rust-lightning
19 months agoCorrect `get_claimable_balance` version info 2022-09-bad-doc-versions
Matt Corallo [Fri, 9 Sep 2022 19:30:30 +0000 (19:30 +0000)]
Correct `get_claimable_balance` version info

5a8ede09fb3c8bbcd8694d94c12dac9ea7485537 updated the documentation
on `get_claimable_balance` to note that if a channel went on-chain
with an LDK version older than 0.0.108 some
counterparty-revoked-output claimable balances my be missing.
However, this failed to account for the fact that we rely on the
entirely-new-in-0.0.111
`confirmed_commitment_tx_counterparty_output` field for some
balances as well.

Thus, the comment should have been in terms of 0.0.111, not
0.0.108.

19 months agoMerge pull request #1701 from TheBlueMatt/2022-09-feature-or
Matt Corallo [Fri, 9 Sep 2022 16:47:25 +0000 (16:47 +0000)]
Merge pull request #1701 from TheBlueMatt/2022-09-feature-or

Fetch InitFeatures from both Channel and Routing Message Handlers

19 months agoMake clear_initial_routing_sync more consistent with other APIs 2022-09-feature-or
Matt Corallo [Wed, 7 Sep 2022 17:43:14 +0000 (17:43 +0000)]
Make clear_initial_routing_sync more consistent with other APIs

19 months agoAdd a new InitFeatures constructor to capture the types of flags
Matt Corallo [Wed, 7 Sep 2022 17:55:01 +0000 (17:55 +0000)]
Add a new InitFeatures constructor to capture the types of flags

When `ChannelMessageHandler` implementations wish to return an
`InitFeatures` which contain all the known flags that are relevant
to channel handling, but not gossip handling, they currently need
to do so by manually constructing an InitFeatures with all known
flags and then clearing the ones they dont want.

Instead of spreading this logic out across the codebase, this
consolidates such construction to one place in features.rs.

19 months agoOR InitFeatures from both Channel and Routing message handlers
Matt Corallo [Wed, 7 Sep 2022 17:35:50 +0000 (17:35 +0000)]
OR InitFeatures from both Channel and Routing message handlers

When we go to send an Init message to new peers, the features we
support are really a combination of all the various features our
different handlers support. This commit captures this concept by
OR'ing our InitFeatures across both our Channel and Routing
handlers.

Note that this also disables setting the `initial_routing_sync`
flag in init messages, as was intended in
e742894492c55802b241eebc585bbd28aa16481b, per the comment added on
`clear_initial_routing_sync`, though this should not be a behavior
change in practice as nodes which support gossip queries ignore the
initial routing sync flag.

19 months agoFetch our `InitFeatures` from `ChannelMessageHandler`
Matt Corallo [Wed, 7 Sep 2022 17:51:16 +0000 (17:51 +0000)]
Fetch our `InitFeatures` from `ChannelMessageHandler`

Like we now do for `NodeFeatures`, this converts to asking our
registered `ChannelMessageHandler` for our `InitFeatures` instead
of hard-coding them to the global LDK known set.

This allows handlers to set different feature bits based on what
our configuration actually supports rather than what LDK supports
in aggregate.

19 months agoMerge pull request #1699 from TheBlueMatt/2022-08-announcement-rework
Matt Corallo [Fri, 9 Sep 2022 00:34:10 +0000 (00:34 +0000)]
Merge pull request #1699 from TheBlueMatt/2022-08-announcement-rework

19 months agoMerge pull request #1704 from TheBlueMatt/2022-09-always-probe-failed
Matt Corallo [Thu, 8 Sep 2022 20:35:15 +0000 (20:35 +0000)]
Merge pull request #1704 from TheBlueMatt/2022-09-always-probe-failed

Dont use PaymentPathFailed a probe fails without making it out

19 months agoAdd a folder to track CHANGELOG entries for the next release 2022-08-announcement-rework
Matt Corallo [Wed, 7 Sep 2022 00:01:12 +0000 (00:01 +0000)]
Add a folder to track CHANGELOG entries for the next release

19 months agoMove `broadcast_node_announcement` to `PeerManager`
Matt Corallo [Tue, 6 Sep 2022 22:34:29 +0000 (22:34 +0000)]
Move `broadcast_node_announcement` to `PeerManager`

Some `NodeFeatures` will, in the future, represent features which
are not enabled by the `ChannelManager`, but by other message
handlers handlers. Thus, it doesn't make sense to determine the
node feature bits in the `ChannelManager`.

The simplest fix for this is to change to generating the
node_announcement in `PeerManager`, asking all the connected
handlers which feature bits they support and simply OR'ing them
together. While this may not be sufficient in the future as it
doesn't consider feature bit dependencies, support for those could
be handled at the feature level in the future.

This commit moves the `broadcast_node_announcement` function to
`PeerHandler` but does not yet implement feature OR'ing.

19 months agoSend channel_{announcement,update} msgs on connection, not timer
Matt Corallo [Tue, 6 Sep 2022 21:30:33 +0000 (21:30 +0000)]
Send channel_{announcement,update} msgs on connection, not timer

When we connect to a new peer, immediately send them any
channel_announcement and channel_update messages for any public
channels we have with other peers. This allows us to stop sending
those messages on a timer when they have not changed and ensures
we are sending messages when we have peers connected, rather than
broadcasting at startup when we have no peers connected.

