Anchor commitments, new payment basepoint #7509

bitromortac · 2021-09-23T08:45:13Z

This PR implements a newer commitment transaction type [1]. Anchor commitments are commitment transactions that attach a dusty output per realized to_local/to_remote output, in order for the commitment transaction to be CPFP-bumpable in a force close scenario. Also HTLC transactions are affected, which are fee-amendable allowing for more inputs via the anyonecanpay sighash. The original proposal of anchor commitments [1] was not safe enough [2] which is why option_anchors_zero_fee_htlc_tx was devised [3]. Our goal is to go directly to this version, to not having to deal with legacy developments. The PR passes unit and regtest tests and is ready for initial review or feedback (see Todos).

Key changes:

The Channel object now carries a has_anchors property which is used to decide on scripts and locktimes, as well as sweeping behavoir.
Adapt commitment transactions for anchors (anchors, to_remote script 1-CSV, fee estimation, output values) (03-transactions)
Adapt HTLC transactions (nsequence numbers, 1 CSV-lock, sighash.ANYONECANPAY+sighash.SINGLE) (03-transactions)
New anchor channels have a payment_basepoint derived from the funding pubkey and a static secret to be sweepable in the event of a force close due to (onchain) backups

Testing:

Reproduce signatures from RFC test vectors for commitment and HTLC transactions (requires different sighash flags) (03-transactions)
Manually tested against lnd v0.11 (last active version of original anchors) - open/close/force close/payment
New integration test framework to run Electrum peer components inside a single interpreter (better debugging and easier introspection), with additional detailed channel usage scenarios. Tests are not running in sequence right now and are still subject to changes. Some tests are redundant and could be removed.

Other:

The watchtower is not updated to sweep to_remote as this is not time critical
Backups don't need an upgrade as we can see at sweep time, whether we deal with a ctx that has anchors

Todo:

Crosscheck RFC
Implement zero fee HTLCs and do fee management at sweep time
Reserve some UTXOs for fee bumping
Compatibility testing with zero-fee-htlc (once zero-fee is implemented)
Wallet upgrading mechanics, compatibilty wallets with old/new channels
Keep tests for both commitment versions (static_remotekey/anchors)?
Make commitment type configurable?
Determine fee for HTLC-transaction bumping
Sweep the anchor outputs (05-onchain) (omitted for this PR, because we still rely on the update_fee message)
Refactor to have explicit commitment and channel types [4] (another PR)

[1] Anchors PR: lightning/bolts#688
[2] https://lists.linuxfoundation.org/pipermail/lightning-dev/2020-September/002796.html
[3] Zero-fee HLTCs: lightning/bolts#824
[4] Channel types: lightning/bolts#880

electrum/lnchannel.py

SomberNight

So many changes :P but looks good overall.
See comments inline.

electrum/lnchannel.py

SomberNight · 2021-09-23T14:36:59Z

electrum/lnpeer.py

@@ -502,6 +502,10 @@ def is_static_remotekey(self):
    def is_upfront_shutdown_script(self):
        return self.features.supports(LnFeatures.OPTION_UPFRONT_SHUTDOWN_SCRIPT_OPT)

+    def use_anchors(self) -> bool:
+        return self.features.supports(LnFeatures.OPTION_ANCHOR_OUTPUTS_OPT) \
+               and self.lnworker.features.supports(LnFeatures.OPTION_ANCHOR_OUTPUTS_OPT)


is it not enough to check self.features; why check self.lnworker.features?

Yes, here I wanted to ask how we want to transition to anchors? The second condition was also a reminder for myself.

Well there are three ~reasonable options I can see:

we require anchors_zero_fee for newly opened channels

we open anchors_zero_fee channels whenever possible, but if the remote does not support it, we can just option static_remotekey channels

we disable anchors_zero_fee by default, gate it with a config key; and change to one of the options above in a later release

I think the decision should depend on a number of things, such as

how many/which nodes on the network will already support anchors_zero_fee,

how confident we are that our code behaves correctly,

how the UX of reserving UTXOs in the wallet looks like,

whether we can batch HTLC txs (not really for fee-minimisation but rather re number of UTXOs we need to reserve)

So atm I am not sure.

electrum/tests/test_lnutil.py

electrum/lnutil.py

electrum/lnsweep.py

electrum/lnutil.py

electrum/lnsweep.py

bitromortac · 2021-09-27T12:47:16Z

I submitted changes via fixup commits, will squash in the end, or should I already?

SomberNight · 2021-09-28T15:21:04Z

I submitted changes via fixup commits, will squash in the end, or should I already?

Feel free to do whatever is convenient.
One remark: if you rebase on master for any reason, it's better to do that in a separate force-push of its own just for the rebase (instead of combining with other changes that would require a force-push). That way, it's easier to see changes in the github UI.

electrum/lnsweep.py

bitromortac · 2021-10-04T10:43:10Z

I submitted changes via fixup commits, will squash in the end, or should I already?

