Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add static invoice creation utils to ChannelManager #3408

Open
wants to merge 13 commits into
base: main
Choose a base branch
from

Conversation

valentinewallace
Copy link
Contributor

  1. Add static invoice creation utilities as part of supporting the async payments BOLTs spec Support async payments in BOLT 12 lightning/bolts#1149.

  2. Take this opportunity to more easily test some code added in Support paying static invoices #3140 that went untested at the time. This is the bulk of the diff.

  3. Address a piece of feedback from Support paying static invoices #3140 regarding InvoiceRequests being unavailable when the time comes to send the async payment, cc Support paying static invoices #3140 (comment)

@jkczyz jkczyz self-requested a review November 14, 2024 18:35
Comment on lines +10066 to +10156
const SECONDS_PER_BLOCK: u32 = 10 * 60;
let relative_expiry_blocks = relative_expiry_seconds / SECONDS_PER_BLOCK;
let max_cltv_expiry = core::cmp::max(relative_expiry_blocks, CLTV_FAR_FAR_AWAY)
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Note for reviewers: I'm not sure this is robust enough with block times being somewhat unreliable. Maybe it should be buffered?

Prior to this patch, we would take() the invoice request stored for
AwaitingInvoice outbound payments when retrying sending the invoice request
onion message. This doesn't work for async payments because we need to keep the
invoice request stored for inclusion in the payment onion. Therefore, clone it
instead of take()ing it.
Prior to this fix, we would attempt to mark outbound async payments as
abandoned but silently fail because they were in state AwaitingInvoice, which
the mark_abandoned utility doesn't currently work for. These payments would
eventually be removed by the remove_stale_payments method, but there would be a
delay in generating the PaymentFailed event.

Move to manually removing the outbound payment entry.
@valentinewallace valentinewallace marked this pull request as ready for review November 14, 2024 20:49
Comment on lines 138 to 142
pub(crate) struct RetryableInvoiceRequest {
pub(crate) invoice_request: InvoiceRequest,
pub(crate) nonce: Nonce,
pub(super) needs_retry: bool,
}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would it be worth not holding on to these for offers that don't support async payments? Can't recall if we can tell from the offer or if it isn't known until the static invoice is received.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

IIRC there is nothing in the offers, just at the invoice request level.

lightning/src/routing/router.rs Outdated Show resolved Hide resolved
Comment on lines +9602 to +9612
let amount_msat = offer.amount().and_then(|amount| {
match amount {
Amount::Bitcoin { amount_msats } => Some(amount_msats),
Amount::Currency { .. } => None
}
});
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is quantity_max allowed for offers paying often-offline nodes? If so, is the amount checked by them in some way? If not, should we prevent building a static invoice from such offers?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

quantity_max is currently allowed. I'm not sure I follow your question though:

If so, is the amount checked by them in some way?

Who is "them"? It looks like the payer checks the amount/quantity when sending an invreq but not sure I'm following.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Does the recipient check that the amount is sufficient for the requested quantity before releasing the preimage?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah seems they will have to, if that isn't done already. Currently we don't support receiving to the static invoices created by these utils (just creating them), so this is definitely something to consider in the follow-up.

Comment on lines +9581 to +9588
for path in message_paths_to_always_online_node {
builder = builder.path(path);
}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If the recipient is online, does the InvoiceRequest get forwarded along? If so, how does the recipient authenticate it?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If the recipient is online, does the InvoiceRequest get forwarded along?

Yep!

how does the recipient authenticate it?

Since the recipient issued the offer themselves, they authenticate it the same way as always-online recipients, i.e. via verify_using_recipient_data, if I'm understanding you.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Right, so I guess that means the recipient must construct these paths since it needs to include the Nonce used to derive the signing keys.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm... but the function is creating and returning the Nonce. I think we'll need to pass it in?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

To be clear, we don't include metadata in the offer when we have blinded paths. Can't recall if we do something different when constructing an offer for use with async payments.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I brought this up during review club. Since payment path contains the offer nonce -- and the sender will include the invoice request with the payment -- we should be able to verify the invoice request is authentic with this nonce. But ISTM we still want the same nonce in the offer's message paths for the normal case when the recipient is online.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah I misunderstood at first, this is a good point.... I'm not sure yet what a good solution would be.

