2017-07-30 20:54:51 +02:00
|
|
|
package lnwallet
|
|
|
|
|
|
|
|
import (
|
|
|
|
"fmt"
|
|
|
|
"sync"
|
|
|
|
"sync/atomic"
|
|
|
|
|
2018-06-05 03:34:16 +02:00
|
|
|
"github.com/btcsuite/btcd/btcec"
|
|
|
|
"github.com/btcsuite/btcd/wire"
|
2019-01-16 15:47:43 +01:00
|
|
|
"github.com/lightningnetwork/lnd/input"
|
2018-07-31 09:17:17 +02:00
|
|
|
"github.com/lightningnetwork/lnd/lnwire"
|
2017-07-30 20:54:51 +02:00
|
|
|
)
|
|
|
|
|
|
|
|
const (
|
|
|
|
// jobBuffer is a constant the represents the buffer of jobs in the two
|
|
|
|
// main queues. This allows clients avoid necessarily blocking when
|
|
|
|
// submitting jobs into the queue.
|
|
|
|
jobBuffer = 100
|
|
|
|
|
|
|
|
// TODO(roasbeef): job buffer pool?
|
|
|
|
)
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// VerifyJob is a job sent to the sigPool sig pool to verify a signature
|
|
|
|
// on a transaction. The items contained in the struct are necessary and
|
|
|
|
// sufficient to verify the full signature. The passed sigHash closure function
|
|
|
|
// should be set to a function that generates the relevant sighash.
|
2017-07-30 20:54:51 +02:00
|
|
|
//
|
|
|
|
// TODO(roasbeef): when we move to ecschnorr, make into batch signature
|
|
|
|
// verification using bos-coster (or pip?).
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
type VerifyJob struct {
|
|
|
|
// PubKey is the public key that was used to generate the purported
|
2017-07-30 20:54:51 +02:00
|
|
|
// valid signature. Note that with the current channel construction,
|
|
|
|
// this public key will likely have been tweaked using the current per
|
|
|
|
// commitment point for a particular commitment transactions.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
PubKey *btcec.PublicKey
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// Sig is the raw signature generated using the above public key. This
|
2017-07-30 20:54:51 +02:00
|
|
|
// is the signature to be verified.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
Sig *btcec.Signature
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// SigHash is a function closure generates the sighashes that the
|
2017-07-30 20:54:51 +02:00
|
|
|
// passed signature is known to have signed.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
SigHash func() ([]byte, error)
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// HtlcIndex is the index of the HTLC from the PoV of the remote
|
2018-04-05 02:41:05 +02:00
|
|
|
// party's update log.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
HtlcIndex uint64
|
2018-04-05 02:41:05 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// Cancel is a channel that should be closed if the caller wishes to
|
2017-07-30 20:54:51 +02:00
|
|
|
// cancel all pending verification jobs part of a single batch. This
|
|
|
|
// channel is to be closed in the case that a single signature in a
|
|
|
|
// batch has been returned as invalid, as there is no need to verify
|
|
|
|
// the remainder of the signatures.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
Cancel chan struct{}
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// ErrResp is the channel that the result of the signature verification
|
2017-07-30 20:54:51 +02:00
|
|
|
// is to be sent over. In the see that the signature is valid, a nil
|
|
|
|
// error will be passed. Otherwise, a concrete error detailing the
|
|
|
|
// issue will be passed.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
ErrResp chan *HtlcIndexErr
|
2018-04-05 02:41:05 +02:00
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// HtlcIndexErr is a special type of error that also includes a pointer to the
|
|
|
|
// original validation job. This error message allows us to craft more detailed
|
2018-04-05 02:41:05 +02:00
|
|
|
// errors at upper layers.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
type HtlcIndexErr struct {
|
2018-04-05 02:41:05 +02:00
|
|
|
error
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
*VerifyJob
|
2017-07-30 20:54:51 +02:00
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// SignJob is a job sent to the sigPool sig pool to generate a valid
|
|
|
|
// signature according to the passed SignDescriptor for the passed transaction.
|
|
|
|
// Jobs are intended to be sent in batches in order to parallelize the job of
|
|
|
|
// generating signatures for a new commitment transaction.