19 months agoImprove debuggability when tests fail due to excess events 2022-09-always-probe-failed
Matt Corallo [Wed, 7 Sep 2022 21:53:13 +0000 (21:53 +0000)]
Improve debuggability when tests fail due to excess events

19 months agoDont use PaymentPathFailed a probe fails without making it out
Matt Corallo [Wed, 7 Sep 2022 21:39:17 +0000 (21:39 +0000)]
Dont use PaymentPathFailed a probe fails without making it out

When we fail to forward a probe HTLC at all and immediately fail it
(e.g. due to the first hop channel closing) we'd previously
spuriously generate only a `PaymentPathFailed` event. This violates
the expected API, as users expect a `ProbeFailed` event instead.

This fixes the oversight by ensuring we generate the correct event.

Thanks to @jkczyz for pointing out this issue.

19 months agoDrop redundant code in `fail_holding_cell_htlcs`
Matt Corallo [Wed, 7 Sep 2022 21:09:50 +0000 (21:09 +0000)]
Drop redundant code in `fail_holding_cell_htlcs`

`fail_holding_cell_htlcs` calls through to
`fail_htlc_backwards_internal` for HTLCs that need to be
failed-backwards but opts to generate its own payment failure
events for `HTLCSource:;OutboundRoute` HTLCs. There is no reason
for that as `fail_htlc_backwards_internal` will also happily
generate (now-)equivalent events for `HTLCSource::OutboundRoute`
HTLCs.

Thus, we can drop the redundant code and always call
`fail_htlc_backwards_internal` for each HTLC in
`fail_holding_cell_htlcs`.

19 months agoMerge pull request #1702 from TheBlueMatt/2022-09-one-hop-retryable
Matt Corallo [Thu, 8 Sep 2022 17:34:04 +0000 (17:34 +0000)]
Merge pull request #1702 from TheBlueMatt/2022-09-one-hop-retryable

Mark failed one-hop HTLCs as retrably

19 months agoMerge pull request #1700 from TheBlueMatt/2022-09-missing-event-deser
Matt Corallo [Thu, 8 Sep 2022 15:41:58 +0000 (15:41 +0000)]
Merge pull request #1700 from TheBlueMatt/2022-09-missing-event-deser

Add missing deserialization of Event::HTLCHandlingFailed

19 months agoAdd missing deserialization of Event::HTLCHandlingFailed 2022-09-missing-event-deser
Matt Corallo [Tue, 6 Sep 2022 22:51:29 +0000 (22:51 +0000)]
Add missing deserialization of Event::HTLCHandlingFailed

17e6c374c513f2eca810fa4e931be65f0d4fc29f added the
`HTLCHandlingFailed` event, including serialization thereof,
however failed to add corresponding deserialization. This corrects
that oversight by adding said deserialization.

Thanks to @wpaulino for catching the oversight.

19 months agoRename `rejected_by_dest` -> `payment_failed_permanently` 2022-09-one-hop-retryable
Matt Corallo [Wed, 7 Sep 2022 20:58:05 +0000 (20:58 +0000)]
Rename `rejected_by_dest` -> `payment_failed_permanently`

The `rejected_by_dest` field of the `PaymentPathFailed` event has
always been a bit of a misnomer, as its really more about retry
than where a payment failed. Now is as good a time as any to
rename it.

19 months agoMark failed counterparty-is-destination HTLCs retryable
Matt Corallo [Wed, 7 Sep 2022 20:02:04 +0000 (20:02 +0000)]
Mark failed counterparty-is-destination HTLCs retryable

When our counterparty is the payment destination and we receive
an `HTLCFailReason::Reason` in `fail_htlc_backwards_internal` we
currently always set `rejected_by_dest` in the `PaymentPathFailed`
event, implying the HTLC should *not* be retried.

There are a number of cases where we use `HTLCFailReason::Reason`,
but most should reasonably be treated as retryable even if our
counterparty was the destination (i.e. `!rejected_by_dest`):
 * If an HTLC times out on-chain, this doesn't imply that the
   payment is no longer retryable, though the peer may well be
   offline so retrying may not be very useful,
 * If a commitment transaction "containing" a dust HTLC is
   confirmed on-chain, this definitely does not imply the payment
   is no longer retryable
 * If the channel we intended to relay over was closed (or
   force-closed) we should retry over another path,
 * If the channel we intended to relay over did not have enough
   capacity we should retry over another path,
 * If we received a update_fail_malformed_htlc message from our
   peer, we likely should *not* retry, however this should be
   exceedingly rare, and appears to nearly never appear in practice

Thus, this commit simply disables the behavior here, opting to
treat all `HTLCFailReason::Reason` errors as retryable.

Note that prior to 93e645daf46f85949ae0edf60d36bf21e9fde8af this
change would not have made sense as it would have resulted in us
retrying the payment over the same channel in some cases, however
we now "blame" our own channel and will avoid it when routing for
the same payment.

