This commit prunes several unused functions from the Db interface and the
underlying implementations. For the most part these are holdovers from
the original sqlite implementation. It also removes the types associated
with those functions since they are no longer needed. The following
functions and types have been removed:
- InvalidateCache
- InvalidateBlockCache
- InvalidateTxCache
- SetDBInsertMode
- InsertMode type and associated constants
- NewIterateBlocks
- BlockIterator interface
The reasons for removing these are broken out below.
- Neither of two current implementations implement these functions nor
does any of the fully functional code using the interface invoke them.
- After contemplating and testing caching of blocks and transactions at
this layer, it doesn't seem to provide any real benefit unless very
specific assumptions about the use case are made. Making those
assumptions can make other use cases worse. For example, assuming a
large cache is harmful to memory-constrained use cases. Leaving it up
to the caller to choose when to cache block and transactions allows much
greater flexibility.
- The DB insert mode was an artifact of the original sqlite implementation
and probably should have only been exposed specifically on the
implementation as opposed through generic interface. If a specific
implementation wishes to provide functionality such as special modes,
that should be done through type assertions.
This commit changes the node index creation to use block headers instead
of full blocks. This speeds up the initial node index generation since it
doesn't require loading a bunch of full blocks at startup.
This commit modifies local variables that are used for more convenient
access to a block's header to use pointers. This avoids copying the
header multiple times.
This commit switches the handleGetHeadersMsg function to make use of the
new FetchBlockHeightBySha and FetchBlockHeaderBySha functions in btcdb.
Also, while here, nuke the header copy which is no longer required due to
the recent btcwire changes.
This commit introduces two new functions to the btcdb.Db interface named
FetchBlockHeightBySha and FetchBlockHeaderBySha.
The FetchBlockHeightBySha function is useful since previously it was only
possible to get the height of block by fetching the entire block with
FetchBlockBySha and pulling the height out of the returned btcutil.Block.
The FetchBlockHeaderBySha function will ultimately make it much more
efficient to fetch block headers. Currently, due to the database design
in the ldb backend, the entire block has to be loaded anyways, so the only
current benefit is to avoid the deserialize on all of the transactions.
However, ultimately btcdb will gain a more efficient backend which can
also avoid reading all of the extra transaction data altogether.
The sqlite3 backend has been deprecated for quite some time. As a result,
it has not been updated with many of the more recent changes which means
the behavior no longer conforms to the interface contract.
This commit introduces two new functions for Blockheader named Serialize
and Deserialize. The functions provide a stable mechanism for serializing
and deserializing block headers to and from disk. The main benefit here
is deserialization of the header since typically only full blocks are
serialized to disk. Then when a header is needed, only the header portion
of the block is read and deserialized.
This commit removes the TxnCount field from the BlockHeader type and
updates the tests accordingly. Note that this change does not affect the
actual wire protocol encoding in any way.
The reason the field has been removed is it really doesn't belong there
even though the wire protocol wiki entry on the official bitcoin wiki
implies it does. The implication is an artifact from the way the
reference implementation serializes headers (MsgHeaders) messages. It
includes the transaction count, which is naturally always 0 for headers,
along with every header. However, in reality, a block header does not
include the transaction count. This can be evidenced by looking at how a
block hash is calculated. It is only up to and including the Nonce field
(a total of 80 bytes).
From an API standpoint, having the field as part of the BlockHeader type
results in several odd cases.
For example, the transaction count for MsgBlocks (the only place that
actually has a real transaction count since MsgHeaders does not) is
available by taking the len of the Transactions slice. As such, having
the extra field in the BlockHeader is really a useless field that could
potentially get out of sync and cause the encode to fail.
Another example is related to deserializing a block header from the
database in order to serve it in response to a getheaders (MsgGetheaders)
request. If a block header is assumed to have the transaction count as a
part of it, then derserializing a block header not only consumes more than
the 80 bytes that actually comprise the header as stated above, but you
then need to change the transaction count to 0 before sending the headers
(MsgHeaders) message. So, not only are you reading and deserializing more
bytes than needed, but worse, you generally have to make a copy of it so
you can change the transaction count without busting cached headers.
This is part 1 of #13.
This commit reduces the initial idle timeout before version negotiation
has happened on a new peer to 30 seconds. Previously it could take 5
minutes due to the general idle timeout.
