691eaf8873 Pass MSG_MORE flag when sending non-final network messages (Matt Whitlock)
Pull request description:
**N.B.:** This is my second attempt at introducing this optimization. #12519 (2018) was closed in deference to switching to doing gathering socket writes using `sendmsg(2)`, which I agree would have superior performance due to fewer syscalls, but that work was apparently abandoned in late 2018. Ever since, Bitcoin Core has continued writing tons of runt packets to the wire. Can we proceed with my halfway solution for now?
----
Since Nagle's algorithm is disabled, each and every call to `send(2)` can potentially generate a separate TCP segment on the wire. This is especially inefficient when sending the tiny header preceding each message payload.
Linux implements a `MSG_MORE` flag that tells the kernel not to push the passed data immediately to the connected peer but rather to collect it in the socket's internal transmit buffer where it can be combined with data from successive calls to `send(2)`. Where available, specify this flag when calling `send(2)` in `CConnman::SocketSendData(CNode &)` if the data buffer being sent is not the last one in `node.vSendMsg`.
ACKs for top commit:
sipa:
ACK 691eaf8873
vasild:
ACK 691eaf8873
Tree-SHA512: 9a7f46bc12edbf78d488f05d1c46760110a24c95af74b627d2604fcd198fa3f511c5956bac36d0034e88c632d432f7d394147e667a11b027af0a30f70a546d70
511aa4f1c7 Add unit test for ChaCha20's new caching (Pieter Wuille)
fb243d25f7 Improve test vectors for ChaCha20 (Pieter Wuille)
93aee8bbda Inline ChaCha20 32-byte specific constants (Pieter Wuille)
62ec713961 Only support 32-byte keys in ChaCha20{,Aligned} (Pieter Wuille)
f21994a02e Use ChaCha20Aligned in MuHash3072 code (Pieter Wuille)
5d16f75763 Use ChaCha20 caching in FastRandomContext (Pieter Wuille)
38eaece67b Add fuzz test for testing that ChaCha20 works as a stream (Pieter Wuille)
5f05b27841 Add xoroshiro128++ PRNG (Martin Leitner-Ankerl)
12ff72476a Make unrestricted ChaCha20 cipher not waste keystream bytes (Pieter Wuille)
6babf40213 Rename ChaCha20::Seek -> Seek64 to clarify multiple of 64 (Pieter Wuille)
e37bcaa0a6 Split ChaCha20 into aligned/unaligned variants (Pieter Wuille)
Pull request description:
This is an alternative to #25354 (by my benchmarking, somewhat faster), subsumes #25712, and adds additional test vectors.
It separates the multiple-of-64-bytes-only "core" logic (which becomes simpler) from a layer around which performs caching/slicing to support arbitrary byte amounts. Both have their uses (in particular, the MuHash3072 code can benefit from multiple-of-64-bytes assumptions), plus the separation results in more readable code. Also, since FastRandomContext effectively had its own (more naive) caching on top of ChaCha20, that can be dropped in favor of ChaCha20's new built-in caching.
I thought about rebasing #25712 on top of this, but the changes before are fairly extensive, so redid it instead.
ACKs for top commit:
ajtowns:
ut reACK 511aa4f1c7
dhruv:
tACK crACK 511aa4f1c7
Tree-SHA512: 3aa80971322a93e780c75a8d35bd39da3a9ea570fbae4491eaf0c45242f5f670a24a592c50ad870d5fd09b9f88ec06e274e8aa3cefd9561d623c63f7198cf2c7
9d3127b11e Add settings.json prune-prev, proxy-prev, onion-prev settings (Ryan Ofsky)
Pull request description:
With #602, if proxy and pruning settings are disabled in the GUI and the GUI is restarted, proxy and prune values are not stored anywhere. So if these settings are enabled in the future, default values will be shown, not previous values.
This PR stores previous values so they will preserved across restarts. I'm not sure I like this behavior because showing default values seems simpler and safer to me. Previous values may just have been set temporarily and may have never actually worked, and it adds some code complexity to store them.