19 months agoMerge pull request #1697 from TheBlueMatt/2022-08-event-docs
valentinewallace [Wed, 7 Sep 2022 02:04:39 +0000 (22:04 -0400)]
Merge pull request #1697 from TheBlueMatt/2022-08-event-docs

Clarify and consolidate event handling requirements

19 months agoMerge pull request #1698 from TheBlueMatt/2022-09-release-names
valentinewallace [Wed, 7 Sep 2022 01:57:16 +0000 (21:57 -0400)]
Merge pull request #1698 from TheBlueMatt/2022-09-release-names

Add release names to CHANGELOG

19 months agoMerge pull request #1691 from TheBlueMatt/2022-08-dust-retry
valentinewallace [Wed, 7 Sep 2022 00:02:50 +0000 (20:02 -0400)]
Merge pull request #1691 from TheBlueMatt/2022-08-dust-retry

Correct payment resolution after on chain failure of dust HTLCs

19 months agoAdd release names to CHANGELOG 2022-09-release-names
Matt Corallo [Tue, 6 Sep 2022 21:17:15 +0000 (21:17 +0000)]
Add release names to CHANGELOG

There have been release names in Github for some time, though some
releases were missed. This adds them to CHANGELOG.

19 months agoClarify and consolidate event handling requirements 2022-08-event-docs
Matt Corallo [Tue, 6 Sep 2022 20:56:24 +0000 (20:56 +0000)]
Clarify and consolidate event handling requirements

We've seen a bit of user confusion about the requirements for event
handling, largely because the idempotency and consistency
requirements weren't super clearly phrased. While we're at it, we
also consolidate some documentation out of the event handling
function onto the trait itself.

Fixes #1675.

19 months agoCorrect payment resolution after on chain failure of dust HTLCs 2022-08-dust-retry
Matt Corallo [Fri, 2 Sep 2022 21:10:43 +0000 (21:10 +0000)]
Correct payment resolution after on chain failure of dust HTLCs

Previously, we wouldn't mark a dust HTLC as permanently resolved if
the commitment transaction went on chain. This resulted in us
always considering the HTLC as pending on restart, when we load the
pending payments set from the monitors.

Fixes #1653.

19 months agoMerge pull request #1657 from TheBlueMatt/2022-08-async-man-update
valentinewallace [Tue, 6 Sep 2022 20:06:06 +0000 (16:06 -0400)]
Merge pull request #1657 from TheBlueMatt/2022-08-async-man-update

Add a background processor which is async

19 months agoMerge pull request #1695 from TheBlueMatt/2022-08-log-chan_update
Matt Corallo [Tue, 6 Sep 2022 19:15:05 +0000 (19:15 +0000)]
Merge pull request #1695 from TheBlueMatt/2022-08-log-chan_update

Ensure we log private channel_updates at a non-GOSSIP log level

19 months agoEnsure we log private channel_updates at a non-GOSSIP log level 2022-08-log-chan_update
Matt Corallo [Mon, 5 Sep 2022 16:28:11 +0000 (16:28 +0000)]
Ensure we log private channel_updates at a non-GOSSIP log level

If we receive a channel_update for one of our private channels, we
will not log the message at the usual TRACE log level as the
message falls into the gossip range. However, for our own channels
they aren't *just* gossip, as we store that info and it changes
how we generate invoices. Thus, we add a log in `ChannelManager`
here at the DEBUG log level.

19 months agoAdd a background processing function that is async. 2022-08-async-man-update
Matt Corallo [Tue, 9 Aug 2022 06:01:10 +0000 (06:01 +0000)]
Add a background processing function that is async.

Adds a method which operates like BackgroundProcessor::start but
instead of functioning by spawning a background thread it is async.

19 months agoAdd a `Future` which can receive manager persistence events
Matt Corallo [Tue, 9 Aug 2022 04:15:21 +0000 (04:15 +0000)]
Add a `Future` which can receive manager persistence events

This allows users who don't wish to block a full thread to receive
persistence events.

The `Future` added here is really just a trivial list of callbacks,
but from that we can build a (somewhat ineffecient)
std::future::Future implementation and can (at least once a mapping
for Box<dyn Trait> is added) include the future in no-std bindings
as well.

Fixes #1595

19 months agoRemove internal references to `persistence` in waker.rs
Matt Corallo [Mon, 29 Aug 2022 21:13:58 +0000 (21:13 +0000)]
Remove internal references to `persistence` in waker.rs

20 months agoMerge pull request #1692 from TheBlueMatt/2022-08-time-goes-backwards
Matt Corallo [Sat, 3 Sep 2022 16:15:33 +0000 (16:15 +0000)]
Merge pull request #1692 from TheBlueMatt/2022-08-time-goes-backwards

Handle monotonic clock going backwards during runtime

20 months agoMerge pull request #1650 from valentinewallace/2022-08-take-onionmsgs-public
Matt Corallo [Fri, 2 Sep 2022 22:20:32 +0000 (22:20 +0000)]
Merge pull request #1650 from valentinewallace/2022-08-take-onionmsgs-public

20 months agoHandle monotonic clock going backwards during runtime 2022-08-time-goes-backwards
Matt Corallo [Fri, 2 Sep 2022 21:57:32 +0000 (21:57 +0000)]
Handle monotonic clock going backwards during runtime

We've had some users complain that `duration_since` is panic'ing
for them. This is possible if the machine being run on is buggy and
the "monotonic clock" goes backwards, which sadly some ancient
systems can do.

