lnd/channeldb/migration_01_to_11/migrations.go
2021-05-07 14:18:56 +02:00

939 lines
28 KiB
Go

package migration_01_to_11
import (
"bytes"
"crypto/sha256"
"encoding/binary"
"fmt"
"github.com/btcsuite/btcd/btcec"
lnwire "github.com/lightningnetwork/lnd/channeldb/migration/lnwire21"
"github.com/lightningnetwork/lnd/kvdb"
)
// MigrateNodeAndEdgeUpdateIndex is a migration function that will update the
// database from version 0 to version 1. In version 1, we add two new indexes
// (one for nodes and one for edges) to keep track of the last time a node or
// edge was updated on the network. These new indexes allow us to implement the
// new graph sync protocol added.
func MigrateNodeAndEdgeUpdateIndex(tx kvdb.RwTx) error {
// First, we'll populating the node portion of the new index. Before we
// can add new values to the index, we'll first create the new bucket
// where these items will be housed.
nodes, err := tx.CreateTopLevelBucket(nodeBucket)
if err != nil {
return fmt.Errorf("unable to create node bucket: %v", err)
}
nodeUpdateIndex, err := nodes.CreateBucketIfNotExists(
nodeUpdateIndexBucket,
)
if err != nil {
return fmt.Errorf("unable to create node update index: %v", err)
}
log.Infof("Populating new node update index bucket")
// Now that we know the bucket has been created, we'll iterate over the
// entire node bucket so we can add the (updateTime || nodePub) key
// into the node update index.
err = nodes.ForEach(func(nodePub, nodeInfo []byte) error {
if len(nodePub) != 33 {
return nil
}
log.Tracef("Adding %x to node update index", nodePub)
// The first 8 bytes of a node's serialize data is the update
// time, so we can extract that without decoding the entire
// structure.
updateTime := nodeInfo[:8]
// Now that we have the update time, we can construct the key
// to insert into the index.
var indexKey [8 + 33]byte
copy(indexKey[:8], updateTime)
copy(indexKey[8:], nodePub)
return nodeUpdateIndex.Put(indexKey[:], nil)
})
if err != nil {
return fmt.Errorf("unable to update node indexes: %v", err)
}
log.Infof("Populating new edge update index bucket")
// With the set of nodes updated, we'll now update all edges to have a
// corresponding entry in the edge update index.
edges, err := tx.CreateTopLevelBucket(edgeBucket)
if err != nil {
return fmt.Errorf("unable to create edge bucket: %v", err)
}
edgeUpdateIndex, err := edges.CreateBucketIfNotExists(
edgeUpdateIndexBucket,
)
if err != nil {
return fmt.Errorf("unable to create edge update index: %v", err)
}
// We'll now run through each edge policy in the database, and update
// the index to ensure each edge has the proper record.
err = edges.ForEach(func(edgeKey, edgePolicyBytes []byte) error {
if len(edgeKey) != 41 {
return nil
}
// Now that we know this is the proper record, we'll grab the
// channel ID (last 8 bytes of the key), and then decode the
// edge policy so we can access the update time.
chanID := edgeKey[33:]
edgePolicyReader := bytes.NewReader(edgePolicyBytes)
edgePolicy, err := deserializeChanEdgePolicy(
edgePolicyReader, nodes,
)
if err != nil {
return err
}
log.Tracef("Adding chan_id=%v to edge update index",
edgePolicy.ChannelID)
// We'll now construct the index key using the channel ID, and
// the last time it was updated: (updateTime || chanID).
var indexKey [8 + 8]byte
byteOrder.PutUint64(
indexKey[:], uint64(edgePolicy.LastUpdate.Unix()),
)
copy(indexKey[8:], chanID)
return edgeUpdateIndex.Put(indexKey[:], nil)
})
if err != nil {
return fmt.Errorf("unable to update edge indexes: %v", err)
}
log.Infof("Migration to node and edge update indexes complete!")
return nil
}
// MigrateInvoiceTimeSeries is a database migration that assigns all existing
// invoices an index in the add and/or the settle index. Additionally, all
// existing invoices will have their bytes padded out in order to encode the
// add+settle index as well as the amount paid.
func MigrateInvoiceTimeSeries(tx kvdb.RwTx) error {
invoices, err := tx.CreateTopLevelBucket(invoiceBucket)
if err != nil {
return err
}
addIndex, err := invoices.CreateBucketIfNotExists(
addIndexBucket,
)
if err != nil {
return err
}
settleIndex, err := invoices.CreateBucketIfNotExists(
settleIndexBucket,
)
if err != nil {
return err
}
log.Infof("Migrating invoice database to new time series format")
// Now that we have all the buckets we need, we'll run through each
// invoice in the database, and update it to reflect the new format
// expected post migration.
// NOTE: we store the converted invoices and put them back into the
// database after the loop, since modifying the bucket within the
// ForEach loop is not safe.
var invoicesKeys [][]byte
var invoicesValues [][]byte
err = invoices.ForEach(func(invoiceNum, invoiceBytes []byte) error {
// If this is a sub bucket, then we'll skip it.
if invoiceBytes == nil {
return nil
}
// First, we'll make a copy of the encoded invoice bytes.
invoiceBytesCopy := make([]byte, len(invoiceBytes))
copy(invoiceBytesCopy, invoiceBytes)
// With the bytes copied over, we'll append 24 additional
// bytes. We do this so we can decode the invoice under the new
// serialization format.
padding := bytes.Repeat([]byte{0}, 24)
invoiceBytesCopy = append(invoiceBytesCopy, padding...)
invoiceReader := bytes.NewReader(invoiceBytesCopy)
invoice, err := deserializeInvoiceLegacy(invoiceReader)
if err != nil {
return fmt.Errorf("unable to decode invoice: %v", err)
}
// Now that we have the fully decoded invoice, we can update
// the various indexes that we're added, and finally the
// invoice itself before re-inserting it.
// First, we'll get the new sequence in the addIndex in order
// to create the proper mapping.
nextAddSeqNo, err := addIndex.NextSequence()
if err != nil {
return err
}
var seqNoBytes [8]byte
byteOrder.PutUint64(seqNoBytes[:], nextAddSeqNo)
err = addIndex.Put(seqNoBytes[:], invoiceNum[:])
if err != nil {
return err
}
log.Tracef("Adding invoice (preimage=%x, add_index=%v) to add "+
"time series", invoice.Terms.PaymentPreimage[:],
nextAddSeqNo)
// Next, we'll check if the invoice has been settled or not. If
// so, then we'll also add it to the settle index.
var nextSettleSeqNo uint64
if invoice.Terms.State == ContractSettled {
nextSettleSeqNo, err = settleIndex.NextSequence()
if err != nil {
return err
}
var seqNoBytes [8]byte
byteOrder.PutUint64(seqNoBytes[:], nextSettleSeqNo)
err := settleIndex.Put(seqNoBytes[:], invoiceNum)
if err != nil {
return err
}
invoice.AmtPaid = invoice.Terms.Value
log.Tracef("Adding invoice (preimage=%x, "+
"settle_index=%v) to add time series",
invoice.Terms.PaymentPreimage[:],
nextSettleSeqNo)
}
// Finally, we'll update the invoice itself with the new
// indexing information as well as the amount paid if it has
// been settled or not.
invoice.AddIndex = nextAddSeqNo
invoice.SettleIndex = nextSettleSeqNo
// We've fully migrated an invoice, so we'll now update the
// invoice in-place.
var b bytes.Buffer
if err := serializeInvoiceLegacy(&b, &invoice); err != nil {
return err
}
// Save the key and value pending update for after the ForEach
// is done.
invoicesKeys = append(invoicesKeys, invoiceNum)
invoicesValues = append(invoicesValues, b.Bytes())
return nil
})
if err != nil {
return err
}
// Now put the converted invoices into the DB.
for i := range invoicesKeys {
key := invoicesKeys[i]
value := invoicesValues[i]
if err := invoices.Put(key, value); err != nil {
return err
}
}
log.Infof("Migration to invoice time series index complete!")