This PR is one way of resolving #596. Other solutions are possible and could be implemented as alternatives.
ACKs for top commit:
hebasto:
ACK 9d3127b11e, tested on Ubuntu 22.04.
vasild:
ACK 9d3127b11e
jarolrod:
tACK 9d3127b11e
Tree-SHA512: 1778d1819443490c880cfd5c1711d9c5ac75ea3ee8440e2f0ced81d293247163a78ae8aba6027215110aec6533bd7dc6472aeead6796bfbd51bf2354e28f24a9
77192c9598 cli: include local ("unreachable") peers in -netinfo table (Matthew Zipkin)
Pull request description:
Closes https://github.com/bitcoin/bitcoin/issues/26579
The `-netinfo` dashboard did not list peers that were connected via "unroutable" networks. This included local peers including local-network peers. Personally, I run one bitcoind instance on my network that is used by other services like Wasabi Wallet and LND running on other machines.
This PR adds an "npr" (not publicly routable) column to the table of networks (ipv4, ipv6, onion, etc) so that every connection to the node is listed, and the totals are accurate as they relate to max inbound and max outbound limits.
Example connecting in regtest mode to one local and one remote peer:
```
Bitcoin Core client v24.99.0-151ce099ea8f-dirty regtest - server 70016/Satoshi:24.99.0/
<-> type net mping ping send recv txn blk hb addrp addrl age id address version
in npr 0 0 90 90 1 1 127.0.0.1:59180 70016/Satoshi:24.99.0/
out manual ipv4 63 63 84 84 3 3 0 143.244.175.41 70016/Satoshi:24.0.1/
ms ms sec sec min min min
ipv4 ipv6 npr total block manual
in 0 0 1 1
out 1 0 0 1 0 1
total 1 0 1 2
Local addresses: n/a
```
ACKs for top commit:
jonatack:
Re-tested ACK 77192c9598
Tree-SHA512: 78aa68bcff0dbaadb5f0604bf023fe8fd921313bd8276d12581f7655c089466a48765f9e123cb31d7f1d294d5ca45fdefdf8aa220466ff738f32414f41099c06
`m_relock_mutex` is introduced so that the passphrase is not
deleted from memory when the timeout provided in
`walletpassphrase` is up, but the wallet is still rescanning.
Wallet passphrases are needed to top up the keypool during a
rescan. The following RPCs need the passphrase when rescanning:
- `importdescriptors`
- `rescanblockchain`
The following RPCs use the information about whether or not the
passphrase is being used to ensure that full rescans are able to
take place:
- `walletlock`
- `encryptwallet`
- `walletpassphrasechange`
3a11adc700 Zero out wallet master key upon lock (John Moffett)
Pull request description:
When an encrypted wallet is locked (for instance via the RPC `walletlock`), the documentation indicates that the key is removed from memory:
b92d609fb2/src/wallet/rpc/encrypt.cpp (L157-L158)
However, the vector (a `std::vector<unsigned char, secure_allocator<unsigned char>>`) is merely _cleared_. As it is a member variable, it also stays in scope as long as the wallet is loaded, preventing the secure allocator from deallocating. This allows the key to persist indefinitely in memory. I confirmed this behavior on my macOS machine by using an open-source third party memory inspector ("Bit Slicer"). I was able to find my wallet's master key in Bit Slicer after unlocking and re-locking my encrypted wallet. I then confirmed the key data was at the address in LLDB.
This PR manually fills the bytes with zeroes before calling `clear()` by using our `memory_cleanse` function, which is designed to prevent the compiler from optimizing it away. I confirmed that it does remove the data from memory on my machine upon locking.
Note: An alternative approach could be to call `vMasterKey.shrink_to_fit()` after the `clear()`, which would trigger the secure allocator's deallocation. However, `shrink_to_fit()` is not _guaranteed_ to actually change the vector's capacity, so I think it's unwise to rely on it.
