lbcwallet/wallet/chainntfns.go

613 lines
19 KiB
Go
Raw Normal View History

// Copyright (c) 2013-2015 The btcsuite developers
// Use of this source code is governed by an ISC
// license that can be found in the LICENSE file.
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
package wallet
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
import (
"bytes"
"fmt"
"strings"
"time"
"github.com/btcsuite/btcd/chaincfg/chainhash"
"github.com/btcsuite/btcd/txscript"
"github.com/btcsuite/btcd/wire"
"github.com/btcsuite/btcwallet/chain"
"github.com/btcsuite/btcwallet/waddrmgr"
"github.com/btcsuite/btcwallet/walletdb"
"github.com/btcsuite/btcwallet/wtxmgr"
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
)
const (
// birthdayBlockDelta is the maximum time delta allowed between our
// birthday timestamp and our birthday block's timestamp when searching
// for a better birthday block candidate (if possible).
birthdayBlockDelta = 2 * time.Hour
)
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 (w *Wallet) handleChainNotifications() {
defer w.wg.Done()
Modernize the RPC server. This is a rather monolithic commit that moves the old RPC server to its own package (rpc/legacyrpc), introduces a new RPC server using gRPC (rpc/rpcserver), and provides the ability to defer wallet loading until request at a later time by an RPC (--noinitialload). The legacy RPC server remains the default for now while the new gRPC server is not enabled by default. Enabling the new server requires setting a listen address (--experimenalrpclisten). This experimental flag is used to effectively feature gate the server until it is ready to use as a default. Both RPC servers can be run at the same time, but require binding to different listen addresses. In theory, with the legacy RPC server now living in its own package it should become much easier to unit test the handlers. This will be useful for any future changes to the package, as compatibility with Core's wallet is still desired. Type safety has also been improved in the legacy RPC server. Multiple handler types are now used for methods that do and do not require the RPC client as a dependency. This can statically help prevent nil pointer dereferences, and was very useful for catching bugs during refactoring. To synchronize the wallet loading process between the main package (the default) and through the gRPC WalletLoader service (with the --noinitialload option), as well as increasing the loose coupling of packages, a new wallet.Loader type has been added. All creating and loading of existing wallets is done through a single Loader instance, and callbacks can be attached to the instance to run after the wallet has been opened. This is how the legacy RPC server is associated with a loaded wallet, even after the wallet is loaded by a gRPC method in a completely unrelated package. Documentation for the new RPC server has been added to the rpc/documentation directory. The documentation includes a specification for the new RPC API, addresses how to make changes to the server implementation, and provides short example clients in several different languages. Some of the new RPC methods are not implementated exactly as described by the specification. These are considered bugs with the implementation, not the spec. Known bugs are commented as such.
2015-06-01 21:57:50 +02:00
chainClient, err := w.requireChainClient()
if err != nil {
log.Errorf("handleChainNotifications called without RPC client")
return
}
sync := func(w *Wallet, birthdayStamp *waddrmgr.BlockStamp) {
// At the moment there is no recourse if the rescan fails for
// some reason, however, the wallet will not be marked synced
// and many methods will error early since the wallet is known
// to be out of date.
err := w.syncWithChain(birthdayStamp)
if err != nil && !w.ShuttingDown() {
log.Warnf("Unable to synchronize wallet to chain: %v", err)
}
}
catchUpHashes := func(w *Wallet, client chain.Interface,
height int32) error {
// TODO(aakselrod): There's a race conditon here, which
// happens when a reorg occurs between the
// rescanProgress notification and the last GetBlockHash
// call. The solution when using btcd is to make btcd
// send blockconnected notifications with each block
// the way Neutrino does, and get rid of the loop. The
// other alternative is to check the final hash and,
// if it doesn't match the original hash returned by
// the notification, to roll back and restart the
// rescan.