Rust addressed this issue in 1.60 by forcing
`Instant::duration_since` to not panic if the machine is buggy
(and time goes backwards), but for users on older rust versions we
do the same by hand here.

20 months agoMove open_zero_conf_channel utility to common test utils
Matt Corallo [Fri, 2 Sep 2022 21:02:13 +0000 (21:02 +0000)]
Move open_zero_conf_channel utility to common test utils

20 months agoLimit OnionMessenger outbound buffer size
Valentine Wallace [Thu, 1 Sep 2022 19:26:17 +0000 (15:26 -0400)]
Limit OnionMessenger outbound buffer size

Drop OMs if they push us over the max OnionMessenger outbound buffer size

20 months agoDon't construct OnionMessage while holding peer lock
Valentine Wallace [Fri, 2 Sep 2022 17:41:41 +0000 (13:41 -0400)]
Don't construct OnionMessage while holding peer lock

20 months agoRefuse to send and forward OMs to disconnected peers
Valentine Wallace [Thu, 4 Aug 2022 15:05:07 +0000 (11:05 -0400)]
Refuse to send and forward OMs to disconnected peers

We also refuse to connect to peers that don't advertise onion message
forwarding support.

20 months agoExpose onion message module as public
Valentine Wallace [Wed, 3 Aug 2022 15:42:54 +0000 (11:42 -0400)]
Expose onion message module as public

And fix warnings

20 months agoMerge pull request #1643 from jurvis/jurvis/2022-07-inflights-htlcs-across-payments
Jeffrey Czyz [Thu, 1 Sep 2022 20:01:30 +0000 (15:01 -0500)]
Merge pull request #1643 from jurvis/jurvis/2022-07-inflights-htlcs-across-payments

Track in-flight HTLCs across payments when routing

20 months agoMake payment tests more realistic
jurvis [Sun, 28 Aug 2022 06:07:50 +0000 (23:07 -0700)]
Make payment tests more realistic

Made sure that every hop has a unique receipient. When we simulate
calling `channel_penalty_msat` in `TestRouter`’s find route, use
actual previous node ids instead of just using the payer’s.

20 months agoKeep track of inflight HTLCs across payments
jurvis [Tue, 30 Aug 2022 05:50:44 +0000 (22:50 -0700)]
Keep track of inflight HTLCs across payments

Added two methods, `process_path_inflight_htlcs` and
`remove_path_inflight_htlcs`, that updates that `payment_cache` map with
path information that may have failed, succeeded, or have been given up
on.

Introduced `AccountForInflightHtlcs`, which will wrap our user-provided
scorer. We move the `S:Score` type parameterization from the `Router` to
`find_route`, so we can use our newly introduced
`AccountForInflightHtlcs`.

`AccountForInflightHtlcs` keeps track of a map of inflight HTLCs by
their short channel id, direction, and give us the value that is being
used up.

This map will in turn be populated prior to calling `find_route`, where
we’ll use `create_inflight_map`, to generate a current map of all
inflight HTLCs based on what was stored in `payment_cache`.

20 months agoMerge pull request #1687 from TheBlueMatt/2022-08-fuzz-msrv
Matt Corallo [Tue, 30 Aug 2022 17:12:47 +0000 (17:12 +0000)]
Merge pull request #1687 from TheBlueMatt/2022-08-fuzz-msrv

Drop honggfuzz `arbitrary` dependency to meet MSRV

20 months agoDrop honggfuzz `arbitrary` dependency to meet MSRV 2022-08-fuzz-msrv
Matt Corallo [Tue, 30 Aug 2022 04:03:47 +0000 (04:03 +0000)]
Drop honggfuzz `arbitrary` dependency to meet MSRV

20 months agoChange `payment_cache` to accept `PaymentInfo`
jurvis [Tue, 30 Aug 2022 05:49:24 +0000 (22:49 -0700)]
Change `payment_cache` to accept `PaymentInfo`

Introduces a new `PaymentInfo` struct that contains both the previous
`attempts` count that was tracked as well as the paths that are also
currently inflight.

20 months agoMerge pull request #1604 from valentinewallace/2022-07-OMs-followup
Jeffrey Czyz [Mon, 29 Aug 2022 17:17:36 +0000 (12:17 -0500)]
Merge pull request #1604 from valentinewallace/2022-07-OMs-followup

Onion messages: add some initial rate limiting

20 months agoImplement buffering onion messages for peers.
Valentine Wallace [Thu, 25 Aug 2022 18:30:29 +0000 (14:30 -0400)]
Implement buffering onion messages for peers.

In this commit, we check if a peer's outbound buffer has room for onion
messages, and if so pulls them from an implementer of a new trait,
OnionMessageProvider.

Makes sure channel messages are prioritized over OMs, and OMs are prioritized
over gossip.

The onion_message module remains private until further rate limiting is added.