## Edit: A little more clarity on why this is an improvement.
Since `mlock`ed memory is guaranteed not to be swapped to disk and our threat model doesn't consider a super-user monitoring the memory in realtime, why is this an improvement? Most importantly, consider hibernation. Even `mlock`ed memory may get written to disk. From the `mlock` [manpage](https://man7.org/linux/man-pages/man2/mlock.2.html):
> (But be aware that the suspend mode on laptops and some desktop computers will save a copy of the system's RAM to disk, regardless of memory locks.)
As far as I can tell, this is true of [Windows](https://web.archive.org/web/20190127110059/https://blogs.msdn.microsoft.com/oldnewthing/20140207-00/?p=1833#:~:text=%5BThere%20does%20not%20appear%20to%20be%20any%20guarantee%20that%20the%20memory%20won%27t%20be%20written%20to%20disk%20while%20locked.%20As%20you%20noted%2C%20the%20machine%20may%20be%20hibernated%2C%20or%20it%20may%20be%20running%20in%20a%20VM%20that%20gets%20snapshotted.%20%2DRaymond%5D) and macOS as well.
Therefore, a user with a strong OS password and a strong wallet passphrase could still have their keys stolen if a thief takes their (hibernated) machine and reads the permanent storage.
ACKs for top commit:
S3RK:
Code review ACK 3a11adc700
achow101:
ACK 3a11adc700
Tree-SHA512: c4e3dab452ad051da74855a13aa711892c9b34c43cc43a45a3b1688ab044e75d715b42843c229219761913b4861abccbcc8d5cb6ac54957d74f6e357f04e8730
561848aaf2 Exercise non-DIRTY spent coins in caches in fuzz test (Pieter Wuille)
59e6828bb5 Add deterministic mode to CCoinsViewCache (Pieter Wuille)
b0ff310840 Add CCoinsViewCache::SanityCheck() and use it in fuzz test (Pieter Wuille)
3c9cea1340 Add simulation-based CCoinsViewCache fuzzer (Pieter Wuille)
Pull request description:
The fuzzer goes through a sequence of operations that get applied to both a real stack of `CCoinsViewCache` objects, and to simulation data, comparing the two at the end.
ACKs for top commit:
jamesob:
re-ACK 561848aaf2
dergoegge:
Code review ACK 561848aaf2
Tree-SHA512: 68634f251fdb39436b128ecba093f651bff12ac11508dc9885253e57fd21efd44edf3b22b0f821c228175ec507df7d46c7f9f5404fc1eb8187fdbd136a5d5ee2
This is a "dumb" way of randomly generating a Miniscript node from
fuzzer input. It defines a strict binary encoding and will always generate
a node defined from the encoding without "helping" to create valid nodes.
It will cut through as soon as it encounters an invalid fragment so
hopefully the fuzzer can tend to learn the encoding and generate valid
nodes with a higher probability.
On a valid generated node a number of invariants are checked, especially
around the satisfactions and testing them against the Script
interpreter.
The node generation and testing is modular in order to later introduce
other ways to generate nodes from fuzzer inputs with minimal code.
Co-Authored-By: Pieter Wuille <pieter@wuille.net>
This is a workaround for Miniscript descriptors containing hash
challenges. For those we can't mock the signature creator without making
OP_EQUAL mockable in the interpreter, so CalculateMaximumInputSize will
always return -1 and outputs for these descriptors would appear
unsolvable while they actually are.
Try to solve a script using the Miniscript satisfier if the legacy
solver fails under P2WSH context. Only solve public key and public key
hash challenges for now.
We don't entirely replace the raw solver and especially rule out trying to
solve CHECKMULTISIG-based multisigs with the Miniscript satisfier since
some features, such as the transaction input combiner, rely on the
specific behaviour of the former.
Cherry-picked and squashed from
https://github.com/sipa/bitcoin/commits/202302_miniscript_improve.
- Explain thresh() and multi() satisfaction algorithms
- Comment on and_v dissatisfaction
- Mark overcomplete thresh() dissats as malleable and explain
- Add comment on unnecessity of Malleable() in and_b dissat
When an encrypted wallet is locked (for instance via the
RPC `walletlock`), the docs indicate that the key is
removed from memory. However, the vector (with a secure
allocator) is merely cleared. This allows the key to persist
indefinitely in memory. Instead, manually fill the bytes with
zeroes before clearing.
2d955ff006 net: add `Ensure{any}Banman` (brunoerg)
Pull request description:
This PR adds `Ensure{any}Banman` functions to avoid code repetition and make it cleaner. Same approach as done with argsman, chainman, connman and others.
ACKs for top commit:
davidgumberg:
ACK [2d955ff](2d955ff006)
Tree-SHA512: 0beb7125312168a3df130c1793a1412ab423ef0f46023bfe2a121630c79df7e55d3d143fcf053bd09e2d96e9385a7a04594635da3e5c6be0c5d3a9cafbe3b631
Use DBParams struct to remove ArgsManager uses from txdb.
To reduce size of this commit, this moves references to gArgs variable out of
txdb.cpp to calling code in chainstate.cpp. But these moves are temporary. The
gArgs references in chainstate.cpp are moved out to calling code in init.cpp in
later commits.
This commit does not change behavior.
Add CoinsViewOptions struct to remove ArgsManager uses from txdb.
To reduce size of this commit, this moves references to gArgs variable out of
txdb.cpp to calling code in validation.cpp. But these moves are temporary. The
gArgs references in validation.cpp are moved out to calling code in init.cpp in
later commits.
This commit does not change behavior.
Add DBParams and DBOptions structs to remove ArgsManager uses from dbwrapper.
To reduce size of this commit, this moves references to gArgs variable out of
dbwrapper.cpp to calling code in txdb.cpp. But these moves are temporary. The
gArgs references in txdb.cpp are moved out to calling code in init.cpp in later
commits.
This commit does not change behavior.
it adds `Ensure{any}Banman` functions to avoid
code repetition and make it cleaner. Similar
approach as done with argsman, chainman, connman
and others.
4de02def84 qt: Persist Mask Values option (Andrew Chow)
Pull request description:
The mask values option is memory only. If a user has enabled this option, it's reasonable to expect that they would want to have it enabled on the next start.
ACKs for top commit:
RandyMcMillan:
tACK 4de02def84
jarolrod:
tACK 4de02def84
pablomartin4btc:
> tACK [4de02de](4de02def84)
john-moffett:
tACK 4de02def84
Tree-SHA512: 247deb78df4911516625bf8b25d752feb480ce30eb31335cf9baeb07b7c6c225fcc37d5c45de62d6e6895ec10c7eefabb15527e3c9723a3b8ddda1e12ebbf46b
faff2ba4f8 Remove reindex special case from the progress bar label (MarcoFalke)
Pull request description:
The user knows which option they passed to the program, so it seems overly verbose to offer the user feedback whether or not they passed `-reindex`. Treat it as `DISK`, like all other cases that are treated as `DISK`:
* `-reindex-chainstate`
* `-loadblock`
ACKs for top commit:
john-moffett:
Re-ACK faff2ba4f8
hebasto:
ACK faff2ba4f8, I have reviewed the code and it looks OK, I agree it can be merged.
Tree-SHA512: 7f110c4beb1451d26f32da3a60150dac91c8a7b8d1c01749017204712b73cc1b77578af492930e4b6704097a73ed051f77bc39d8f60e0ff15a797a201805312e
c9ba4f9ecb test: Add test for file system permissions (Hennadii Stepanov)
581f16ef34 Apply default umask in `SetupEnvironment()` (Hennadii Stepanov)
8a6219e543 Remove `-sysperms` option (Hennadii Stepanov)
Pull request description:
On master (1e7564eca8) docs say:
```
$ ./src/bitcoind -help | grep -A 3 sysperms
-sysperms
Create new files with system default permissions, instead of umask 077
(only effective with disabled wallet functionality)
```
Basing on that, one could expect that running `bitcoind` first time will create data directory and `wallets/` subdirectory with safe 0700 permissions.
