-
Notifications
You must be signed in to change notification settings - Fork 714
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Masternodes] Missing cs main locks in CalculateScore and GetMasternodeInputAge #1791
[Masternodes] Missing cs main locks in CalculateScore and GetMasternodeInputAge #1791
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looking good.
Few non-blocking nits and redundant checks (unrelated, can be addressed in a followup PR).
Note: for future refactoring. In places like GetBlockHash
we duplicate the implementation of the function GetChainTip()
(defined and used in the miner).
We should move it to a more accessible file, and not duplicate those lines every time we need the tip index.
|
||
if (mapCacheBlockHashes.count(nBlockHeight)) { | ||
hash = mapCacheBlockHashes[nBlockHeight]; | ||
return true; | ||
} | ||
|
||
const CBlockIndex* BlockLastSolved = chainActive.Tip(); | ||
const CBlockIndex* BlockReading = chainActive.Tip(); | ||
const CBlockIndex* BlockLastSolved = tipIndex; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we really need another variable for this?
Can't we directly define BlockLastSolved
at the start (inside the cs_main
block), instead of tipIndex
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yep, not needed at all, same as for the other variable BlockReading
that can be removed.
I just didn't want to modify much code around this area for now (it needs its own focused PR). Just made the needed changes to synchronize chainActive
access .
|
||
if (BlockLastSolved == NULL || BlockLastSolved->nHeight == 0 || chainActive.Tip()->nHeight + 1 < nBlockHeight) return false; | ||
if (BlockLastSolved == nullptr || BlockLastSolved->nHeight == 0 || tipHeight + 1 < nBlockHeight) return false; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Here BlockLastSolved
is just tipindex
. If it is null, we've already returned (line 31).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yep. Not part of this PR focus. Can be tackled in another one.
if (chainActive.Tip() == NULL) return UINT256_ZERO; | ||
{ | ||
LOCK(cs_main); | ||
if (chainActive.Tip() == NULL) return UINT256_ZERO; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: use nullptr
instead of NULL
(or !chainActive.Tip()
).
int nBlocksAgo = 0; | ||
if (nBlockHeight > 0) nBlocksAgo = (chainActive.Tip()->nHeight + 1) - nBlockHeight; | ||
if (nBlockHeight > 0) nBlocksAgo = (tipHeight + 1) - nBlockHeight; | ||
assert(nBlocksAgo >= 0); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I know that this was already like that, and this PR is not changing the behaviour, but, since we are at it, why this assert?
nBlocksAgo < 0
iff
(tipHeight + 1) < nBlockHeight
in which case we've already returned (line 47)
CBlockIndex *pindex = chainActive.Tip(); | ||
if (!pindex) return false; | ||
tipHeight = pindex->nHeight; | ||
tipIndex = mapBlockIndex[pindex->GetBlockHash()]; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: better to use mapBlockIndex.at()
, which throws an exception if (for whatever reason) the key is not found.
LOCK(cs_main); | ||
CBlockIndex *pindex = chainActive.Tip(); | ||
if (!pindex) return false; | ||
tipHeight = pindex->nHeight; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: we can initialize this after the LOCK
code-block, as
int tipHeight = tipIndex->nHeight
(...or just remove it, and use tipIndex->nHeight
directly).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
utACK a14aba3
Merging... |
Github-Pull: PIVX-Project#1791 Rebased-From: 224c8fd
…without cs_main fix. Github-Pull: PIVX-Project#1791 Rebased-From: a14aba3
dc0ed71 [BUG][GUI] Don't append cold-stake records twice (random-zebra) 91adcd7 masternode: CalculateScore and GetBlockHash accessing to chainActive without cs_main fix. (furszy) eeb112f GetMasternodeInputAge: Missing cs_main lock (furszy) 52ec12d refactor: decouple decompose coinstake from TransactionRecord::decomposeTransaction. (furszy) 51a8389 [BUG] Properly copy fCoinStake memeber between CTxInUndo and CCoins Github-Pull: #1796 Rebased-From: acf757b (random-zebra) 5c3caa4 lock cs_main for Misbehaving (furszy) 5c629d8 openssl.org dependency download link changed (CryptoDev-Project) 11aa63c Do not try to resend transactions if the node is importing or in IBD. (furszy) c59b62e [Core] Remove BIP30 check (random-zebra) 3f05eba include missing atomic to make CMake linux happy. (furszy) f0cfd88 Make the cs_sendProcessing a LOCK instead of a TRY_LOCK (Matt Corallo) d38e28c Split CNode::cs_vSend: message processing and message sending (Matt Corallo) 7561b18 Remove double brackets in addrman (Matt Corallo) 63c629b Fix AddrMan locking (Matt Corallo) f78cec4 Make fDisconnect an std::atomic (Matt Corallo) ec56cf5 net: fix a few cases where messages were sent rather than dropped upon disconnection (furszy) 7697085 Acquire cs_main lock before cs_wallet during wallet initialization (random-zebra) c796593 Remove bogus assert on number of oubound connections. (Matt Corallo) 7521065 wallet: Ignore MarkConflict if block hash is not known (random-zebra) 72042ac Move disconnectBlocks logging to use __func__ instead of hardcoded method name. (furszy) 0f66764 Simplify DisconnectBlock arguments/return value (furszy) cca4a8c Split logic to undo txin's off DisconnectBlock (furszy) 0f883e6 Cleaner disconnectBlock flow for coinbase and zerocoin txs. (furszy) a7cbc46 Move UndoWriteToDisk() and UndoReadFromDisk() to anon namespace (random-zebra) de5a405 Decouple CBlockUndo from CDiskBlockPos (jtimon) 671143c Decouple miner.o and txmempool.o from CTxUndo (random-zebra) 9419228 Decouple CCoins from CTxInUndo (jtimon) Pull request description: Backports the following PRs to the `4.2` branch: #1746 #1735 #1767 #1769 #1781 #1780 #1775 #1783 #1750 #1785 #1796 #1776 #1791 ACKs for top commit: furszy: Added #1805. utACK dc0ed71. random-zebra: utACK dc0ed71 Tree-SHA512: 0e59c9e751597b1b6f9a419e117946cec468dffdb921c882a44e5770ecb1a36b7d3d25f8c7f97d48bb3d59f136492842c08969901512d957a17bfaec6aece449
def064f [Refactor] Cleanup usage of chainActive.{Tip(),Height()} in a few places (random-zebra) 026dcc9 [Refactor] Move GetChainTip from miner to main (random-zebra) Pull request description: Refactor a few places in masternode/rpc code where `chainActive.Tip()` or `chainActive.Height()` is used. Make use of the function `GetChainTip()`, moved from miner to main, which returns a more reliable pointer (from mapBlockIndex, instead of chainActive). Remove redundant checks in masternode's `GetBlockHash` (e.g. see comments in #1791). ACKs for top commit: furszy: re-re-ACK def064f. Tree-SHA512: 03a4e50b3846ee3f91fd529ebe7d629400c553c64ff033fe40901f96bc106dba237bf344e7c4ac0092cfd206f7c7ea3de003a91e8f82e27a6bd416923d006f90
In
CalculateScore
,GetMasternodeInputAge
andGetBlockHash
we hadchainActive
accessed without cs_main being held. This PR fixes it.No functional changes. Can be added to 4.2.1 back ports list as well.