lbcwallet/chain/chain.go

363 lines
9.7 KiB
Go
Raw Normal View History

Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
/*
* Copyright (c) 2013, 2014 Conformal Systems LLC <info@conformal.com>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
package chain
import (
"errors"
"sync"
"time"
"github.com/btcsuite/btcnet"
"github.com/btcsuite/btcrpcclient"
"github.com/btcsuite/btcutil"
"github.com/btcsuite/btcwallet/keystore"
"github.com/btcsuite/btcwallet/txstore"
"github.com/btcsuite/btcwire"
"github.com/btcsuite/btcws"
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
)
type Client struct {
*btcrpcclient.Client
netParams *btcnet.Params
enqueueNotification chan interface{}
dequeueNotification chan interface{}
currentBlock chan *keystore.BlockStamp
// Notification channels regarding the state of the client. These exist
// so other components can listen in on chain activity. These are
// initialized as nil, and must be created by calling one of the Listen*
// methods.
connected chan bool
notificationLock sync.Locker
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
quit chan struct{}
wg sync.WaitGroup
started bool
quitMtx sync.Mutex
}
func NewClient(net *btcnet.Params, connect, user, pass string, certs []byte, disableTLS bool) (*Client, error) {
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
client := Client{
netParams: net,
enqueueNotification: make(chan interface{}),
dequeueNotification: make(chan interface{}),
currentBlock: make(chan *keystore.BlockStamp),
notificationLock: new(sync.Mutex),
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
quit: make(chan struct{}),
}
ntfnCallbacks := btcrpcclient.NotificationHandlers{
OnClientConnected: client.onClientConnect,
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
OnBlockConnected: client.onBlockConnected,
OnBlockDisconnected: client.onBlockDisconnected,
OnRecvTx: client.onRecvTx,
OnRedeemingTx: client.onRedeemingTx,
OnRescanFinished: client.onRescanFinished,
OnRescanProgress: client.onRescanProgress,
}
conf := btcrpcclient.ConnConfig{
Host: connect,
Endpoint: "ws",
User: user,
Pass: pass,
Certificates: certs,
DisableConnectOnNew: true,
DisableTLS: disableTLS,
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
}
c, err := btcrpcclient.New(&conf, &ntfnCallbacks)
if err != nil {
return nil, err
}
client.Client = c
return &client, nil
}
func (c *Client) Start() error {
err := c.Connect(5) // attempt connection 5 tries at most
if err != nil {
return err
}
// Verify that the server is running on the expected network.
net, err := c.GetCurrentNet()
if err != nil {
c.Disconnect()
return err
}
if net != c.netParams.Net {
c.Disconnect()
return errors.New("mismatched networks")
}
c.quitMtx.Lock()
c.started = true
c.quitMtx.Unlock()
c.wg.Add(1)
go c.handler()
return nil
}
func (c *Client) Stop() {
c.quitMtx.Lock()
defer c.quitMtx.Unlock()
select {
case <-c.quit:
default:
close(c.quit)
c.Client.Shutdown()
if !c.started {
close(c.dequeueNotification)
}
}
}
func (c *Client) WaitForShutdown() {
c.Client.WaitForShutdown()
c.wg.Wait()
}
func (c *Client) Notifications() <-chan interface{} {
return c.dequeueNotification
}
func (c *Client) BlockStamp() (*keystore.BlockStamp, error) {
select {
case bs := <-c.currentBlock:
return bs, nil
case <-c.quit:
return nil, errors.New("disconnected")
}
}
// Notification types. These are defined here and processed from from reading
// a notificationChan to avoid handling these notifications directly in
// btcrpcclient callbacks, which isn't very Go-like and doesn't allow
// blocking client calls.
type (
BlockConnected keystore.BlockStamp
BlockDisconnected keystore.BlockStamp
RecvTx struct {
Tx *btcutil.Tx // Index is guaranteed to be set.
Block *txstore.Block // nil if unmined
}
RedeemingTx struct {
Tx *btcutil.Tx // Index is guaranteed to be set.