But that is not the case:
```
$ stat .bitcoin | grep id
Access: (0775/drwxrwxr-x) Uid: ( 1000/ hebasto) Gid: ( 1000/ hebasto)
$ stat .bitcoin/wallets | grep id
Access: (0775/drwxrwxr-x) Uid: ( 1000/ hebasto) Gid: ( 1000/ hebasto)
```
Both directories, in fact, are created with system default permissions.
With this PR:
```
$ stat .bitcoin/wallets | grep id
Access: (0700/drwx------) Uid: ( 1000/ hebasto) Gid: ( 1000/ hebasto)
$ stat .bitcoin/wallets | grep id
Access: (0700/drwx------) Uid: ( 1000/ hebasto) Gid: ( 1000/ hebasto)
```
---
This PR:
- is alternative to bitcoin/bitcoin#13389
- fixes bitcoin/bitcoin#15902
- fixes bitcoin/bitcoin#22595
- closes bitcoin/bitcoin#13371
- reverts bitcoin/bitcoin#4286
Changes in behavior: removed `-sysperms` command-line argument / configure option. The related discussions are here:
- https://github.com/bitcoin/bitcoin/pull/13389#issuecomment-395306690
- https://github.com/bitcoin/bitcoin/pull/13389#issuecomment-539906114
- https://github.com/bitcoin/bitcoin/pull/13389#discussion_r279160472
If users rely on non-default access permissions, they could use `chmod`.
ACKs for top commit:
john-moffett:
ACK c9ba4f9ecb
willcl-ark:
ACK c9ba4f9ecb
Tree-SHA512: 96c745339e6bd0e4d7bf65daf9a721e2e1945b2b0ab74ca0f66576d0dc358b5de8eb8cdb89fe2160f3b19c39d2798bb8b291784316085dc73a27102d3415bd57
935acdcc79 refactor: modernize the implementation of uint256.* (pasta)
Pull request description:
- Constructors of uint256 to utilize Span instead of requiring a std::vector
- converts m_data into a std::array
- Prefers using `WIDTH` instead of `sizeof(m_data)`
- make all the things constexpr
- replace C style functions with c++ equivalents
- memset -> std::fill
This may also be replaced by std::memset, but I think that std::fill is more idiomatic of modern c++ and readable.
- memcpy -> std::copy
Note: In practice, implementations of std::copy avoid multiple assignments and use bulk copy functions such as std::memmove if the value type is TriviallyCopyable and the iterator types satisfy LegacyContiguousIterator. (https://en.cppreference.com/w/cpp/algorithm/copy)
This could also likely be replaced by std::memcpy, but as said above, I believe the using std::copy is the more c++ way to do anything and is almost guaranteed to compile to the same asm
- memcmp -> std::memcmp
ACKs for top commit:
achow101:
ACK 935acdcc79
hebasto:
Approach ACK 935acdcc79.
aureleoules:
reACK 935acdcc79
john-moffett:
ACK 935acdcc79
stickies-v:
Approach ACK 935acdcc7
Tree-SHA512: 4f1ba54ff2198eea0e505d41e73d552c84c60f6878d5c85a94a8ab57f39afc94ef8d79258e7afd01fa84ec2a99f4404bb877eecd671f65e1ee9273f3129fc650
b8032293e6 Remove use of snprintf and simplify (John Moffett)
Pull request description:
These are the only remaining uses of `snprintf` in our project, and they can cause unexpected issues -- for example, see https://github.com/bitcoin/bitcoin/issues/27014. Change them to use our `ToString` (which uses a locale-independent version of `std::to_string`) to convert an `int` to `std::string`. Also remove resulting unused parts of `StringContentsSerializer`.
Closes https://github.com/bitcoin/bitcoin/issues/27014
ACKs for top commit:
Sjors:
tACK b8032293e6, fixes #27014.
