5e7cc4144b test: add unit test for CConnman::AddedNodesContain() (Jon Atack)
cc62716920 p2p: do not make automatic outbound connections to addnode peers (Jon Atack)
Pull request description:
to allocate our limited outbound slots correctly, and to ensure addnode
connections benefit from their intended protections.
Our addnode logic usually connects the addnode peers before the automatic
outbound logic does, but not always, as a connection race can occur. If an
addnode peer disconnects us and if it was the only one from its network, there
can be a race between reconnecting to it with the addnode thread, and it being
picked as automatic network-specific outbound peer. Or our internet connection
or router or the addnode peer could be temporarily offline, and then return
online during the automatic outbound thread. Or we could add a new manual peer
using the addnode RPC at that time.
The race can be more apparent when our node doesn't know many peers, or with
networks like cjdns that currently have few bitcoin peers.
When an addnode peer is connected as an automatic outbound peer and is the only
connection we have to a network, it can be protected by our new outbound
eviction logic and persist in the "wrong role".
Finally, there does not seem to be a reason to make block-relay or short-lived
feeler connections to addnode peers, as the addnode logic will ensure we connect
to them if they are up, within the addnode connection limit.
Fix these issues by checking if the address is an addnode peer in our automatic
outbound connection logic.
ACKs for top commit:
mzumsande:
Tested ACK 5e7cc4144b
brunoerg:
utACK 5e7cc4144b
vasild:
ACK 5e7cc4144b
guggero:
utACK 5e7cc4144b
Tree-SHA512: 2438c3ec92e98aebca2a0da960534e4655a9c6e1192a24a085fc01326d95cdb1b67d8c44e4ee706bc1d8af8564126d446a21b5579dcbec61bdea5fce2f0115ee
007d6f0e85 test: fix `AddNode` unit test failure on OpenBSD (Sebastian Falbesoner)
Pull request description:
On OpenBSD 7.4, the following check of the unit test `test_addnode_getaddednodeinfo_and_connection_detection` currently fails:
```
BOOST_CHECK(!connman->AddNode({/*m_added_node=*/"127.1", /*m_use_v2transport=*/true}));
```
The reason for that is that this OS seemingly doesn't support the IPv4 shorthand notation with omitted zero-bytes:
```
$ ping 127.1
ping: no address associated with name
```
As a simple fix, this PR skips the check for this with a pre-processor #if. On NetBSD and FreeBSD, `127.1` is resolved correctly to localhost and hence the test passes (thanks to vasild for verifying on the latter!).
ACKs for top commit:
vasild:
ACK 007d6f0e85
Tree-SHA512: 8ab8393c490e1ecc140e8ff74f6fa4d26d0dd77e6a77a241cd198314b8c5afee7422f95351ca05f4c1742433dab77016a8ccb8d28062f8edd4b703a918a2bbda
If alignment of the PoolAllocator would be insufficient, then the test would fail. This also catches the issue with ARM 32bit,
where int64_t is aligned to 8 bytes but void* is aligned to 4 bytes. The test adds a check to ensure the pool has allocated
a minimum number of chunks
This changes the PoolAllocator to default the alignment to the given type. This makes the code simpler, and most importantly
fixes a bug on ARM 32bit that caused OOM: The class CTxOut has a member CAmount which is an int64_t and on ARM 32bit int64_t
are 8 byte aligned which is larger than the pointer alignment of 4 bytes. So for CCoinsMap to be able to use the pool, we
need to use the alignment of the member instead of just alignof(void*).
83986f464c Include version.h in fewer places (Anthony Towns)
c7b61fd61b Convert some CDataStream to DataStream (Anthony Towns)
1410d300df serialize: Drop useless version param from GetSerializeSize() (Anthony Towns)
bf574a7501 serialize: drop GetSerializeSizeMany (Anthony Towns)
efa9eb6d7c serialize: Drop nVersion from [C]SizeComputer (Anthony Towns)
Pull request description:
Drops the version field from `GetSerializeSize()`, simplifying the code in various places. Also drop `GetSerializeSizeMany()` (as just removing the version parameter could result in silent bugs) and remove unnecessary instances of `#include <version.h>`.
ACKs for top commit:
maflcko:
ACK 83986f464c📒
theuni:
ACK 83986f464c.
Tree-SHA512: 36617b6dfbb1b4b0afbf673e905525fc6d623d3f568d3f86e3b9d4f69820db97d099e83a88007bfff881f731ddca6755ebf1549e8d8a7762437dfadbf434c62e
faa25718b3 fuzz: AutoFile with XOR (MarcoFalke)
fab5cb9066 fuzz: Reduce LIMITED_WHILE limit for file fuzzing (MarcoFalke)
fa5388fad3 fuzz: Remove FuzzedAutoFileProvider (MarcoFalke)
Pull request description:
This should help to get fuzz coverage for https://maflcko.github.io/b-c-cov/fuzz.coverage/src/streams.cpp.gcov.html
Also, remove unused code and fix a timeout bug.