20 months agoImplement OnionMessageProvider for OnionMessenger
Valentine Wallace [Mon, 22 Aug 2022 16:22:22 +0000 (12:22 -0400)]
Implement OnionMessageProvider for OnionMessenger

20 months agoPeerManager: bump the read pause limit
Valentine Wallace [Tue, 16 Aug 2022 18:33:10 +0000 (14:33 -0400)]
PeerManager: bump the read pause limit

...to make sure we can still get channel messages out after enqueuing some big gossip messages.

20 months agoAdd boilerplate for sending and receiving onion messages in PeerManager
Valentine Wallace [Sat, 6 Aug 2022 04:33:48 +0000 (00:33 -0400)]
Add boilerplate for sending and receiving onion messages in PeerManager

Adds the boilerplate needed for PeerManager and OnionMessenger to work
together, with some corresponding docs and misc updates mostly due to the
PeerManager public API changing.

20 months agoMerge pull request #1683 from valentinewallace/2022-08-multiqueue-peerman
Matt Corallo [Thu, 25 Aug 2022 21:30:15 +0000 (21:30 +0000)]
Merge pull request #1683 from valentinewallace/2022-08-multiqueue-peerman

20 months agoMerge pull request #1673 from TheBlueMatt/2022-08-may-learn-preimage-balance
Matt Corallo [Thu, 25 Aug 2022 20:26:30 +0000 (20:26 +0000)]
Merge pull request #1673 from TheBlueMatt/2022-08-may-learn-preimage-balance

Expose a `Balance` for inbound HTLCs even without a preimage

20 months agoSeparate gossip broadcasts into their own queue in PeerManager
Valentine Wallace [Wed, 24 Aug 2022 22:04:58 +0000 (18:04 -0400)]
Separate gossip broadcasts into their own queue in PeerManager

This allows us to better prioritize channel messages over gossip broadcasts and
lays groundwork for rate limiting onion messages more simply, since they won't
be competing with gossip broadcasts for space in the main message queue.

20 months agoPeerMan: rename drop_gossip util to be more accurate
Valentine Wallace [Thu, 25 Aug 2022 17:58:49 +0000 (13:58 -0400)]
PeerMan: rename drop_gossip util to be more accurate

It's more accurate to name it as dropping gossip broadcasts, as it won't drop
all gossip.

20 months agoPeerMan: fix bug in drop_gossip util
Valentine Wallace [Thu, 25 Aug 2022 17:56:19 +0000 (13:56 -0400)]
PeerMan: fix bug in drop_gossip util

Fixes a flipped bool that was introduced in
4a1ee5f9a984c9b0c0892025d624ade734337b1a

20 months agoRename MaybeClaimableHTLCAwaitingTimeout for consistency 2022-08-may-learn-preimage-balance
Matt Corallo [Wed, 24 Aug 2022 22:16:04 +0000 (22:16 +0000)]
Rename MaybeClaimableHTLCAwaitingTimeout for consistency

As we now have a MaybePreimageClaimableHTLC, its more consistent
to rename `MaybeClaimableHTLCAwaitingTimeout` to
`MaybeTimeoutClaimableHTLC`.

20 months agoExpose a `Balance` for inbound HTLCs even without a preimage
Matt Corallo [Wed, 17 Aug 2022 20:15:23 +0000 (20:15 +0000)]
Expose a `Balance` for inbound HTLCs even without a preimage

If we don't currently have the preimage for an inbound HTLC, that
does not guarantee we can never claim it, but instead only that we
cannot claim it unless we receive the preimage from the channel we
forwarded the channel out on.

Thus, we cannot consider a channel to have no claimable balances if
the only remaining output on the commitment ransaction is an
inbound HTLC for which we do not have the preimage, as we may be
able to claim it in the future.

This commit addresses this issue by adding a new `Balance` variant
- `MaybePreimageClaimableHTLCAwaitingTimeout`, which is generated
until the HTLC output is spent.

Fixes #1620

20 months agoMerge pull request #1682 from tnull/2022-08-export-all-log-levels
Matt Corallo [Wed, 24 Aug 2022 18:41:46 +0000 (18:41 +0000)]
Merge pull request #1682 from tnull/2022-08-export-all-log-levels

20 months agoMerge pull request #1652 from valentinewallace/2022-08-reply-paths
Jeffrey Czyz [Wed, 24 Aug 2022 14:55:20 +0000 (09:55 -0500)]
Merge pull request #1652 from valentinewallace/2022-08-reply-paths

Onion messages: support reply paths

20 months agoExport and document all `log` macros.
Elias Rohrer [Wed, 24 Aug 2022 11:59:58 +0000 (13:59 +0200)]
Export and document all `log` macros.

Previously, only `log_error` and `log_trace` macros have been exported.
This change exports the macros of all log levels, which enables them to
be used downstream.

20 months agoSupport sending and receiving reply paths
Valentine Wallace [Fri, 5 Aug 2022 22:03:12 +0000 (18:03 -0400)]
Support sending and receiving reply paths

20 months agoFix bug in onion message payload decode
Valentine Wallace [Fri, 5 Aug 2022 21:50:02 +0000 (17:50 -0400)]
Fix bug in onion message payload decode

Previously, we were decoding payload lengths as a VarInt. Per the spec, this is
wrong -- it should be decoded as a BigSize.  This bug also exists in our
payment payload decoding, to be fixed separately.

