223081ece6 scripted-diff: rename block and undo functions for consistency (Lőrinc)
baaa3b2846 refactor,blocks: remove costly asserts and modernize affected logs (Lőrinc)
fa39f27a0f refactor,blocks: deduplicate block's serialized size calculations (Lőrinc)
dfb2f9d004 refactor,blocks: inline `WriteBlockToDisk` (Lőrinc)
42bc491465 refactor,blocks: inline `UndoWriteToDisk` (Lőrinc)
86b85bb11f bench: add SaveBlockBench (Lőrinc)
34f9a0157a refactor,bench: rename bench/readblock.cpp to bench/readwriteblock.cpp (Lőrinc)
Pull request description:
`UndoWriteToDisk` and `WriteBlockToDisk` were delegating a subset of their functionality to single-use methods that didn't optimally capture a meaningful chunk of the algorithm, resulting in calculating things twice (serialized size, header size).
This change inlines the awkward methods (asserting that all previous behavior was retained), and in separate commits makes the usages less confusing.
Besides making the methods slightly more intuitive, the refactorings reduce duplicate calculations as well.
The speed difference is insignificant for now (~0.5% for the new `SaveBlockToDiskBench`), but are a cleanup for follow-ups such as https://github.com/bitcoin/bitcoin/pull/31539
ACKs for top commit:
ryanofsky:
Code review ACK 223081ece6. Since last review, "Save" was renamed to "Write", uint32_t references were dropped, some log statements and comments were improved as suggested, and a lot of tweaks made to commits and commit messages which should make this easier to review.
hodlinator:
ACK 223081ece6
TheCharlatan:
ACK 223081ece6
andrewtoth:
ACK 223081ece6
Tree-SHA512: 951bc8ad3504c510988afd95c561e3e259c6212bd14f6536fe56e8eb5bf5c35c32a368bbdb1d5aea1acc473d7e5bd9cdcde02008a148b05af1f955e413062d5c
86d7135e36 [p2p] only attempt 1p1c when both txns provided by the same peer (glozow)
f7658d9b14 [cleanup] remove p2p_inv from AddTxAnnouncement (glozow)
063c1324c1 [functional test] getorphantxs reflects multiple announcers (glozow)
0da693f7e1 [functional test] orphan handling with multiple announcers (glozow)
b6ea4a9afe [p2p] try multiple peers for orphan resolution (glozow)
1d2e1d709c [refactor] move creation of unique_parents to helper function (glozow)
c6893b0f0b [txdownload] remove unique_parents that we already have (glozow)
163aaf285a [fuzz] orphanage multiple announcer functions (glozow)
22b023b09d [unit test] multiple orphan announcers (glozow)
96c1a822a2 [unit test] TxOrphanage EraseForBlock (glozow)
04448ce32a [txorphanage] add GetTx so that orphan vin can be read (glozow)
e810842acd [txorphanage] support multiple announcers (glozow)
62a9ff1870 [refactor] change type of unique_parents to Txid (glozow)
6951ddcefd [txrequest] GetCandidatePeers (glozow)
Pull request description:
Part of #27463.
(Transaction) **orphan resolution** is a process that kicks off when we are missing UTXOs to validate an unconfirmed transaction. We currently request missing parents by txid; BIP 331 also defines a way to [explicitly request ancestors](https://github.com/bitcoin/bips/blob/master/bip-0331.mediawiki#handle-orphans-better).
Currently, when we find that a transaction is an orphan, we only try to resolve it with the peer who provided the `tx`. If this doesn't work out (e.g. they send a `notfound` or don't respond), we do not try again. We actually can't, because we've already forgotten who else could resolve this orphan (i.e. all the other peers who announced the transaction).
What is wrong with this? It makes transaction download less reliable, particularly for 1p1c packages which must go through orphan resolution in order to be downloaded.
Can we fix this with BIP 331 / is this "duct tape" before the real solution?