ACKs for top commit:
dergoegge:
ACK faa25718b3
Tree-SHA512: 56f1e6fd5cb2b66ffd9a7d9c09c9b8e396be3e7485feb03b35b6bd3c48e624fdaed50b472e4ffec21f09efb5e949d7ee32a13851849c9140b6b4cf25917dd7ac
to allocate our limited outbound slots correctly, and to ensure addnode
connections benefit from their intended protections.
Our addnode logic usually connects the addnode peers before the automatic
outbound logic does, but not always, as a connection race can occur. If an
addnode peer disconnects us and if it was the only one from its network, there
can be a race between reconnecting to it with the addnode thread, and it being
picked as automatic network-specific outbound peer. Or our internet connection
or router, or the addnode peer, could be temporarily offline, and then return
online during the automatic outbound thread. Or we could add a new manual peer
using the addnode RPC at that time.
The race can be more apparent when our node doesn't know many peers, or with
networks like cjdns that currently have few bitcoin peers.
When an addnode peer is connected as an automatic outbound peer and is the only
connection we have to a network, it can be protected by our new outbound
eviction logic and persist in the "wrong role".
Examples on mainnet using logging added in the same pull request:
2023-08-12T14:51:05.681743Z [opencon] [net.cpp:1949] [ThreadOpenConnections]
[net:debug] Not making automatic network-specific outbound-full-relay connection
to i2p peer selected for manual (addnode) connection: [geh...odq.b32.i2p]:0
2023-08-13T03:59:28.050853Z [opencon] [net.cpp:1949] [ThreadOpenConnections]
[net:debug] Not making automatic block-relay-only connection to onion peer
selected for manual (addnode) connection: kpg...aid.onion:8333
2023-08-13T16:21:26.979052Z [opencon] [net.cpp:1949] [ThreadOpenConnections]
[net:debug] Not making automatic network-specific outbound-full-relay connection
to cjdns peer selected for manual (addnode) connection: [fcc...8ce]:8333
2023-08-14T20:43:53.401271Z [opencon] [net.cpp:1949] [ThreadOpenConnections]
[net:debug] Not making automatic network-specific outbound-full-relay connection
to cjdns peer selected for manual (addnode) connection: [fc7...59e]:8333
2023-08-15T00:10:01.894147Z [opencon] [net.cpp:1949] [ThreadOpenConnections]
[net:debug] Not making automatic feeler connection to i2p peer selected for
manual (addnode) connection: geh...odq.b32.i2p:8333
Finally, there does not seem to be a reason to make block-relay or short-lived
feeler connections to addnode peers, as the addnode logic will ensure we connect
to them if they are up, within the addnode connection limit.
Fix these issues by checking if the address is an addnode peer in our automatic
outbound connection logic.
43de4d3630 doc: fix typos (Sjors Provoost)
Pull request description:
This PR fixes typos found by lint-spelling.py using codespell 2.2.6.
Our CI linter job uses codespell 2.2.5 and found fewer typos that I did locally. In any case it's happy now.
ACKs for top commit:
pablomartin4btc:
re ACK 43de4d3630
Tree-SHA512: c032fe86cb49c924a468385653b31f309a9db68c478d70335bba3e65a1ff3826abe80284fe00a090ab5a509e1edbf17e476f6922fb15d055e50f1103dad2ccb0
6a917918b7 fuzz: allow fake and duplicate inputs in tx_package_eval target (Greg Sanders)
a0626ccdad fuzz: allow reaching MempoolAcceptResult::ResultType::DIFFERENT_WITNESS in tx_package_eval target (Greg Sanders)
Pull request description:
Exercises `DIFFERENT_WITNESS` by using "blank" WSH() and allowing witness to determine wtxid, and attempts to make invalid/duplicate inputs.
ACKs for top commit:
dergoegge:
Coverage looks good to me ACK 6a917918b7
Tree-SHA512: db894f5f5b81c6b454874baf11f296462832285f41ccb09f23c0db92b9abc98f8ecacd72fc8f60dc92cb7947f543a2e55bed2fd210b0e8ca7c7d5389d90b14af
Protocol version is no longer needed to work out the serialized size
of objects so drop that information from CSizeComputer and rename the
class to SizeComputer.
a0c254c13a Drop CHashWriter (Anthony Towns)
c94f7e5b1c Drop OverrideStream (Anthony Towns)
6e9e4e6130 Use ParamsWrapper for witness serialization (Anthony Towns)
Pull request description:
Choose whether witness is included in transaction serialization via serialization parameter rather than the stream version. See #25284 and #19477 for previous context.