For context, message_paths_to_always_online_node terminate at the always-online node. Currently in LDK the recipient expects to receive the invreq either over a blinded path from the offer or to their node id but with offer_metadata set (IIUC), but neither of those are supported in this PR atm.

For more context, I think the always-online node is expected to be a direct peer of the async receiver, otherwise an onion message round-trip would be incurred to check whether the recipient is online when the always-online node receives the invreq.

So it seems like our two options here are to either set the offer metadata for async receive offers (which may be a privacy leak), or for the recipient to provide the always-online node with a one-hop blinded path to send the invreq over, separately from their static invoice. Either of these options should allow the recipient to verify the invreq if they happen to be online at the time.

Any thoughts on this?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wonder how big that privacy leak is in practice? If we assume most BPs to async-recipients are to an LSP and probably 1-hop BPs for reliability, then its probably nothing (as such BPs are always to async-recipients). If we assume they're sometimes 2-hops, then its probably not huge but is at least something non-zero. It may also just be unintuitive that this leak exists and users might not notice it.

That said, how does the LSP currently pass the invreq to the recipient? Presumably there's some offer -> node registration process so the LSP has a DB with (offer, node, static_invoice) tuples? Replacing the node with a one-hop BP doesn't sound like all that much work, I think, depending on what the API from LDK looks like?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That said, how does the LSP currently pass the invreq to the recipient? Presumably there's some offer -> node registration process so the LSP has a DB with (offer, node, static_invoice) tuples? Replacing the node with a one-hop BP doesn't sound like all that much work, I think, depending on what the API from LDK looks like?

Yep. I think I agree the blinded path route is the way to go. It's more flexible in case async receivers prefer their offer's corresponding always-online node to be a hop or two away (IMO this would introduce too much payment latency in practice, but who knows), and avoids increasing the size of the offer via offer_metadata.

We may want a separate method on ChannelManager to create this blinded path, though it's actually already supported via MessageRouter since the recipient has the offer nonce. So I may punt on hashing out the details of this for now until LSP async payments support is more fleshed out.

Copy link
Collaborator

@TheBlueMatt TheBlueMatt left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice, thanks!

lightning/src/ln/outbound_payment.rs Outdated Show resolved Hide resolved
///
/// [`Offer`]: crate::offers::offer::Offer
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct AsyncBolt12OfferContext {
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

#3427 applies here too, though not sure if it needs to be addressed in this PR.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Tracking this on #2298. I'm inclined to build on Jeff's solution to #3427 to make sure we align with the rest of the offers code unless there are objections.

@valentinewallace valentinewallace mentioned this pull request Nov 26, 2024
22 tasks
Useful for creating payment paths for static invoices which are typically
amount-less.
Will be useful for static invoices' blinded paths, which may have long
expiries. Rather than having a default max_cltv_expiry, we now base it
on the invoice expiry.
This context is stored in the blinded payment paths we put in static invoices
and is useful to authenticate payments over these paths to the recipient.

We can't reuse Bolt12OfferContext for this because we don't have access to the
invoice request fields at static invoice creation time.
@@ -203,6 +203,15 @@ impl PaymentPurpose {
payment_context: context,
}
},
Some(PaymentContext::AsyncBolt12Offer(_context)) => {
debug_assert!(false, "Receiving async payments is not yet supported");
Copy link
Contributor Author

@valentinewallace valentinewallace Nov 27, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think due in part to #3427 this debug_assert may be reachable. I have a patch in an upcoming PR to change this method to return a Result (to handle the case where the sender doesn't include an invoice request when async paying) so I may just add that diff to this commit.

Edit: added 3a18f5a.

This context is included in static invoice's blinded message paths, provided
back to us in HeldHtlcAvailable onion messages for blinded path authentication.
In future work, we will check if this context is valid and respond with a
ReleaseHeldHtlc message to release the upstream payment if so.

We also add creation methods for the hmac used for authenticating the blinded path
using the static invoice's corresponding offer id.
We can't use our regular offer creation util for receiving async payments
because the recipient can't be relied on to be online to service
invoice_requests.

Therefore, add a new offer creation util that is parameterized by blinded
message paths to another node on the network that *is* always-online and can
serve static invoices on behalf of the often-offline recipient.

Also add a utility for creating static invoices corresponding to these offers.
See new utils' docs and BOLTs PR 1149 for more info.
Since adding support for creating static invoices from ChannelManager, it's
easier to test these failure cases that went untested when we added support for
paying static invoices.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants