rust-lightning
13 months agoInclude a route hint for public, not-yet-announced channels 2023-02-6conf-pub-hints
Matt Corallo [Thu, 9 Feb 2023 19:02:04 +0000 (19:02 +0000)]
Include a route hint for public, not-yet-announced channels

If we have a public channel which doesn't yet have six
confirmations the network can't possibly know about it as we cannot
have announced it yet. However, because we refuse to include
route-hints if we have any public channels, we will generate
invoices that no one can pay.

Thus, if we have any public, not-yet-announced channels, include
them as a route-hint.

14 months agoMerge pull request #2022 from alecchendev/2023-02-cache-peer-nodeid
Matt Corallo [Wed, 8 Feb 2023 23:50:59 +0000 (23:50 +0000)]
Merge pull request #2022 from alecchendev/2023-02-cache-peer-nodeid

Store a cached `NodeId` for each `Peer`

14 months agoCache `NodeId` by converting `their_node_id` to tuple
Alec Chen [Wed, 8 Feb 2023 17:55:16 +0000 (11:55 -0600)]
Cache `NodeId` by converting `their_node_id` to tuple

This is done to avoid redundantly serializing peer node
ids when forwarding gossip messages in
`PeerManager::forward_broadcast_msg`.

14 months agoMerge pull request #2016 from alecchendev/2023-02-gossip-msg-pubkey-to-nodeid
Matt Corallo [Tue, 7 Feb 2023 20:18:19 +0000 (20:18 +0000)]
Merge pull request #2016 from alecchendev/2023-02-gossip-msg-pubkey-to-nodeid

Swap gossip message `PublicKey` for `NodeId`

14 months agoMerge pull request #2007 from valentinewallace/2023-02-remove-invpayer-add-utils
Matt Corallo [Tue, 7 Feb 2023 17:37:38 +0000 (17:37 +0000)]
Merge pull request #2007 from valentinewallace/2023-02-remove-invpayer-add-utils

14 months agoSwap `PublicKey` for `NodeId` in `UnsignedNodeAnnouncement`
Alec Chen [Mon, 6 Feb 2023 21:33:02 +0000 (15:33 -0600)]
Swap `PublicKey` for `NodeId` in `UnsignedNodeAnnouncement`

Also swaps `PublicKey` for `NodeId` in `get_next_node_announcement`
and `InitSyncTracker` to avoid unnecessary deserialization that came
from changing `UnsignedNodeAnnouncement`.

14 months agoSwap `PublicKey` for `NodeId` in `UnsignedChannelAnnouncement`
Alec Chen [Mon, 6 Feb 2023 20:41:18 +0000 (14:41 -0600)]
Swap `PublicKey` for `NodeId` in `UnsignedChannelAnnouncement`

Adds the macro `get_pubkey_from_node_id`
to parse `PublicKey`s back from `NodeId`s for signature
verification, as well as `make_funding_redeemscript_from_slices`
to avoid parsing back and forth between types.

14 months agoExpand ChannelManager::send_spontaneous_payment_with_retry docs
Valentine Wallace [Mon, 6 Feb 2023 21:30:51 +0000 (16:30 -0500)]
Expand ChannelManager::send_spontaneous_payment_with_retry docs

14 months agoRemove InvoicePayer and replace with util methods
Valentine Wallace [Fri, 3 Feb 2023 01:08:00 +0000 (20:08 -0500)]
Remove InvoicePayer and replace with util methods

15 months agoMerge pull request #1873 from jurvis/jurvis/2022-11-expose-pending-payments
Matt Corallo [Fri, 3 Feb 2023 19:16:02 +0000 (19:16 +0000)]
Merge pull request #1873 from jurvis/jurvis/2022-11-expose-pending-payments

Expose pending payments through `ChannelManager`

15 months agoMerge pull request #2002 from valentinewallace/2023-01-keysend-retries
valentinewallace [Fri, 3 Feb 2023 19:09:00 +0000 (14:09 -0500)]
Merge pull request #2002 from valentinewallace/2023-01-keysend-retries

Support auto-retrying keysend payments in `ChannelManager`

15 months agoMerge pull request #1996 from valentinewallace/2023-01-migrate-payment-scoring
Matt Corallo [Fri, 3 Feb 2023 18:44:13 +0000 (18:44 +0000)]
Merge pull request #1996 from valentinewallace/2023-01-migrate-payment-scoring

Score payment paths from events in `BackgroundProcessor`

15 months agoExpose pending payments through ChannelManager
jurvis [Mon, 5 Dec 2022 01:24:08 +0000 (17:24 -0800)]
Expose pending payments through ChannelManager

Adds a new method, `list_recent_payments ` to `ChannelManager` that
returns an array of `RecentPaymentDetails` containing the payment
status (Fulfilled/Retryable/Abandoned) and its total amount across all
paths.

