mirrors/lightning-bolts
1
0
Fork 0
mirror of https://github.com/lightning/bolts.git synced 2024-11-19 01:50:03 +01:00
Code Issues Projects Releases Wiki Activity
46fcc15756
lightning-bolts/05-onchain.md

29 KiB
Raw Blame History

BOLT #5: Recommendations for On-chain Transaction Handling

Abstract

Lightning allows for two parties (nodes A and B) to conduct transactions off-chain by giving each of them a cross-signed commitment transaction, which describes the current state of the channel (basically, the current balance). This commitment transaction is updated every time a new payment is made and is spendable at all times.

There are three ways a channel can end:

  1. The good way (mutual close): at some point nodes A and B agree to close the channel. They generate a closing transaction (which is similar to a commitment transaction, but without any pending payments) and publish it on the blockchain (see BOLT #2: Channel Close).
  2. The bad way (unilateral close): something goes wrong, possibly without evil intent on either side. Perhaps one party crashed, for instance. One side publishes its latest commitment transaction.
  3. The ugly way (revoked transaction close): one of the parties deliberately tries to cheat, by publishing an outdated commitment transaction (presumably, a prior version, which is more in its favor).

Because Lightning is designed to be trustless, there is no risk of loss of funds in any of these three cases; provided that the situation is properly handled. The goal of this document is to explain exactly how a node should react when it encounters any of the above situations, on-chain.

Table of Contents

General Nomenclature

Any unspent output is considered to be unresolved and can be resolved as detailed in this document. Usually this is accomplished by spending it with another resolving transaction. Although, sometimes simply noting the output for later wallet spending is sufficient, in which case the transaction containing the output is considered to be its own resolving transaction.

Outputs that are resolved are considered irrevocably resolved once the remote's resolving transaction is included in a block at least 100 deep, on the most-work blockchain. 100 blocks is far greater than the longest known Bitcoin fork and is the same wait time used for confirmations of miners' rewards (see Reference Implementation).

Requirements

A node:

  • once it has broadcast a funding transaction OR sent a commitment signature for a commitment transaction that contains an HTLC output:
    • until all outputs are irrevocably resolved:
      • MUST monitor the blockchain for transactions that spend any output that is NOT irrevocably resolved.
  • MUST resolve all outputs, as specified below.
  • MUST be prepared to resolve outputs multiple times, in case of blockchain reorganizations.
  • upon the funding transaction being spent, if the channel is NOT already closed:
    • SHOULD fail the channel.
    • MAY send a descriptive error packet.
  • SHOULD ignore invalid transactions.

Rationale

Once a local node has some funds at stake, monitoring the blockchain is required to ensure the remote node does not close unilaterally.

Invalid transactions (e.g. bad signatures) can be generated by anyone, (and will be ignored by the blockchain anyway), so they should not trigger any action.

Commitment Transaction

Nodes A and B each hold a commitment transaction. Each of these commitment transactions has four types of outputs:

  1. A's main output: Zero or one output, to pay to A's commitment key.
  2. B's main output: Zero or one output, to pay to B's commitment key.
  3. A's offered HTLCs: Zero or more pending payments (HTLCs), to pay B in return for a payment preimage.
  4. B's offered HTLCs: Zero or more pending payments (HTLCs), to pay A in return for a payment preimage.

To incentivize nodes A and B to cooperate, an OP_CHECKSEQUENCEVERIFY relative timeout encumbers node A's outputs (in A's commitment transaction) and node B's outputs (in B's commitment transaction). So for example, if node A publishes its commitment transaction, it will have to wait to claim its own funds, whereas node B will have immediate access to its own funds. As a consequence, the two commitment transactions are not identical, but they are (usually) symmetrical.

See BOLT #3: Commitment Transaction for more details.

Failing a Channel

Although closing a channel can be accomplished in several ways, the most efficient is preferred.

Various error cases involve closing a channel. The requirements for sending error messages to peers are specified in BOLT #1: The error Message.

Requirements

A node:

  • if a local commitment transaction has NOT ever contained a to_local or HTLC output:
    • MAY simply forget the channel.
  • otherwise:
    • if the current commitment transaction does NOT contain to_local or other HTLC outputs:
      • MAY simply wait for the remote node to close the channel.
      • until the remote node closes:
        • MUST NOT forget the channel.
    • otherwise:
      • if it has received a valid closing_signed message that includes a sufficient fee:
        • SHOULD use this fee to perform a mutual close.
      • otherwise:
        • MUST use the last commitment transaction, for which it has a signature, to perform a unilateral close.

