is present, this allows for skipping the cache
re-balance during initialization in the normal non-assumeutxo case.
It should be obvious that a wait is not needed if the tip does not
match.
Also, remove a comment that the blockTip notification was only meant for
the "UI". (It is used by other stuff for a long time)
The comparison of m_best_invalid with the tip of the respective chainstate
makes no sense for the background chainstate, and can lead to incorrect
error messages.
7942951e3f Remove unused g_best_block (Ryan Ofsky)
e3a560ca68 rpc: use waitTipChanged for longpoll (Ryan Ofsky)
460687a09c Remove unused CRPCSignals (Sjors Provoost)
dca923150e Replace RPCNotifyBlockChange with waitTipChanged() (Sjors Provoost)
2a40ee1121 rpc: check for negative timeout arg in waitfor* (Sjors Provoost)
de7c855b3a rpc: recommend -rpcclienttimeout=0 for waitfor* (Sjors Provoost)
77ec072925 rpc: fix waitfornewblock description (Sjors Provoost)
285fe9fb51 rpc: add test for waitforblock and waitfornewblock (Sjors Provoost)
b94b27cf05 Add waitTipChanged to Mining interface (Sjors Provoost)
7eccdaf160 node: Track last block that received a blockTip notification (Sjors Provoost)
ebb8215f23 Rename getTipHash() to getTip() and return BlockRef (Sjors Provoost)
89a8f74bbb refactor: rename BlockKey to BlockRef (Sjors Provoost)
Pull request description:
This continues the work in #30200 so that a future Stratum v2 Template Provider (see #29432) can avoid accessing node internals. It needs to know when a new block arrives in order to push new templates to connected clients.
`waitTipChanged()` uses a new kernel notification `notifications().m_tip_block_mutex`, which this PR also introduces (a previous version used `g_best_block`).
In order to ensure the new method works as intended, the `waitfornewblock`, `waitforblock` and `waitforblockheight` RPC methods are refactored to use it. This allows removing `RPCNotifyBlockChange`.
There's a commit to add (direct) tests for the methods that are about to be refactored:
- `waitfornewblock` was already implicitly tested by `feature_shutdown.py`.
- `waitforblockheight` by `feature_coinstatsindex.py` and `example_test.py`
This PR renames `getTipHash()` to `getTip()` and returns a `BlockRef` (renamed from `BlockKey`) so that callers can use either the height or hash.
The later commits make trivial improvements to the `waitfor*` RPC calls (not needed for this PR).
The `waitTipChanged()` method could probably also be used for the longpoll functionality in `getblocktemplate`, but I'm a bit reluctant to touch that.
`RPCServer::OnStarted` no longer does anything and `RPCServer::OnStopped` merely prints a log statement. They were added in #5711 as a refactor. This PR drops them entirely.
Finally `g_best_block` is also dropped.
ACKs for top commit:
achow101:
ACK 7942951e3f
ryanofsky:
Code review ACK 7942951e3f. Just rebased since last review
TheCharlatan:
Re-ACK 7942951e3f
Tree-SHA512: a5559446b4000c95e07aad33284b7ee2e57aafd87e1ae778b3825d59689566d047a8047e47a10f76e6e341e7dc72fd265a65afbc0a9c011d17c4cafd55031837
Because AssumeUTXO nodes prioritize tip synchronization, they relay their local
address through the network before completing the background chain sync.
This, combined with the advertising of full-node service (NODE_NETWORK), can
result in an honest peer in IBD connecting to the AssumeUTXO node (while syncing)
and requesting an historical block the node does not have. This behavior leads to
an abrupt disconnection due to perceived unresponsiveness (lack of response)
from the AssumeUTXO node.
This lack of response occurs because nodes ignore getdata requests when they do
not have the block data available (further discussion can be found in PR 30385).
Fix this by refraining from signaling full-node service support while the
background chain is being synced. During this period, the node will only
signal 'NODE_NETWORK_LIMITED' support. Then, full-node ('NODE_NETWORK')
support will be re-enabled once the background chain sync is completed.
a2955f0979 validation: Use span for ImportBlocks paths (TheCharlatan)
20515ea3f5 validation: Use span for CalculateClaimedHeadersWork (TheCharlatan)
52575e96e7 validation: Use span for ProcessNewBlockHeaders (TheCharlatan)
Pull request description:
Makes it friendlier for potential future users of the kernel library if they do not store the headers in a std::vector, but can guarantee contiguous memory.
Take this opportunity to also change the argument of ImportBlocks previously taking a `std::vector` to a `std::span`.