15 months agoScore payment paths in BackgroundProcessor
Valentine Wallace [Fri, 3 Feb 2023 16:25:20 +0000 (11:25 -0500)]
Score payment paths in BackgroundProcessor

15 months agoUse TestScorer in BackgroundProcessor testing
Valentine Wallace [Fri, 3 Feb 2023 16:14:53 +0000 (11:14 -0500)]
Use TestScorer in BackgroundProcessor testing

15 months agoSpontaneous payments: make preimage construction more concise
Valentine Wallace [Fri, 3 Feb 2023 15:44:31 +0000 (10:44 -0500)]
Spontaneous payments: make preimage construction more concise

15 months agoSupport spontaneous payment retries in ChannelManager
Valentine Wallace [Tue, 24 Jan 2023 17:15:40 +0000 (12:15 -0500)]
Support spontaneous payment retries in ChannelManager

15 months agoStore keysend preimage in outbound payments
Valentine Wallace [Tue, 24 Jan 2023 16:31:50 +0000 (11:31 -0500)]
Store keysend preimage in outbound payments

This sets us up for spontaneous payment retries in ChannelManager.

Currently, retrying spontaneous payments is broken in InvoicePayer because it
does not include the keysend preimage on retry.

15 months agoMerge pull request #1994 from TheBlueMatt/2023-01-1916-followups
Matt Corallo [Thu, 2 Feb 2023 23:05:07 +0000 (23:05 +0000)]
Merge pull request #1994 from TheBlueMatt/2023-01-1916-followups

1916 Followups Part 1

15 months agoReduce the code in the commitment_signed_dance macro 2023-01-1916-followups
Matt Corallo [Sat, 28 Jan 2023 05:25:19 +0000 (05:25 +0000)]
Reduce the code in the commitment_signed_dance macro

This should marginally reduce compile times for the tests by
reducing the total volume of code across the tests in the lightning
crate.

15 months agoUse only the failed amount when retrying payments, not the full amt
Matt Corallo [Sat, 28 Jan 2023 02:14:26 +0000 (02:14 +0000)]
Use only the failed amount when retrying payments, not the full amt

When we landed the initial in-`ChannelManager` payment retries, we
stored the `RouteParameters` in the payment info, and then re-use
it as-is for new payments. `RouteParameters` is intended to store
the information specific to the *route*, `PaymentParameters` exists
to store information specific to a payment.

Worse, because we don't recalculate the amount stored in the
`RouteParameters` before fetching a new route with it, we end up
attempting to retry the full payment amount, rather than only the
failed part.

This issue brought to you by having redundant data in
datastructures, part 5,001.

15 months agoMove retry-limiting to `retry_payment_with_route`
Matt Corallo [Fri, 27 Jan 2023 23:01:39 +0000 (23:01 +0000)]
Move retry-limiting to `retry_payment_with_route`

The documentation for `Retry` is very clear that it counts the
number of failed paths, not discrete retries. When we added
retries internally in `ChannelManager`, we switched to counting
the number if discrete retries, even if multiple paths failed and
were replace with multiple MPP HTLCs.

Because we are now rewriting retries, we take this opportunity to
reduce the places where retries are analyzed, specifically a good
chunk of code is removed from `pay_internal`.

Because we now retry multiple failed paths with one single retry,
we keep the new behavior, updating the docs on `Retry` to describe
the new behavior.

15 months agoLog more information when retrying a payment attempt fails
Matt Corallo [Fri, 27 Jan 2023 22:59:28 +0000 (22:59 +0000)]
Log more information when retrying a payment attempt fails

15 months agoTest the `RouteParameters` passed to `TestRouter`
Matt Corallo [Fri, 27 Jan 2023 20:31:10 +0000 (20:31 +0000)]
Test the `RouteParameters` passed to `TestRouter`

`TestRouter` allows us to simply select the `Route` that will be
returned in the next `find_route` call, but it does so without any
checking of what was *requested* for the call. This makes it a
somewhat dubious test utility as it very helpfully ensures we
ignore errors in the routes we're looking for.

Instead, we require users of `TestRouter` pass a `RouteParameters`
to `expect_find_route`, which we compare against the requested
parameters passed to `find_route`.

15 months agoUse the `PaymentParameter` final CLTV delta over `RouteParameters`
Matt Corallo [Fri, 27 Jan 2023 20:28:20 +0000 (20:28 +0000)]
Use the `PaymentParameter` final CLTV delta over `RouteParameters`

`PaymentParams` is all about the parameters for a payment, i.e. the
parameters which are static across all the paths of a paymet.
`RouteParameters` is about the information specific to a given
`Route` (i.e. a set of paths, among multiple potential sets of
paths for a payment). The CLTV delta thus doesn't belong in
`RouterParameters` but instead in `PaymentParameters`.

Worse, because `RouteParameters` is built from the information in
the last hops of a `Route`, when we deliberately inflate the CLTV
delta in path-finding, retries of the payment will have the final
CLTV delta double-inflated as it inflates starting from the final
CLTV delta used in the last attempt.

When we calculate the `final_cltv_expiry_delta` to put in the
`RouteParameters` returned via events after a payment failure, we
should re-use the new one in the `PaymentParameters`, rather than
the one that was in the route itself.