Upcoming reply path tests caught this bug because we hadn't encoded a payload
greater than 253 before, so we hadn't hit the problem that VarInts are encoded
as little-endian whereas BigSizes are encoded as big-endian.

20 months agoOM functional tests: update util to take nodes by reference
Valentine Wallace [Fri, 5 Aug 2022 21:45:43 +0000 (17:45 -0400)]
OM functional tests: update util to take nodes by reference

And fix one test to be uniform with the others

20 months agoMerge pull request #1679 from TheBlueMatt/2022-08-no-double-access
Arik [Tue, 23 Aug 2022 20:11:36 +0000 (13:11 -0700)]
Merge pull request #1679 from TheBlueMatt/2022-08-no-double-access

Avoid querying the chain for outputs for channels we already have

20 months agoAvoid querying the chain for outputs for channels we already have 2022-08-no-double-access
Matt Corallo [Sun, 21 Aug 2022 20:34:22 +0000 (20:34 +0000)]
Avoid querying the chain for outputs for channels we already have

If we receive a ChannelAnnouncement message but we already have the
channel, there's no reason to do a chain lookup. Instead of
immediately calling the user-provided `chain::Access` when handling
a ChannelAnnouncement, we first check if we have the corresponding
channel in the graph.

Note that if we do have the corresponding channel but it was not
previously checked against the blockchain, we should still check
with the `chain::Access` and update if necessary.

20 months agoMerge pull request #1681 from dunxen/2022-08-githubactionsfix
Jeffrey Czyz [Tue, 23 Aug 2022 14:22:48 +0000 (09:22 -0500)]
Merge pull request #1681 from dunxen/2022-08-githubactionsfix

Workaround GH Actions issue for 1.45

20 months agoWorkaround GH Actions issue for 1.45
Duncan Dean [Tue, 23 Aug 2022 07:20:35 +0000 (09:20 +0200)]
Workaround GH Actions issue for 1.45

20 months agoMerge pull request #1677 from ok300/ok300-patch-1
Matt Corallo [Fri, 19 Aug 2022 22:18:59 +0000 (22:18 +0000)]
Merge pull request #1677 from ok300/ok300-patch-1

Bump bitcoin_hashes to v0.11

20 months agoMerge pull request #1676 from arik-so/2022-08-gossip-error-message-fix
Matt Corallo [Fri, 19 Aug 2022 19:03:08 +0000 (19:03 +0000)]
Merge pull request #1676 from arik-so/2022-08-gossip-error-message-fix

Fix script order in gossip key mismatch error.

20 months agoMerge pull request #1575 from NicolaLS/hash-invoice
Matt Corallo [Fri, 19 Aug 2022 17:45:11 +0000 (17:45 +0000)]
Merge pull request #1575 from NicolaLS/hash-invoice

20 months agoFix script order in gossip key mismatch error.
Arik Sosman [Fri, 19 Aug 2022 17:07:55 +0000 (10:07 -0700)]
Fix script order in gossip key mismatch error.

20 months agoderive Hash for Invoice
NicolaLS [Mon, 27 Jun 2022 12:41:46 +0000 (14:41 +0200)]
derive Hash for Invoice

20 months agoBump bitcoin_hashes to v0.11
ok300 [Fri, 19 Aug 2022 14:40:19 +0000 (16:40 +0200)]
Bump bitcoin_hashes to v0.11

20 months agoMerge pull request #1670 from TheBlueMatt/2022-08-mon-size-guidelines
Matt Corallo [Thu, 18 Aug 2022 16:33:20 +0000 (16:33 +0000)]
Merge pull request #1670 from TheBlueMatt/2022-08-mon-size-guidelines

Provide guidance on ChannelMonitorUpdate serialized size

20 months agoMerge pull request #1495 from TheBlueMatt/2022-04-all-claimables
Matt Corallo [Wed, 17 Aug 2022 22:50:49 +0000 (22:50 +0000)]
Merge pull request #1495 from TheBlueMatt/2022-04-all-claimables

Expose counterparty-revoked-outputs in `get_claimable_balance`

20 months agoProvide guidance on ChannelMonitorUpdate serialized size 2022-08-mon-size-guidelines
Matt Corallo [Mon, 15 Aug 2022 19:30:32 +0000 (19:30 +0000)]
Provide guidance on ChannelMonitorUpdate serialized size

Users need to make decisions about storage sizing and we need to
have advice on the maximum size of various things users need to
store. ChannelMonitorUpdates are likely the worst case of this,
they're usually at max a few KB, but can get up to a few hundred
KB for commitment transactions that have 400+ HTLCs pending.

We had one user report an update (likely) going over 400 KiB, which
isn't immediately obvious to me is practical, but its within a few
multiples of trivially-reachable sizes, so its likely that did
occur. To be on the safe side, we simply recommend users ensure
they can support "upwards of 1 MiB" here.