Feel free to do whatever is convenient. One remark: if you rebase on master for any reason, it's better to do that in a separate force-push of its own just for the rebase (instead of combining with other changes that would require a force-push). That way, it's easier to see changes in the github UI.

I squashed the fixups and rebased on master separately.

There's a new commit, which switches to the anchors-zero-htlc-fee type. I have tested this together with the latest LND master. The only issue for now is how to deal with password protected wallets, where we are not able to fee-bump HTLCs automatically (or bump commitment transactions once we phase out update_fee).

bitromortac · 2021-10-12T13:42:52Z

As discussed, for now we only enable anchor channels via a config setting (not exposed in the GUI), storing the wallet password in lnworker for dynamic signing. A wallet can only be loaded with anchor channels enabled, if it is not a watch-only wallet. Also, the anchors config option cannot be disabled, if a wallet already has anchor channels.

This way we can test more in production and we can explore further how we want to handle UTXO reservation.

bitromortac · 2021-10-15T10:32:34Z

I reorganized the commits a bit and think this is ready for a second pass. Tests for anchor outputs are disabled right now (can be controlled via TEST_ANCHOR_CHANNELS), but they should all pass.

I think we could postpone the reservation of UTXOs for now, as we keep anchors disabled.

Edit: somehow the CI regtest hangs, but passes locally

electrum/gui/qt/installwizard.py

electrum/lnworker.py

electrum/tests/test_lnchannel.py

SomberNight · 2021-11-01T15:56:12Z

electrum/tests/regtest/regtest.sh

@@ -2,6 +2,7 @@
 export HOME=~
 set -eu

+TEST_ANCHOR_CHANNELS=False


Same as for the unit tests: I think we should test by default for both possibilities.
One approach: some refactoring where most code of each test is inside a function that is parameterised with anchors_enabled, and the functions are called twice (one for each setting).

Done, I'm setting an environment variable in subprocess.Popen. Not sure if updating of the environment dict is the best approach.

electrum/tests/integration.py

electrum/lnsweep.py

bitromortac · 2021-11-10T12:53:07Z

After the recent update all tests passed, but I had to run the watchtower test twice (needs further investigation). I'm still working on the HTLC issue we discussed.

* add anchor ln features * peer.use_anchors is added * channel.has_anchors is added

* in order to be able to sweep to_remote in an onchain backup scenario we need to retain the private key for the payment_basepoint * to facilitate the above, we open a channel derived from a static secret (tied to the wallet seed), the static_payment_key combined with the funding pubkey (multisig_key), which we can restore from the channel closing transaction

* changes the htlc outputs' witness script to have a csv lock of 1 * send signatures for remote ctx with ANYONECANPAY|SINGLE * refactor htlc weight (useful for zero-fee-htlc)

* to_remote has now an additional csv lock of 1 * anchor outputs are added if to_local/remote outputs are present * funder balance is reduced to accomodate anchors

* sweep to_remote output, as this is now a p2wsh (previously internal wallet address) * sweep htlc outputs with new scripts

* add a method for backups to sweep to_remote * to_remote sweeping needs the payment_basepoint's private key to sign the sweep transaction * we restore the private key from our funding multisig pubkey (pubished with the closing transaction) and a static payment key secret * lower the final balance of the backup regtest, which is due to additional sweep transactions

source: https://tabler-icons.io/anchor

* anchor channels can be activated via `enable_anchor_channels` config option * for anchor-enabled wallets, we store the wallet password, because UTXOs need to be added to commitment or HTLC transactions dynamically

* testing of anchor channels is controlled via TEST_ANCHOR_CHANNELS * rewrite tests in test_lnchannel.py

* tests are kept variable via TEST_ANCHOR_CHANNELS * add funds to bob to be able to bump htlc transaction

* sets the weight of htlc transactions to zero, thereby putting a zero fee for the htlc transactions * add inputs to htlc-tx for fee bumping * switches feature flags * disable anchor test vectors, which are now partially invalid

* in test_lnutil, patch htlc weight to pass original anchor commitment test vectors * activate tests for both commitment types

naming scheme: tx(s)_our/their_ctx/htlctx_output-description * functinon names are shortened to whether a single (tx) or several sweep transactions (txs) are generated * functions are ordered by their purpose (our/their ctx related)

Due to anchor channel's sighash.SINGLE and sighash.ANYONECANPAY, several HTLC-transactions can be combined. This means we must watch for revoked outputs in the HTLC transaction not only at index 0 but at any index. Also, any input can now contain preimages which we have to extract.