return nil
}
// MigrateInvoiceTimeSeriesOutgoingPayments is a follow up to the
// migrateInvoiceTimeSeries migration. As at the time of writing, the
// OutgoingPayment struct embeddeds an instance of the Invoice struct. As a
// result, we also need to migrate the internal invoice to the new format.
func MigrateInvoiceTimeSeriesOutgoingPayments(tx kvdb.RwTx) error {
payBucket := tx.ReadWriteBucket(paymentBucket)
if payBucket == nil {
return nil
}
log.Infof("Migrating invoice database to new outgoing payment format")
// We store the keys and values we want to modify since it is not safe
// to modify them directly within the ForEach loop.
var paymentKeys [][]byte
var paymentValues [][]byte
err := payBucket.ForEach(func(payID, paymentBytes []byte) error {
log.Tracef("Migrating payment %x", payID[:])
// The internal invoices for each payment only contain a
// populated contract term, and creation date, as a result,
// most of the bytes will be "empty".
// We'll calculate the end of the invoice index assuming a
// "minimal" index that's embedded within the greater
// OutgoingPayment. The breakdown is:
// 3 bytes empty var bytes, 16 bytes creation date, 16 bytes
// settled date, 32 bytes payment pre-image, 8 bytes value, 1
// byte settled.
endOfInvoiceIndex := 1 + 1 + 1 + 16 + 16 + 32 + 8 + 1
// We'll now extract the prefix of the pure invoice embedded
// within.
invoiceBytes := paymentBytes[:endOfInvoiceIndex]
// With the prefix extracted, we'll copy over the invoice, and
// also add padding for the new 24 bytes of fields, and finally
// append the remainder of the outgoing payment.
paymentCopy := make([]byte, len(invoiceBytes))
copy(paymentCopy[:], invoiceBytes)
padding := bytes.Repeat([]byte{0}, 24)
paymentCopy = append(paymentCopy, padding...)
paymentCopy = append(
paymentCopy, paymentBytes[endOfInvoiceIndex:]...,
)
// At this point, we now have the new format of the outgoing
// payments, we'll attempt to deserialize it to ensure the
// bytes are properly formatted.
paymentReader := bytes.NewReader(paymentCopy)
_, err := deserializeOutgoingPayment(paymentReader)
if err != nil {
return fmt.Errorf("unable to deserialize payment: %v", err)
}
// Now that we know the modifications was successful, we'll
// store it to our slice of keys and values, and write it back
// to disk in the new format after the ForEach loop is over.
paymentKeys = append(paymentKeys, payID)
paymentValues = append(paymentValues, paymentCopy)
return nil
})
if err != nil {
return err
}
// Finally store the updated payments to the bucket.
for i := range paymentKeys {
key := paymentKeys[i]
value := paymentValues[i]
if err := payBucket.Put(key, value); err != nil {
return err
}
}
log.Infof("Migration to outgoing payment invoices complete!")
return nil
}
// MigrateEdgePolicies is a migration function that will update the edges
// bucket. It ensure that edges with unknown policies will also have an entry
// in the bucket. After the migration, there will be two edge entries for
// every channel, regardless of whether the policies are known.
func MigrateEdgePolicies(tx kvdb.RwTx) error {
nodes := tx.ReadWriteBucket(nodeBucket)
if nodes == nil {
return nil
}
edges := tx.ReadWriteBucket(edgeBucket)
if edges == nil {
return nil
}
edgeIndex := edges.NestedReadWriteBucket(edgeIndexBucket)
if edgeIndex == nil {
return nil
}
// checkKey gets the policy from the database with a low-level call
// so that it is still possible to distinguish between unknown and
// not present.
checkKey := func(channelId uint64, keyBytes []byte) error {
var channelID [8]byte
byteOrder.PutUint64(channelID[:], channelId)
_, err := fetchChanEdgePolicy(edges,
channelID[:], keyBytes, nodes)
if err == ErrEdgeNotFound {
log.Tracef("Adding unknown edge policy present for node %x, channel %v",
keyBytes, channelId)
err := putChanEdgePolicyUnknown(edges, channelId, keyBytes)
if err != nil {
return err
}
return nil
}
return err
}
// Iterate over all channels and check both edge policies.