20 months agoMerge pull request #1648 from valentinewallace/2022-08-fuzz-onion-msgs
valentinewallace [Wed, 17 Aug 2022 19:48:29 +0000 (15:48 -0400)]
Merge pull request #1648 from valentinewallace/2022-08-fuzz-onion-msgs

Onion messages: add fuzz testing

20 months agoMerge pull request #1666 from TheBlueMatt/2022-08-fix-script-check
Matt Corallo [Wed, 17 Aug 2022 19:34:26 +0000 (19:34 +0000)]
Merge pull request #1666 from TheBlueMatt/2022-08-fix-script-check

Correct the on-chain script checked in gossip verification

20 months agoCorrect the on-chain script checked in gossip verification 2022-08-fix-script-check
Matt Corallo [Sat, 13 Aug 2022 17:29:06 +0000 (17:29 +0000)]
Correct the on-chain script checked in gossip verification

The `bitcoin_key_1` and `bitcoin_key_2` fields in
`channel_announcement` messages are sorted according to node_ids
rather than the keys themselves, however the on-chain funding
script is sorted according to the bitcoin keys themselves. Thus,
with some probability, we end up checking that the on-chain script
matches the wrong script and rejecting the channel announcement.

The correct solution is to use our existing channel funding script
generation function which ensure we always match what we generate.

This was found in testing the Java bindings, where a test checks
that retunring the generated funding script in `chain::Access`
results in the constructed channel ending up in our network graph.

20 months agoFuzz test onion messages
Valentine Wallace [Sun, 24 Jul 2022 18:47:31 +0000 (14:47 -0400)]
Fuzz test onion messages

Also update the fuzz ChaCha20Poly1305 to not mark as finished after a single
encrypt_in_place.  This is because more bytes may still need to be encrypted,
causing us to panic at the assertion that finished == false when we go to
encrypt more.

Also fix unused_mut warning in messenger + add log on OM forward for testing

20 months agoMerge pull request #1660 from TheBlueMatt/2022-08-cleanup-ratelimits
Matt Corallo [Tue, 16 Aug 2022 04:43:02 +0000 (04:43 +0000)]
Merge pull request #1660 from TheBlueMatt/2022-08-cleanup-ratelimits

Backfill gossip without buffering directly in LDK

20 months agoMove per-HTLC logic out of get_claimable_balances into a helper 2022-04-all-claimables
Matt Corallo [Sat, 16 Jul 2022 20:41:45 +0000 (20:41 +0000)]
Move per-HTLC logic out of get_claimable_balances into a helper

Val suggested this as an obvious cleanup to separate per_HTLC logic
from the total commitment transaction logic, separating the large
function into two.

20 months agoExpose counterparty-revoked-outputs in `get_claimable_balance`
Matt Corallo [Tue, 24 May 2022 23:57:56 +0000 (23:57 +0000)]
Expose counterparty-revoked-outputs in `get_claimable_balance`

This uses the various new tracking added in the prior commits to
expose a new `Balance` type - `CounterpartyRevokedOutputClaimable`.

Some nontrivial work is required, however, as we now have to track
HTLC outputs as spendable in a transaction that comes *after* an
HTLC-Success/HTLC-Timeout transaction, which we previously didn't
need to do. Thus, we have to check if an
`onchain_events_awaiting_threshold_conf` event spends a commitment
transaction's HTLC output while walking events. Further, because
we now need to track HTLC outputs after the
HTLC-Success/HTLC-Timeout confirms, and because we have to track
the counterparty's `to_self` output as a contentious output which
could be claimed by either party, we have to examine the
`OnchainTxHandler`'s set of outputs to spend when determining if
certain outputs are still spendable.

Two new tests are added which test various different transaction
formats, and hopefully provide good test coverage of the various
revoked output paths.

20 months agoDrop uneccessary `if {...; bool}` pattern in PeerManager 2022-08-cleanup-ratelimits
Matt Corallo [Mon, 15 Aug 2022 19:38:45 +0000 (19:38 +0000)]
Drop uneccessary `if {...; bool}` pattern in PeerManager

20 months agoReplace manual iteration with for loops in outbound gossip sync
Matt Corallo [Mon, 15 Aug 2022 18:08:13 +0000 (18:08 +0000)]
Replace manual iteration with for loops in outbound gossip sync

20 months agoSimplify claimable balance trivially with consistent accessors
Matt Corallo [Wed, 13 Jul 2022 01:20:09 +0000 (01:20 +0000)]
Simplify claimable balance trivially with consistent accessors

20 months agoScan `onchain_events_awaiting_threshold_conf` once in balance calc
Matt Corallo [Tue, 17 May 2022 20:45:17 +0000 (20:45 +0000)]
Scan `onchain_events_awaiting_threshold_conf` once in balance calc

Instead of a series of different
`onchain_events_awaiting_threshold_conf.iter()...` calls to scan
for HTLC status in balance calculation, pull them all out into one
`for ... { match ... }` to do it once and simplify the code
somewhat.

20 months agoTrack the txid that resolves HTLCs even after resolution completes
Matt Corallo [Sat, 21 May 2022 01:11:52 +0000 (01:11 +0000)]
Track the txid that resolves HTLCs even after resolution completes

We need this information when we look up if we still need to spend
a revoked output from an HTLC-Success/HTLC-Timeout transaction for
balance calculation.

20 months agoTrack HTLC-Success/HTLC-Timeout claims of revoked outputs
Matt Corallo [Thu, 19 May 2022 01:50:37 +0000 (01:50 +0000)]
Track HTLC-Success/HTLC-Timeout claims of revoked outputs

When a counterparty broadcasts a revoked commitment transaction,
followed immediately by HTLC-Success/-Timeout spends thereof, we'd
like to have an `onchain_events_awaiting_threshold_conf` entry
for them.