log.Infof("Catching up block hashes to height %d, this"+
" might take a while", height)
err := walletdb.Update(w.db, func(tx walletdb.ReadWriteTx) error {
ns := tx.ReadWriteBucket(waddrmgrNamespaceKey)
startBlock := w.Manager.SyncedTo()
for i := startBlock.Height + 1; i <= height; i++ {
hash, err := client.GetBlockHash(int64(i))
if err != nil {
return err
}
header, err := chainClient.GetBlockHeader(hash)
if err != nil {
return err
}
bs := waddrmgr.BlockStamp{
Height: i,
Hash: *hash,
Timestamp: header.Timestamp,
}
err = w.Manager.SetSyncedTo(ns, &bs)
if err != nil {
return err
}
}
return nil
})
if err != nil {
log.Errorf("Failed to update address manager "+
"sync state for height %d: %v", height, err)
}
log.Info("Done catching up block hashes")
return err
}
for {
select {
case n, ok := <-chainClient.Notifications():
if !ok {
return
}
var notificationName string
var err error
switch n := n.(type) {
case chain.ClientConnected:
// Before attempting to sync with our backend,
// we'll make sure that our birthday block has
// been set correctly to potentially prevent
// missing relevant events.
birthdayStore := &walletBirthdayStore{
db: w.db,
manager: w.Manager,
}
birthdayBlock, err := birthdaySanityCheck(
chainClient, birthdayStore,
)
if err != nil && !waddrmgr.IsError(err, waddrmgr.ErrBirthdayBlockNotSet) {
err := fmt.Errorf("unable to sanity "+
"check wallet birthday block: %v",
err)
log.Error(err)
panic(err)
}
go sync(w, birthdayBlock)
case chain.BlockConnected:
err = walletdb.Update(w.db, func(tx walletdb.ReadWriteTx) error {
return w.connectBlock(tx, wtxmgr.BlockMeta(n))
})
notificationName = "blockconnected"
case chain.BlockDisconnected:
err = walletdb.Update(w.db, func(tx walletdb.ReadWriteTx) error {
return w.disconnectBlock(tx, wtxmgr.BlockMeta(n))
})
notificationName = "blockdisconnected"
case chain.RelevantTx:
err = walletdb.Update(w.db, func(tx walletdb.ReadWriteTx) error {
return w.addRelevantTx(tx, n.TxRecord, n.Block)
})
notificationName = "recvtx/redeemingtx"
case chain.FilteredBlockConnected:
// Atomically update for the whole block.
if len(n.RelevantTxs) > 0 {
err = walletdb.Update(w.db, func(
tx walletdb.ReadWriteTx) error {
var err error
for _, rec := range n.RelevantTxs {
err = w.addRelevantTx(tx, rec,
n.Block)
if err != nil {
return err
}
}
return nil
})
}
notificationName = "filteredblockconnected"
// The following require some database maintenance, but also
// need to be reported to the wallet's rescan goroutine.
case *chain.RescanProgress:
err = catchUpHashes(w, chainClient, n.Height)
notificationName = "rescanprogress"
select {
case w.rescanNotifications <- n:
case <-w.quitChan():
return
}
case *chain.RescanFinished:
err = catchUpHashes(w, chainClient, n.Height)
notificationName = "rescanprogress"
w.SetChainSynced(true)
select {
case w.rescanNotifications <- n:
case <-w.quitChan():
return
}
}
if err != nil {
// On out-of-sync blockconnected notifications, only
// send a debug message.
errStr := "Failed to process consensus server " +
"notification (name: `%s`, detail: `%v`)"
if notificationName == "blockconnected" &&
strings.Contains(err.Error(),
"couldn't get hash from database") {
log.Debugf(errStr, notificationName, err)
} else {
log.Errorf(errStr, notificationName, err)
}
}
case <-w.quit:
return
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
}
}
}
// connectBlock handles a chain server notification by marking a wallet
// that's currently in-sync with the chain server as being synced up to
// the passed block.