Due to malleability of HTLC-transactions, we can't send presigned justice transactions for the second-stage HTLC transactions, which is why we now send first-stage justice transactions for anchor channels.

bitromortac · 2021-11-18T13:40:57Z

After the recent update all tests passed, but I had to run the watchtower test twice (needs further investigation). I'm still working on the HTLC issue we discussed.

I tried to add all the edge cases concerning HTLC transaction batching. Do we need an explicit test for this? I also added a test for the watchtower in case of anchor outputs with an HTLC in the ctx (see integration.TestLightningABC.test_watchtower_htlc)

ecdsa · 2022-03-09T17:07:16Z

Note: I'm going to rebase this in a new branch. Since it is not possible to change the source branch of a PR, I guess this means that the conversation will have to be continued in a new PR

ecdsa · 2022-09-26T07:24:15Z

new branch, rebased on current master: /~https://github.com/spesmilo/electrum/commits/anchor_commitments_2022

changes:

the wallet password is not stored anymore. Instead, we will send funds to a new sequence of addresses, derived from ln_xprv.
the commit that was both renaming and reordering functions in lnsweep has been split into two commits, so that changes can actually be reviewed. (I checked that the reordering commit does only that)

ecdsa · 2024-06-09T09:20:00Z

new rebase: /~https://github.com/spesmilo/electrum/commits/anchor_commitments_2024

SomberNight · 2024-10-20T10:04:34Z

@ecdsa Could you make a separate PR for your rebase? I cannot usefully add comments to it without that.
(note to self: lnchannel.sweep_address looks broken in the rebase)

ecdsa · 2024-10-20T13:53:38Z

opened PR #9264

ecdsa · 2024-11-20T10:54:18Z

electrum/lnchannel.py

+
+            # initiator pays for anchor outputs
+            if sender == initiator and self.has_anchors():
+                max_send_msat -= 2 * FIXED_ANCHOR_SAT


note: needs to be multiplied by 1000

ecdsa · 2024-11-26T09:23:56Z

commits were reworked and merged in #9264

ecdsa reviewed Sep 23, 2021

View reviewed changes

electrum/lnchannel.py Show resolved Hide resolved

SomberNight reviewed Sep 23, 2021

View reviewed changes

SomberNight reviewed Sep 28, 2021

View reviewed changes

electrum/lnsweep.py Outdated Show resolved Hide resolved

bitromortac force-pushed the 2109-anchor-commitments-backup branch 2 times, most recently from 03b789b to ee2a5d7 Compare October 4, 2021 10:38

bitromortac force-pushed the 2109-anchor-commitments-backup branch from ee2a5d7 to a07f27a Compare October 12, 2021 13:34

bitromortac force-pushed the 2109-anchor-commitments-backup branch 2 times, most recently from 4ddbc84 to e70a918 Compare October 15, 2021 10:28

bitromortac force-pushed the 2109-anchor-commitments-backup branch from e70a918 to 811d31d Compare October 19, 2021 15:35

SomberNight reviewed Nov 1, 2021

View reviewed changes

bitromortac force-pushed the 2109-anchor-commitments-backup branch from 811d31d to 20b9e98 Compare November 10, 2021 12:49

bitromortac added 13 commits November 18, 2021 14:28

prepare a channel to have anchors

7926219

* add anchor ln features * peer.use_anchors is added * channel.has_anchors is added

lnutil: update ctx fee calculation for anchors

a3cd1fd

lnchannel+lnutil: change htlc output, send new sig

85c6a83

* changes the htlc outputs' witness script to have a csv lock of 1 * send signatures for remote ctx with ANYONECANPAY|SINGLE * refactor htlc weight (useful for zero-fee-htlc)

lnutil+lnchannel: add anchors, adapt to_remote

7816110

* to_remote has now an additional csv lock of 1 * anchor outputs are added if to_local/remote outputs are present * funder balance is reduced to accomodate anchors

tests: add anchor commitment test vectors from rfc

3500246

lnsweep: update sweeps to_remote and htlcs

81d7556

* sweep to_remote output, as this is now a p2wsh (previously internal wallet address) * sweep htlc outputs with new scripts

lnwatcher: renaming and comments for clarity

8a7ea74

lnwatcher: fix early cltv-locked claiming

63d0a6a

lnwatcher: add field for onchain htlc settlement control

f456024

qt: add anchor channel icon

1d89e4e

source: https://tabler-icons.io/anchor

enable anchor outputs via config option

205d6cb

* anchor channels can be activated via `enable_anchor_channels` config option * for anchor-enabled wallets, we store the wallet password, because UTXOs need to be added to commitment or HTLC transactions dynamically

bitromortac added 7 commits November 18, 2021 14:28

unit tests: test anchors in lnpeer and lnchannel

6326817

* testing of anchor channels is controlled via TEST_ANCHOR_CHANNELS * rewrite tests in test_lnchannel.py