15 months agoMerge pull request #1976 from tnull/2023-01-expose-impl-writeable-msg
Matt Corallo [Wed, 1 Feb 2023 20:24:37 +0000 (20:24 +0000)]
Merge pull request #1976 from tnull/2023-01-expose-impl-writeable-msg

Expose `impl_writeable_msg`

15 months agoMove the final CLTV delta to `PaymentParameters` from `RouteParams`
Matt Corallo [Fri, 27 Jan 2023 19:24:52 +0000 (19:24 +0000)]
Move the final CLTV delta to `PaymentParameters` from `RouteParams`

`PaymentParams` is all about the parameters for a payment, i.e. the
parameters which are static across all the paths of a paymet.
`RouteParameters` is about the information specific to a given
`Route` (i.e. a set of paths, among multiple potential sets of
paths for a payment). The CLTV delta thus doesn't belong in
`RouterParameters` but instead in `PaymentParameters`.

Worse, because `RouteParameters` is built from the information in
the last hops of a `Route`, when we deliberately inflate the CLTV
delta in path-finding, retries of the payment will have the final
CLTV delta double-inflated as it inflates starting from the final
CLTV delta used in the last attempt.

By moving the CLTV delta to `PaymentParameters` we avoid this
issue, leaving only the sought amount in the `RouteParameters`.

15 months agoMerge pull request #1998 from tnull/2023-01-no-none-in-channel-relevant-txids
Matt Corallo [Wed, 1 Feb 2023 17:48:59 +0000 (17:48 +0000)]
Merge pull request #1998 from tnull/2023-01-no-none-in-channel-relevant-txids

Only return previously confirmed Txids from CM's `get_relevant_txids()`

15 months agoAdd version note in `Confirm` docs
Elias Rohrer [Tue, 31 Jan 2023 23:15:46 +0000 (17:15 -0600)]
Add version note in `Confirm` docs

While now `ChannelManager` will only return previously confirmed
transactions, we can't ensure the same for `ChainMonitor`, as we need to
maintain backwards compatibility with version prior to 0.0.113, at which
we started tracking the block hash in `ChannelMonitor`s. We therefore
add a note to the docs stating that users need to track confirmations on
their own for channels created prior to 0.0.113.

15 months agoReturn only `Some(block_hash)` in CM rel. txids
Elias Rohrer [Tue, 31 Jan 2023 23:07:31 +0000 (17:07 -0600)]
Return only `Some(block_hash)` in CM rel. txids

As of now the `Confirm::get_relevant_txids()` docs state that it won't
return any transactions for which we hadn't previously seen a
confirmation. To align its functionality a bit more with the docs, at
least for `ChannelManager`, we only return values for which we had
registered a confirmation block hash before.

15 months agoExpose Retry enum from channelmanager module
Valentine Wallace [Tue, 31 Jan 2023 20:28:49 +0000 (15:28 -0500)]
Expose Retry enum from channelmanager module

15 months agoMerge pull request #1997 from Kurtsley/fix-clean-up-documentation
Matt Corallo [Tue, 31 Jan 2023 18:06:55 +0000 (18:06 +0000)]
Merge pull request #1997 from Kurtsley/fix-clean-up-documentation

Clean up documentation in the create_phantom_invoice functions

15 months agoClean up documentation in the create_phantom_invoice functions
Kurtsley [Tue, 31 Jan 2023 03:09:17 +0000 (21:09 -0600)]
Clean up documentation in the create_phantom_invoice functions

15 months agoMerge pull request #1961 from TheBlueMatt/2023-01-expose-hist-buckets
Matt Corallo [Tue, 31 Jan 2023 00:38:14 +0000 (00:38 +0000)]
Merge pull request #1961 from TheBlueMatt/2023-01-expose-hist-buckets

Expose historical bucket data via new accessors

15 months agoMerge pull request #1972 from jkczyz/2023-01-bolt12-spec-updates
Matt Corallo [Mon, 30 Jan 2023 23:36:20 +0000 (23:36 +0000)]
Merge pull request #1972 from jkczyz/2023-01-bolt12-spec-updates

BOLT 12 spec updates

15 months agoExpose historical bucket data via new accessors 2023-01-expose-hist-buckets
Matt Corallo [Mon, 16 Jan 2023 23:23:13 +0000 (23:23 +0000)]
Expose historical bucket data via new accessors

Users should be able to view the data we use to score channels, so
this exposes that data in new accessors.

Fixes #1854.

15 months agoCalc decayed buckets to decide if we have valid historical points
Matt Corallo [Mon, 16 Jan 2023 23:18:39 +0000 (23:18 +0000)]
Calc decayed buckets to decide if we have valid historical points

When we're calculating if, once we apply the unupdated decays, the
historical data tracker has enough data to assign a score, we
previously calculated the decayed points while walking the buckets
as we don't use the decayed buckets anyway (to avoid losing
precision). That is fine, except that as written it decayed
individual buckets additional times.

Instead, here we actually calculate the full set of decayed buckets
and use those to decide if we have valid points. This adds some
additional stack space and may in fact be slower, but will be
useful in the next commit and shouldn't be a huge change.

