Blinded pathfinding groundwork #2146

valentinewallace · 2023-04-03T22:24:33Z

This lays some groundwork commits for #2120. Split out because we'll need some more interface changes and I don't want to hold up the serialization breakage portion, which we want to land for 0.0.115.

valentinewallace · 2023-04-03T22:34:17Z

Oops, benchmark made me realize that the scoring methods need updates too. Will push in a bit.

codecov-commenter · 2023-04-04T01:08:16Z

Codecov Report

Patch coverage: 84.70% and project coverage change: +0.30 🎉

Comparison is base (b8ed4d2) 91.39% compared to head (b131634) 91.70%.

📣 This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2146      +/-   ##
==========================================
+ Coverage   91.39%   91.70%   +0.30%     
==========================================
  Files         104      104              
  Lines       51431    53043    +1612     
  Branches    51431    53043    +1612     
==========================================
+ Hits        47007    48643    +1636     
+ Misses       4424     4400      -24

Impacted Files	Coverage Δ
lightning/src/lib.rs	`67.74% <ø> (ø)`
lightning/src/offers/invoice_request.rs	`93.40% <ø> (ø)`
lightning/src/offers/offer.rs	`91.31% <ø> (+1.42%)`	⬆️
lightning/src/offers/refund.rs	`96.70% <ø> (+2.33%)`	⬆️
lightning/src/offers/test_utils.rs	`96.72% <ø> (ø)`
lightning/src/onion_message/packet.rs	`75.53% <ø> (+0.78%)`	⬆️
lightning/src/routing/router.rs	`95.20% <ø> (-0.15%)`	⬇️
lightning/src/util/ser.rs	`89.42% <ø> (-0.20%)`	⬇️
lightning/src/events/mod.rs	`34.19% <5.88%> (+0.17%)`	⬆️
lightning/src/ln/functional_test_utils.rs	`92.72% <60.00%> (-0.02%)`	⬇️
... and 19 more

... and 35 files with indirect coverage changes

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

valentinewallace · 2023-04-05T17:26:13Z

Rebased.

lightning/src/util/ser.rs

lightning/src/ln/channelmanager.rs

lightning/src/routing/router.rs

TheBlueMatt

Generally lgtm, note that we'll want to squash down the commits that fix things after the path introduction to avoid having broken commits in the history. To make it simpler, I wonder if it isn't easier to split the vec->path commit into one commit that does that transition without adding the blinded_tail field, and then a separate one that does.

lightning/src/routing/router.rs

TheBlueMatt · 2023-04-07T17:58:22Z

lightning/src/routing/router.rs

 			}
 		}
+		if blinded_tails.is_some() && blinded_tails.as_ref().unwrap().len() != self.paths.len() {
+			return Err(io::Error::new(io::ErrorKind::Other, "Missing blinded tail for path (if blinded tails are included, there must be 1 set per path)"))


Can we just make the serialization format support this by pushing the option onto the vec? I mean I guess we hopefully won't ever actually use this, but in general all of our in-memory structs have a strong guarantee that they will never fail to serialize as long as the underlying Write doesn't fail a push.

I'm assuming we don't want to write anything unless a blinded tail is actually present, since it's an even tlv. So we could do what you're suggesting, but that would unnecessarily allocate a vec of Nones in most cases, and then we'd have to check that none of them are Some before writing, which all seems unfortunate(?).

I think it would be ok to just remove this check, since we have strong in-memory guarantees as you say?

On second thought, let's support writing the Nones? We don't have to always allocate - we can allocate the required number when we find the first blinded_tail set. The issue here is we dont have any in-memory guarantees that there arent some paths that have a blinded tail and others that dont, and while its kinda nonsense, its also the case that someone could construct such a route and we could pay it without issue.

lightning/src/routing/router.rs

TheBlueMatt · 2023-04-07T18:04:23Z

Also needs rebase, sorry :(.

jkczyz

Concept ACK

lightning/src/routing/router.rs

lightning/src/routing/scoring.rs

valentinewallace · 2023-04-12T19:36:53Z

Rebased.

lightning/src/routing/router.rs

TheBlueMatt · 2023-04-18T20:22:31Z

lightning/src/routing/router.rs

+
+/// A path in a [`Route`] to the payment recipient. Must always be at least length one. While the
+/// maximum length of any given path is variable, keeping the length of any path less than or equal
+/// to 19 should currently ensure it is viable.


Is this true with blinded paths, we have to give each hop a pubkey right, so the max length is much shorter.