regtest: adapt to anchor channels

cfb4a10

* tests are kept variable via TEST_ANCHOR_CHANNELS * add funds to bob to be able to bump htlc transaction

anchors: switch to zero-fee-htlcs

421f32d

* sets the weight of htlc transactions to zero, thereby putting a zero fee for the htlc transactions * add inputs to htlc-tx for fee bumping * switches feature flags * disable anchor test vectors, which are now partially invalid

tests: tests for both anchors and old ctx types

3f9c530

* in test_lnutil, patch htlc weight to pass original anchor commitment test vectors * activate tests for both commitment types

sweep: rename sweep creation functions and reorder

f9210b4

naming scheme: tx(s)_our/their_ctx/htlctx_output-description * functinon names are shortened to whether a single (tx) or several sweep transactions (txs) are generated * functions are ordered by their purpose (our/their ctx related)

watchtower: only send first-stage HTLC justice txs

5cb97c0

Due to malleability of HTLC-transactions, we can't send presigned justice transactions for the second-stage HTLC transactions, which is why we now send first-stage justice transactions for anchor channels.

bitromortac force-pushed the 2109-anchor-commitments-backup branch from 20b9e98 to 5cb97c0 Compare November 18, 2021 13:35

bitromortac force-pushed the 2109-anchor-commitments-backup branch 2 times, most recently from 0633154 to 5cb97c0 Compare November 25, 2021 08:30

bitromortac mentioned this pull request Nov 25, 2021

integration tests: all Electrum components inside an interpreter #7577

Draft

4 tasks

bitromortac mentioned this pull request Jan 19, 2022

lightning: implement channel types #7636

Merged

2 tasks

ecdsa mentioned this pull request Sep 14, 2023

daemon: store wallet password in memory when wallet is loaded #8612

Closed

SomberNight linked an issue Jan 18, 2024 that may be closed by this pull request

implement LN anchor outputs #6588

Open

ecdsa mentioned this pull request Oct 20, 2024

Anchor commitments 2024 #9264

Merged

ecdsa reviewed Nov 20, 2024

View reviewed changes

ecdsa closed this Nov 26, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Anchor commitments, new payment basepoint #7509

Anchor commitments, new payment basepoint #7509

bitromortac commented Sep 23, 2021 •

edited

Loading

SomberNight left a comment

SomberNight Sep 23, 2021

bitromortac Sep 27, 2021

SomberNight Sep 28, 2021

bitromortac commented Sep 27, 2021

SomberNight commented Sep 28, 2021 •

edited

Loading

bitromortac commented Oct 4, 2021

bitromortac commented Oct 12, 2021

bitromortac commented Oct 15, 2021 •

edited

Loading

SomberNight Nov 1, 2021

bitromortac Nov 10, 2021

bitromortac commented Nov 10, 2021

bitromortac commented Nov 18, 2021

ecdsa commented Mar 9, 2022

ecdsa commented Sep 26, 2022 •

edited

Loading

ecdsa commented Jun 9, 2024

SomberNight commented Oct 20, 2024

ecdsa commented Oct 20, 2024

ecdsa Nov 20, 2024

ecdsa commented Nov 26, 2024

Anchor commitments, new payment basepoint #7509

Anchor commitments, new payment basepoint #7509

Conversation

bitromortac commented Sep 23, 2021 • edited Loading

SomberNight left a comment

Choose a reason for hiding this comment

SomberNight Sep 23, 2021

Choose a reason for hiding this comment

bitromortac Sep 27, 2021

Choose a reason for hiding this comment

SomberNight Sep 28, 2021

Choose a reason for hiding this comment

bitromortac commented Sep 27, 2021

SomberNight commented Sep 28, 2021 • edited Loading

bitromortac commented Oct 4, 2021

bitromortac commented Oct 12, 2021

bitromortac commented Oct 15, 2021 • edited Loading

SomberNight Nov 1, 2021

Choose a reason for hiding this comment

bitromortac Nov 10, 2021

Choose a reason for hiding this comment

bitromortac commented Nov 10, 2021

bitromortac commented Nov 18, 2021

ecdsa commented Mar 9, 2022

ecdsa commented Sep 26, 2022 • edited Loading

ecdsa commented Jun 9, 2024

SomberNight commented Oct 20, 2024

ecdsa commented Oct 20, 2024

ecdsa Nov 20, 2024

Choose a reason for hiding this comment

ecdsa commented Nov 26, 2024

bitromortac commented Sep 23, 2021 •

edited

Loading

SomberNight commented Sep 28, 2021 •

edited

Loading

bitromortac commented Oct 15, 2021 •

edited

Loading

ecdsa commented Sep 26, 2022 •

edited

Loading