The websocket.Server used the by websocket.Handler type automatically adds
a handshake function which prevents connections when the Origin header is
not set. Not all clients send this information and we already require
authentication headers as the auth mechanism anyways.
The websocket package assumes binary blobs if []byte is used. Since we're
using JSON-RPC for all the websocket communications, it should be text
based.
This commit changes the websocket Send/Receive and associated channels to
strings accordingly.
This removes the last notification that was being sent unsolicited.
Since it is no longer needed, the code to duplicate notifications to
all clients has been removed.
Previously the code performed a database query for every checkpoint (going
backwards) to find the latest known checkpoint on every block. This was
particularly noticabled near the beginning of the block chain when there
are still several checkpoints that haven't been reached yet.
This commit changes the logic to cache the latest known checkpoint while
keeping track of when it needs to be updated once a new later known
checkpoint has been reached.
While here, also add a log message when a checkpoint has been reached and
verified.
The previous script validation logic entailed starting up a hard-coded
number of goroutines to process the transaction scripts in parallel. In
particular, one goroutine (up to 8 max) was started per transaction in a
block and another one was started for each input script pair in the
each transaction. This resulted in 64 goroutines simultaneously running
scripts and verifying cryptographic signatures. This could easily lead to
the overall system feeling sluggish.
Further the previous design could also result in bursty behavior since the
number of inputs to a transaction as well as its complexity can vary
widely between transactions. For example, starting 2 goroutines (one to
process the transaction and one for actual script pair validation) to
verify a transaction with a single input was not desirable.
Finally, the previous design validated all transactions and inputs
regardless of a failure in one of the other scripts. This really didn't
have a big impact since it's quite rare that blocks with invalid
verifications are being processed, but it was a potential way DoS vector.
This commit changes the logic in a few ways to improve things:
- The max number of validation goroutines is now based on the number of
cores in the system
- All transaction inputs from all transactions in the block are collated
into a single list which is fed through the aforementioned validation
goroutines
- The validation CPU usage is much more consistent due to the collation of
inputs
- A validation error in any goroutine immediately stops validation of all
remaining inputs
- The errors have been improved to include context about what tx script
pair failed as opposed to showing the information as a warning
This closesconformal/btcd#59.
The RPC server was performing some of the shutdown logic in the wrong
order, that is, logging the the server has shut down, waiting for all
server goroutines to finish, and then closing a channel to notify
server goroutines to stop. These three items have been reversed to
fix a hang where goroutines currently being waited on had not shut
down because they did not receive the notification.
While here, the server waitgroup was incremented for a goroutine that
was running without it, another select statement was added to stop a
duplicate close (which never occured last commit when I added the
select statements), and the "stopping rescan" logging was moved to
debug to make the ^C shutdown logging nicer.
It was pointed out in #76 that if you arrived to the Update section
of the README without seeing the Installation section, the requirement for
Go 1.2 is easy to miss. This commit builds the requirement in the
Installation section and adds it to the Updating section as well to
hopefully make it more noticable.
The addblock utility was originally written as a quick debug tool during
initial development to populate blocks into the database. However, now
that it has been designated as the standard way to import bootstrap.dat
(and indeed block data files in general), it was lacking a few features
such as properly checking against the chain rules and known good
checkpoints.
This commit reworks and improves the utility in several ways:
- Imported blocks are now checked against the chain rules including
checkpoints to ensure they match the known good chain
- The utility now properly shuts down after processing all blocks
- Attempting to import orphan blocks (blocks which build off a block you
don't yet have in the database) returns an error
- Blocks that are already known are now skipped instead of causing an
error which means you can stop and restart the import mid-way through
without issues or start it after you've already downloaded a
portion of the chain
- The block height is no longer assumed to start at 0 which means input
files that start later in the chain will work properly so long as you
already have the chain at least up to the point of the block just before
the first one in the input file
- Improved error handling and reporting
- How often the progress display is shown is now configurable
- Statistics about how many blocks were processed, imported, and already
known are now displayed after the input file has been fully processed
This resolves comments made in #60.
The warning about a missing config file should only be shown after all
other configuration has succeeded so it's not shown when there are invalid
options specified. Also add a comment about it where its intended
placement is for the future.