Rationale

Since dust_limit_satoshis is supposed to prevent creation of uneconomic outputs (which would otherwise remain forever, unspent on the blockchain), all commitment transaction outputs MUST be spent.

In the early stages of a channel, it's common for one side to have little or no funds in the channel; in this case, having nothing at stake, a node need not consume resources monitoring the channel state.

There exists a bias towards preferring mutual closes over unilateral closes, because outputs of the former are unencumbered by a delay and are directly spendable by wallets. In addition, mutual close fees tend to be less exaggerated than those of commitment transactions. So, the only reason not to use the signature from closing_signed would be if the fee offered was too small for it to be processed.

Mutual Close Handling

A mutual close transaction resolves the funding transaction output.

In the case of a mutual close, a node need not do anything else, as it has already agreed to the output, which is sent to its specified scriptpubkey (see BOLT #2: Closing initiation: shutdown).

Unilateral Close Handling: Local Commitment Transaction

This is the first of two cases involving unilateral closes. In this case, a node discovers its local commitment transaction, which resolves the funding transaction output.

However, a node cannot claim funds from the outputs of a unilateral close that it initiated, until the OP_CHECKSEQUENCEVERIFY delay has passed (as specified by the remote node's to_self_delay field). Where relevant, this situation is noted below.

Requirements

A node:

  • upon discovering its local commitment transaction:
    • SHOULD spend the to_local output to a convenient address.
    • MUST wait until the OP_CHECKSEQUENCEVERIFY delay has passed (as specified by the remote node's to_self_delay field) before spending the output.
      • Note: if the output is spent (as recommended), the output is resolved by the spending transaction, otherwise it is considered resolved by the commitment transaction itself.
    • MAY ignore the to_remote output.
      • Note: No action is required by the local node, as to_remote is considered resolved by the commitment transaction itself.
    • MUST handle HTLCs offered by itself as specified in HTLC Output Handling: Local Commitment, Local Offers.
    • MUST handle HTLCs offered by the remote node as specified in HTLC Output Handling: Local Commitment, Remote Offers.

Rationale

Spending the to_local output avoids having to remember the complicated witness script, associated with that particular channel, for later spending.

The to_remote output is entirely the business of the remote node, and can be ignored.

HTLC Output Handling: Local Commitment, Local Offers

Each HTLC output can only be spent by either a local offerer, by using the HTLC-timeout transaction after it's timed out, or a remote recipient, if it has the payment preimage.

There can be HTLCs which are not represented by an output: either because they were trimmed as dust, or because the transaction has only been partially committed.

The HTLC has timed out once the depth of the latest block is equal to or greater than the HTLC cltv_expiry.

Requirements

A node:

  • if the commitment transaction HTLC output is spent using the payment preimage, the output is considered irrevocably resolved:
    • MUST extract the payment preimage from the transaction input witness.
  • if the commitment transaction HTLC output has timed out and hasn't been resolved:
    • MUST resolve the output by spending it using the HTLC-timeout transaction.
    • once the resolving transaction has reached reasonable depth:
      • MUST fail the corresponding incoming HTLC (if any).
      • MUST resolve the output of that HTLC-timeout transaction.
      • SHOULD resolve the HTLC-timeout transaction by spending it to a convenient address.
        • Note: if the output is spent (as recommended), the output is resolved by the spending transaction, otherwise it is considered resolved by the commitment transaction itself.
      • MUST wait until the OP_CHECKSEQUENCEVERIFY delay has passed (as specified by the remote node's open_channel to_self_delay field) before spending that HTLC-timeout output.
  • for any committed HTLC that does NOT have an output in this commitment transaction:
    • once the commitment transaction has reached reasonable depth:
      • MUST fail the corresponding incoming HTLC (if any).
    • if no valid commitment transaction contains an output corresponding to the HTLC.
      • MAY fail the corresponding incoming HTLC sooner.

Rationale

The payment preimage either serves to prove payment (when the offering node originated the payment) or to redeem the corresponding incoming HTLC from another peer (when the offering node is forwarding the payment). Once a node has extracted the payment, it no longer cares about the fate of the HTLC-spending transaction itself.

In cases where both resolutions are possible (e.g. when a node receives payment success after timeout), either interpretation is acceptable; it is the responsibility of the recipient to spend it before this occurs.

The local HTLC-timeout transaction needs to be used to time out the HTLC to prevent the remote node fulfilling it and claiming the funds, and before we can back-fail any corresponding incoming HTLC, using update_fail_htlc (presumably with reason permanent_channel_failure) as detailed in BOLT 02. If the incoming HTLC is on-chain too, we simply wait for it to timeout: there's no way to signal early failure.