ACKs for top commit:
stickies-v:
re-ACK a2955f0979 - no changes except further walking the ~file~ path of modernizing variable names.
maflcko:
ACK a2955f0979🕑
achow101:
ACK a2955f0979
danielabrozzoni:
ACK a2955f0979
Tree-SHA512: 8b07f4ad26e270b65600d1968cd78847b85caca5bfbb83fd9860389f26656b1d9a40b85e0990339f50403d18cedcd2456990054f3b8b0bedce943e50222d2709
When the tracepoint was introduced in 8f37f5c2a5,
the connect_block duration was passed in microseconds `µs`.
By starting to use steady clock in fabf1cdb20
this changed to nanoseconds `ns`. As the test only checked if the
duration value is `> 0` as a plausibility check, this went unnoticed.
I detected this when setting up monitoring for block validation time
as part of the Great Consensus Cleanup Revival discussion.
This change casts the duration explicitly to nanoseconds (as it has been
nanoseconds for the last three releases; switching back now would 'break'
the broken API again; there don't seem to be many users affected), updates
the documentation and adds a check for an upper bound to the tracepoint
interface tests. The upper bound is quite lax as mining the block takes
much longer than connecting the empty test block. It's however able to
detect incorrect duration units passed.
Makes it friendlier for potential future users of the kernel library if
they do not store the headers in a std::vector, but can guarantee
contiguous memory.
Makes it friendlier for potential future users of the kernel library if
they do not store the headers in a std::vector, but can guarantee
contiguous memory.
In 6bfa26048d the testnet4 timewarp
attack fix block time variation was increased from the Great
Consensus Cleanup value of 600s to 7200s on the thesis that this
allows miners to always create blocks with the current time. Sadly,
doing so does allow for some nonzero inflation, even if not a huge
amount.
While it could be that some hardware ignores the timestamp provided
to it over Stratum and forces the block header timestamp to the
current time, I'm not aware of any such hardware, and it would also
likely suffer from random invalid blocks due to relying on NTP
anyway, making its existence highly unlikely.
This leaves the only concern being pools, but most of those rely on
work generated by Bitcoin Core (in one way or another, though when
spy mining possibly not), and it seems likely that they will also
not suffer any lost work. While its possible that a pool does
generate invalid work due to spy mining or otherwise custom logic,
it seems unlikely that a substantial portion of hashrate would do
so, making the difference somewhat academic (any pool that screws
this up will only do so once and the network would come out just
fine).
Further, while we may end up deciding these assumptions were
invalid and we should instead use 7200s, it seems prudent to try
with the value we "want" on testnet4, giving us the ability to
learn if the compatibility concerns are an issue before we go to
mainnet.
00618e8745 assumeutxo: Drop block height from metadata (Fabian Jahr)
Pull request description:
Fixes https://github.com/bitcoin/bitcoin/issues/30514 which has more context and shows how the issue can be reproduced. Since the value in question is removed, there is no test to add to reproduce anything.
This is an alternative approach to #30516 with much of the [code being suggested there](https://github.com/bitcoin/bitcoin/pull/30516#discussion_r1689146902).
ACKs for top commit:
maflcko:
re-ACK 00618e8745🎌
achow101:
ACK 00618e8745
theStack:
Code-review ACK 00618e8745
ismaelsadeeq:
Re-ACK 00618e8745
mzumsande:
ACK 00618e8745
Tree-SHA512: db9575247bae838ad7742a27a216faaf55bb11e022f9afdd05752bb09bbf9614717d0ad64304ff5722a16bf41d8dea888af544e4ae26dcaa528c1add0269a4a8
92c1d7d1f8 validation: Use MAX_TIMEWARP constant as testnet4 timewarp defense delta (Fabian Jahr)
4b2fad502e doc: Add release notes for 29775 (Fabian Jahr)
f7cc97313b doc: Align deprecation warnings (Fabian Jahr)
1163b08378 chainparams: Add initial minimum chain work for Testnet4 (Fabian Jahr)
Pull request description:
This completes follow-ups left open in #29775.
- Adds release notes
- Addresses the [misalignment](https://github.com/bitcoin/bitcoin/pull/29775#discussion_r1706982102) in deprecation warnings and hints at the intention to remove support for Testnet3.
- Adds initial minimum chainwork for Testnet4.
- Use the `MAX_TIMEWARP` constant as the timewarp defense delta, equal to `MAX_FUTURE_BLOCK_TIME`.
ACKs for top commit:
Sjors:
ACK 92c1d7d1f8
achow101:
ACK 92c1d7d1f8
tdb3:
re ACK 92c1d7d1f8
Tree-SHA512: 7ebdac7809f96231f75ca62706af59cd1ed27f713a4c7be5e2ad69fae95832b146b3ea23c712fb03b412da1deda7e8a5dae55bb2bbd2dcfd9f926e85c2a72666
2925bd537c refactor: use c++20 std::views::reverse instead of reverse_iterator.h (stickies-v)
Pull request description:
C++20 introduces [`std::ranges::views::reverse`](https://en.cppreference.com/w/cpp/ranges/reverse_view), which allows us to drop our own `reverse_iterator.h` implementation and also makes it easier to chain views (even though I think we currently don't use this).