ACKs for top commit:
maflcko:
re-ACK a0c254c13a🐜
theuni:
ACK a0c254c13a
Tree-SHA512: 8fd5cadfd84c5128e36c34a51fb94fdccd956280e7f65b7d73c512d6a9cdb53cdd3649de99ffab5322bd34be26cb95ab4eb05932b3b9de9c11d85743f50dcb13
d799ea26ed doc: rewrite explanation for -par= (fanquake)
Pull request description:
The negative bound for script threads comes from the machine which generates the man pages, so may only be correct for that machine. Any other placeholder value will also be wrong for some machines. Fix this be removing the value. This also fixes help2man incorrectly bolding the value, as if it were a paramater.
Closes #28850.
ACKs for top commit:
maflcko:
lgtm ACK d799ea26ed
theStack:
ACK d799ea26ed
Tree-SHA512: 2eec0086faf4cc64bbf46b22949662f84d8546d2322c3d507fc44a4e1f64d228a2901af4fa4535c0771e3e14600be8308fc5dbd407b66ae6ae4f8878d8372c0a
1e5b86171e test: Add test for array serialization (TheCharlatan)
d49d198840 refactor: Initialize magic bytes in constructor initializer (TheCharlatan)
Pull request description:
This is a followup-PR for #28423
* Initialize magic bytes in constructor
* Add a small unit test for serializing arrays.
ACKs for top commit:
sipa:
utACK 1e5b86171e
maflcko:
lgtm ACK 1e5b86171e
Tree-SHA512: 0f58d2332dc501ca9fd419f40ed4f977c83dce0169e9a0eee1ffc9f8daa2d2ef7e7df18205ba076f55d90ae6c4a20d2b51ab303150d38470a962bcc58a66f6e7
fa6b053b5c mempool: persist with XOR (MarcoFalke)
Pull request description:
Currently the `mempool.dat` file stores data received from remote peers as-is. This may be problematic when a program other than Bitcoin Core tries to interpret them by accident. For example, an anti-virus program or other program may scan the file and move it into quarantine, or delete it, or corrupt it.
While the local wallet is expected to re-submit any pending transactions, unrelated transactions may be missing from the mempool after a restart. This may cause fee estimates to be off, or may cause block relay to be slower.
Fix this, similar to https://github.com/bitcoin/bitcoin/pull/6650, by rolling a random XOR pattern over the dat file when writing or reading it.
Obviously this can only protect against programs that accidentally and unintentionally are trying to mess with the dat file. Any program that intentionally wants to mess with the dat file can still trivially do so.
ACKs for top commit:
achow101:
re-ACK fa6b053b5c
glozow:
reACK fa6b053b5c
ismaelsadeeq:
ACK fa6b053b5c
Tree-SHA512: ded2ce3d81bc944b828263534e3178a1e45a914fe8e024f4a14c6561a73e301820944ecc75dd704b3d4221a7a3a5c0597ccab79546250c1197609ee981fe324e
bbbbdb0cd5 ci: Add filesystem lint check (MarcoFalke)
fada2f9110 refactor: Replace <filesystem> with <util/fs.h> (MarcoFalke)
Pull request description:
Using `std::filesystem` is problematic:
* There is a `fs` namespace wrapper for it. So having two ways to achieve the same is confusing.
* Not using the `fs` wrapper is dangerous and buggy, because it disables known bugs by deleting problematic functions.
Fix all issues by removing use of it and adding a linter to avoid using it again in the future.
ACKs for top commit:
TheCharlatan:
ACK bbbbdb0cd5
fanquake:
ACK bbbbdb0cd5🦀
Tree-SHA512: 0e2d49742b08eb2635e6fce41485277cb9c40fe20b81017c391d3472a43787db1278a236825714ca1e41c9d2f59913865cfb0c649e3c8ab1fb598c849f80c660
3b70f7b615 doc: fix broken doc/design/multiprocess.md links after #24352 (Ryan Ofsky)
6d43aad742 span: Make Span template deduction guides work in SFINAE context (Ryan Ofsky)
8062c3bdb9 util: Add ArgsManager SetConfigFilePath method (Ryan Ofsky)
441d00c60f interfaces: Rename CalculateBumpFees methods to be compatible with capn'proto (Ryan Ofsky)
156f49d682 interfaces: Change getUnspentOutput return type to avoid multiprocess segfault (Ryan Ofsky)
4978754c00 interfaces: Add schedulerMockForward method so mockscheduler RPC can work across processes (Ryan Ofsky)
924327eaf3 interfaces: Fix const virtual method that breaks multiprocess support (Ryan Ofsky)
82a379eca8 streams: Add SpanReader ignore method (Russell Yanofsky)
Pull request description:
This is a collection of small changes to interfaces and code which were needed as part of multiprocess PR #10102, but have been moved here to make that PR smaller.