This does so using the `HTLCSpendConfirmation` entry, giving it
(slightly) new meaning. Because all existing uses of
`HTLCSpendConfirmation` already check if the relevant commitment
transaction is revoked first, this should be trivially backwards
compatible.

We will ultimately figure out if something is being spent via the
`OnchainTxHandler`, but to do so we need to look up the output via
the HTLC transaction txid, which this allows us to do.

20 months agoFix off-by-one in test_onchain_htlc_claim_reorg_remote_commitment
Matt Corallo [Tue, 17 May 2022 23:57:52 +0000 (23:57 +0000)]
Fix off-by-one in test_onchain_htlc_claim_reorg_remote_commitment

The test intended to disconnect a transaction previously connected
but didn't disconnect enough blocks to do so, leading to it
confirming two conflicting transactions.

In the next few commits this will become an assertion failure.

20 months agoClean up nesting in `get_counterparty_output_claim_info`
Matt Corallo [Thu, 14 Jul 2022 21:23:39 +0000 (21:23 +0000)]
Clean up nesting in `get_counterparty_output_claim_info`

20 months agoTrack counterparty payout info in counterparty commitment txn
Matt Corallo [Sat, 30 Apr 2022 20:29:31 +0000 (20:29 +0000)]
Track counterparty payout info in counterparty commitment txn

When handling a revoked counterparty commitment transaction which
was broadcast on-chain, we occasionally need to look up which
output (and its value) was to the counterparty (the `to_self`
output). This will allow us to generate `Balance`s for the user for
the revoked output.

20 months agoStore the full event transaction in `OnchainEvent` structs
Matt Corallo [Fri, 13 May 2022 05:11:14 +0000 (05:11 +0000)]
Store the full event transaction in `OnchainEvent` structs

When we see a transaction which generates some `OnchainEvent`, its
useful to have the full transaction around for later analysis.
Specifically, it lets us check the list of outputs which were spent
in the transaction, allowing us to look up, e.g. which HTLC
outpoint was spent in a transaction.

This will be used in a few commits to do exactly that - figure out
which HTLC a given `OnchainEvent` corresponds with.

20 months agoMerge pull request #1663 from tnull/2022-08-drop-register-output-return
Matt Corallo [Mon, 15 Aug 2022 23:10:32 +0000 (23:10 +0000)]
Merge pull request #1663 from tnull/2022-08-drop-register-output-return

Drop return value from `Filter::register_output`

20 months agoBackfill gossip without buffering directly in LDK
Matt Corallo [Tue, 9 Aug 2022 21:26:16 +0000 (21:26 +0000)]
Backfill gossip without buffering directly in LDK

Instead of backfilling gossip by buffering (up to) ten messages at
a time, only buffer one message at a time, as the peers' outbound
socket buffer drains. This moves the outbound backfill messages out
of `PeerHandler` and into the operating system buffer, where it
arguably belongs.

Not buffering causes us to walk the gossip B-Trees somewhat more
often, but avoids allocating vecs for the responses. While its
probably (without having benchmarked it) a net performance loss, it
simplifies buffer tracking and leaves us with more room to play
with the buffer sizing constants as we add onion message forwarding
which is an important win.

Note that because we change how often we check if we're out of
messages to send before pinging, we slightly change how many
messages are exchanged at once, impacting the
`test_do_attempt_write_data` constants.

20 months agoClarify 'should' vs 'will' in `get_relevant_txids`
Elias Rohrer [Thu, 11 Aug 2022 12:33:22 +0000 (14:33 +0200)]
Clarify 'should' vs 'will' in `get_relevant_txids`

20 months agoDrop return value from `Filter::register_output`
Elias Rohrer [Thu, 11 Aug 2022 12:27:45 +0000 (14:27 +0200)]
Drop return value from `Filter::register_output`

This commit removes the return value from `Filter::register_output` as
creating a suitable value almost always entails blocking operations
(e.g., lookups via network request), which however conflicts with the
requirement that user calls should avoid blocking calls at all cost.

Removing the return value also rendered quite a bit of test code for
dependent transaction handling superfluous, which is therefore also
removed with this commit.

20 months agoMerge pull request #1659 from valentinewallace/2022-08-fix-payload-decode
valentinewallace [Mon, 15 Aug 2022 18:32:47 +0000 (14:32 -0400)]
Merge pull request #1659 from valentinewallace/2022-08-fix-payload-decode

Fix payment onion payload decode

20 months agoFix fuzzer-found underflow
Valentine Wallace [Fri, 5 Aug 2022 17:15:57 +0000 (13:15 -0400)]
Fix fuzzer-found underflow

20 months agoRename HighZeroBytesDroppedVarInt to HighZeroBytesDroppedBigSize
Valentine Wallace [Sun, 7 Aug 2022 03:06:19 +0000 (23:06 -0400)]
Rename HighZeroBytesDroppedVarInt to HighZeroBytesDroppedBigSize

As observed by @wpaulino, this struct encodes its bytes as big-endian,
therefore it's a BigSize, not a VarInt.