If an HTLC is too small to appear in any commitment transaction, it can be safely failed immediately. Otherwise, if a HTLC isn't in local commitment transaction a node needs to make sure that a blockchain reorganization or race does not switch to a commitment transaction which does contain it before the node fails it, hence the wait. The requirement that the incoming HTLC be failed before its own timeout still applies as an upper bound.

HTLC Output Handling: Local Commitment, Remote Offers

Each HTLC output can only be spent by us, the recipient, using the HTLC-success transaction, which we can only populate if we have the payment preimage. If we don't have the preimage (and don't discover it), it's the offerer's responsibility to spend the HTLC output once it's timed out.

There are actually several possible cases for an offered HTLC:

  1. The offerer is not irrevocably committed to it. The recipient won't normally know the preimage here, since it won't forward HTLCs until they're fully committed. So using the preimage would reveal that this recipient is the final hop, so it's best to allow the HTLC to time out in this case.
  2. The offerer is irrevocably committed to the offered HTLC, but the recipient hasn't yet committed to an outgoing HTLC. In this case the recipient can either forward or timeout.
  3. The recipient has committed to an outgoing HTLC for the offered one. In this case the recipient has to use the preimage if it receives it from the outgoing HTLC, otherwise it will lose funds by making an outgoing payment without redeeming the incoming one.

Requirements

If the node receives (or already knows) a payment preimage for an unresolved HTLC output it was offered for which it has committed to an outgoing HTLC, it MUST resolve the output by spending it using the HTLC-success transaction, and MUST resolve the output of that HTLC-success transaction. Otherwise, if the remote node is not irrevocably committed to the HTLC, it MUST NOT resolve the output by spending it.

A node SHOULD resolve that HTLC-success transaction output by spending it to a convenient address. If the output is spent (as recommended), the output is resolved by the spending transaction, otherwise it is considered resolved by the commitment transaction itself.

A node MUST wait until the OP_CHECKSEQUENCEVERIFY delay has passed (as specified by the remote node's open_channel to_self_delay field) before spending that HTLC-success transaction output.

If not otherwise resolved, once the HTLC output has expired, it is considered irrevocably resolved.

Unilateral Close Handling: Remote Commitment Transaction

The remote node's commitment transaction resolves the funding transaction output.

There are no delays constraining node behavior in this case, so it's simpler for a node to handle than the case in which it discovers its local commitment transaction.

Requirements

When a node discovers a commitment transaction from the remote node:

  1. to_remote: No action is required; this is a simple P2WPKH output to us. This output is considered resolved by the commitment transaction itself.
  2. to_local: No action required; this is a payment to them. This output is considered resolved by the commitment transaction.
  3. HTLCs offered by the local node: See "HTLC Output Handling: Remote Commitment, Local Offers" below.
  4. HTLCs offered by the remote node: See "HTLC Output Handling: Remote Commitment, Remote Offers" below.

A node MUST handle the broadcast of any valid commitment transaction from the remote node in this way; if it is unable to do so it MUST warn about lost funds.

Rationale

Note that there can be more than one valid, unrevoked commitment transaction after a signature has been received via commitment_signed and before the corresponding revoke_and_ack. Either commitment can serve as the remote node's commitment transaction, hence the requirement to handle both.

In the case of data loss, a node can reach a state where it doesn't recognize all of the remote node's commitment transaction HTLC outputs. It can tell this has happened because the commitment number will be greater than expected, and because it has signed the transaction. If both nodes support option-data-loss-protect the node will know the peer's per_commitment_point and thus be able to derive its own remotekey for the transaction and salvage its own funds (but not the HTLCs).

HTLC Output Handling: Remote Commitment, Local Offers

Each HTLC output can only be spent by us, the offerer, after it's timed out, or by them, the recipient, if they have the payment preimage.

The HTLC output has timed out once the depth of the latest block is equal or greater than the HTLC cltv_expiry.

There can be HTLCs which are not represented by an output: either because they were trimmed as dust, or in the case where the remote node has two valid commitment transactions, and the HTLCs differ in each.

Requirements

If the commitment transaction HTLC output is spent using the payment preimage, the output is considered irrevocably resolved, and the node MUST extract the payment preimage from the HTLC-success transaction input witness.

If the commitment transaction HTLC output has timed out and not been resolved, the node MUST resolve the output by spending it to a convenient address.

