CWallet::MarkConflicted may acquire the cs_main lock after
CWalletDB::LoadWallet acquires the cs_wallet lock during wallet initialization.
(CWalletDB::LoadWallet calls ReadKeyValue which calls CWallet::LoadToWallet
which calls CWallet::MarkConflicted). This is the opposite order that cs_main
and cs_wallet locks are acquired in the rest of the code, and so leads to
POTENTIAL DEADLOCK DETECTED errors if bitcoin is built with -DDEBUG_LOCKORDER.
This commit changes CWallet::LoadWallet (which calls CWalletDB::LoadWallet) to
acquire both locks in the standard order. It also fixes some tests that were
acquiring wallet and main locks out of order and failed with the new locking in
CWallet::LoadWallet.
Error was reported by Luke Dashjr <luke-jr@utopios.org> in
https://botbot.me/freenode/bitcoin-core-dev/msg/90244330/
This assures that we don't overwrite a random file called
`wallet.backup` that happens to be in the current directory. It also
assures that the temporary file will be cleaned up.
Noticed by Evan Klitzke, came up in discussion here:
https://github.com/bitcoin/bitcoin/pull/10880#discussion_r128460722
Start importwallet rescans at the first block with timestamp greater or equal
to the wallet birthday instead of the last block with timestamp less or equal.
This fixes an edge case bug where importwallet could fail to start the rescan
early enough if there are blocks with decreasing timestamps or multiple blocks
with the same timestamp.
An off-by-one-block bug in importmulti rescan logic could cause it to return
success in an edge case even when a rescan was not successful. The case where
this would happen is if there were multiple blocks in a row with the same
GetBlockTimeMax() value, and the last block was scanned successfully, but one
or more of the earlier blocks was not readable.
Abstract database handle from explicit strFilename into
CWalletDBWrapper.
Also move CWallet::Backup to db.cpp - as it deals with representation
details this is a database specific operation.
Change ScanForWalletTransactions return value so it is possible to distinguish
scans that skip reading every block (due to the nTimeFirstKey optimization)
from scans that fail while reading the chainActive.Tip() block. Return value is
now non-null in the non-failing case.
This change doesn't affect any user-visible behavior, it is only an internal
API improvement. The only code currently using the ScanForWalletTransactions
return value is in importmulti, and importmulti always calls
ScanForWalletTransactions with a pindex pointing to the first block in
chainActive whose block time is >= (nLowestTimestamp - 7200), while
ScanForWalletTransactions would only return null without reading blocks when
pindex and every block after it had a block time < (nTimeFirstKey - 7200).
These conditions could never happen at the same time because nTimeFirstKey <=
nLowestTimestamp.
I'm planning to make a more substantial API improvement in the future (making
ScanForWalletTransactions private and exposing a higher level rescan method to
RPC code), but Matt Corallo <git@bluematt.me> pointed out this odd behavior
introduced by e2e2f4c "Return errors from importmulti if complete rescans are
not successful" yesterday, so I'm following up now to get rid of badness
introduced by that merge.
Warnings introduced by commit e2e2f4c "Return errors from importmulti if
complete rescans are not successful" and reported by Pavel Janík
<Pavel@Janik.cz> in https://github.com/bitcoin/bitcoin/pull/9773 and
https://github.com/bitcoin/bitcoin/pull/9827
wallet/test/wallet_tests.cpp: In member function ‘void wallet_tests::rescan::test_method()’:
wallet/test/wallet_tests.cpp:377:17: warning: declaration of ‘wallet’ shadows a global declaration [-Wshadow]
CWallet wallet;
Change CCrypter to use vectors with secure allocator instead of buffers
on in the object itself which will end up on the stack. This avoids
having to call LockedPageManager to lock stack memory pages to prevent the
memory from being swapped to disk. This is wasteful.
There are only a few uses of `insecure_random` outside the tests.
This PR replaces uses of insecure_random (and its accompanying global
state) in the core code with an FastRandomContext that is automatically
seeded on creation.
This is meant to be used for inner loops. The FastRandomContext
can be in the outer scope, or the class itself, then rand32() is used
inside the loop. Useful e.g. for pushing addresses in CNode or the fee
rounding, or randomization for coin selection.
As a context is created per purpose, thus it gets rid of
cross-thread unprotected shared usage of a single set of globals, this
should also get rid of the potential race conditions.
- I'd say TxMempool::check is not called enough to warrant using a special
fast random context, this is switched to GetRand() (open for
discussion...)
- The use of `insecure_rand` in ConnectThroughProxy has been replaced by
an atomic integer counter. The only goal here is to have a different
credentials pair for each connection to go on a different Tor circuit,
it does not need to be random nor unpredictable.
- To avoid having a FastRandomContext on every CNode, the context is
passed into PushAddress as appropriate.
There remains an insecure_random for test usage in `test_random.h`.
practicalswift