err := edgeIndex.ForEach(func(chanID, edgeInfoBytes []byte) error {
infoReader := bytes.NewReader(edgeInfoBytes)
edgeInfo, err := deserializeChanEdgeInfo(infoReader)
if err != nil {
return err
}
for _, key := range [][]byte{edgeInfo.NodeKey1Bytes[:],
edgeInfo.NodeKey2Bytes[:]} {
if err := checkKey(edgeInfo.ChannelID, key); err != nil {
return err
}
}
return nil
})
if err != nil {
return fmt.Errorf("unable to update edge policies: %v", err)
}
log.Infof("Migration of edge policies complete!")
return nil
}
// PaymentStatusesMigration is a database migration intended for adding payment
// statuses for each existing payment entity in bucket to be able control
// transitions of statuses and prevent cases such as double payment
func PaymentStatusesMigration(tx kvdb.RwTx) error {
// Get the bucket dedicated to storing statuses of payments,
// where a key is payment hash, value is payment status.
paymentStatuses, err := tx.CreateTopLevelBucket(paymentStatusBucket)
if err != nil {
return err
}
log.Infof("Migrating database to support payment statuses")
circuitAddKey := []byte("circuit-adds")
circuits := tx.ReadWriteBucket(circuitAddKey)
if circuits != nil {
log.Infof("Marking all known circuits with status InFlight")
err = circuits.ForEach(func(k, v []byte) error {
// Parse the first 8 bytes as the short chan ID for the
// circuit. We'll skip all short chan IDs are not
// locally initiated, which includes all non-zero short
// chan ids.
chanID := binary.BigEndian.Uint64(k[:8])
if chanID != 0 {
return nil
}
// The payment hash is the third item in the serialized
// payment circuit. The first two items are an AddRef
// (10 bytes) and the incoming circuit key (16 bytes).
const payHashOffset = 10 + 16
paymentHash := v[payHashOffset : payHashOffset+32]
return paymentStatuses.Put(
paymentHash[:], StatusInFlight.Bytes(),
)
})
if err != nil {
return err
}
}
log.Infof("Marking all existing payments with status Completed")
// Get the bucket dedicated to storing payments
bucket := tx.ReadWriteBucket(paymentBucket)
if bucket == nil {
return nil
}
// For each payment in the bucket, deserialize the payment and mark it
// as completed.
err = bucket.ForEach(func(k, v []byte) error {
// Ignores if it is sub-bucket.
if v == nil {
return nil
}
r := bytes.NewReader(v)
payment, err := deserializeOutgoingPayment(r)
if err != nil {
return err
}
// Calculate payment hash for current payment.
paymentHash := sha256.Sum256(payment.PaymentPreimage[:])
// Update status for current payment to completed. If it fails,
// the migration is aborted and the payment bucket is returned
// to its previous state.
return paymentStatuses.Put(paymentHash[:], StatusSucceeded.Bytes())
})
if err != nil {
return err
}
log.Infof("Migration of payment statuses complete!")
return nil
}
// MigratePruneEdgeUpdateIndex is a database migration that attempts to resolve
// some lingering bugs with regards to edge policies and their update index.
// Stale entries within the edge update index were not being properly pruned due
// to a miscalculation on the offset of an edge's policy last update. This
// migration also fixes the case where the public keys within edge policies were
// being serialized with an extra byte, causing an even greater error when
// attempting to perform the offset calculation described earlier.
func MigratePruneEdgeUpdateIndex(tx kvdb.RwTx) error {
// To begin the migration, we'll retrieve the update index bucket. If it
// does not exist, we have nothing left to do so we can simply exit.
edges := tx.ReadWriteBucket(edgeBucket)
if edges == nil {
return nil
}
edgeUpdateIndex := edges.NestedReadWriteBucket(edgeUpdateIndexBucket)
if edgeUpdateIndex == nil {
return nil
}
// Retrieve some buckets that will be needed later on. These should
// already exist given the assumption that the buckets above do as
// well.