For any committed HTLC that does not have an output in this commitment transaction, the node MUST fail the corresponding incoming HTLC (if any) once the commitment transaction has reached reasonable depth, and MAY fail it sooner if no valid commitment transaction contains an output corresponding to the HTLC.

Rationale

If the commitment transaction belongs to the remote node, the only way to spend the HTLC output using a payment preimage is for them to use the HTLC-success transaction.

The payment preimage either serves to prove payment (when the offering node originated the payment) or to redeem the corresponding incoming HTLC from another peer (when the offering node is forwarding the payment). Once a node has extracted the payment, it no longer cares about the fate of the HTLC-spending transaction itself.

In cases where both resolutions are possible (e.g., when a node receives payment success after timeout), either interpretation is acceptable; it is the responsibility of the recipient to spend it before this occurs.

We need to spend the HTLC output once it has timed out to prevent them using the HTLC-success transaction, and before we can back-fail any corresponding incoming HTLC, using update_fail_htlc (presumably with reason permanent_channel_failure) as detailed in BOLT 02. If the incoming HTLC is on-chain too, we simply wait for it to timeout: there's no way to signal early failure.

If an HTLC is too small to appear in any commitment transaction, it can be safely failed immediately. Otherwise, if a HTLC isn't in the local commitment transaction a node needs to make sure that a blockchain reorganization or race does not switch to a commitment transaction which does contain it before the node fails it, hence the wait. The requirement that the incoming HTLC be failed before its own timeout still applies as an upper bound.

HTLC Output Handling: Remote Commitment, Remote Offers

Each HTLC output can only be spent by us, the recipient, using the payment preimage. If we don't have the preimage (and don't discover it), it's the offerer's responsibility to spend the HTLC output once it's timed out.

We can only spend remote HTLC outputs if we have the payment preimage. If we don't have the preimage (and don't discover it), it's the remote's responsibility to spend the HTLC output once it's timed out.

There are actually several possible cases for an offered HTLC:

  1. The offerer is not irrevocably committed to it. The recipient won't normally know the preimage here, since it won't forward HTLCs until they're fully committed. So using the preimage would reveal that this recipient is the final hop, so it's best to allow the HTLC to time out in this case.
  2. The offerer is irrevocably committed to the offered HTLC, but the recipient hasn't yet committed to an outgoing HTLC. In this case the recipient can either forward or timeout.
  3. The recipient has committed to an outgoing HTLC for the offered one. In this case the recipient has to use the preimage if it receives it from the outgoing HTLC, otherwise it will lose funds by making an outgoing payment without redeeming the incoming one.

Requirements

If the node receives (or already knows) a payment preimage for an unresolved HTLC output it was offered for which it has committed to an outgoing HTLC, it MUST resolve the output by spending it to a convenient address. Otherwise, if the remote node is not irrevocably committed to the HTLC, it MUST NOT resolve the output by spending it.

If not otherwise resolved, once the HTLC output has expired, it is considered irrevocably resolved.

Revoked Transaction Close Handling

If a node tries to broadcast old state, the remote node can use the revocation key to claim all the funds.

Requirements

A node MUST NOT broadcast a commitment transaction for which it has exposed the revocation key.

If a node discovers a commitment transaction for which it has a revocation key, that resolves the funding transaction output.

A node MUST resolve all unresolved outputs as follows:

  1. A's main output: No action is required; this is a simple P2WPKH output. This output is considered resolved by the commitment transaction.
  2. B's main output: The node MUST resolve this by spending using the revocation key.
  3. A's offered HTLCs: The node MUST resolve these in one of three ways:
  • spending using the payment revocation
  • spending using any transaction once the HTLC timeout has passed
  • by noting B's HTLC-success transaction if B publishes it
  1. B's offered HTLCs: The node MUST resolve these in one of three ways:
  • spending using the payment revocation
  • spending using the payment preimage if known
  • by noting B's HTLC-timeout transaction if B publishes it
  1. B's HTLC-timeout transaction: The node MUST resolve this by spending using the revocation key.
  2. B's HTLC-success transaction: The node MUST resolve this by spending using the revocation key. The node SHOULD extract the payment preimage from the transaction input witness if not already known.

The node MAY use a single transaction to resolve all the outputs, but MUST handle its transactions being invalidated by HTLC transactions.

Rationale

A single transaction that resolves all the outputs will be under the standard size limit thanks to the 483 HTLC-per-party limit (see BOLT #2).

