511aa4f1c7508f15cab8d7e58007900ad6fd3d5d Add unit test for ChaCha20's new caching (Pieter Wuille)
fb243d25f754da8f01793b41e2d225b917f3e5d7 Improve test vectors for ChaCha20 (Pieter Wuille)
93aee8bbdad808b7009279b67470d496cc26b936 Inline ChaCha20 32-byte specific constants (Pieter Wuille)
62ec713961ade7b58e90c905395558a41e8a59f0 Only support 32-byte keys in ChaCha20{,Aligned} (Pieter Wuille)
f21994a02e1cc46d41995581b54222abc655be93 Use ChaCha20Aligned in MuHash3072 code (Pieter Wuille)
5d16f757639e2cc6e81db6e07bc1d5dd74abca6c Use ChaCha20 caching in FastRandomContext (Pieter Wuille)
38eaece67b1bc37b2f502348c5d7537480a34346 Add fuzz test for testing that ChaCha20 works as a stream (Pieter Wuille)
5f05b27841af0bed1b6e7de5f46ffe33e5919e4d Add xoroshiro128++ PRNG (Martin Leitner-Ankerl)
12ff72476ac0dbf8add736ad3fb5fad2eeab156c Make unrestricted ChaCha20 cipher not waste keystream bytes (Pieter Wuille)
6babf402130a8f3ef3058594750aeaa50b8f5044 Rename ChaCha20::Seek -> Seek64 to clarify multiple of 64 (Pieter Wuille)
e37bcaa0a6dbb334ab6e817efcb609ccee6edc39 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 511aa4f1c7508f15cab8d7e58007900ad6fd3d5d
dhruv:
tACK crACK 511aa4f1c7
Tree-SHA512: 3aa80971322a93e780c75a8d35bd39da3a9ea570fbae4491eaf0c45242f5f670a24a592c50ad870d5fd09b9f88ec06e274e8aa3cefd9561d623c63f7198cf2c7
fa6c114ae604571435e8c4d25906a8b6d5b9984c test: Add sanitizer suppressions for AMD EPYC CPUs (MarcoFalke)
Pull request description:
Currently the ci system only runs on intel cpus (and some arm devices), but it won't run on CPUs `Using the 'shani(1way,2way)' SHA256 implementation` (excerpt from debug log).
For reference, google cloud CPUs (which is what Cirrus CI uses) print `Using the 'sse4(1way),sse41(4way),avx2(8way)' SHA256 implementation`
The traceback I got:
```
crypto/sha256_shani.cpp:87:18: runtime error: unsigned integer overflow: 0 - 1 cannot be represented in type 'size_t' (aka 'unsigned long')
#0 0x55c0000e95ec in sha256_shani::Transform(unsigned int*, unsigned char const*, unsigned long) /root/bitcoin/ci/scratch/build/bitcoin-x86_64-pc-linux-gnu/src/crypto/sha256_shani.cpp:87:18
#1 0x55bfffb926f8 in (anonymous namespace)::SelfTest() /root/bitcoin/ci/scratch/build/bitcoin-x86_64-pc-linux-gnu/src/crypto/sha256.cpp:517:9
#2 0x55bfffb906ed in SHA256AutoDetect[abi:cxx11]() /root/bitcoin/ci/scratch/build/bitcoin-x86_64-pc-linux-gnu/src/crypto/sha256.cpp:626:5
#3 0x55bfff87ab97 in BasicTestingSetup::BasicTestingSetup(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::vector<char const*, std::allocator<char const*> > const&) /root/bitcoin/ci/scratch/build/bitcoin-x86_64-pc-linux-gnu/src/test/util/setup_common.cpp:104:5
#4 0x55bffe885877 in main /root/bitcoin/ci/scratch/build/bitcoin-x86_64-pc-linux-gnu/src/qt/test/test_main.cpp:52:27
#5 0x7f20c3bf60b2 in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x270b2)
#6 0x55bffe7a5f6d in _start (/root/bitcoin/ci/scratch/build/bitcoin-x86_64-pc-linux-gnu/src/qt/test/test_bitcoin-qt+0x1d00f6d)
SUMMARY: UndefinedBehaviorSanitizer: unsigned-integer-overflow crypto/sha256_shani.cpp:87:18 in
ACKs for top commit:
laanwj:
Anyhow ACK fa6c114ae604571435e8c4d25906a8b6d5b9984c
Tree-SHA512: 968a1d28eedec58c337b1323862f583cb1bcd78c5f03396940b9ab53ded12f8c6652877909aba05ee5586532137418fd817ff979bd7bef6e07856094f9d7f9b1
0ccb3addf68067200892963521a92713c4667a63 tests: Remove no longer needed UBSan suppression (float-divide-by-zero in validation.cpp) (practicalswift)
Pull request description:
Remove no longer needed UBSan suppression.