Block *txstore.Block // nil if unmined
}
RescanProgress struct {
Hash *btcwire.ShaHash
Height int32
Time time.Time
}
RescanFinished struct {
Hash *btcwire.ShaHash
Height int32
Time time.Time
}
)
// parseBlock parses a btcws definition of the block a tx is mined it to the
// Block structure of the txstore package, and the block index. This is done
// here since btcrpcclient doesn't parse this nicely for us.
func parseBlock(block *btcws.BlockDetails) (blk *txstore.Block, idx int, err error) {
if block == nil {
return nil, btcutil.TxIndexUnknown, nil
}
blksha, err := btcwire.NewShaHashFromStr(block.Hash)
if err != nil {
return nil, btcutil.TxIndexUnknown, err
}
blk = &txstore.Block{
Height: block.Height,
Hash: *blksha,
Time: time.Unix(block.Time, 0),
}
return blk, block.Index, nil
}
func (c *Client) onClientConnect() {
log.Info("Established websocket RPC connection to btcd")
c.notifyConnected(true)
}
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
func (c *Client) onBlockConnected(hash *btcwire.ShaHash, height int32) {
c.enqueueNotification <- BlockConnected{Hash: hash, Height: height}
}
func (c *Client) onBlockDisconnected(hash *btcwire.ShaHash, height int32) {
c.enqueueNotification <- BlockDisconnected{Hash: hash, Height: height}
}
func (c *Client) onRecvTx(tx *btcutil.Tx, block *btcws.BlockDetails) {
var blk *txstore.Block
index := btcutil.TxIndexUnknown
if block != nil {
var err error
blk, index, err = parseBlock(block)
if err != nil {
// Log and drop improper notification.
log.Errorf("recvtx notification bad block: %v", err)
return
}
}
tx.SetIndex(index)
c.enqueueNotification <- RecvTx{tx, blk}
}
func (c *Client) onRedeemingTx(tx *btcutil.Tx, block *btcws.BlockDetails) {
var blk *txstore.Block
index := btcutil.TxIndexUnknown
if block != nil {
var err error
blk, index, err = parseBlock(block)
if err != nil {
// Log and drop improper notification.
log.Errorf("redeemingtx notification bad block: %v", err)
Remove account support, fix races on btcd connect. This commit is the result of several big changes being made to the wallet. In particular, the "handshake" (initial sync to the chain server) was quite racy and required proper synchronization. To make fixing this race easier, several other changes were made to the internal wallet data structures and much of the RPC server ended up being rewritten. First, all account support has been removed. The previous Account struct has been replaced with a Wallet structure, which includes a keystore for saving keys, and a txstore for storing relevant transactions. This decision has been made since it is the opinion of myself and other developers that bitcoind accounts are fundamentally broken (as accounts implemented by bitcoind support both arbitrary address groupings as well as moving balances between accounts -- these are fundamentally incompatible features), and since a BIP0032 keystore is soon planned to be implemented (at which point, "accounts" can return as HD extended keys). With the keystore handling the grouping of related keys, there is no reason have many different Account structs, and the AccountManager has been removed as well. All RPC handlers that take an account option will only work with "" (the default account) or "*" if the RPC allows specifying all accounts. Second, much of the RPC server has been cleaned up. The global variables for the RPC server and chain server client have been moved to part of the rpcServer struct, and the handlers for each RPC method that are looked up change depending on which components have been set. Passthrough requests are also no longer handled specially, but when the chain server is set, a handler to perform the passthrough will be returned if the method is not otherwise a wallet RPC. The notification system for websocket clients has also been rewritten so wallet components can send notifications through channels, rather than requiring direct access to the RPC server itself, or worse still, sending directly to a websocket client's send channel. In the future, this will enable proper registration of notifications, rather than unsolicited broadcasts to every connected websocket client (see issue #84). Finally, and the main reason why much of this cleanup was necessary, the races during intial sync with the chain server have been fixed. Previously, when the 'Handshake' was run, a rescan would occur which would perform modifications to Account data structures as notifications were received. Synchronization was provided with a single binary semaphore which serialized all access to wallet and account data. However, the Handshake itself was not able to run with this lock (or else notifications would block), and many data races would occur as both notifications were being handled. If GOMAXPROCS was ever increased beyond 1, btcwallet would always immediately crash due to invalid addresses caused by the data races on startup. To fix this, the single lock for all wallet access has been replaced with mutexes for both the keystore and txstore. Handling of btcd notifications and client requests may now occur simultaneously. GOMAXPROCS has also been set to the number of logical CPUs at the beginning of main, since with the data races fixed, there's no reason to prevent the extra parallelism gained by increasing it. Closes #78. Closes #101. Closes #110.