BIP 331 (receiver-initiated ancestor package relay) is also based on the idea that there is an orphan that needs resolution, but it's just a new way of communicating information. It's not inherently more honest; you can request ancestor package information and get a `notfound`. So ancestor package relay still requires some kind of procedure for retrying when an orphan resolution attempt fails. See the #27742 implementation which builds on this orphan resolution tracker to keep track of what packages to download (it just isn't rebased on this exact branch). The difference when using BIP 331 is that we request `ancpkginfo` and then `pkgtxns` instead of the parent txids.
Zooming out, we'd like orphan handling to be:
- Bandwidth-efficient: don't have too many requests out at once. As already implemented today, transaction requests for orphan parents and regular download both go through the `TxRequestTracker` so that we don't have duplicate requests out.
- Not vulnerable to censorship: don't give up too easily, use all candidate peers. See e.g. https://bitcoincore.org/en/2024/07/03/disclose_already_asked_for/
- Load-balance between peers: don't overload peers; use all peers available. This is also useful for when we introduce per-peer orphan protection, since each peer will have limited slots.
The approach taken in this PR is to think of each peer who announces an orphan as a potential "orphan resolution candidate." These candidates include:
- the peer who sent us the orphan tx
- any peers who announced the orphan prior to us downloading it
- any peers who subsequently announce the orphan after we have started trying to resolve it
For each orphan resolution candidate, we treat them as having "announced" all of the missing parents to us at the time of receipt of this orphan transaction (or at the time they announced the tx if they do so after we've already started tracking it as an orphan). We add the missing parents as entries to `m_txrequest`, incorporating the logic of typical txrequest processing, which means we prefer outbounds, try not to have duplicate requests in flight, don't overload peers, etc.
ACKs for top commit:
marcofleon:
Code review ACK 86d7135e36
instagibbs:
reACK 86d7135e36
dergoegge:
Code review ACK 86d7135e36
mzumsande:
ACK 86d7135e36
Tree-SHA512: 618d523b86e60c3ea039e88326d50db4e55e8e18309c6a20e8f2b10ed9e076f1de0315c335fd3b8abdabcc8b53cbceb66fb59147d05470ea25b83a2b4bd9c877
This is not a pure refactor:
1. It slightly changes the log messages, as reflected in the test changes
2. It adds the IP address to all disconnect logging (when fLogIPs is set)
e80e4c6ff9 validation: Remove RECENT_CONSENSUS_CHANGE validation result (TheCharlatan)
Pull request description:
The *_RECENT_CONSENSUS_CHANGE variants in the validation result enumerations were always unused. They seem to have been kept around speculatively for a soft fork after segwit, however they were never used for taproot either. This points at them not having a clear purpose. Based on the original pull requests' comments their usage was never entirely clear:
https://github.com/bitcoin/bitcoin/pull/11639#issuecomment-370234133https://github.com/bitcoin/bitcoin/pull/15141#discussion_r271039747
Since they are part of the validation interface and need to be exposed by the kernel library keeping them around may also be confusing to future users of the library.
ACKs for top commit:
sipa:
ACK e80e4c6ff9
naumenkogs:
ACK e80e4c6ff9
dergoegge:
ACK e80e4c6ff9
ajtowns:
ACK e80e4c6ff9
Tree-SHA512: 0af17c4435bb1b5a4f43600da30545cbbe95a7d642419cabdefabfb82b9335d92262c1c48be7ca2f2a024078ae9447161228b6f951d2f508a51159a31947fb54
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
The *_RECENT_CONSENSUS_CHANGE variants in the validation result
enumerations were always unused. They seem to have been kept around
speculatively for a soft fork after segwit, however they were never used
for taproot either. This points at them not having a clear purpose.
Based on the original pull requests' comments their usage was never
entirely clear:
https://github.com/bitcoin/bitcoin/pull/11639#issuecomment-370234133https://github.com/bitcoin/bitcoin/pull/15141#discussion_r271039747
Since they are part of the validation interface and need to exposed by
the kernel library keeping them around may also be confusing to future
users of the library.
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.