edgeIndex, err := edges.CreateBucketIfNotExists(edgeIndexBucket)
if err != nil {
return fmt.Errorf("error creating edge index bucket: %s", err)
}
if edgeIndex == nil {
return fmt.Errorf("unable to create/fetch edge index " +
"bucket")
}
nodes, err := tx.CreateTopLevelBucket(nodeBucket)
if err != nil {
return fmt.Errorf("unable to make node bucket")
}
log.Info("Migrating database to properly prune edge update index")
// We'll need to properly prune all the outdated entries within the edge
// update index. To do so, we'll gather all of the existing policies
// within the graph to re-populate them later on.
var edgeKeys [][]byte
err = edges.ForEach(func(edgeKey, edgePolicyBytes []byte) error {
// All valid entries are indexed by a public key (33 bytes)
// followed by a channel ID (8 bytes), so we'll skip any entries
// with keys that do not match this.
if len(edgeKey) != 33+8 {
return nil
}
edgeKeys = append(edgeKeys, edgeKey)
return nil
})
if err != nil {
return fmt.Errorf("unable to gather existing edge policies: %v",
err)
}
log.Info("Constructing set of edge update entries to purge.")
// Build the set of keys that we will remove from the edge update index.
// This will include all keys contained within the bucket.
var updateKeysToRemove [][]byte
err = edgeUpdateIndex.ForEach(func(updKey, _ []byte) error {
updateKeysToRemove = append(updateKeysToRemove, updKey)
return nil
})
if err != nil {
return fmt.Errorf("unable to gather existing edge updates: %v",
err)
}
log.Infof("Removing %d entries from edge update index.",
len(updateKeysToRemove))
// With the set of keys contained in the edge update index constructed,
// we'll proceed in purging all of them from the index.
for _, updKey := range updateKeysToRemove {
if err := edgeUpdateIndex.Delete(updKey); err != nil {
return err
}
}
log.Infof("Repopulating edge update index with %d valid entries.",
len(edgeKeys))
// For each edge key, we'll retrieve the policy, deserialize it, and
// re-add it to the different buckets. By doing so, we'll ensure that
// all existing edge policies are serialized correctly within their
// respective buckets and that the correct entries are populated within
// the edge update index.
for _, edgeKey := range edgeKeys {
edgePolicyBytes := edges.Get(edgeKey)
// Skip any entries with unknown policies as there will not be
// any entries for them in the edge update index.
if bytes.Equal(edgePolicyBytes[:], unknownPolicy) {
continue
}
edgePolicy, err := deserializeChanEdgePolicy(
bytes.NewReader(edgePolicyBytes), nodes,
)
if err != nil {
return err
}
_, err = updateEdgePolicy(tx, edgePolicy)
if err != nil {
return err
}
}
log.Info("Migration to properly prune edge update index complete!")
return nil
}
// MigrateOptionalChannelCloseSummaryFields migrates the serialized format of
// ChannelCloseSummary to a format where optional fields' presence is indicated
// with boolean markers.
func MigrateOptionalChannelCloseSummaryFields(tx kvdb.RwTx) error {
closedChanBucket := tx.ReadWriteBucket(closedChannelBucket)
if closedChanBucket == nil {
return nil
}
log.Info("Migrating to new closed channel format...")
// We store the converted keys and values and put them back into the
// database after the loop, since modifying the bucket within the
// ForEach loop is not safe.
var closedChansKeys [][]byte
var closedChansValues [][]byte
err := closedChanBucket.ForEach(func(chanID, summary []byte) error {
r := bytes.NewReader(summary)
// Read the old (v6) format from the database.
c, err := deserializeCloseChannelSummaryV6(r)
if err != nil {
return err
}
// Serialize using the new format, and put back into the
// bucket.
var b bytes.Buffer
if err := serializeChannelCloseSummary(&b, c); err != nil {
return err
}
// Now that we know the modifications was successful, we'll
// Store the key and value to our slices, and write it back to
// disk in the new format after the ForEach loop is over.