|
|
|
|
type SignJob struct {
|
|
|
|
// SignDesc is intended to be a full populated SignDescriptor which
|
2017-07-30 20:54:51 +02:00
|
|
|
// encodes the necessary material (keys, witness script, etc) required
|
|
|
|
// to generate a valid signature for the specified input.
|
2019-01-16 15:47:43 +01:00
|
|
|
SignDesc input.SignDescriptor
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// Tx is the transaction to be signed. This is required to generate the
|
2017-07-30 20:54:51 +02:00
|
|
|
// proper sighash for the input to be signed.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
Tx *wire.MsgTx
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// OutputIndex is the output index of the HTLC on the commitment
|
2018-04-05 02:41:05 +02:00
|
|
|
// transaction being signed.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
OutputIndex int32
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// Cancel is a channel that should be closed if the caller wishes to
|
2017-07-30 20:54:51 +02:00
|
|
|
// abandon all pending sign jobs part of a single batch.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
Cancel chan struct{}
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// Resp is the channel that the response to this particular SignJob
|
2017-07-30 20:54:51 +02:00
|
|
|
// will be sent over.
|
|
|
|
//
|
|
|
|
// TODO(roasbeef): actually need to allow caller to set, need to retain
|
|
|
|
// order mark commit sig as special
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
Resp chan SignJobResp
|
2017-07-30 20:54:51 +02:00
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// SignJobResp is the response to a sign job. Both channels are to be read in
|
2017-07-30 20:54:51 +02:00
|
|
|
// order to ensure no unnecessary goroutine blocking occurs. Additionally, both
|
|
|
|
// channels should be buffered.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
type SignJobResp struct {
|
|
|
|
// Sig is the generated signature for a particular SignJob In the case
|
2017-07-30 20:54:51 +02:00
|
|
|
// of an error during signature generation, then this value sent will
|
|
|
|
// be nil.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
Sig lnwire.Sig
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// Err is the error that occurred when executing the specified
|
2017-07-30 20:54:51 +02:00
|
|
|
// signature job. In the case that no error occurred, this value will
|
|
|
|
// be nil.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
Err error
|
2017-07-30 20:54:51 +02:00
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// TODO(roasbeef); fix description
|
|
|
|
|
|
|
|
// SigPool is a struct that is meant to allow the current channel state
|
|
|
|
// machine to parallelize all signature generation and verification. This
|
|
|
|
// struct is needed as _each_ HTLC when creating a commitment transaction
|
|
|
|
// requires a signature, and similarly a receiver of a new commitment must
|
|
|
|
// verify all the HTLC signatures included within the CommitSig message. A pool
|
|
|
|
// of workers will be maintained by the sigPool. Batches of jobs (either
|
|
|
|
// to sign or verify) can be sent to the pool of workers which will
|
|
|
|
// asynchronously perform the specified job.
|
|
|
|
type SigPool struct {
|
2018-06-01 00:41:41 +02:00
|
|
|
started uint32 // To be used atomically.
|
|
|
|
stopped uint32 // To be used atomically.
|
2017-07-30 20:54:51 +02:00
|
|
|
|
2019-01-16 15:47:43 +01:00
|
|
|
signer input.Signer
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
verifyJobs chan VerifyJob
|
|
|
|
signJobs chan SignJob
|
2017-07-30 20:54:51 +02:00
|
|
|
|
|
|
|
wg sync.WaitGroup
|
|
|
|
quit chan struct{}
|
|
|
|
|
|
|
|
numWorkers int
|
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// NewSigPool creates a new signature pool with the specified number of
|
2017-07-30 20:54:51 +02:00
|
|
|
// workers. The recommended parameter for the number of works is the number of
|
|
|
|
// physical CPU cores available on the target machine.
|
2019-01-16 15:47:43 +01:00
|
|
|
func NewSigPool(numWorkers int, signer input.Signer) *SigPool {
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
return &SigPool{
|
2017-07-30 20:54:51 +02:00
|
|
|
signer: signer,
|
|
|
|
numWorkers: numWorkers,
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
verifyJobs: make(chan VerifyJob, jobBuffer),
|
|
|
|
signJobs: make(chan SignJob, jobBuffer),
|
2017-07-30 20:54:51 +02:00
|
|
|
quit: make(chan struct{}),
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// Start starts of all goroutines that the sigPool sig pool needs to
|
|
|
|
// carry out its duties.