ACKs for top commit:
achow101:
ACK 2925bd537c
maflcko:
ACK 2925bd537c🎷
Tree-SHA512: 567666ec44af5d1beb7a271836bcc89c4c577abc77f522fcc18bc6d4de516ae9b0df766d0bfa6dd217569e6878331c2aee1d9815620860375e3510dad7fed476
589db872e1 validation: don't erase coins cache on prune flushes (Andrew Toth)
0e8918755f Add linked-list test to CCoinsViewCache::SanityCheck (Pieter Wuille)
05cf4e1875 coins: move Sync logic to CoinsViewCacheCursor (Andrew Toth)
7825b8b9ae coins: pass linked list of flagged entries to BatchWrite (Andrew Toth)
a14edada8a test: add cache entry linked list tests (Andrew Toth)
24ce37cb86 coins: track flagged cache entries in linked list (Andrew Toth)
58b7ed156d coins: call ClearFlags in CCoinsCacheEntry destructor (Andrew Toth)
8bd3959fea refactor: require self and sentinel parameters for AddFlags (Andrew Toth)
75f36d241d refactor: add CoinsCachePair alias (Andrew Toth)
f08faeade2 refactor: move flags to private uint8_t and rename to m_flags (Andrew Toth)
4e4fb4cbab refactor: disallow setting flags in CCoinsCacheEntry constructors (Andrew Toth)
8737c0cefa refactor: encapsulate flags setting with AddFlags and ClearFlags (Andrew Toth)
9715d3bf1e refactor: encapsulate flags get access for all other checks (Andrew Toth)
df34a94e57 refactor: encapsulate flags access for dirty and fresh checks (Andrew Toth)
Pull request description:
Since https://github.com/bitcoin/bitcoin/pull/17487 we no longer need to clear the coins cache when syncing to disk. A warm coins cache significantly speeds up block connection, and only needs to be fully flushed when nearing the `dbcache` limit.
For frequent pruning flushes there's no need to empty the cache and kill connect block speed. However, simply using `Sync` in place of `Flush` actually slows down a pruned full IBD with a high `dbcache` value. This is because as the cache grows, sync takes longer since every coin in the cache is scanned to check if it's dirty. For frequent prune flushes and a large cache this constant scanning starts to really slow IBD down, and just emptying the cache on every prune becomes faster.
To fix this, we can add two pointers to each cache entry and construct a doubly linked list of dirty entries. We can then only iterate through all dirty entries on each `Sync`, and simply clear the pointers after.
With this approach a full IBD with `dbcache=16384` and `prune=550` was 32% faster than master. For default `dbcache=450` speedup was ~9%. All benchmarks were run with `stopatheight=800000`.
| | prune | dbcache | time | max RSS | speedup |
|-----------:|----------:|------------:|--------:|-------------:|--------------:|
| master | 550 | 16384 | 8:52:57 | 2,417,464k | - |
| branch | 550 | 16384 | 6:01:00 | 16,216,736k | 32% |
| branch | 550 | 450 | 8:05:08 | 2,818,072k | 8.8% |
| master | 10000 | 5000 | 8:19:59 | 2,962,752k | - |
| branch | 10000 | 5000| 5:56:39 | 6,179,764k | 28.8% |
| master | 0 | 16384 | 4:51:53 | 14,726,408k | - |
| branch | 0 | 16384 | 4:43:11 | 16,526,348k | 2.7% |
| master | 0 | 450 | 7:08:07 | 3,005,892k | - |
| branch | 0 | 450 | 6:57:24 | 3,013,556k |2.6%|
While the 2 pointers add memory to each cache entry, it did not slow down IBD. For non-pruned IBD results were similar for this branch and master. When I performed the initial IBD, the full UTXO set could be held in memory when using the max `dbcache` value. For non-pruned IBD with max `dbcache` to tip ended up using 12% more memory, but it was also 2.7% faster somehow. For smaller `dbcache` values the `dbcache` limit is respected so does not consume more memory, and the potentially more frequent flushes were not significant enough to cause any slowdown.
For reviewers, the commits in order do the following:
First 4 commits encapsulate all accesses to `flags` on cache entries, and then the 5th makes `flags` private.
Commits `refactor: add CoinsCachePair alias` to `coins: call ClearFlags in CCoinsCacheEntry destructor` create the linked list head nodes and cache entry self references and pass them into `AddFlags`.
Commit `coins: track flagged cache entries in linked list` actually adds the entries into a linked list when they are flagged DIRTY or FRESH and removes them from the linked list when they are destroyed or the flags are cleared manually. However, the linked list is not yet used anywhere.
Commit `test: add cache entry linked list tests` adds unit tests for the linked list.
