If the fuzz target is failing due to a channel force-close, the
immediately-visible error is that we're signing a stale state. This
is because the ChannelMonitorUpdateStep::ChannelForceClosed event
results in a signature in the test clone which was deserialized
using a OnlyReadsKeysInterface. Instead, we need to deserialize
using the full KeysInterface instance.
Previously, if we got disconnected from a peer while there were
HTLCs pending forwarding in the holding cell, we'd clear them and
fail them all backwards. This is largely fine, but since we now
have support for handling such HTLCs on reconnect, we might as
well not, instead relying on our timeout logic to fail them
backwards if it takes too long to forward them.
Because we may now generate a monitor update during
get_and_clear_pending_msg_events calls, we need to ensure we
re-serialize the relevant ChannelManager before attempting to
reload it, if such a monitor update occurred.
If there is no pending channel update messages when monitor updating
is restored (though there may be an RAA to send), and we're
connected to our peer and not awaiting a remote RAA, we need to
free anything in our holding cell.
However, we don't want to immediately free the holding cell during
channel_monitor_updated as it presents a somewhat bug-prone case of
reentrancy:
a) it would re-enter user code around a monitor update while being
called from user code notifying us of the same monitor being
updated, making deadlocs very likely (in fact, our fuzzers
would have a bug here!),
b) the re-entrancy only occurs in a very rare case, making it
likely users will not hit it in testing, only deadlocking in
production.
Thus, we add a holding-cell-free pass over each channel in
get_and_clear_pending_msg_events. This fits up nicely with the
anticipated bug - users almost certainly need to process new
network messages immediately after monitor updating has been
restored to send messages which were not sent originally when the
monitor updating was paused.
Without this, chanmon_fail_consistency was able to find a stuck
condition where we sit on an HTLC failure in our holding cell and
don't ever handle it (at least until we have other actions to take
which empty the holding cell).
Both break_chan_entry and try_chan_entry do almost identical work,
only differing on if they `break` or `return` in response to an
error. Because we will now also need an option to do neither, we
break out the common code into a shared `convert_chan_err` macro.
Because of the merge between peer reconnection and channel monitor
updating channel restoration code, we now sometimes generate
(somewhat spurious) announcement signatures when restoring channel
monitor updating. This should not result in a fuzzing failure.
This fails chanmon_consistency on IgnoreError error events and on
messages left over to be sent to a just-disconnected peer, which
should have been drained.
These should never appear, so consider them a fuzzer fail case.
The channel restoration code in channel monitor updating and peer
reconnection both do incredibly similar things, and there is
little reason to have them be separate. Sadly because they require
holding a lock with a reference to elements in the lock, its not
practical to make them utility functions, so instead we introduce
a two-step macro here which will eventually be used for both.
Because we still support pre-NLL Rust, the macro has to be in two
parts - one which runs with the channel_state lock, and one which
does not.
Our fuzz tests previously only printed the log output of the first
fuzz test case to fail. This commit changes that (with lots of
auto-generated updates) to ensure we print all log outputs.
We currently generate duplicative PaymentFailed/PaymentSent events
in two cases:
a) If we receive a update_fulfill_htlc message, followed by a
disconnect, then a resend of the same update_fulfill_htlc
message, we will generate a PaymentSent event for each message.
b) When a Channel is closed, any outbound HTLCs which were relayed
through it are simply dropped when the Channel is. From there,
the ChannelManager relies on the ChannelMonitor having a copy of
the relevant fail-/claim-back data and processes the HTLC
fail/claim when the ChannelMonitor tells it to.
If, due to an on-chain event, an HTLC is failed/claimed, and
then we serialize the ChannelManager, but do not re-serialize
the relevant ChannelMonitor, we may end up getting a duplicative
event.
In order to provide the expected consistency, we add explicit
tracking of pending outbound payments using their unique
session_priv field which is generated when the payment is sent.
Then, before generating PaymentFailed/PaymentSent events, we check
that the session_priv for the payment is still pending.
Thix fixes#209.
To avoid caller data struct storing HTLC-related information when
a revokeable output is claimed on top of a commitment/second-stage
HTLC transactions, we split `keysinterface::sign_justice_transaction`
in two new halves `keysinterfaces::sign_justice_revoked_output` and
`keysinterfaces::sign_justice_revoked_htlc`.
Further, this split offers more flexibility to signer policy as a
commitment revokeable output might be of a value far more significant
than HTLC ones.
When we had a event which caused us to set the persist flag in a
PersistenceNotifier in between wait calls, we will still wait,
potentially not persisting a ChannelManager when we should.
Worse, for wait_timeout, this caused us to always wait up to the
timeout, but then always return true that a persistence is needed.
Instead, we simply check the persist flag before waiting, returning
immediately if it is set.
Currently, when a user calls `ChannelManager::timer_tick_occurred`
we always set the persister's update flag to true. This results in
a ChannelManager persistence after each timer tick, even when
nothing happened.
Instead, we add a new flag to `PersistenceNotifierGuard` to
indicate if we should skip setting the update flag.
Currently, we only send an update_channel message after
disconnecting a peer and waiting some time. We do not send a
followup when the peer has been reconnected for some time.
This changes that behavior to make the disconnect and reconnect
channel updates symmetric, and also simplifies the state machine
somewhat to make it more clear.
Finally, it serializes the current announcement state so that we
usually know when we need to send a new update_channel.
Early sample testing showed multiple users hitting
EWOULDBLOCK/EAGAIN waiting for an initial response from Bitcoin
Core while it was doing some long operation (eg UTXO cache
flushing). Instead of only waiting 5 seconds for each attempt, we
now wait a full two minutes, but only for the first header
response, not each byte.
