This commit adds a check during router's startup and fails the inflight
HTLCs if they are routing using channels unknown to us. The channels are
unknown because they are already closed, usually long time ago.
Adds a utility function to be able to compute the outgoing routing
amount from the incoming amount by taking inbound and outbound fees into
account. The discussion was contributed by user feelancer21, see
f6f05fa930.
This commit breaks the ChannelConstraints structure into two
sub-structures that reflect the fundamental differences in how
these parameters are used. On its face it may not seem necessary,
however the distinction introduced here is relevant for how we
will be implementing the Dynamic Commitments proposal.
We shift the duty of determining the policies to the backward pass as
the forward pass will only be responsible for finding the corrected
receiver amount.
Note that this is not a pure refactor as demonstrated in the test, as
the forward pass doesn't select new policies anymore, which is less
flexible and doesn't lead to the highest possible receiver amount. This
is however neccessary as we otherwise won't be able to compute
forwarding amounts involving inbound fees and this edge case is unlikely
to occur, because we search for a min amount for a route that was most
likely constructed for a larger amount.
Continue adding some complexity behind the BlindedPaymentPathSet. What
we do here is add a new IntroNodeOnlyPath method. The assumption we
make here is: If multiple blinded paths are provided to us in an invoice
but one of those paths only includes an intro node, then there is no
point in looking at any other path since we know that the intro node is
the destination node. So in such a case, we would have discarded any
other path in the `NewBlindedPaymentPathSet` constructor. So then we
would only have a single blinded path made up of an introduction node
only. In this specific case, in the `newRoute` function, no edge passed
to the function would have a blindedPayment associated with it (since
there are no blinded hops in this case). So we will have a case where
`blindedPathSet` passed to `newRoute` is not nil but `blindedPayment` is
nil since nonce was extacted from any edge. If this happens then we can
assume that this is the Intro-Node-Only situation described above. And
so we grabe the associated payment from the path set.
Instead of needing to remember how to handle the FinalCLTV value of a
blinded payment path at various points in the code base, we hide the
logic behind a unified FinalCLTVDelta method on the blinded path.
If multiple blinded paths are provided, they will each have a different
pub key for the destination node. This makes using our existing
pathfinding logic tricky since it depends on having a single destination
node (characterised by a single pub key). We want to re-use this logic.
So what we do is swap out the pub keys of the destinaion hop with a
pseudo target pub key. This will then be used during pathfinding. Later
on once a path is found, we will swap the real destination keys back in
so that onion creation can be done.
This commit introduces a new type, `BlindedPaymentPathSet`. For now, it
holds only a single `BlindedPayment` but eventually it will hold and
manage a set of blinded payments provided for a specific payment. To
make the PR easier to follow though, we start off just letting it hold a
single one and do some basic replacements.
We split up the functionality in getRouteUnifiers into checking that all
edges exist via getEdgeUnifiers and then add a backward pass that will
be responsible for determining the sender amount.
We remove the node pub key from the error string, as in route building
this is duplicate info, which can be determined from the input keys,
further it's not available in the backward pass anymore.
We refactor the BuildRoute test to use the require library and add a
test case for a max HTLC violation on the last hop.
The time lock weight for a hop is supposed to be proportional to the
amount that is sent/locked, but in a previous change we switched to the
net amount, where inbound fees aren't yet applied. This is corrected in
this commit.
When iterating edges, pathfinding checks early whether using an edge
would violate the requested total fee limit for a route. This check is
done on the net amount (an amount the inbound fee is calculated with).
However, a possible next hop's fee discount leads to a reduction in fees
and as such using the net amount leads to assuming a higher cumulative
fee than the route really has, excluding the path erroneously. We
perform the fee limit check on the amount to send, which includes both
inbound and outbound fees. This should be possible as the first hop's
outbound fee is zero and therefore doesn't have to be checked in the
end.
This commit introduces more sophisticated code for selecting dummy hop
policy values for dummy hops in blinded paths.
For the case where the path does contain real hops, the dummy hop policy
values are derived by taking the average of those hop polices. For the
case where there are no real hops (in other words, we are the
introduction node), we use the default policy values used for normal
ChannelUpdates but then for the MaxHTLC value, we take the average of
all our open channel capacities.
Make various sender side adjustments so that a sender is able to send an
MP payment to a single blinded path without actually including an MPP
record in the payment.
Add a `FindBlindedPaths` method to the `ChannelRouter` which will use
the new `findBlindedPaths` function to get a set of candidate blinded
path routes. It then uses mission control to select the best of these
paths.
Note that as of this commit, the MC data we get from these queries won't
mean much since we wont have data about a channel in the direction
towards us. But we do this now in preparation for a future PR which will
start writing mission control success pairs for successful receives from
blinded route paths.
This commit adds a new function, `findBlindedPaths`, that does a depth
first search from the target node to find a set of blinded paths to the
target node given the set of restrictions. This function will select and
return any candidate path. A candidate path is a path to the target node
with a size determined by the given hop number constraints where all the
nodes on the path signal the route blinding feature _and_ the
introduction node for the path has more than one public channel. Any
filtering of paths based on payment value or success probabilities is
left to the caller.
This commit is a large refactor that moves over various responsibilities
from the ChannelRouter to the graph.Builder. These include all graph
related tasks such as:
- graph pruning
- validation of new network updates & persisting new updates
- notifying topology update clients of any changes.
This is a large commit but:
- many of the files are purely moved from `routing` to `graph`
- the business logic put in the graph Builder is copied exactly as is
from the ChannelRouter with one exception:
- The ChannelRouter just needs to be able to call the Builder's
`ApplyChannelUpdate` method. So this is now exported and provided to
the ChannelRouter as a config option.
- The trickiest part was just moving over the test code since quite a
bit had to be duplicated.
In preparation for adding a clean Graph DB interface, we create a
version of FetchLightningNode that doesnt allow a caller to provide in a
transaction.
In this commit, we completely remove the Router's dependence on a Graph
source that requires a `kvdb.RTx`. In so doing, we are more prepared for
a future where the Graph source is backed by different DB structure such
as pure SQL.
The two areas affected here are: the ChannelRouter's graph access that
it uses for pathfinding. And the SessionSource's graph access that it
uses for payments.
The ChannelRouter gets given a Graph and the SessionSource is given a
GraphSessionFactory which it can use to create a new session. Behind the
scenes, this will acquire a kvdb.RTx that will be used for calls to the
Graph's `ForEachNodeChannel` method.
In preparation for structs outside of the `routing` package implementing
this interface, export `routingGraph` and rename it to `Graph` so as to
avoid stuttering.
In this commit, we further reduce the routingGraph interface and this
time we make it more node-agnostic so that it can be backed by any graph
and not one with a concept of "sourceNode".
