Instead of killing a connection when the receive buffer overflows,
just temporarily halt receiving before that happens. Also, no
matter what, always allow at least one full message in the receive
buffer (otherwise blocks larger than the configured buffer size
would pause indefinitely).
It is possible to have a wallet.dat file without any bestblock
record at all (if created offline, for example), which - when
loaded into a client with a up-to-date chain - does no rescan and
shows no transactions.
Also make sure to write the current best block after a rescan, so
it isn't necessary twice.
Every block index entry currently requires a separately-allocated
CBigNum. By replacing them with uint256, it's just 32 bytes extra
in CBlockIndex itself.
This should save us a few megabytes in RAM, and less allocation
overhead.
This introduces the concept of the 'sync node', which is the one we
asked for missing blocks. In case the sync node goes away, a new one
will be selected.
For now, the heuristic is very simple, but it can easily be extended
later to add better policies.
It seems there were two mechanisms for assessing whether a CNode
was still in use: a refcount and a release timestamp. The latter
seems to have been there for a long time, as a safety mechanism.
However, this timer also keeps CNode objects alive for far longer
than necessary after disconnects, potentially opening up a DoS
window.
This commit removes the timestamp-based mechanism, and replaces
it with an assert(nRefCount >= 0), to verify that the refcounting
is indeed correctly working.
As these were not updated when 'backporting' the 225430 checkpoint
into head.
Additionally, also report verification progress in debug.log, and
tweak the sigcheck-verification-speed-factor a bit.
There exists a per-message-processed send buffer overflow protection,
where processing is halted when the send buffer is larger than the
allowed maximum.
This protection does not apply to individual items, however, and
getdata has the potential for causing large amounts of data to be
sent. In case several hundreds of blocks are requested in one getdata,
the send buffer can easily grow 50 megabytes above the send buffer
limit.
This commit breaks up the processing of getdata requests, remembering
them inside a CNode when too many are requested at once.
* Change CNode::vRecvMsg to be a deque instead of a vector (less copying)
* Make sure to acquire cs_vRecvMsg in CNode::CloseSocketDisconnect (as it
may be called without that lock).
-dbcache was originally used to set the maximum buffer size in the
BDB environment, and was later changed to set the chainstate cache
and leveldb caches. No need to use it for BDB now that only the
wallet remains there.
This should reduce memory allocation (but not necessarily memory
usage) a bit.
* Pass txid's to CCoinsView functions by reference instead of by value
* Add a method to swap CCoins, and use it in some places to avoid a
allocating copy + destruct.
* Optimize CCoinsViewCache::FetchCoins to do only a single search
through the backing map.
By specifying -txindex when initializing the database, a txid-to-diskpos
index is maintained in the blktree database. This database is used to
help answering getrawtransaction() RPC queries, when enabled.
Changing the -txindex value requires a -reindex; the client will abort
at startup if the database and the specified -txindex mismatch.
Client (SPV) mode never got implemented entirely, and whatever part was already
working, is likely not been tested (or even executed at all) for the past two
years. This removes it entirely.
If we want an SPV implementation, I think we should first get the block chain
data structures to be encapsulated in a class implementing a standard interface,
and then writing an alternate implementation with SPV semantics.
Since block validation happens in parallel, multiple threads may be
accessing the signature cache simultaneously. To prevent contention:
* Turn the signature cache lock into a shared mutex
* Make reading from the cache only acquire a shared lock
* Let block validations not store their results in the cache
* During block verification (when parallelism is requested), script
check actions are stored instead of being executed immediately.
* After every processed transactions, its signature actions are
pushed to a CScriptCheckQueue, which maintains a queue and some
synchronization mechanism.
* Two or more threads (if enabled) start processing elements from
this queue,
* When the block connection code is finished processing transactions,
it joins the worker pool until the queue is empty.
As cs_main is held the entire time, and all verification must be
finished before the block continues processing, this does not reach
the best possible performance. It is a less drastic change than
some more advanced mechanisms (like doing verification out-of-band
entirely, and rolling back blocks when a failure is detected).
The -par=N flag controls the number of threads (1-16). 0 means auto,
and is the default.
-checklevel gets a new meaning:
0: verify blocks can be read from disk (like before)
1: verify (contextless) block validity (like before)
2: verify undo files can be read and have good checksums
3: verify coin database is consistent with the last few blocks
(close to level 6 before)
4: verify all validity rules of the last few blocks
Level 3 is the new default, as it's reasonably fast. As level 3 and
4 are implemented using an in-memory rollback of the database, they
are limited to as many blocks as possible without exceeding the
limits set by -dbcache. The default of -dbcache=25 allows for some
150-200 blocks to be rolled back.
In case an error is found, the application quits with a message
instructing the user to restart with -reindex. Better instructions,
and automatic recovery (when possible) or automatic reindexing are
left as future work.
When the coin database is out of date with the block database, the
best block in it is automatically switched to. This reconnection
process can take time, so allow it to be interrupted.
This also stops block connection as soon as shutdown is requested,
leading to a faster shutdown.
This problem is like earth (mostly harmless). After/during a
-reindex, it means the statistics about the last block file
reported in debug.log are always of blk00000.dat instead of the
last file. Apart from that, it means a few more database entries
need to be read when finding a file to append to the first time.
Pieter Wuille