func (w *Wallet) connectBlock(dbtx walletdb.ReadWriteTx, b wtxmgr.BlockMeta) error {
addrmgrNs := dbtx.ReadWriteBucket(waddrmgrNamespaceKey)
bs := waddrmgr.BlockStamp{
Height: b.Height,
Hash: b.Hash,
Timestamp: b.Time,
}
err := w.Manager.SetSyncedTo(addrmgrNs, &bs)
if err != nil {
return 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
Modernize the RPC server. This is a rather monolithic commit that moves the old RPC server to its own package (rpc/legacyrpc), introduces a new RPC server using gRPC (rpc/rpcserver), and provides the ability to defer wallet loading until request at a later time by an RPC (--noinitialload). The legacy RPC server remains the default for now while the new gRPC server is not enabled by default. Enabling the new server requires setting a listen address (--experimenalrpclisten). This experimental flag is used to effectively feature gate the server until it is ready to use as a default. Both RPC servers can be run at the same time, but require binding to different listen addresses. In theory, with the legacy RPC server now living in its own package it should become much easier to unit test the handlers. This will be useful for any future changes to the package, as compatibility with Core's wallet is still desired. Type safety has also been improved in the legacy RPC server. Multiple handler types are now used for methods that do and do not require the RPC client as a dependency. This can statically help prevent nil pointer dereferences, and was very useful for catching bugs during refactoring. To synchronize the wallet loading process between the main package (the default) and through the gRPC WalletLoader service (with the --noinitialload option), as well as increasing the loose coupling of packages, a new wallet.Loader type has been added. All creating and loading of existing wallets is done through a single Loader instance, and callbacks can be attached to the instance to run after the wallet has been opened. This is how the legacy RPC server is associated with a loaded wallet, even after the wallet is loaded by a gRPC method in a completely unrelated package. Documentation for the new RPC server has been added to the rpc/documentation directory. The documentation includes a specification for the new RPC API, addresses how to make changes to the server implementation, and provides short example clients in several different languages. Some of the new RPC methods are not implementated exactly as described by the specification. These are considered bugs with the implementation, not the spec. Known bugs are commented as such.
2015-06-01 21:57:50 +02:00
// Notify interested clients of the connected block.
//
// TODO: move all notifications outside of the database transaction.
w.NtfnServer.notifyAttachedBlock(dbtx, &b)
return nil
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
}
// disconnectBlock handles a chain server reorganize by rolling back all
// block history from the reorged block for a wallet in-sync with the chain
// server.
func (w *Wallet) disconnectBlock(dbtx walletdb.ReadWriteTx, b wtxmgr.BlockMeta) error {
addrmgrNs := dbtx.ReadWriteBucket(waddrmgrNamespaceKey)
txmgrNs := dbtx.ReadWriteBucket(wtxmgrNamespaceKey)
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
if !w.ChainSynced() {
return nil
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
}
// Disconnect the removed block and all blocks after it if we know about
// the disconnected block. Otherwise, the block is in the future.
if b.Height <= w.Manager.SyncedTo().Height {
hash, err := w.Manager.BlockHash(addrmgrNs, b.Height)
if err != nil {
return err
}
if bytes.Equal(hash[:], b.Hash[:]) {
bs := waddrmgr.BlockStamp{
Height: b.Height - 1,
}
hash, err = w.Manager.BlockHash(addrmgrNs, bs.Height)
if err != nil {
return err
}
b.Hash = *hash
client := w.ChainClient()
header, err := client.GetBlockHeader(hash)
if err != nil {
return err
}
bs.Timestamp = header.Timestamp
err = w.Manager.SetSyncedTo(addrmgrNs, &bs)
if err != nil {
return err
}
err = w.TxStore.Rollback(txmgrNs, b.Height)
if err != nil {
return 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
}
}
Modernize the RPC server. This is a rather monolithic commit that moves the old RPC server to its own package (rpc/legacyrpc), introduces a new RPC server using gRPC (rpc/rpcserver), and provides the ability to defer wallet loading until request at a later time by an RPC (--noinitialload). The legacy RPC server remains the default for now while the new gRPC server is not enabled by default. Enabling the new server requires setting a listen address (--experimenalrpclisten). This experimental flag is used to effectively feature gate the server until it is ready to use as a default. Both RPC servers can be run at the same time, but require binding to different listen addresses. In theory, with the legacy RPC server now living in its own package it should become much easier to unit test the handlers. This will be useful for any future changes to the package, as compatibility with Core's wallet is still desired. Type safety has also been improved in the legacy RPC server. Multiple handler types are now used for methods that do and do not require the RPC client as a dependency. This can statically help prevent nil pointer dereferences, and was very useful for catching bugs during refactoring. To synchronize the wallet loading process between the main package (the default) and through the gRPC WalletLoader service (with the --noinitialload option), as well as increasing the loose coupling of packages, a new wallet.Loader type has been added. All creating and loading of existing wallets is done through a single Loader instance, and callbacks can be attached to the instance to run after the wallet has been opened. This is how the legacy RPC server is associated with a loaded wallet, even after the wallet is loaded by a gRPC method in a completely unrelated package. Documentation for the new RPC server has been added to the rpc/documentation directory. The documentation includes a specification for the new RPC API, addresses how to make changes to the server implementation, and provides short example clients in several different languages. Some of the new RPC methods are not implementated exactly as described by the specification. These are considered bugs with the implementation, not the spec. Known bugs are commented as such.
