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.
This fixes an potential overflow which existed prior to this patchset.
If CuckooCache::cache<Element, Hash>::setup_bytes is called with a
`size_t bytes` which, when divided by sizeof(Element), does not fit into
an uint32_t, the implicit conversion to uint32_t in the call to setup
will result in an overflow.
At least on x86_64, this overflow is possible:
static_assert(std::numeric_limits<size_t>::max() / 32 <= std::numeric_limits<uint32_t>::max());
static_assert(std::numeric_limits<size_t>::max() / 4 <= std::numeric_limits<uint32_t>::max());
This commit detects such cases and signals to callers that the `size_t
bytes` input is too large.
1. -maxsigcachesize is a DEBUG_ONLY option
2. Almost 7 years has passed since its semantics change in
830e3f3d02 from "number of entries" to
"number of mebibytes"
3. A std::new_handler was added to the codebase after the original PR
which introduced this limit, which will terminate immediately instead
of causing trouble by being caught somewhere unexpected.
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.
Improve readability of code, simplify future scripted diff cleanup PRs, and be
more consistent with naming for GetBoolArg.
This will also be useful for replacing runtime settings type checking
with compile time checking.
-BEGIN VERIFY SCRIPT-
git grep -l GetArg | xargs sed -i 's/GetArg(\([^)]*\( [0-9]\+\|-1\|port\|BaseParams().RPCPort()\|Params().GetDefaultPort()\|_TIMEOUT\|Height\|_WORKQUEUE\|_THREADS\|_CONNECTIONS\|LIMIT\|SigOp\|Bytes\|_VERSION\|_AGE\|_CHECKS\|Checks() ? 1 : 0\|_BANTIME\|Cache\|BLOCKS\|LEVEL\|Weight\|Version\|BUFFER\|TARGET\|WEIGHT\|TXN\|TRANSACTIONS\|ADJUSTMENT\|i64\|Size\|nDefault\|_EXPIRY\|HEIGHT\|SIZE\|SNDHWM\|_TIME_MS\)\))/GetIntArg(\1)/g'
-END VERIFY SCRIPT-
Co-authored-by: Hennadii Stepanov <32963518+hebasto@users.noreply.github.com>
Bitcoin allocates 32M for signature caching by default, split
between a signature cache and a script validity cache (see
Core #10192). Since 0.14 we have added an additional 32M for
rangeproof caching *and* an additional 32M for surjectionproof
caching.
These cache entries cost a bit over 32 bytes, so these are room
for a million entries....or 4Gb of rangeproofs and 300M of
surjection proofs.
Presumably we did not intend to triple memory usage relative to
Core to deal with some champagne problem in which our mempool is
overfilled ten times with pure rangeproofs. So put the total
default cache size back to 32M. This should have no performance
hit under realistic circumstances and should reduce CI OOM failures.
On my system we now use 50M rather than 110M during the fedpeg
test; we still use 18M that Core does not by having three additional
global secp contexts (one in blind.cpp, one in pegins.cpp, one in
confidential_validation.cpp) but we can settle that in a future
commit.
This enables the schnorrsig module in libsecp256k1, adds the relevant types
and functions to src/pubkey, as well as in higher-level `SignatureChecker`
classes. The (verification side of the) BIP340 test vectors is also added.
In preparation for adding Schnorr versions of `CheckSig`, `VerifySignature`, and
`ComputeEntry`, give them an ECDSA specific name.
-BEGIN VERIFY SCRIPT-
sed -i 's/CheckSig(/CheckECDSASignature(/g' $(git grep -l CheckSig ./src)
sed -i 's/VerifySignature(/VerifyECDSASignature(/g' $(git grep -l VerifySignature ./src)
sed -i 's/ComputeEntry(/ComputeEntryECDSA(/g' $(git grep -l ComputeEntry ./src)
-END VERIFY SCRIPT-
Use salted hasher instead of nonce in Script Execution Cache
Don't read more than 32 bytes from GetRand
Apply g_* naming convention to scriptExecutionCache in validation.cpp
Fully apply g_* naming convention to scriptCacheHasher
Write same uint256 nonce twice for cache hash rather than calling getrand twice
Use salted hasher instead of nonce in sigcache
Use salted hasher instead of nonce in Script Execution Cache
Don't read more than 32 bytes from GetRand
Apply g_* naming convention to scriptExecutionCache in validation.cpp
Fully apply g_* naming convention to scriptCacheHasher
Write same uint256 nonce twice for cache hash rather than calling getrand twice
This adds a new CuckooCache in validation, caching whether all of a
transaction's scripts were valid with a given set of script flags.
Unlike previous attempts at caching an entire transaction's
validity, which have nearly universally introduced consensus
failures, this only caches the validity of a transaction's
scriptSigs. As these are pure functions of the transaction and
data it commits to, this should be much safer.
This is somewhat duplicative with the sigcache, as entries in the
new cache will also have several entries in the sigcache. However,
the sigcache is kept both as ATMP relies on it and because it
prevents malleability-based DoS attacks on the new higher-level
cache. Instead, the -sigcachesize option is re-used - cutting the
sigcache size in half and using the newly freed memory for the
script execution cache.
Transactions which match the script execution cache never even have
entries in the script check thread's workqueue created.
Note that the cache is indexed only on the script execution flags
and the transaction's witness hash. While this is sufficient to
make the CScriptCheck() calls pure functions, this introduces
dependancies on the mempool calculating things such as the
PrecomputedTransactionData object, filling the CCoinsViewCache, etc
in the exact same way as ConnectBlock. I belive this is a reasonable
assumption, but should be noted carefully.
In a rather naive benchmark (reindex-chainstate up to block 284k
with cuckoocache always returning true for contains(),
-assumevalid=0 and a very large dbcache), this connected blocks
~1.7x faster.
- If the -maxsigcachesize parameter is set to zero, setup a minimum sized
sigcache (2 elements) rather than segfaulting.
- Handle maxsigcachesize being negative
- Handle maxsigcachesize being too large