15 months agoUpdate docs regarding payment path privacy
Jeffrey Czyz [Wed, 25 Jan 2023 05:43:14 +0000 (23:43 -0600)]
Update docs regarding payment path privacy

15 months agoDisallow offer_metadata in Refund
Jeffrey Czyz [Thu, 19 Jan 2023 00:58:20 +0000 (18:58 -0600)]
Disallow offer_metadata in Refund

The offer_metadata was optional but is redundant with invreq_metadata
(i.e., payer_metadata) for refunds. It is now disallowed in the spec and
was already unsupported by RefundBuilder.

15 months agoAllow quantity in Refund
Jeffrey Czyz [Wed, 18 Jan 2023 23:29:31 +0000 (17:29 -0600)]
Allow quantity in Refund

The spec always allowed this but the reason was unclear. It's useful if
the refund is for an invoice paid for offer where a quantity was given
in the request. The description in the refund would be from the offer,
which may have given a unit for each item. So allowing a quantity makes
it clear how many items the refund is for.

15 months agoSupport explicit quantity_max = 1 in Offer
Jeffrey Czyz [Wed, 18 Jan 2023 22:44:16 +0000 (16:44 -0600)]
Support explicit quantity_max = 1 in Offer

The spec was modified to allow setting offer_quantity_max explicitly to
one. This is to support a use case where more than one item is supported
but only one item is left in the inventory. Introduce a Quantity::One
variant to replace Quantity::Bounded(1) so the later can be used for the
explicit setting.

15 months agoMerge pull request #1985 from Kurtsley/chore-no-std-phantom-invoice
Matt Corallo [Mon, 30 Jan 2023 21:01:51 +0000 (21:01 +0000)]
Merge pull request #1985 from Kurtsley/chore-no-std-phantom-invoice

Remove std::SystemTime from create_phantom_invoice, ref #1978

15 months agoRemove std::SystemTime from create_phantom_invoice, ref #1978
Kurtsley [Wed, 25 Jan 2023 06:40:56 +0000 (00:40 -0600)]
Remove std::SystemTime from create_phantom_invoice, ref #1978

Replace current_timestamp call with no-std duration_from_epoch

15 months agoMerge pull request #1974 from danielgranhao/speed-up-secure-random-byte-gen
Matt Corallo [Thu, 26 Jan 2023 23:13:06 +0000 (23:13 +0000)]
Merge pull request #1974 from danielgranhao/speed-up-secure-random-byte-gen

15 months agoUse `Chacha20` in `get_secure_random_bytes()`
Daniel Granhão [Thu, 26 Jan 2023 10:32:14 +0000 (10:32 +0000)]
Use `Chacha20` in `get_secure_random_bytes()`

15 months agoMerge pull request #1991 from TheBlueMatt/2023-01-fix-comment
Matt Corallo [Thu, 26 Jan 2023 19:51:32 +0000 (19:51 +0000)]
Merge pull request #1991 from TheBlueMatt/2023-01-fix-comment

Remove stale comment in test

15 months agoMerge pull request #1986 from TheBlueMatt/2023-01-monitor-eq
Matt Corallo [Thu, 26 Jan 2023 19:11:49 +0000 (19:11 +0000)]
Merge pull request #1986 from TheBlueMatt/2023-01-monitor-eq

Implement PartialEq for ChannelMonitor

15 months agoRemove stale comment in test 2023-01-fix-comment
Matt Corallo [Thu, 26 Jan 2023 17:39:55 +0000 (17:39 +0000)]
Remove stale comment in test

This should have been done in 7dcbf2cd1c4de29b7c32165ca3d6ac3c47f1
but was not.

15 months agoAdd `get_secure_random_bytes()` benchmark
Daniel Granhão [Wed, 25 Jan 2023 20:55:09 +0000 (20:55 +0000)]
Add `get_secure_random_bytes()` benchmark

15 months agoMerge pull request #1984 from TheBlueMatt/2023-01-test-robust
Matt Corallo [Thu, 26 Jan 2023 04:02:31 +0000 (04:02 +0000)]
Merge pull request #1984 from TheBlueMatt/2023-01-test-robust

Make `test_duplicate_payment_hash_one_failure_one_success` robust

15 months agoRemove the `ChannelMonitor` secp context 2023-01-monitor-eq
Matt Corallo [Thu, 26 Jan 2023 02:21:31 +0000 (02:21 +0000)]
Remove the `ChannelMonitor` secp context

`ChannelMonitor` indirectly already has a context - the
`OnchainTxHandler` has one. This makes it trivial to remove the
existing one, so we do so for a free memory usage reduction.

15 months agoImplement `PartialEq` for `ChannelMonitor`
Matt Corallo [Thu, 26 Jan 2023 02:23:08 +0000 (02:23 +0000)]
Implement `PartialEq` for `ChannelMonitor`

It turns out `#[derive(PartialEq)]` will automatically bound the
`PartialEq` implementation by any bounds on the struct also being
`PartialEq`. This means to use an auto-derived `ChannelMonitor`
`PartialEq` the `EcdsaSigner` used must also be `PartialEq`, but
for the use-cases we have today for a `ChannelMonitor` `PartialEq`
it doesn't really matter - we use it internally in tests and
downstream users wanted similar test-only usage.

Fixes #1912.