closedChansKeys = append(closedChansKeys, chanID)
closedChansValues = append(closedChansValues, b.Bytes())
return nil
})
if err != nil {
return fmt.Errorf("unable to update closed channels: %v", err)
}
// Now put the new format back into the DB.
for i := range closedChansKeys {
key := closedChansKeys[i]
value := closedChansValues[i]
if err := closedChanBucket.Put(key, value); err != nil {
return err
}
}
log.Info("Migration to new closed channel format complete!")
return nil
}
var messageStoreBucket = []byte("message-store")
// MigrateGossipMessageStoreKeys migrates the key format for gossip messages
// found in the message store to a new one that takes into consideration the of
// the message being stored.
func MigrateGossipMessageStoreKeys(tx kvdb.RwTx) error {
// We'll start by retrieving the bucket in which these messages are
// stored within. If there isn't one, there's nothing left for us to do
// so we can avoid the migration.
messageStore := tx.ReadWriteBucket(messageStoreBucket)
if messageStore == nil {
return nil
}
log.Info("Migrating to the gossip message store new key format")
// Otherwise we'll proceed with the migration. We'll start by coalescing
// all the current messages within the store, which are indexed by the
// public key of the peer which they should be sent to, followed by the
// short channel ID of the channel for which the message belongs to. We
// should only expect to find channel announcement signatures as that
// was the only support message type previously.
msgs := make(map[[33 + 8]byte]*lnwire.AnnounceSignatures)
err := messageStore.ForEach(func(k, v []byte) error {
var msgKey [33 + 8]byte
copy(msgKey[:], k)
msg := &lnwire.AnnounceSignatures{}
if err := msg.Decode(bytes.NewReader(v), 0); err != nil {
return err
}
msgs[msgKey] = msg
return nil
})
if err != nil {
return err
}
// Then, we'll go over all of our messages, remove their previous entry,
// and add another with the new key format. Once we've done this for
// every message, we can consider the migration complete.
for oldMsgKey, msg := range msgs {
if err := messageStore.Delete(oldMsgKey[:]); err != nil {
return err
}
// Construct the new key for which we'll find this message with
// in the store. It'll be the same as the old, but we'll also
// include the message type.
var msgType [2]byte
binary.BigEndian.PutUint16(msgType[:], uint16(msg.MsgType()))
newMsgKey := append(oldMsgKey[:], msgType[:]...)
// Serialize the message with its wire encoding.
var b bytes.Buffer
if _, err := lnwire.WriteMessage(&b, msg, 0); err != nil {
return err
}
if err := messageStore.Put(newMsgKey, b.Bytes()); err != nil {
return err
}
}
log.Info("Migration to the gossip message store new key format complete!")
return nil
}
// MigrateOutgoingPayments moves the OutgoingPayments into a new bucket format
// where they all reside in a top-level bucket indexed by the payment hash. In
// this sub-bucket we store information relevant to this payment, such as the
// payment status.
//
// Since the router cannot handle resumed payments that have the status
// InFlight (we have no PaymentAttemptInfo available for pre-migration
// payments) we delete those statuses, so only Completed payments remain in the
// new bucket structure.
func MigrateOutgoingPayments(tx kvdb.RwTx) error {
log.Infof("Migrating outgoing payments to new bucket structure")
oldPayments := tx.ReadWriteBucket(paymentBucket)
// Return early if there are no payments to migrate.
if oldPayments == nil {
log.Infof("No outgoing payments found, nothing to migrate.")
return nil
}
newPayments, err := tx.CreateTopLevelBucket(paymentsRootBucket)
if err != nil {
return err
}
// Helper method to get the source pubkey. We define it such that we
// only attempt to fetch it if needed.
sourcePub := func() ([33]byte, error) {
var pub [33]byte
nodes := tx.ReadWriteBucket(nodeBucket)
if nodes == nil {
return pub, ErrGraphNotFound
}
selfPub := nodes.Get(sourceKey)
if selfPub == nil {
return pub, ErrSourceNodeNotSet
}
copy(pub[:], selfPub[:])
return pub, nil
}
err = oldPayments.ForEach(func(k, v []byte) error {
// Ignores if it is sub-bucket.