The float divide-by-zero in `validation.cpp` was fixed by instagibbs in ec30a79f1c430cc7fbda37e5d747b0b31b262fa5 (#15283).
ACKs for top commit:
MarcoFalke:
ACK 0ccb3addf68067200892963521a92713c4667a63
Tree-SHA512: 89a4f4b7371fa5725d9f801cee7ebbd17523f66017c9acfa813657dcb8d837f42209eff44ce9e5d48296a630bab9599d75f10024a0c7da7defb228f4eae3392a
5e146022daa4336de94447e5b8e5418296286927 wallet: fix scanning progress calculation for single block range (Sebastian Falbesoner)
Pull request description:
If the blockchain is rescanned for a single block (i.e. start and stop hashes are equal, and with that also the estimated start/stop verification progress values) the progress calculation could lead to a NaN value caused by a division by zero (0.0/0.0), resulting in an invalid JSON result for the `getwalletinfo` RPC. This PR fixes this behaviour by setting the progress to zero in that special case. Fixes#20297.
The behaviour can easily be reproduced by continuously running single block rescans in an endless loop, e.g. via
```bash
#!/bin/bash
while true
do
bitcoin-cli rescanblockchain $(bitcoin-cli getblockcount)
done
```
and at the same time perform some `getwalletinfo` RPCs.
On the master branch, this leads to frequent invalid responses (tested on mainchain):
```
$ bitcoin-cli getwalletinfo
error: couldn't parse reply from server
$ curl --user `cat ~/.bitcoin/.cookie` --data-binary '{"jsonrpc": "1.0", "id": "curltest", "method": "getwalletinfo", "params": []}' -H 'content-type: text/plain;' http://127.0.0.1:8332/
{"result":{"walletname":"","walletversion":169900,"format":"bdb","balance":0.00000000,"unconfirmed_balance":0.00000000,"immature_balance":0.00000000,"txcount":0,"keypoololdest":1603677276,"keypoolsize":1000,"hdseedid":"3196e33ecb47c7130e6ca60f2f895f9259860dca","keypoolsize_hd_internal":1000,"paytxfee":0.00000000,"private_keys_enabled":true,"avoid_reuse":false,"scanning":{"duration":0,"progress":},"descriptors":false},"error":null,"id":"curltest"}
```
(note that missing value for "progress" in the JSON result).
On the PR branch, the behaviour doesn't occur anymore.
ACKs for top commit:
MarcoFalke:
review ACK 5e146022daa4336de94447e5b8e5418296286927
promag:
Core review ACK 5e146022daa4336de94447e5b8e5418296286927.
Tree-SHA512: f0e6aad5a6cd08b36c5fe820fff0ef26663229b39169a4dbe757f3c795a41cf5c69c9dc90efe7515675ae1059307f8971123781a0514d10704123a6f28b125ab
9c891b64ffd14bc8216dbd5eb60816043af265b6 net: initialize nMessageSize to max uint32_t instead of -1 (eugene)
Pull request description:
nMessageSize is uint32_t and is set to -1. This will warn with `-fsanitize=implicit-integer-sign-change` when V1TransportDeserializer calls into the ctor. This pull initializes nMessageSize to `numeric_limits<uint32_t>::max()` instead and removes the ubsan suppression.
ACKs for top commit:
laanwj:
Code review ACK 9c891b64ffd14bc8216dbd5eb60816043af265b6
promag:
Code review ACK 9c891b64ffd14bc8216dbd5eb60816043af265b6.
Tree-SHA512: f05173d9553a01d207a5a7f8ff113d9e11354c50b494a67d44d3931c151581599a9da4e28f40edd113f4698ea9115e6092b2a5b7329c841426726772076c1493
a3abeec33a6ae903e514c7a7b6f587b7c17288a0 policy/fees: remove a floating-point division by zero (Antoine Poinsot)
c36869bbf6a38626833b4aea53be024c48ede475 policy/fees: unify some duplicated for loops (Antoine Poinsot)
569d92a4d2924a1f6d50775980b591552f6372e7 policy/fees: small readability improvements (Antoine Poinsot)
5b8cb35621891b681f9b49a9de5f6d8da4ccdecc policy/fee: remove requireGreater parameter in EstimateMedianVal() (Antoine Poinsot)
dba8196b447b6a85be66890db70928100e867d8b policy/fees: correct decay explanation comments (Antoine Poinsot)
Pull request description:
This (*does not* change behaviour and) cleans up a bit of unused code in `CBlockPolicyEstimator` and friends, and slightly improves readability of the rest (comment correction etc.). The last commit is a small reformatting one which I could not resist but am happy to remove at will.
ACKs for top commit:
jnewbery:
utACK a3abeec33a6ae903e514c7a7b6f587b7c17288a0
MarcoFalke:
ACK a3abeec33a6ae903e514c7a7b6f587b7c17288a0 💹
ariard:
Code Review ACK a3abeec.