All of these changes are refactoring changes which do not affect behavior of current code
---
This PR is part of the [process separation project](https://github.com/bitcoin/bitcoin/issues/28722).
ACKs for top commit:
achow101:
ACK 3b70f7b615
naumenkogs:
ACK 3b70f7b615
maflcko:
re-ACK 3b70f7b615🎆
Tree-SHA512: 2368772b887056ad8a9f84c299cfde76ba45943770e3b5353130580900afa9611302195b899ced7b6e303b11f053ff204cae7c28ff4e12c55562fcc81119ba4c
The negative bound for script threads comes from the machine which
generates the man pages, so may only be correct for that machine. Any
other placeholder value will also be wrong for some machines. Fix this
be removing the value. This also fixes help2man incorrectly bolding the
value, as if it were a paramater.
Closes #28850.
fca0a8938e ci: remove "--exclude banman" for fuzzing in mac (brunoerg)
f9b286353f fuzz: call lookup functions before calling `Ban` (brunoerg)
Pull request description:
Fixes #27924
To not have any discrepancy, it's required to call lookup functions before calling `Ban`. If we don't do it, the assertion `assert(banmap == banmap_read);` may fail because `BanMapFromJson` will call `LookupSubNet` and cause the discrepancy between the banned and the loaded one. It happens especially in MacOS (#27924).
Also, calling lookup functions before banning is what RPC `setban` does.
ACKs for top commit:
maflcko:
lgtm ACK fca0a8938e
dergoegge:
ACK fca0a8938e
Tree-SHA512: a3d635088a556df4507e65542157f10b41d4f87dce42927b58c3b812f262f4544b6b57f3384eef1097ffdd7c32b8dd1556aae201254960cbfbf48d45551200f7
4dd94ca18f [refactor] remove access to mapTx in validation_block_tests (TheCharlatan)
d0cd2e804e [refactor] rewrite BlockAssembler inBlock and failedTx as sets of txids (glozow)
55b0939cab scripted-diff: rename vTxHashes to txns_randomized (TheCharlatan)
a03aef9cec [refactor] rewrite vTxHashes as a vector of CTransactionRef (glozow)
938643c3b2 [refactor] remove access to mapTx in validation.cpp (glozow)
333367a940 [txmempool] make CTxMemPoolEntry::lockPoints mutable (glozow)
1bf4855016 [refactor] use CheckPackageLimits for checkChainLimits (glozow)
dbc5bdbf59 [refactor] remove access to mapTx.find in mempool_tests.cpp (glozow)
f80909e7a3 [refactor] remove access to mapTx in blockencodings_tests.cpp (glozow)
8892d6b744 [refactor] remove access to mapTx from rpc/mempool.cpp (glozow)
fad61aa561 [refactor] get wtxid from entry instead of vTxHashes (glozow)
9cd8cafb77 [refactor] use exists() instead of mapTx.find() (glozow)
14804699e5 [refactor] remove access to mapTx from policy/rbf.cpp (glozow)
1c6a73abbd [refactor] Add helper for retrieving mempool entry (TheCharlatan)
453b4813eb [refactor] Add helper for iterating through mempool entries (stickies-v)
Pull request description:
Motivation
* It seems preferable to use stdlib data structures instead of boost if they can achieve close to the same thing.
* Code external to mempool should ideally use its public helper methods instead of accessing `mapTx` or its iterators directly.
* Reduce the number of complex boost multi index type interactions
* Also see #28335 for further context/motivation. This PR together with #28385 simplifies that one.
Overview of things done in this PR:
* Make `vTxHashes` a vector of transaction references instead of a pair of transaction hash and iterator. The trade off here is that the data is retrieved on the fly with `GetEntry` instead of being cached in `vTxHashes`.
* Introduce `GetEntry` helper method to replace the more involved `GetIter` where applicable
* Replace `mapTx` access with `CTxMemPool` helper methods
* Simplify `checkChainLimits` call in `node/interfaces.cpp`
* Make `CTxMemPoolEntry`s `lockPoints`mutable such that they can be changed with a const iterator directly instead of going through `mapTx`
* Make `BlockAssembler`'s `inBlock` and `failedTx` sets of transaction hashes.
ACKs for top commit:
glozow:
reACK 4dd94ca
maflcko:
re-ACK 4dd94ca18f👝
stickies-v:
re-ACK 4dd94ca18f
Tree-SHA512: c4d043f2186e4fde337591883fac66cade3058173987b49502bd65cecf69207a3df1077f6626809652ab63230013167b7f39a2b39f1c5166959e5495df57065f
Allows calling UpdateLockPoints() with a (const) txiter. Note that this
was already possible for caller using mapTx.modify(txiter). The point
here is to not be accessing mapTx when doing so.
The behavior is the same as CalculateMemPoolAncestors. The only
difference is the string returned, and the string is discarded anyway
since checkChainLimits only cares about pass/fail.