2015-06-01 21:57:50 +02:00
// Notify interested clients of the disconnected block.
w.NtfnServer.notifyDetachedBlock(&b.Hash)
return nil
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 (w *Wallet) addRelevantTx(dbtx walletdb.ReadWriteTx, rec *wtxmgr.TxRecord, block *wtxmgr.BlockMeta) error {
addrmgrNs := dbtx.ReadWriteBucket(waddrmgrNamespaceKey)
txmgrNs := dbtx.ReadWriteBucket(wtxmgrNamespaceKey)
2015-04-06 21:03:24 +02:00
// At the moment all notified transactions are assumed to actually be
// relevant. This assumption will not hold true when SPV support is
// added, but until then, simply insert the transaction because there
// should either be one or more relevant inputs or outputs.
err := w.TxStore.InsertTx(txmgrNs, rec, block)
2015-04-06 21:03:24 +02:00
if err != nil {
return err
}
// Check every output to determine whether it is controlled by a wallet
// key. If so, mark the output as a credit.
for i, output := range rec.MsgTx.TxOut {
_, addrs, _, err := txscript.ExtractPkScriptAddrs(output.PkScript,
w.chainParams)
2015-04-06 21:03:24 +02:00
if err != nil {
// Non-standard outputs are skipped.
continue
}
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
for _, addr := range addrs {
ma, err := w.Manager.Address(addrmgrNs, addr)
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
if err == nil {
2015-04-06 21:03:24 +02:00
// TODO: Credits should be added with the
// account they belong to, so wtxmgr is able to
// track per-account balances.
err = w.TxStore.AddCredit(txmgrNs, rec, block, uint32(i),
2015-04-06 21:03:24 +02:00
ma.Internal())
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
if err != nil {
return err
}
err = w.Manager.MarkUsed(addrmgrNs, addr)
2015-04-06 21:03:24 +02:00
if err != nil {
return err
}
log.Debugf("Marked address %v used", addr)
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
continue
}
2015-04-06 21:03:24 +02:00
// Missing addresses are skipped. Other errors should
// be propagated.
if !waddrmgr.IsError(err, waddrmgr.ErrAddressNotFound) {
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 err
}
}
}
Modernize the RPC server. This is a rather monolithic commit that moves the old RPC server to its own package (rpc/legacyrpc), introduces a new RPC server using gRPC (rpc/rpcserver), and provides the ability to defer wallet loading until request at a later time by an RPC (--noinitialload). The legacy RPC server remains the default for now while the new gRPC server is not enabled by default. Enabling the new server requires setting a listen address (--experimenalrpclisten). This experimental flag is used to effectively feature gate the server until it is ready to use as a default. Both RPC servers can be run at the same time, but require binding to different listen addresses. In theory, with the legacy RPC server now living in its own package it should become much easier to unit test the handlers. This will be useful for any future changes to the package, as compatibility with Core's wallet is still desired. Type safety has also been improved in the legacy RPC server. Multiple handler types are now used for methods that do and do not require the RPC client as a dependency. This can statically help prevent nil pointer dereferences, and was very useful for catching bugs during refactoring. To synchronize the wallet loading process between the main package (the default) and through the gRPC WalletLoader service (with the --noinitialload option), as well as increasing the loose coupling of packages, a new wallet.Loader type has been added. All creating and loading of existing wallets is done through a single Loader instance, and callbacks can be attached to the instance to run after the wallet has been opened. This is how the legacy RPC server is associated with a loaded wallet, even after the wallet is loaded by a gRPC method in a completely unrelated package. Documentation for the new RPC server has been added to the rpc/documentation directory. The documentation includes a specification for the new RPC API, addresses how to make changes to the server implementation, and provides short example clients in several different languages. Some of the new RPC methods are not implementated exactly as described by the specification. These are considered bugs with the implementation, not the spec. Known bugs are commented as such.