if v == nil {
return nil
}
// Read the old payment format.
r := bytes.NewReader(v)
payment, err := deserializeOutgoingPayment(r)
if err != nil {
return err
}
// Calculate payment hash from the payment preimage.
paymentHash := sha256.Sum256(payment.PaymentPreimage[:])
// Now create and add a PaymentCreationInfo to the bucket.
c := &PaymentCreationInfo{
PaymentHash: paymentHash,
Value: payment.Terms.Value,
CreationDate: payment.CreationDate,
PaymentRequest: payment.PaymentRequest,
}
var infoBuf bytes.Buffer
if err := serializePaymentCreationInfo(&infoBuf, c); err != nil {
return err
}
sourcePubKey, err := sourcePub()
if err != nil {
return err
}
// Do the same for the PaymentAttemptInfo.
totalAmt := payment.Terms.Value + payment.Fee
rt := Route{
TotalTimeLock: payment.TimeLockLength,
TotalAmount: totalAmt,
SourcePubKey: sourcePubKey,
Hops: []*Hop{},
}
for _, hop := range payment.Path {
rt.Hops = append(rt.Hops, &Hop{
PubKeyBytes: hop,
AmtToForward: totalAmt,
})
}
// Since the old format didn't store the fee for individual
// hops, we let the last hop eat the whole fee for the total to
// add up.
if len(rt.Hops) > 0 {
rt.Hops[len(rt.Hops)-1].AmtToForward = payment.Terms.Value
}
// Since we don't have the session key for old payments, we
// create a random one to be able to serialize the attempt
// info.
priv, _ := btcec.NewPrivateKey(btcec.S256())
s := &PaymentAttemptInfo{
PaymentID: 0, // unknown.
SessionKey: priv, // unknown.
Route: rt,
}
var attemptBuf bytes.Buffer
if err := serializePaymentAttemptInfoMigration9(&attemptBuf, s); err != nil {
return err
}
// Reuse the existing payment sequence number.
var seqNum [8]byte
copy(seqNum[:], k)
// Create a bucket indexed by the payment hash.
bucket, err := newPayments.CreateBucket(paymentHash[:])
// If the bucket already exists, it means that we are migrating
// from a database containing duplicate payments to a payment
// hash. To keep this information, we store such duplicate
// payments in a sub-bucket.
if err == kvdb.ErrBucketExists {
pHashBucket := newPayments.NestedReadWriteBucket(paymentHash[:])
// Create a bucket for duplicate payments within this
// payment hash's bucket.
dup, err := pHashBucket.CreateBucketIfNotExists(
paymentDuplicateBucket,
)
if err != nil {
return err
}
// Each duplicate will get its own sub-bucket within
// this bucket, so use their sequence number to index
// them by.
bucket, err = dup.CreateBucket(seqNum[:])
if err != nil {
return err
}
} else if err != nil {
return err
}
// Store the payment's information to the bucket.
err = bucket.Put(paymentSequenceKey, seqNum[:])
if err != nil {
return err
}
err = bucket.Put(paymentCreationInfoKey, infoBuf.Bytes())
if err != nil {
return err
}
err = bucket.Put(paymentAttemptInfoKey, attemptBuf.Bytes())
if err != nil {
return err
}
err = bucket.Put(paymentSettleInfoKey, payment.PaymentPreimage[:])
if err != nil {
return err
}
return nil
})
if err != nil {
return err
}
// To continue producing unique sequence numbers, we set the sequence
// of the new bucket to that of the old one.
seq := oldPayments.Sequence()
if err := newPayments.SetSequence(seq); err != nil {
return err
}
// Now we delete the old buckets. Deleting the payment status buckets
// deletes all payment statuses other than Complete.
err = tx.DeleteTopLevelBucket(paymentStatusBucket)
if err != nil && err != kvdb.ErrBucketNotFound {
return err
}
// Finally delete the old payment bucket.
err = tx.DeleteTopLevelBucket(paymentBucket)
if err != nil && err != kvdb.ErrBucketNotFound {
return err
}
log.Infof("Migration of outgoing payment bucket structure completed!")
return nil
}