Note that if a single transaction is used it may be invalidated as node B broadcasts HTLC-timeout and HTLC-success transactions, but the requirement of persistence, until all outputs are irrevocably resolved, should cover this. [FIXME: May have to divide and conquer here, since they may be able to delay us long enough to avoid successful penalty spend? ]

Penalty Transactions Weight Calculation

There are three different scripts for penalty transactions, with the following witnesses weight (details of the computation are in Appendix A):

to_local_penalty_witness: 160 bytes
offered_htlc_penalty_witness: 243 bytes
accepted_htlc_penalty_witness: 249 bytes

The penalty txinput itself takes 41 bytes, which has a weight of 164, making the weight of each input:

to_local_penalty_input_weight: 324 bytes
offered_htlc_penalty_input_weight: 407 bytes
accepted_htlc_penalty_input_weight: 413 bytes

The rest of the penalty transaction takes 4+1+1+8+1+34+4=53 bytes of non-witness data, assuming that it has a pay-to-witness-script-hash (the largest standard output script), in addition to a 2-byte witness header.

In addition to outputs being swept under as penalty, the node MAY also sweep the to_remote output of the commitment transaction (e.g. to reduce the total amount paid in fees). Doing so requires the inclusion of a p2wpkh witness and additional txinput, resulting in an additional 108 + 164 = 272 bytes.

In a worst case scenario, a node holds only incoming HTLCs, and the HTLC-timeout transactions are not published, which forces the node to spend from the commitment transaction.

With a maximum standard weight of 400000, the maximum number of HTLCs that can be swept in a single transaction:

max_num_htlcs = (400000 - 324 - 272 - (4 * 53) - 2) / 413 = 966

This allows 483 HTLCs in each direction (with both to_local and to_remote outputs) to still be resolved in a single penalty transaction. Note that even if the to_remote output is not swept, the resulting max_num_htlcs is 967, which yields the same unidirectional limit of 483 HTLCs.

General Requirements

A node SHOULD report an error to the operator if it discovers a transaction spend the funding transaction output which does not fall into one of these categories (mutual close, unilateral close, or revoked transaction close). Such a transaction implies its private key has leaked, and funds may be lost.

A node MAY simply watch the contents of the most-work chain for transactions, or MAY watch for (valid) broadcast transactions a.k.a mempool. Considering mempool transactions should cause lower latency for HTLC redemption, but on-chain HTLCs should be such an unusual case that speed cannot be considered critical.

Appendix A: Expected Weights

Expected Weight of the to_local Penalty Transaction Witness

As described in BOLT #3, the witness for this transaction is:

<sig> 1 { OP_IF <key> OP_ELSE to_self_delay OP_CSV OP_DROP <key> OP_ENDIF OP_CHECKSIG }

The expected weight is calculated as follows:

to_local_script: 83 bytes
    - OP_IF: 1 byte
        - OP_DATA: 1 byte (revocationkey length)
        - revocationkey: 33 bytes
        - OP_CHECKSIG: 1 byte
    - OP_ELSE: 1 byte
        - OP_DATA: 1 byte (localkey length)
        - localkey: 33 bytes
        - OP_CHECKSIG_VERIFY: 1 byte
        - OP_DATA: 1 byte (delay length)
        - delay: 8 bytes
        - OP_CHECKSEQUENCEVERIFY: 1 byte
    - OP_ENDIF: 1 byte

to_local_penalty_witness: 160 bytes
    - number_of_witness_elements: 1 byte
    - revocation_sig_length: 1 byte
    - revocation_sig: 73 bytes
    - one_length: 1 byte
    - witness_script_length: 1 byte
    - witness_script (to_local_script)

Expected Weight of the offered_htlc Penalty Transaction Witness

The expected weight is calculated as follows (some calculations have already been made in BOLT #3):

offered_htlc_script: 133 bytes

offered_htlc_penalty_witness: 243 bytes
    - number_of_witness_elements: 1 byte
    - revocation_sig_length: 1 byte
    - revocation_sig: 73 bytes
    - revocation_key_length: 1 byte
    - revocation_key: 33 bytes
    - witness_script_length: 1 byte
    - witness_script (offered_htlc_script)

Expected Weight of the accepted_htlc Penalty Transaction Witness

The expected weight is calculated as follows (some calculations have already been made in BOLT #3):

accepted_htlc_script: 139 bytes

accepted_htlc_penalty_witness: 249 bytes
    - number_of_witness_elements: 1 byte
    - revocation_sig_length: 1 byte
    - revocation_sig: 73 bytes
    - revocation_key_length: 1 byte
    - revocation_key: 33 bytes
    - witness_script_length: 1 byte
    - witness_script (accepted_htlc_script)

Authors

[FIXME:]

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.