2015-06-01 21:57:50 +02:00
// Send notification of mined or unmined transaction to any interested
// clients.
//
// TODO: Avoid the extra db hits.
if block == nil {
details, err := w.TxStore.UniqueTxDetails(txmgrNs, &rec.Hash, nil)
Modernize the RPC server. This is a rather monolithic commit that moves the old RPC server to its own package (rpc/legacyrpc), introduces a new RPC server using gRPC (rpc/rpcserver), and provides the ability to defer wallet loading until request at a later time by an RPC (--noinitialload). The legacy RPC server remains the default for now while the new gRPC server is not enabled by default. Enabling the new server requires setting a listen address (--experimenalrpclisten). This experimental flag is used to effectively feature gate the server until it is ready to use as a default. Both RPC servers can be run at the same time, but require binding to different listen addresses. In theory, with the legacy RPC server now living in its own package it should become much easier to unit test the handlers. This will be useful for any future changes to the package, as compatibility with Core's wallet is still desired. Type safety has also been improved in the legacy RPC server. Multiple handler types are now used for methods that do and do not require the RPC client as a dependency. This can statically help prevent nil pointer dereferences, and was very useful for catching bugs during refactoring. To synchronize the wallet loading process between the main package (the default) and through the gRPC WalletLoader service (with the --noinitialload option), as well as increasing the loose coupling of packages, a new wallet.Loader type has been added. All creating and loading of existing wallets is done through a single Loader instance, and callbacks can be attached to the instance to run after the wallet has been opened. This is how the legacy RPC server is associated with a loaded wallet, even after the wallet is loaded by a gRPC method in a completely unrelated package. Documentation for the new RPC server has been added to the rpc/documentation directory. The documentation includes a specification for the new RPC API, addresses how to make changes to the server implementation, and provides short example clients in several different languages. Some of the new RPC methods are not implementated exactly as described by the specification. These are considered bugs with the implementation, not the spec. Known bugs are commented as such.
2015-06-01 21:57:50 +02:00
if err != nil {
log.Errorf("Cannot query transaction details for notification: %v", err)
}
// It's possible that the transaction was not found within the
// wallet's set of unconfirmed transactions due to it already
// being confirmed, so we'll avoid notifying it.
//
// TODO(wilmer): ideally we should find the culprit to why we're
// receiving an additional unconfirmed chain.RelevantTx
// notification from the chain backend.
if details != nil {
w.NtfnServer.notifyUnminedTransaction(dbtx, details)
Modernize the RPC server. This is a rather monolithic commit that moves the old RPC server to its own package (rpc/legacyrpc), introduces a new RPC server using gRPC (rpc/rpcserver), and provides the ability to defer wallet loading until request at a later time by an RPC (--noinitialload). The legacy RPC server remains the default for now while the new gRPC server is not enabled by default. Enabling the new server requires setting a listen address (--experimenalrpclisten). This experimental flag is used to effectively feature gate the server until it is ready to use as a default. Both RPC servers can be run at the same time, but require binding to different listen addresses. In theory, with the legacy RPC server now living in its own package it should become much easier to unit test the handlers. This will be useful for any future changes to the package, as compatibility with Core's wallet is still desired. Type safety has also been improved in the legacy RPC server. Multiple handler types are now used for methods that do and do not require the RPC client as a dependency. This can statically help prevent nil pointer dereferences, and was very useful for catching bugs during refactoring. To synchronize the wallet loading process between the main package (the default) and through the gRPC WalletLoader service (with the --noinitialload option), as well as increasing the loose coupling of packages, a new wallet.Loader type has been added. All creating and loading of existing wallets is done through a single Loader instance, and callbacks can be attached to the instance to run after the wallet has been opened. This is how the legacy RPC server is associated with a loaded wallet, even after the wallet is loaded by a gRPC method in a completely unrelated package. Documentation for the new RPC server has been added to the rpc/documentation directory. The documentation includes a specification for the new RPC API, addresses how to make changes to the server implementation, and provides short example clients in several different languages. Some of the new RPC methods are not implementated exactly as described by the specification. These are considered bugs with the implementation, not the spec. Known bugs are commented as such.