Commit `coins: pass linked list of flagged entries to BatchWrite` uses the linked list to iterate through DIRTY entries instead of using the entire coins cache.
Commit `validation: don't erase coins cache on prune flushes` uses `Sync` instead of `Flush` for pruning flushes, so the cache is no longer cleared.
Inspired by [this comment](https://github.com/bitcoin/bitcoin/pull/15265#issuecomment-457720636).
Fixes https://github.com/bitcoin/bitcoin/issues/11315.
ACKs for top commit:
paplorinc:
ACK 589db872e1
sipa:
reACK 589db872e1
achow101:
ACK 589db872e1
mzumsande:
re-ACK 589db872e1
Tree-SHA512: 23b2bc01c83edacb5b39aa60bb0b766de9a74ce17f0c59bf13b97b4328a7b758ad9aff6581c3ca88e2973f7658380651530d497444f48d6e22ea0bfc51cc921d
fa530ec543 rpc: Return precise loadtxoutset error messages (MarcoFalke)
faa5c86dbf refactor: Use untranslated error message in ActivateSnapshot (MarcoFalke)
Pull request description:
The error messages should never happen in normal operation. However, if
they do, they are helpful to return to the user to debug the issue. For
example, to notice a truncated file.
This fixes https://github.com/bitcoin/bitcoin/issues/28621
Also includes a minor refactor commit.
ACKs for top commit:
fjahr:
Code review ACK fa530ec543
ryanofsky:
Code review ACK fa530ec543, just adjusting error messages a little since last review. (Thanks!)
Tree-SHA512: 224968c9b13d082ca2ed1f6a8fcc5f51ff16d6c96bd38c3679699505b54337b99cccaf7a8474391f6b11f9ccb101977b4e626898c1217eae95802e290cf105f1
The error messages should never happen in normal operation. However, if
they do, they are helpful to return to the user to debug the issue. For
example, to notice a truncated file.
c85accecaf [refactor] delete EraseTxNoLock, just use EraseTx (glozow)
6ff84069a5 remove obsoleted TxOrphanage::m_mutex (glozow)
61745c7451 lock m_recent_confirmed_transactions using m_tx_download_mutex (glozow)
723ea0f9a5 remove obsoleted hashRecentRejectsChainTip (glozow)
18a4355250 update recent_rejects filters on ActiveTipChange (glozow)
36f170d879 add ValidationInterface::ActiveTipChange (glozow)
3eb1307df0 guard TxRequest and rejection caches with new mutex (glozow)
Pull request description:
See #27463 for full project tracking.
This contains the first few commits of #30110, which require some thinking about thread safety in review.
- Introduce a new `m_tx_download_mutex` which guards the transaction download data structures including `m_txrequest`, the rolling bloom filters, and `m_orphanage`. Later this should become the mutex guarding `TxDownloadManager`.
- `m_txrequest` doesn't need to be guarded using `cs_main` anymore
- `m_recent_confirmed_transactions` doesn't need its own lock anymore
- `m_orphanage` doesn't need its own lock anymore
- Adds a new `ValidationInterface` event, `ActiveTipChanged`, which is a synchronous callback whenever the tip of the active chainstate changes.
- Flush `m_recent_rejects` and `m_recent_rejects_reconsiderable` on `ActiveTipChanged` just once instead of checking the tip every time `AlreadyHaveTx` is called. This should speed up calls to that function (no longer comparing a block hash each time) and removes the need to lock `cs_main` every time it is called.
Motivation:
- These data structures need synchronization. While we are holding `m_tx_download_mutex`, these should hold:
- a tx hash in `m_txrequest` is not also in `m_orphanage`
- a tx hash in `m_txrequest` is not also in `m_recent_rejects` or `m_recent_confirmed_transactions`
- In the future, orphan resolution tracking should also be synchronized. If a tx has an entry in the orphan resolution tracker, it is also in `m_orphanage`, and not in `m_txrequest`, etc.
- Currently, `cs_main` is used to e.g. sync accesses to `m_txrequest`. We should not broaden the scope of things it locks.
- Currently, we need to know the current chainstate every time we call `AlreadyHaveTx` so we can decide whether we should update it. Every call compares the current tip hash with `hashRecentRejectsChainTip`. It is more efficient to have a validation interface callback that updates the rejection filters whenever the chain tip changes.
ACKs for top commit:
instagibbs:
reACK c85accecaf
dergoegge:
Code review ACK c85accecaf
theStack:
Light code-review ACK c85accecaf
hebasto:
ACK c85accecaf, I have reviewed the code and it looks OK.
Tree-SHA512: c3bd524b5de1cafc9a10770dadb484cc479d6d4c687d80dd0f176d339fd95f73b85cb44cb3b6b464d38a52e20feda00aa2a1da5a73339e31831687e4bd0aa0c5