|
|
|
|
func (s *SigPool) Start() error {
|
2017-07-30 20:54:51 +02:00
|
|
|
if !atomic.CompareAndSwapUint32(&s.started, 0, 1) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
for i := 0; i < s.numWorkers; i++ {
|
|
|
|
s.wg.Add(1)
|
|
|
|
go s.poolWorker()
|
|
|
|
}
|
|
|
|
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// Stop signals any active workers carrying out jobs to exit so the sigPool can
|
|
|
|
// gracefully shutdown.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
func (s *SigPool) Stop() error {
|
2017-07-30 20:54:51 +02:00
|
|
|
if !atomic.CompareAndSwapUint32(&s.stopped, 0, 1) {
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
close(s.quit)
|
|
|
|
s.wg.Wait()
|
|
|
|
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// poolWorker is the main worker goroutine within the sigPool sig pool.
|
|
|
|
// Individual batches are distributed amongst each of the active workers. The
|
|
|
|
// workers then execute the task based on the type of job, and return the
|
|
|
|
// result back to caller.
|
|
|
|
func (s *SigPool) poolWorker() {
|
2017-07-30 20:54:51 +02:00
|
|
|
defer s.wg.Done()
|
|
|
|
|
|
|
|
for {
|
|
|
|
select {
|
|
|
|
|
|
|
|
// We've just received a new signature job. Given the items
|
|
|
|
// contained within the message, we'll craft a signature and
|
|
|
|
// send the result along with a possible error back to the
|
|
|
|
// caller.
|
|
|
|
case sigMsg := <-s.signJobs:
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
rawSig, err := s.signer.SignOutputRaw(
|
|
|
|
sigMsg.Tx, &sigMsg.SignDesc,
|
|
|
|
)
|
2017-07-30 20:54:51 +02:00
|
|
|
if err != nil {
|
|
|
|
select {
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case sigMsg.Resp <- SignJobResp{
|
|
|
|
Sig: lnwire.Sig{},
|
|
|
|
Err: err,
|
2017-07-30 20:54:51 +02:00
|
|
|
}:
|
|
|
|
continue
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case <-sigMsg.Cancel:
|
2017-07-30 20:54:51 +02:00
|
|
|
continue
|
2017-11-11 23:59:49 +01:00
|
|
|
case <-s.quit:
|
|
|
|
return
|
2017-07-30 20:54:51 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-01-31 04:55:39 +01:00
|
|
|
sig, err := lnwire.NewSigFromRawSignature(rawSig)
|
2017-07-30 20:54:51 +02:00
|
|
|
select {
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case sigMsg.Resp <- SignJobResp{
|
|
|
|
Sig: sig,
|
|
|
|
Err: err,
|
2017-07-30 20:54:51 +02:00
|
|
|
}:
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case <-sigMsg.Cancel:
|
2017-07-30 20:54:51 +02:00
|
|
|
continue
|
2017-11-11 23:59:49 +01:00
|
|
|
case <-s.quit:
|
|
|
|
return
|
2017-07-30 20:54:51 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// We've just received a new verification job from the outside
|
|
|
|
// world. We'll attempt to construct the sighash, parse the
|
|
|
|
// signature, and finally verify the signature.