Tree-SHA512: c903977e654711929decafe8887d0de13b38a340d7082875acc5d41950d834dcfde074e9cabecaf5f9a760f62c34322297b4b156af29761650ef5803b1a54b59
82f895d7b5 Update nanobench to version v4.3.10 (Martin Leitner-Ankerl)
Pull request description:
Nothing has changed that would affect Bitcoin's usage of nanobench.
Here is a detailed list of the changes
* Plenty of clang-tidy updates
* documentation updates
* faster Rng::shuffle
* Enable perf counters on older kernels
* Raise default minimum epoch time to 1ms (doesn't effect bitcoin's usage)
* Add support for custom information per benchmark
ACKs for top commit:
hebasto:
ACK 82f895d7b5, I've reviewed the code, all related changes from #26642 have been implemented.
Tree-SHA512: 942518398809a2794617a347ab8182b784a8e822e84de5af078b2531eabb438412d687cac22a21936585e60e07138a89b41c28c9750744c05a3d1053f55cad01
fe683f3524 log: Log VerifyDB Progress over multiple lines (Martin Zumsande)
61431e3a57 validation: Skip VerifyDB checks of level >=3 if dbcache is too small (Martin Zumsande)
Pull request description:
This is the first two commits from #25574, leaving out all changes to `-verifychain` error-handling :
- The Problem of [25563](https://github.com/bitcoin/bitcoin/issues/25563) is that when we skip blocks at level 3 due to an insufficient dbcache (skipping some `DisconnectBlock()` calls), we would still attempt the level 4 checks, attempting to reconnect a block that was never disconnected, leading to an assert in `ConnectBlock()`.
Fix this by not attempting level 4 checks in this case.
- Logging of verification progress is now split over multiple lines. This is more verbose, but now each update has its own timestamp, and other threads logging concurrently will no longer lead to mangled output.
This can be tested with a small `dbcache` value, for example:
`bitcoind -signet -dbcache=10`
`bitcoin-cli -signet verifychain 4 1000`
Fixes #25563
ACKs for top commit:
MarcoFalke:
review ACK fe683f3524 🗄
john-moffett:
ACK fe683f3524
Tree-SHA512: 3e2e0f8b73cbc518a0fa17912c1956da437787aab95001c110b01048472e0dfe4783c44df22bd903d198069dd2f6b02bfdf74e0b934c7a776f144c2e86cb818a
Silently emitting an error makes it a bit harder to debug. Instead,
print a helpful log message to point the developer in the right
direction.
Alternatively this could have been implemented by just removing the
recipe echo suppression (@), but the subsequent make output became too
noisy.
When generating new files as part of the Makefile the recipe is
sometimes suppressed with $(AM_V_GEN) and sometimes with `@`. We should
prefer $(AM_V_GEN), since this also prints the lines in silent mode.
This is arguably more in style with the current recipe echoing.
Before:
Generated test/data/script_tests.json.h
Now:
GEN test/data/script_tests.json.h
A side effect of this change is that the recipe for generating build.h
is now echoed on each make run. Arguably this makes its generation more
transparent.
6699d850e4 doc: release notes for #27037 (Antoine Poinsot)
dfc9acbf01 rpc: decode Miniscript descriptor when possible in decodescript (Antoine Poinsot)
Pull request description:
The descriptor inference logic would previously always use a dummy signing provider and would never analyze the witness script of a P2WSH scriptPubKey.
It's often not possible to infer a Miniscript only from the onchain Script, but it was such a low hanging fruit that it's probably worth having it?
Fixes https://github.com/bitcoin/bitcoin/issues/27007. I think it also closes https://github.com/bitcoin/bitcoin/issues/25606.
ACKs for top commit:
instagibbs:
ACK 6699d850e4
achow101:
ACK 6699d850e4
sipa:
utACK 6699d850e4
Tree-SHA512: e592bf1ad45497e7bd58c26b33cd9d05bb3007f1e987bee773d26013c3824e1b394fe4903809d80997d5ba66616cc79d77850cd7e7f847a0efb2211c59466982