2014-07-09 05:17:38 +02:00
return
}
}
tx.SetIndex(index)
c.enqueueNotification <- RedeemingTx{tx, blk}
}
func (c *Client) onRescanProgress(hash *btcwire.ShaHash, height int32, blkTime time.Time) {
c.enqueueNotification <- &RescanProgress{hash, height, blkTime}
}
func (c *Client) onRescanFinished(hash *btcwire.ShaHash, height int32, blkTime time.Time) {
c.enqueueNotification <- &RescanFinished{hash, height, blkTime}
}
// handler maintains a queue of notifications and the current state (best
// block) of the chain.
func (c *Client) handler() {
hash, height, err := c.GetBestBlock()
if err != nil {
close(c.quit)
c.wg.Done()
}
bs := &keystore.BlockStamp{Hash: hash, Height: height}
// TODO: Rather than leaving this as an unbounded queue for all types of
// notifications, try dropping ones where a later enqueued notification
// can fully invalidate one waiting to be processed. For example,
// blockconnected notifications for greater block heights can remove the
// need to process earlier blockconnected notifications still waiting
// here.
var notifications []interface{}
enqueue := c.enqueueNotification
var dequeue chan interface{}
var next interface{}
out:
for {
select {
case n, ok := <-enqueue:
if !ok {
// If no notifications are queued for handling,
// the queue is finished.
if len(notifications) == 0 {
break out
}
// nil channel so no more reads can occur.
enqueue = nil
continue
}
if len(notifications) == 0 {
next = n
dequeue = c.dequeueNotification
}
notifications = append(notifications, n)
case dequeue <- next:
if n, ok := next.(BlockConnected); ok {
bs = (*keystore.BlockStamp)(&n)
}
notifications[0] = nil
notifications = notifications[1:]
if len(notifications) != 0 {
next = notifications[0]
} else {
// If no more notifications can be enqueued, the
// queue is finished.
if enqueue == nil {
break out
}
dequeue = nil
}
case c.currentBlock <- bs:
case <-c.quit:
break out
}
}
close(c.dequeueNotification)
c.wg.Done()
}
// ErrDuplicateListen is returned for any attempts to listen for the same
// notification more than once. If callers must pass along a notifiation to
// multiple places, they must broadcast it themself.
var ErrDuplicateListen = errors.New("duplicate listen")
type noopLocker struct{}
func (noopLocker) Lock() {}
func (noopLocker) Unlock() {}
// ListenConnected returns a channel that passes the current connection state
// of the client. This will be automatically sent to when the client is first
// connected, as well as the current state whenever NotifyConnected is
// forcibly called.
//
// If this is called twice, ErrDuplicateListen is returned.
func (c *Client) ListenConnected() (<-chan bool, error) {
c.notificationLock.Lock()
defer c.notificationLock.Unlock()
if c.connected != nil {
return nil, ErrDuplicateListen
}
c.connected = make(chan bool)
c.notificationLock = noopLocker{}
return c.connected, nil
}
func (c *Client) notifyConnected(connected bool) {
c.notificationLock.Lock()
if c.connected != nil {
c.connected <- connected
}
c.notificationLock.Unlock()
}
// NotifyConnected sends the channel notification for a connected or
// disconnected client. This is exported so it can be called by other
// packages which require notifying the current connection state.
//
// TODO: This shouldn't exist, but the current notification API requires it.
func (c *Client) NotifyConnected() {
connected := !c.Client.Disconnected()
c.notifyConnected(connected)
}