|
|
|
|
case verifyMsg := <-s.verifyJobs:
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
sigHash, err := verifyMsg.SigHash()
|
2017-07-30 20:54:51 +02:00
|
|
|
if err != nil {
|
|
|
|
select {
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case verifyMsg.ErrResp <- &HtlcIndexErr{
|
2018-04-05 02:41:05 +02:00
|
|
|
error: err,
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
VerifyJob: &verifyMsg,
|
2018-04-05 02:41:05 +02:00
|
|
|
}:
|
2017-07-30 20:54:51 +02:00
|
|
|
continue
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case <-verifyMsg.Cancel:
|
2017-07-30 20:54:51 +02:00
|
|
|
continue
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
rawSig := verifyMsg.Sig
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
if !rawSig.Verify(sigHash, verifyMsg.PubKey) {
|
2017-07-30 20:54:51 +02:00
|
|
|
err := fmt.Errorf("invalid signature "+
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
"sighash: %x, sig: %x", sigHash,
|
|
|
|
rawSig.Serialize())
|
|
|
|
|
2017-07-30 20:54:51 +02:00
|
|
|
select {
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case verifyMsg.ErrResp <- &HtlcIndexErr{
|
2018-04-05 02:41:05 +02:00
|
|
|
error: err,
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
VerifyJob: &verifyMsg,
|
2018-04-05 02:41:05 +02:00
|
|
|
}:
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case <-verifyMsg.Cancel:
|
2017-11-11 23:59:49 +01:00
|
|
|
case <-s.quit:
|
|
|
|
return
|
2017-07-30 20:54:51 +02:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
select {
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case verifyMsg.ErrResp <- nil:
|
|
|
|
case <-verifyMsg.Cancel:
|
2017-11-11 23:59:49 +01:00
|
|
|
case <-s.quit:
|
|
|
|
return
|
2017-07-30 20:54:51 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// The sigPool sig pool is exiting, so we will as well.
|
2017-07-30 20:54:51 +02:00
|
|
|
case <-s.quit:
|
|
|
|
return
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
// SubmitSignBatch submits a batch of signature jobs to the sigPool. The
|
|
|
|
// response and cancel channels for each of the SignJob's are expected to be
|
|
|
|
// fully populated, as the response for each job will be sent over the
|
|
|
|
// response channel within the job itself.
|
|
|
|
func (s *SigPool) SubmitSignBatch(signJobs []SignJob) {
|
2017-07-30 20:54:51 +02:00
|
|
|
for _, job := range signJobs {
|
|
|
|
select {
|
|
|
|
case s.signJobs <- job:
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case <-job.Cancel:
|
2017-07-30 20:54:51 +02:00
|
|
|
// TODO(roasbeef): return error?
|
2018-02-01 15:32:57 +01:00
|
|
|
case <-s.quit:
|
|
|
|
return
|
2017-07-30 20:54:51 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// SubmitVerifyBatch submits a batch of verification jobs to the sigPool. For
|
|
|
|
// each job submitted, an error will be passed into the returned channel
|
|
|
|
// denoting if signature verification was valid or not. The passed cancelChan
|
|
|
|
// allows the caller to cancel all pending jobs in the case that they wish to
|
|
|
|
// bail early.
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
func (s *SigPool) SubmitVerifyBatch(verifyJobs []VerifyJob,
|
|
|
|
cancelChan chan struct{}) <-chan *HtlcIndexErr {
|
2017-07-30 20:54:51 +02:00
|
|
|
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
errChan := make(chan *HtlcIndexErr, len(verifyJobs))
|
2017-07-30 20:54:51 +02:00
|
|
|
|
|
|
|
for _, job := range verifyJobs {
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
job.Cancel = cancelChan
|
|
|
|
job.ErrResp = errChan
|
2017-07-30 20:54:51 +02:00
|
|
|
|
|
|
|
select {
|
|
|
|
case s.verifyJobs <- job:
|
multi: replace per channel sigPool with global daemon level sigPool
In this commit, we remove the per channel `sigPool` within the
`lnwallet.LightningChannel` struct. With this change, we ensure that as
the number of channels grows, the number of gouroutines idling in the
sigPool stays constant. It's the case that currently on the daemon, most
channels are likely inactive, with only a hand full actually
consistently carrying out channel updates. As a result, this change
should reduce the amount of idle CPU usage, as we have less active
goroutines in select loops.
In order to make this change, the `SigPool` itself has been publicly
exported such that outside callers can make a `SigPool` and pass it into
newly created channels. Since the sig pool now lives outside the
channel, we were also able to do away with the Stop() method on the
channel all together.
Finally, the server is the sub-system that is currently responsible for
managing the `SigPool` within lnd.
2018-12-15 01:35:07 +01:00
|
|
|
case <-job.Cancel:
|
2017-07-30 20:54:51 +02:00
|
|
|
return errChan
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return errChan
|
|
|
|
}
|