15 months agoMerge pull request #1916 from valentinewallace/2022-11-chanman-payment-retries
valentinewallace [Thu, 26 Jan 2023 02:09:13 +0000 (21:09 -0500)]
Merge pull request #1916 from valentinewallace/2022-11-chanman-payment-retries

`ChannelManager` Payment Retries

15 months agoMake `test_duplicate_payment_hash_one_failure_one_success` robust 2023-01-test-robust
Matt Corallo [Wed, 25 Jan 2023 02:56:13 +0000 (02:56 +0000)]
Make `test_duplicate_payment_hash_one_failure_one_success` robust

`test_duplicate_payment_hash_one_failure_one_success` currently
fails if the "wrong" HTLC is picked to be claimed. Given the HTLCs
are identical, there's no way to figure out which we should claim.
The test instead relies on a magic value - the first one is the
right one....unless we change our CSPRNG implementation. When we
try to do so, the test randomly fails.

Here we change one HTLC to a lower amount so we can figure out
which transaction to broadcast to make the test robust against
CSPRNG changes.

15 months agoImplement `PartialEq`/`Eq` for `Event`s
Matt Corallo [Thu, 26 Jan 2023 01:51:43 +0000 (01:51 +0000)]
Implement `PartialEq`/`Eq` for `Event`s

15 months agoMerge pull request #1948 from alecchendev/custom-fail-back-err
Matt Corallo [Wed, 25 Jan 2023 23:24:49 +0000 (23:24 +0000)]
Merge pull request #1948 from alecchendev/custom-fail-back-err

Allow specifying an error when failing back HTLC

15 months agoMerge pull request #1799 from TheBlueMatt/2022-10-heap-nerdsnipe
Matt Corallo [Wed, 25 Jan 2023 23:19:13 +0000 (23:19 +0000)]
Merge pull request #1799 from TheBlueMatt/2022-10-heap-nerdsnipe

Router Optimizations

15 months agoAdd test_fail_htlc_backwards_with_reason
Alec Chen [Tue, 17 Jan 2023 04:50:58 +0000 (22:50 -0600)]
Add test_fail_htlc_backwards_with_reason

Add a test for newly added function failing back a basic payment
and ensuring the intended failure code and data are sent back
to the peer.

15 months agoAdd FailureCode enum and ChannelManager::fail_htlc_backwards_with_reason
Alec Chen [Wed, 25 Jan 2023 18:27:59 +0000 (12:27 -0600)]
Add FailureCode enum and ChannelManager::fail_htlc_backwards_with_reason

FailureCode is used to specify which error code and data to send
to peers when failing back an HTLC.

ChannelManager::fail_htlc_backwards_with_reason
allows a user to specify the error code and
corresponding data to send to peers when failing back an HTLC.
This function is mentioned in Event::PaymentClaimable docs.
ChannelManager::get_htlc_fail_reason_from_failure_code was also
added to assist with this function.

15 months agoDisambiguate send_payment_internal from pay_internal
Valentine Wallace [Tue, 24 Jan 2023 03:45:59 +0000 (22:45 -0500)]
Disambiguate send_payment_internal from pay_internal

15 months agoPayment retries: copy tests from InvoicePayer
Valentine Wallace [Wed, 4 Jan 2023 23:32:12 +0000 (18:32 -0500)]
Payment retries: copy tests from InvoicePayer

As part of migrating payment retries from InvoicePayer to ChannelManager,
several tests don't need a rewrite and can be pretty much copied and pasted.

15 months agoTest ChannelManager automatic retries
Valentine Wallace [Mon, 19 Dec 2022 05:38:54 +0000 (00:38 -0500)]
Test ChannelManager automatic retries

15 months agoSupport sending payments with a retry strategy in ChannelManager
Valentine Wallace [Mon, 19 Dec 2022 05:19:47 +0000 (00:19 -0500)]
Support sending payments with a retry strategy in ChannelManager

15 months agoRetry HTLCs in process_pending_htlc_forwards
Valentine Wallace [Sat, 7 Jan 2023 00:39:40 +0000 (19:39 -0500)]
Retry HTLCs in process_pending_htlc_forwards

15 months agoRemove AllPathsFailed outbounds at send_payment_internal callsites instead
Valentine Wallace [Mon, 19 Dec 2022 04:29:45 +0000 (23:29 -0500)]
Remove AllPathsFailed outbounds at send_payment_internal callsites instead

This makes it easier to retry payments if all paths fail on initial send, in
in which case we'll want to hold off on removing the pending payment

15 months agoGenerate PendingHTLCsForwardable upon retryable payment
Valentine Wallace [Fri, 6 Jan 2023 23:39:51 +0000 (18:39 -0500)]
Generate PendingHTLCsForwardable upon retryable payment

15 months agoDecode onion fail outside of outbound_payments lock
Valentine Wallace [Tue, 24 Jan 2023 19:26:57 +0000 (14:26 -0500)]
Decode onion fail outside of outbound_payments lock

It's not ideal to do all this computation while the lock is held. We also want
to decode the failure *before* taking the lock, so we can store the failed scid
in the relevant outbound for retry in the next commit(s).

