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