Tree-SHA512: b7620bcd23a2ffa8f7ed859467868fc0f6488279e3ee634f6d408872cb866ad086a037e8ace76599a05b7e9c07768adf5016b0ae782d153196b9c030db4c34a5
fa40e48c50d8ccf42ce5e66c12390e2ed4b60e75 ci: Remove unparseable lines from supp file for old xenial clang tsan (MarcoFalke)
fa1bfc476c9208a4c412c8ca74d05f52bb47766f ci: ubsan report_error_type=1 and add suppressions (MarcoFalke)
fa69cef13e5aab8264339eb3d50a9e89d59efd87 test: Print stderr when subprocess fails (MarcoFalke)
2222c305866a77065ab5be24c1c252bae252bb59 test: Use char instead of unsigned char (MarcoFalke)
faa8023ce9a47b282e1fac3ca8b3a7bb0042935a ci: Bump to clang-8 for asan build to avoid segfaults on ppc64le (MarcoFalke)
Pull request description:
Use clang-8 instead of default clang (which is clang-6 on Bionic) to avoid spurious segfaults when running the ci system on ppc64le
ACKs for top commit:
practicalswift:
ACK fa40e48c50d8ccf42ce5e66c12390e2ed4b60e75 assuming Travis is happy -- diff looks correct :)
Tree-SHA512: f4f26232d3a0ef38da245869340f723d279a3db9823befbc735fb5a00024dae041c7306d7ae55d2488e6f86aa96cdea155b007aefb561fba505141e8dbc717dc
0616138a0797cf68ad869906c36cf0767e20b313 tests: Remove no longer needed UBSan suppressions (issues fixed). Add documentation. (practicalswift)
Pull request description:
Remove no longer needed UBSan suppressions (issues fixed). Add documentation.
This PR is the CI-only subset of #17208 (which touches code).
From a fuzzing perspective it would be really nice to be able to run UBSan with as few suppressions as possible :)
Top commit has no ACKs.
Tree-SHA512: a926ab3e80e12a805af110fbff470cdc61ef4db536919a5b8896ea8b70f761114a52d9b1c0f48b11c1d48338351bf2e003e01ce60c613612f26ba298dcc29cd9
5f26855f109af53a336d5f98ed0ae584e7a31f84 test: Remove ubsan alignment suppressions (Wladimir J. van der Laan)
9d933ef9191417b4b7d29eaa3c3a571f814acc8e prevector: avoid misaligned member accesses (Anthony Towns)
Pull request description:
Ensure prevector data is appropriately aligned. Earlier discussion in #17530.
**Edit laanwj**: In contrast to #17530, it does this without increase in size of any of the coin cache data structures (x86_64, clang)
| Struct | (size,align) before | (size,align) after |
| ------------- | ------------- | ------- |
| Coin | 48, 8 | 48, 8 |
| CCoinsCacheEntry | 56, 8 | 56, 8 |
| CScript | 32, 1 | 32, 8 |
ACKs for top commit:
laanwj:
ACK 5f26855f109af53a336d5f98ed0ae584e7a31f84
practicalswift:
ACK 5f26855f109af53a336d5f98ed0ae584e7a31f84
jonatack:
ACK 5f26855f109af53a336d5f98ed0ae584e7a31f84
Tree-SHA512: 98d112d6856f683d5b212410b73f3071d2994f1efb046a2418a35890aa1cf1aa7c96a960fc2e963fa15241e861093c1ea41951cf5b4b5431f88345eb1dd0a98a
4f4993fe2a Remove UBSan suppression (practicalswift)
958e1a307e streams: Remove unused seek(size_t) (practicalswift)
Pull request description:
Fix broken `streams_vector_reader` test. Remove unused `seek(size_t)`.
Before this change the test `streams_vector_reader` triggered an unintended unsigned integer wraparound. It tried so seek using a negative value in `reader.seek(-6)`.
Changes in this PR:
* Fix broken `VectorReader::seek(size_t)` test case
* Remove unused `seek(size_t)`
Tree-SHA512: 6c6affd680626363eef9e496748f2f86a522325abab9d6b13161f41125cdc29ceb36c2c1509c90b8ff108d606df7629e55e094cc2b6253b05a892b81ce176b71
fa7d36b8e7 test: Move UBSAN suppressions to test/sanitizer_suppressions/ubsan (MarcoFalke)
fa36d4e456 travis: --disable-hardening for xenial thread sanitizer (MarcoFalke)
89bf196c88 travis: Run thread sanitizer (MarcoFalke)
Pull request description:
On unit tests only for now. Disabled for the gui unit tests and all functional tests.
Tree-SHA512: 56f7d3b44e7cb68c76a2dc5abd85658955b1c2188932e988667c5a1cbcdd6be995d37bb949d62c6eb08a4aebfc43ff0370b7da1719d4e4f322a3495c1941a5e0