15 months agoStore retry data in PendingOutboundPayment::Retryable
Valentine Wallace [Sun, 18 Dec 2022 23:22:56 +0000 (18:22 -0500)]
Store retry data in PendingOutboundPayment::Retryable

Used in upcoming commit(s) to automatically retry HTLCs in ChannelManager

15 months agoClean up `compute_fees` and add a saturating variant 2022-10-heap-nerdsnipe
Matt Corallo [Wed, 25 Jan 2023 17:42:20 +0000 (17:42 +0000)]
Clean up `compute_fees` and add a saturating variant

Often when we call `compute_fees` we really just want it to
saturate and we deal with `u64::max_value` later. In that case,
we're much better off doing the saturating in the `compute_fees` as
it can use CMOVs rather than branching at each step and then
`unwrap_or`ing at the callsite.

15 months agoAdd a fuzzer to check that `IndexedMap` is equivalent to `BTreeMap`
Matt Corallo [Thu, 19 Jan 2023 20:24:22 +0000 (20:24 +0000)]
Add a fuzzer to check that `IndexedMap` is equivalent to `BTreeMap`

15 months agoSwap `IndexedMap` implementation for a `HashMap`+B-Tree
Matt Corallo [Thu, 19 Jan 2023 17:59:10 +0000 (17:59 +0000)]
Swap `IndexedMap` implementation for a `HashMap`+B-Tree

Our network graph has to be iterable in a deterministic order and
with the ability to iterate over a specific range. Thus,
historically, we've used a `BTreeMap` to do the iteration. This is
fine, except our map needs to also provide high performance lookups
in order to make route-finding fast. Sadly, `BTreeMap`s are quite
slow due to the branching penalty.

Here we replace the implementation of our `IndexedMap` with a
`HashMap` to store the elements itself and a `BTreeSet` to store
the keys set in sorted order for iteration.

As of this commit on the same hardware as the above few commits,
the benchmark results are:

```
test routing::router::benches::generate_mpp_routes_with_probabilistic_scorer ... bench: 109,544,993 ns/iter (+/- 27,553,574)
test routing::router::benches::generate_mpp_routes_with_zero_penalty_scorer  ... bench:  81,164,590 ns/iter (+/- 55,422,930)
test routing::router::benches::generate_routes_with_probabilistic_scorer     ... bench:  34,726,569 ns/iter (+/- 9,646,345)
test routing::router::benches::generate_routes_with_zero_penalty_scorer      ... bench:  22,772,355 ns/iter (+/- 9,574,418)
```

15 months agoAdd a new `IndexedMap` type and use it in network graph storage
Matt Corallo [Tue, 25 Oct 2022 03:50:07 +0000 (03:50 +0000)]
Add a new `IndexedMap` type and use it in network graph storage

Our network graph has to be iterable in a deterministic order and
with the ability to iterate over a specific range. Thus,
historically, we've used a `BTreeMap` to do the iteration. This is
fine, except our map needs to also provide high performance lookups
in order to make route-finding fast. Sadly, `BTreeMap`s are quite
slow due to the branching penalty.

Here we replace the `BTreeMap`s in the scorer with a dummy wrapper.
In the next commit the internals thereof will be replaced with a
`HashMap`-based implementation.

15 months agoDrop A* implementation in the router for simple Dijkstra's
Matt Corallo [Tue, 25 Oct 2022 03:15:03 +0000 (03:15 +0000)]
Drop A* implementation in the router for simple Dijkstra's

As evidenced by the previous commit, it appears our A* router
does worse than a more naive approach. This isn't super surpsising,
as the A* heuristic calculation requires a map lookup, which is
relatively expensive.

```
test routing::router::benches::generate_mpp_routes_with_probabilistic_scorer ... bench: 169,991,943 ns/iter (+/- 30,838,048)
test routing::router::benches::generate_mpp_routes_with_zero_penalty_scorer  ... bench: 122,144,987 ns/iter (+/- 61,708,911)
test routing::router::benches::generate_routes_with_probabilistic_scorer     ... bench:  48,546,068 ns/iter (+/- 10,379,642)
test routing::router::benches::generate_routes_with_zero_penalty_scorer      ... bench:  32,898,557 ns/iter (+/- 14,157,641)
```

15 months agoMerge pull request #1983 from valentinewallace/2022-01-update-contributingmd
Matt Corallo [Wed, 25 Jan 2023 01:59:52 +0000 (01:59 +0000)]
Merge pull request #1983 from valentinewallace/2022-01-update-contributingmd

15 months agoMerge pull request #1878 from dunxen/2022-11-config-minfinalcltvexpiry
Matt Corallo [Tue, 24 Jan 2023 22:35:55 +0000 (22:35 +0000)]
Merge pull request #1878 from dunxen/2022-11-config-minfinalcltvexpiry

Allow configurable `min_final_cltv_expiry_delta`

15 months agocontributing.md: fix link and link to CoC
Valentine Wallace [Tue, 24 Jan 2023 22:33:14 +0000 (17:33 -0500)]
contributing.md: fix link and link to CoC

15 months agoMerge pull request #1967 from arik-so/2023-01-rename-signer-traits
Arik [Tue, 24 Jan 2023 21:20:26 +0000 (13:20 -0800)]
Merge pull request #1967 from arik-so/2023-01-rename-signer-traits

Split out `EcdsaChannelSigner` method from `BaseSign`, and rename it to `ChannelSigner`

15 months agoParameterize add_new_pending_payment with retry strategy and route params
Valentine Wallace [Sun, 18 Dec 2022 22:13:47 +0000 (17:13 -0500)]
Parameterize add_new_pending_payment with retry strategy and route params

15 months agoCopy Retry from invoice module to outbound_payment module
Valentine Wallace [Sun, 18 Dec 2022 22:02:17 +0000 (17:02 -0500)]
Copy Retry from invoice module to outbound_payment module

Also configure it such that in std tests, it will use SinceEpoch instead of
Instant so time can be manually advanced.

