At the end of our `monitor_tests`, which test `ChannelMonitor`
`SpendableOutputs` and claimable `Balance`s, add new checks that
ensure that, if we're using the new
`ConnectStyle::HighlyRedundantTransactionsFirstSkippingBlocks`, we
can replay the full chain without getting redundant events or
`Balance`s.
In many complexity-reduced implementations of chain syncing using
esplora `transactions_confirmed` may be called redundantly for
transactions which were already confirmed. To ensure this is
idempotent we add two new `ConnectionStyle`s in our tests which
(a) call `transactions_confirmed` twice for each call, ensuring
simple idempotency is ensured and (b) call `transactions_confirmed`
once for each historical block every time we're connecting a new
block, ensuring we're fully idempotent even if every call is
repeated constantly.
In order to actually behave correctly this requires a simple
already-confirmed check in `ChannelMonitor`, which is included.
If no network graph is provided to the `BackgroundProcessor`, we
log every time the processor loop goes around (at least every
100ms, if not more) which fille up logs with useless indications
that we have no network graph.
For Windows build only, the
`TestPersister::chain_sync_monitor_persistences` lock has a lock order
before the `ChannelManager::per_peer_state` lock. This fix ensures that
the `per_peer_state` lock isn't held before the
`TestPersister::chain_sync_monitor_persistences` lock is acquired.
As the `channel_state` lock will be removed, we prepare for that by
flipping the lock order for `pending_inbound_payments` and
`pending_outbound_payments` locks to before the `channel_state` lock.
Now that `handle_channel_resumption` can't fail, the error handling
in `post_handle_chan_restoration` is now dead code. Removing it
makes `post_handle_chan_restoration` only a single block, so here
we simply remove the macro and inline the single block into the two
places the macro was used.
In c70bd1f, we implemented tracking HTLCs by adding path information
for pending HTLCs to `InvoicePayer`’s `payment_cache` when receiving
specific events.
Since we can now track inflight HTLCs entirely within ChannelManager,
there is no longer a need for this to exist.
In `ChannelMonitor`s, when a transaction containing a spend of a
revoked remote output reaches 6 confs, we may have no other
tracking of that txid remaining. Thus, if we see that transaction
again (because a user duplicatively confirms it), we'll generate a
redundant spendable output event for it.
Here we simply explicitly track all txids of transactions which
confirm with a spendable output, allowing us to check this
condition in the next commit.
In `ChannelMonitor`, if we see a `transaction_unconfirmed` for a
transaction we last saw in a block at height X, we shouldn't
*only* remove the `onchain_events_awaiting_threshold_conf` entry
for the given tx but rather for all transactions that we last saw
at height >= X.
This avoids any potential `onchain_events_awaiting_threshold_conf`
inconsistencies due to the order in whcih users mark transactions
unconfirmed (which the `chain::Confirm` docs do not currently set
any requirements on).
This also matches the `OnchainTxHandler` behavior, which does the
same lookup.
In the next commit we'll add some checks that redundant
transactions aren't confirmed in different blocks, which would
cause test_htlc_ignore_latest_remote_commitment to fail. Here we
fix it to avoid the issue.
BOLT 12 messages are limited to a range of TLV record types. Refactor
decode_tlv_stream into a decode_tlv_stream_range macro for limiting
which types are parsed. Requires a SeekReadable trait for rewinding when
a type outside of the range is seen. This allows for composing TLV
streams of different ranges.
Updates offer parsing accordingly and adds a test demonstrating failure
if a type outside of the range is included.
Add common bech32 parsing for BOLT 12 messages. The encoding is similar
to bech32 only without a checksum and with support for continuing
messages across multiple parts.
Messages implementing Bech32Encode are parsed into a TLV stream, which
is converted to the desired message content while performing semantic
checks. Checking after conversion allows for more elaborate checks of
data composed of multiple TLV records and for more meaningful error
messages.
The parsed bytes are also saved to allow creating messages with mirrored
data, even if TLV records are unknown.
There is no reason anymore for `handle_chan_restoration_locked` to
be a macro, and our long-term desire is to move away from macros as
they substantially bloat our compilation time (and binary size).
Thus, we simply remove `handle_chan_restoration_locked` here and
turn it into a function.
When we process a `channel_reestablish` message we free the HTLC
update holding cell as things may have changed while we were
disconnected. However, some time ago, to handle freeing from the
holding cell when a monitor update completes, we added a holding
cell freeing check in `get_and_clear_pending_msg_events`. This
leaves the in-`channel_reestablish` holding cell clear redundant,
as doing it immediately or is `get_and_clear_pending_msg_events` is
not a user-visible difference.
Thus, we remove the redundant code here, substantially simplifying
`handle_chan_restoration_locked` while we're at it.
Asserting that specific log entries were printed isn't all that
useful, we should really be focusing on the expected messages (or,
when a monitor udpate fails, the lack thereof). In the next commit
one of these log checks would otherwise break due to the particular
time a monitor update fails changing, but I also plan on reworking
the montior update flows substantially soon, breaking lots of them.
As it was previously omitted, we clarify here starting from which version users can expect the `user_channel_id` to be randomized for inbound channels.
LND nodes have very broken fee estimators, causing them to suggest
feerates that don't even meet a current mempool minimum feerate
when fees go up over the course of hours. This can cause us to
reject their feerate estimates as they're not high enough, even
though their new feerate is higher than what we had already (which
is the feerate we'll use to broadcast a closing transaction). This
implies we force-close the channel and broadcast something with a
feerate lower than our counterparty was offering.
Here we simply accept such feerates as they are better than what we
had. We really should also close the channel, but only after we
get their signature on the new feerate. That should happen by
checking channel feerates every time we see a new block so is
orthogonal to this code.
Ultimately the fix is anchor outputs plus package-based relay in
Bitcoin Core, however we're still quite some ways from that, so
worth needlessly closing channels for now.
When we mark a future as complete, if the user is using the
`std::future::Future` impl to get notified, we shouldn't just
assume we have completed the `Future` when we call the `Waker`. A
`Future` may have been `drop`'d at that point (or may not be
`poll`'d again) even though we wake the `Waker`.
Because we now have a `callbacks_made` flag, we can fix this rather
trivially, simply not setting the flag until the `Future` is
`poll`'d `Complete`.
When we return from one of the wait functions in `Notifier`, we
should also ensure that the next `Future` doesn't start in the
`complete` state, as we have already notified the user, as far as
we're concerned.
This is technically a regression from the previous commit, but as
it is a logically separate change it is in its own commit.
If a `Notifier` gets `notify()`ed and the a `Future` is fetched,
even though the `Future` is marked completed from the start and
the user may pass callbacks which are called, we'll never wipe the
needs-notify bit in the `Notifier`.
The solution is to keep track of the `FutureState` in the returned
`Future` even though its `complete` from the start, adding a new
flag in the `FutureState` which indicates callbacks have been made
and checking that flag when waiting or returning a second `Future`.