2015-06-01 21:57:50 +02:00
}
} else {
details, err := w.TxStore.UniqueTxDetails(txmgrNs, &rec.Hash, &block.Block)
Modernize the RPC server. This is a rather monolithic commit that moves the old RPC server to its own package (rpc/legacyrpc), introduces a new RPC server using gRPC (rpc/rpcserver), and provides the ability to defer wallet loading until request at a later time by an RPC (--noinitialload). The legacy RPC server remains the default for now while the new gRPC server is not enabled by default. Enabling the new server requires setting a listen address (--experimenalrpclisten). This experimental flag is used to effectively feature gate the server until it is ready to use as a default. Both RPC servers can be run at the same time, but require binding to different listen addresses. In theory, with the legacy RPC server now living in its own package it should become much easier to unit test the handlers. This will be useful for any future changes to the package, as compatibility with Core's wallet is still desired. Type safety has also been improved in the legacy RPC server. Multiple handler types are now used for methods that do and do not require the RPC client as a dependency. This can statically help prevent nil pointer dereferences, and was very useful for catching bugs during refactoring. To synchronize the wallet loading process between the main package (the default) and through the gRPC WalletLoader service (with the --noinitialload option), as well as increasing the loose coupling of packages, a new wallet.Loader type has been added. All creating and loading of existing wallets is done through a single Loader instance, and callbacks can be attached to the instance to run after the wallet has been opened. This is how the legacy RPC server is associated with a loaded wallet, even after the wallet is loaded by a gRPC method in a completely unrelated package. Documentation for the new RPC server has been added to the rpc/documentation directory. The documentation includes a specification for the new RPC API, addresses how to make changes to the server implementation, and provides short example clients in several different languages. Some of the new RPC methods are not implementated exactly as described by the specification. These are considered bugs with the implementation, not the spec. Known bugs are commented as such.
2015-06-01 21:57:50 +02:00
if err != nil {
log.Errorf("Cannot query transaction details for notification: %v", err)
}
// We'll only notify the transaction if it was found within the
// wallet's set of confirmed transactions.
if details != nil {
w.NtfnServer.notifyMinedTransaction(dbtx, details, block)
Modernize the RPC server. This is a rather monolithic commit that moves the old RPC server to its own package (rpc/legacyrpc), introduces a new RPC server using gRPC (rpc/rpcserver), and provides the ability to defer wallet loading until request at a later time by an RPC (--noinitialload). The legacy RPC server remains the default for now while the new gRPC server is not enabled by default. Enabling the new server requires setting a listen address (--experimenalrpclisten). This experimental flag is used to effectively feature gate the server until it is ready to use as a default. Both RPC servers can be run at the same time, but require binding to different listen addresses. In theory, with the legacy RPC server now living in its own package it should become much easier to unit test the handlers. This will be useful for any future changes to the package, as compatibility with Core's wallet is still desired. Type safety has also been improved in the legacy RPC server. Multiple handler types are now used for methods that do and do not require the RPC client as a dependency. This can statically help prevent nil pointer dereferences, and was very useful for catching bugs during refactoring. To synchronize the wallet loading process between the main package (the default) and through the gRPC WalletLoader service (with the --noinitialload option), as well as increasing the loose coupling of packages, a new wallet.Loader type has been added. All creating and loading of existing wallets is done through a single Loader instance, and callbacks can be attached to the instance to run after the wallet has been opened. This is how the legacy RPC server is associated with a loaded wallet, even after the wallet is loaded by a gRPC method in a completely unrelated package. Documentation for the new RPC server has been added to the rpc/documentation directory. The documentation includes a specification for the new RPC API, addresses how to make changes to the server implementation, and provides short example clients in several different languages. Some of the new RPC methods are not implementated exactly as described by the specification. These are considered bugs with the implementation, not the spec. Known bugs are commented as such.
2015-06-01 21:57:50 +02:00
}
}
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 nil
}
// chainConn is an interface that abstracts the chain connection logic required
// to perform a wallet's birthday block sanity check.
type chainConn interface {
// GetBestBlock returns the hash and height of the best block known to
// the backend.
GetBestBlock() (*chainhash.Hash, int32, error)
// GetBlockHash returns the hash of the block with the given height.