15 months agoAdd new payment type and metadata bytes
Duncan Dean [Thu, 24 Nov 2022 20:31:16 +0000 (22:31 +0200)]
Add new payment type and metadata bytes

Adds two new payment `Method`s for identifying payments with custom
`min_final_cltv_expiry_delta` as payments with LDK or user payment
hashes.

The `min_final_cltv_expiry_delta` value is packed into the first 2
bytes of the expiry timestamp in the payment secret metadata.

15 months agoAdd `min_final_cltv_expiry` parameter to invoice utils
Duncan Dean [Mon, 21 Nov 2022 12:53:52 +0000 (14:53 +0200)]
Add `min_final_cltv_expiry` parameter to invoice utils

All utility functions for invoice construction will now also accept an
Option<>al `min_final_cltv_expiry_delta` which is useful for things like
swaps etc. The `min_final_cltv_expiry_delta` will default back to
`MIN_FINAL_CLTV_EXPIRY_DELTA` if `None` is provided.

15 months agoAdd `_delta` suffix to `min_final_cltv_expiry`
Duncan Dean [Tue, 29 Nov 2022 10:47:00 +0000 (12:47 +0200)]
Add `_delta` suffix to `min_final_cltv_expiry`

This matches the spec and helps avoid any confusion around
naming. We're also then consistent with `cltv_expiry` in an HTLC being
the actual block height value for the CLTV and not a delta.

15 months agoMake `Channel` and `ChannelManager` less particular about their `Signer` type.
Arik Sosman [Fri, 20 Jan 2023 00:41:15 +0000 (16:41 -0800)]
Make `Channel` and `ChannelManager` less particular about their `Signer` type.

15 months agoRename `Sign` to `WriteableEcdsaChannelSigner`.
Arik Sosman [Thu, 19 Jan 2023 23:51:38 +0000 (15:51 -0800)]
Rename `Sign` to `WriteableEcdsaChannelSigner`.

15 months agoSeparate channel-type-agnostic methods into `ChannelSigner` trait.
Arik Sosman [Wed, 18 Jan 2023 04:33:54 +0000 (20:33 -0800)]
Separate channel-type-agnostic methods into `ChannelSigner` trait.

15 months agoRename `BaseSign` to `EcdsaChannelSigner`.
Arik Sosman [Wed, 18 Jan 2023 00:30:32 +0000 (16:30 -0800)]
Rename `BaseSign` to `EcdsaChannelSigner`.

15 months ago`impl Display for NodeId`
Matt Corallo [Mon, 23 Jan 2023 23:33:19 +0000 (23:33 +0000)]
`impl Display for NodeId`

`NodeId` is a public key, there's not much reason to not implement
`Display` for it and only `Debug`.

15 months agoMerge pull request #1926 from jkczyz/2022-12-invoice
Matt Corallo [Mon, 23 Jan 2023 23:18:09 +0000 (23:18 +0000)]
Merge pull request #1926 from jkczyz/2022-12-invoice

BOLT 12 `invoice` encoding and building

15 months agoCopy PaymentAttempts from invoice module to outbound_payment module
Valentine Wallace [Sun, 18 Dec 2022 21:53:57 +0000 (16:53 -0500)]
Copy PaymentAttempts from invoice module to outbound_payment module

15 months agoExpose `impl_writeable_msg`
Elias Rohrer [Fri, 20 Jan 2023 21:21:49 +0000 (15:21 -0600)]
Expose `impl_writeable_msg`

15 months agoFix unused_imports warning in no-std tests
Jeffrey Czyz [Thu, 19 Jan 2023 04:14:35 +0000 (22:14 -0600)]
Fix unused_imports warning in no-std tests

15 months agoInvoice parsing tests
Jeffrey Czyz [Mon, 16 Jan 2023 05:59:49 +0000 (23:59 -0600)]
Invoice parsing tests

Tests for checking invoice semantics when parsing invoice bytes as
defined by BOLT 12.

15 months agoInvoice building tests
Jeffrey Czyz [Mon, 16 Jan 2023 04:26:52 +0000 (22:26 -0600)]
Invoice building tests

Tests for checking invoice message semantics when building an invoice as
defined by BOLT 12.

