There are 3 pieces of data that are maintained on disk. The actual block
and undo data, the block index (which can refer to positions on disk),
and the chainstate (which refers to the best block hash).
Earlier, there was no guarantee that blocks were written to disk before
block index entries referring to them were written. This commit introduces
dirty flags for block index data, and delays writing entries until the actual
block data is flushed.
With this stricter ordering in writes, it is now safe to not always flush
after every block, so there is no need for the IsInitialBlockDownload()
check there - instead we just write whenever enough time has passed or
the cache size grows too large. Also updating the wallet's best known block
is delayed until this is done, otherwise the wallet may end up referring to an
unknown block.
In addition, only do a write inside the block processing loop if necessary
(because of cache size exceeded). Otherwise, move the writing to a point
after processing is done, after relaying.
Many changes:
* Do not use 'getblocks', but 'getheaders', and use it to build a headers tree.
* Blocks are fetched in parallel from all available outbound peers, using a
limited moving window. When one peer stalls the movement of the window, it is
disconnected.
* No more orphan blocks. At all. We only ever request a block for which we have
verified the headers, and store it to disk immediately. This means that a
disk-fill attack would require PoW.
* Require protocol version 31800 for every peer (released in december 2010).
* No more syncnode (we sync from everyone we can, though limited to 1 during
initial *headers* sync).
* Introduce some extra named constants, comments and asserts.
There is only one message passed to AbortNode() that makes sense to
translate to the user specifically: Disk space is low. For the others
show a generic message and refer to debug.log for details.
Reduces the number of confusing jargon translation messages.
There is no reason to store thousands of orphan transactions;
normally an orphan's parents will either be broadcast or
mined reasonably quickly.
This pull drops the maximum number of orphans from 10,000 down
to 100, and adds a command-line option (-maxorphantx) that is
just like -maxorphanblocks to override the default.
Thanks to Pieter Wuille for most of the work on this commit.
I did not fixup the overhaul commit, because a rebase conflicted
with "remove fields of ser_streamplaceholder".
I prefer not to risk making a mistake while resolving it.
The implementation of each class' serialization/deserialization is no longer
passed within a macro. The implementation now lies within a template of form:
template <typename T, typename Stream, typename Operation>
inline static size_t SerializationOp(T thisPtr, Stream& s, Operation ser_action, int nType, int nVersion) {
size_t nSerSize = 0;
/* CODE */
return nSerSize;
}
In cases when codepath should depend on whether or not we are just deserializing
(old fGetSize, fWrite, fRead flags) an additional clause can be used:
bool fRead = boost::is_same<Operation, CSerActionUnserialize>();
The IMPLEMENT_SERIALIZE macro will now be a freestanding clause added within
class' body (similiar to Qt's Q_OBJECT) to implement GetSerializeSize,
Serialize and Unserialize. These are now wrappers around
the "SerializationOp" template.
Luke Dashjr