GetBlockHash(int64) (*chainhash.Hash, error)
// GetBlockHeader returns the header for the block with the given hash.
GetBlockHeader(*chainhash.Hash) (*wire.BlockHeader, error)
}
// birthdayStore is an interface that abstracts the wallet's sync-related
// information required to perform a birthday block sanity check.
type birthdayStore interface {
// Birthday returns the birthday timestamp of the wallet.
Birthday() time.Time
// BirthdayBlock returns the birthday block of the wallet. The boolean
// returned should signal whether the wallet has already verified the
// correctness of its birthday block.
BirthdayBlock() (waddrmgr.BlockStamp, bool, error)
// SetBirthdayBlock updates the birthday block of the wallet to the
// given block. The boolean can be used to signal whether this block
// should be sanity checked the next time the wallet starts.
//
// NOTE: This should also set the wallet's synced tip to reflect the new
// birthday block. This will allow the wallet to rescan from this point
// to detect any potentially missed events.
SetBirthdayBlock(waddrmgr.BlockStamp) error
}
// walletBirthdayStore is a wrapper around the wallet's database and address
// manager that satisfies the birthdayStore interface.
type walletBirthdayStore struct {
db walletdb.DB
manager *waddrmgr.Manager
}
var _ birthdayStore = (*walletBirthdayStore)(nil)
// Birthday returns the birthday timestamp of the wallet.
func (s *walletBirthdayStore) Birthday() time.Time {
return s.manager.Birthday()
}
// BirthdayBlock returns the birthday block of the wallet.
func (s *walletBirthdayStore) BirthdayBlock() (waddrmgr.BlockStamp, bool, error) {
var (
birthdayBlock waddrmgr.BlockStamp
birthdayBlockVerified bool
)
err := walletdb.View(s.db, func(tx walletdb.ReadTx) error {
var err error
ns := tx.ReadBucket(waddrmgrNamespaceKey)
birthdayBlock, birthdayBlockVerified, err = s.manager.BirthdayBlock(ns)
return err
})
return birthdayBlock, birthdayBlockVerified, err
}
// SetBirthdayBlock updates the birthday block of the wallet to the
// given block. The boolean can be used to signal whether this block
// should be sanity checked the next time the wallet starts.
//
// NOTE: This should also set the wallet's synced tip to reflect the new
// birthday block. This will allow the wallet to rescan from this point
// to detect any potentially missed events.
func (s *walletBirthdayStore) SetBirthdayBlock(block waddrmgr.BlockStamp) error {
return walletdb.Update(s.db, func(tx walletdb.ReadWriteTx) error {
ns := tx.ReadWriteBucket(waddrmgrNamespaceKey)
err := s.manager.SetBirthdayBlock(ns, block, true)
if err != nil {
return err
}
return s.manager.SetSyncedTo(ns, &block)
})
}
// birthdaySanityCheck is a helper function that ensures a birthday block
// correctly reflects the birthday timestamp within a reasonable timestamp
// delta. It's intended to be run after the wallet establishes its connection
// with the backend, but before it begins syncing. This is done as the second
// part to the wallet's address manager migration where we populate the birthday
// block to ensure we do not miss any relevant events throughout rescans.
// waddrmgr.ErrBirthdayBlockNotSet is returned if the birthday block has not
// been set yet.
func birthdaySanityCheck(chainConn chainConn,
birthdayStore birthdayStore) (*waddrmgr.BlockStamp, error) {
// We'll start by fetching our wallet's birthday timestamp and block.
birthdayTimestamp := birthdayStore.Birthday()
birthdayBlock, birthdayBlockVerified, err := birthdayStore.BirthdayBlock()
if err != nil {
return nil, err
}
// If the birthday block has already been verified to be correct, we can
// exit our sanity check to prevent potentially fetching a better
// candidate.
if birthdayBlockVerified {
log.Debugf("Birthday block has already been verified: "+
"height=%d, hash=%v", birthdayBlock.Height,
birthdayBlock.Hash)
return &birthdayBlock, nil
}
log.Debugf("Starting sanity check for the wallet's birthday block "+
"from: height=%d, hash=%v", birthdayBlock.Height,
birthdayBlock.Hash)
// Now, we'll need to determine if our block correctly reflects our
// timestamp. To do so, we'll fetch the block header and check its
// timestamp in the event that the birthday block's timestamp was not
// set (this is possible if it was set through the migration, since we
// do not store block timestamps).