15 months agoExpand invoice module docs and include an example
Jeffrey Czyz [Wed, 18 Jan 2023 15:54:13 +0000 (09:54 -0600)]
Expand invoice module docs and include an example

15 months agoUse SystemTime::now() for Invoice creation time
Jeffrey Czyz [Fri, 13 Jan 2023 05:02:39 +0000 (23:02 -0600)]
Use SystemTime::now() for Invoice creation time

For std builds, Invoice::created_at can be automatically set upon
construction using SystemTime::now() offset by SystemTime::UNIX_EPOCH.
Change InvoiceRequest::respond_with and Refund::respond_with to only
take a created_at parameter in no-std builds.

15 months agoBuilder for creating invoices for refunds
Jeffrey Czyz [Tue, 20 Dec 2022 15:33:11 +0000 (09:33 -0600)]
Builder for creating invoices for refunds

Add a builder for creating invoices for a refund and required fields.
Other settings are optional and duplicative settings will override
previous settings. Building produces a semantically valid `invoice`
message for the refund, which then can be signed with the key associated
with the provided signing pubkey.

15 months agoBuilder for creating invoices for offers
Jeffrey Czyz [Tue, 20 Dec 2022 04:23:39 +0000 (22:23 -0600)]
Builder for creating invoices for offers

Add a builder for creating invoices for an offer from a given request
and required fields. Other settings are optional and duplicative
settings will override previous settings. Building produces a
semantically valid `invoice` message for the offer, which then can be
signed with the key associated with the offer's signing pubkey.

15 months agoInvoice encoding and parsing
Jeffrey Czyz [Mon, 12 Sep 2022 14:30:06 +0000 (09:30 -0500)]
Invoice encoding and parsing

Define an interface for BOLT 12 `invoice` messages. The underlying
format consists of the original bytes and the parsed contents.

The bytes are later needed for serialization. This is because it must
mirror all the `offer` and `invoice_request` TLV records, including
unknown ones, which aren't represented in the contents.

Invoices may be created for an Offer (from an InvoiceRequest) or for a
Refund. The primary difference is how the signing pubkey is given -- by
the writer of the offer or the reader of the refund.

15 months agoUpdate min-inbound-fee values on `NetworkGraph` load
Matt Corallo [Thu, 19 Jan 2023 18:24:30 +0000 (18:24 +0000)]
Update min-inbound-fee values on `NetworkGraph` load

Historically we've had various bugs in keeping the
`lowest_inbound_channel_fees` field in `NodeInfo` up-to-date as we
go. This leaves the A* routing less efficient as it can't prune
hops as aggressively.

In order to get accurate benchmarks, this commit updates the
minimum-inbound-fees field on load. This is not the most efficient
way of doing so, but suffices for fetching benchmarks and will be
removed in the coming commits.

Note that this is *slower* than the non-updating version in the
previous commit. While I haven't dug into this incredibly deeply,
the graph snapshot in use has min-fee info for only 9,618 of
20,818 nodes. Thus, it is my guess that with the graph snapshot
as-is the branch predictor is able to largely remove the A*
heuristic lookups, but with this change it is forced to wait for
A* heuristic map lookups to complete, causing a performance
regression.

```
test routing::router::benches::generate_mpp_routes_with_probabilistic_scorer ... bench: 182,980,059 ns/iter (+/- 32,662,047)
test routing::router::benches::generate_mpp_routes_with_zero_penalty_scorer  ... bench: 151,170,457 ns/iter (+/- 75,351,011)
test routing::router::benches::generate_routes_with_probabilistic_scorer     ... bench:  58,187,277 ns/iter (+/- 11,606,440)
test routing::router::benches::generate_routes_with_zero_penalty_scorer      ... bench:  41,210,193 ns/iter (+/- 18,103,320)
```

15 months agoUpdate the lightning graph snapshot used in benchmarks
Matt Corallo [Thu, 19 Jan 2023 04:41:02 +0000 (04:41 +0000)]
Update the lightning graph snapshot used in benchmarks

The previous copy was more than one and a half years old, the
lightning network has changed a lot since!

As of this commit, performance on my Xeon W-10885M with a
SK hynix Gold P31 storing a BTRFS volume is as follows:

```
test ln::channelmanager::bench::bench_sends                                  ... bench:   5,896,492 ns/iter (+/- 512,421)
test routing::gossip::benches::read_network_graph                            ... bench: 1,645,740,604 ns/iter (+/- 47,611,514)
test routing::gossip::benches::write_network_graph                           ... bench: 234,870,775 ns/iter (+/- 8,301,775)
test routing::router::benches::generate_mpp_routes_with_probabilistic_scorer ... bench: 166,155,032 ns/iter (+/- 30,206,162)
test routing::router::benches::generate_mpp_routes_with_zero_penalty_scorer  ... bench: 136,843,661 ns/iter (+/- 67,111,218)
test routing::router::benches::generate_routes_with_probabilistic_scorer     ... bench:  52,954,598 ns/iter (+/- 11,360,547)
test routing::router::benches::generate_routes_with_zero_penalty_scorer      ... bench:  37,598,126 ns/iter (+/- 17,262,519)
test bench::bench_sends                                                      ... bench:  37,760,922 ns/iter (+/- 5,179,123)
test bench::bench_reading_full_graph_from_file                               ... bench:      25,615 ns/iter (+/- 1,149)
```