candidate := birthdayBlock
header, err := chainConn.GetBlockHeader(&candidate.Hash)
if err != nil {
return nil, fmt.Errorf("unable to get header for block hash "+
"%v: %v", candidate.Hash, err)
}
candidate.Timestamp = header.Timestamp
// We'll go back a day worth of blocks in the chain until we find a
// block whose timestamp is below our birthday timestamp.
heightDelta := int32(144)
for birthdayTimestamp.Before(candidate.Timestamp) {
// If the birthday block has reached genesis, then we can exit
// our search as there exists no data before this point.
if candidate.Height == 0 {
break
}
// To prevent requesting blocks out of range, we'll use a lower
// bound of the first block in the chain.
newCandidateHeight := int64(candidate.Height - heightDelta)
if newCandidateHeight < 0 {
newCandidateHeight = 0
}
// Then, we'll fetch the current candidate's hash and header to
// determine if it is valid.
hash, err := chainConn.GetBlockHash(newCandidateHeight)
if err != nil {
return nil, fmt.Errorf("unable to get block hash for "+
"height %d: %v", candidate.Height, err)
}
header, err := chainConn.GetBlockHeader(hash)
if err != nil {
return nil, fmt.Errorf("unable to get header for "+
"block hash %v: %v", candidate.Hash, err)
}
candidate.Hash = *hash
candidate.Height = int32(newCandidateHeight)
candidate.Timestamp = header.Timestamp
log.Debugf("Checking next birthday block candidate: "+
"height=%d, hash=%v, timestamp=%v",
candidate.Height, candidate.Hash,
candidate.Timestamp)
}
// To ensure we have a reasonable birthday block, we'll make sure it
// respects our birthday timestamp and it is within a reasonable delta.
// The birthday has already been adjusted to two days in the past of the
// actual birthday, so we'll make our expected delta to be within two
// hours of it to account for the network-adjusted time and prevent
// fetching more unnecessary blocks.
_, bestHeight, err := chainConn.GetBestBlock()
if err != nil {
return nil, err
}
timestampDelta := birthdayTimestamp.Sub(candidate.Timestamp)
for timestampDelta > birthdayBlockDelta {
// We'll determine the height for our next candidate and make
// sure it is not out of range. If it is, we'll lower our height
// delta until finding a height within range.
newHeight := candidate.Height + heightDelta
if newHeight > bestHeight {
heightDelta /= 2
// If we've exhausted all of our possible options at a
// later height, then we can assume the current birthday
// block is our best estimate.
if heightDelta == 0 {
break
}
continue
}
// We'll fetch the header for the next candidate and compare its
// timestamp.
hash, err := chainConn.GetBlockHash(int64(newHeight))
if err != nil {
return nil, fmt.Errorf("unable to get block hash for "+
"height %d: %v", candidate.Height, err)
}
header, err := chainConn.GetBlockHeader(hash)
if err != nil {
return nil, fmt.Errorf("unable to get header for "+
"block hash %v: %v", hash, err)
}
log.Debugf("Checking next birthday block candidate: "+
"height=%d, hash=%v, timestamp=%v", newHeight, hash,
header.Timestamp)
// If this block has exceeded our birthday timestamp, we'll look
// for the next candidate with a lower height delta.
if birthdayTimestamp.Before(header.Timestamp) {
heightDelta /= 2
// If we've exhausted all of our possible options at a
// later height, then we can assume the current birthday
// block is our best estimate.
if heightDelta == 0 {
break
}
continue
}
// Otherwise, this is a valid candidate, so we'll check to see
// if it meets our expected timestamp delta.
candidate.Hash = *hash
candidate.Height = newHeight
candidate.Timestamp = header.Timestamp
timestampDelta = birthdayTimestamp.Sub(header.Timestamp)
}
// At this point, we've found a new, better candidate, so we'll write it
// to disk.
log.Debugf("Found a new valid wallet birthday block: height=%d, hash=%v",
candidate.Height, candidate.Hash)
if err := birthdayStore.SetBirthdayBlock(candidate); err != nil {
return nil, err
}
return &candidate, nil
}