Ah I meant that path.len() should be <=19, which includes blinded hops. Will clarify.

But my point is that its not 19, as each blinded hop is substantially longer than non-blinded ones, the maximum length is variable. cc #2201

we have to give each hop a pubkey right

I believe we only need to give the intro node's payload a (blinding point) pubkey. But there is more data in blinded hop payloads, yeah.

I updated the comment to caveat that 19 only applies to paths without a blinded tail. Do you want to go through updating MAX_PATH_LENGTH_ESTIMATE and other docs updates in this PR, or save this for later addressing #2201?

We can address in 2201, I think, we're not generating such paths yet anyway.

TheBlueMatt · 2023-04-18T20:32:19Z

lightning/src/routing/router.rs

 			}
 		}
+		if blinded_tails.is_some() && blinded_tails.as_ref().unwrap().len() != self.paths.len() {
+			return Err(io::Error::new(io::ErrorKind::Other, "Missing blinded tail for path (if blinded tails are included, there must be 1 set per path)"))


On second thought, let's support writing the Nones? We don't have to always allocate - we can allocate the required number when we find the first blinded_tail set. The issue here is we dont have any in-memory guarantees that there arent some paths that have a blinded tail and others that dont, and while its kinda nonsense, its also the case that someone could construct such a route and we could pay it without issue.

lightning/src/ln/channelmanager.rs

TheBlueMatt · 2023-04-18T20:48:15Z

lightning/src/routing/router.rs

 		let mut max_path_offset = payment_params.max_total_cltv_expiry_delta - path_total_cltv_expiry_delta;
 		max_path_offset = cmp::max(
 			max_path_offset - (max_path_offset % MEDIAN_HOP_CLTV_EXPIRY_DELTA),
 			max_path_offset % MEDIAN_HOP_CLTV_EXPIRY_DELTA);
 		shadow_ctlv_expiry_delta_offset = cmp::min(shadow_ctlv_expiry_delta_offset, max_path_offset);

 		// Add 'shadow' CLTV offset to the final hop
-		if let Some(last_hop) = path.last_mut() {
+		if let Some(tail) = path.blinded_tail.as_mut() {
+			tail.final_cltv_expiry_delta = tail.final_cltv_expiry_delta


Shouldn't we skip adding it to the last hop in this case?

For the current docs to be accurate, we would need to mirror the cltv_expiry_delta in path.hops.last() and BlindedTail for 1-hop blinded paths. We could add a special case to the docs and make it 0 in the last hop, if you prefer.

Oh oh, right, because the last hop includes the blinded CLTV delta, okay, nvm.

Okay, wait, I'm confused on this again - the way I read the current docs the blinded path's final_cltv_expiry_delta is the final cltv expiry delta the recipient wants to receive, and the cltv_expiry_delta of the last hop is the difference between that and the introduction point's cltv expiry. Thus, if we only want to change the last hop's cltv (and have it bubble up through the route) we should just change the blinded path part.

Ah, yeah, you're right. Updating this and the docs. I had the last hop as containing the entire blinded path's cltv delta, including the recipient's, but that doesn't make sense.

Btw, I'm not entirely sure how to get the final_cltv_expiry_delta for the last hop. It appears that eclair computes it as 0 + random_shadow_offset. The route blinding spec indicates that a max_cltv_expiry should be provided by the receiver in the BOLT12 invoice, but that doesn't seem to be a thing in the offers spec.

Ah! I guess the cltv delta in the BlindedPayInfo includes the recipient, then? Let's just remove the cltv delta from the blinded tail, then?

valentinewallace · 2023-04-20T14:16:00Z

Squashed and rebased.

lightning/src/routing/router.rs

lightning/src/util/macro_logger.rs

lightning/src/events/mod.rs

jkczyz · 2023-04-20T20:56:49Z

lightning/src/routing/router.rs

+				if let Some(blinded_tail) = &path.blinded_tail {
+					if blinded_tails.is_empty() {
+						blinded_tails = Vec::with_capacity(path.hops.len());
+						for _ in 0..idx {
+							blinded_tails.push(None);
+						}
+					}
+					blinded_tails.push(Some(blinded_tail));
+				} else if !blinded_tails.is_empty() { blinded_tails.push(None); }


Should it be the case that if one path has a blinded tail then all should? Should we fail if that's not the case?

Originally I required all-or-nothing for Path BlindedTails, but ended up allowing it because even though we'll never construct such Route's, we'd have no issue paying them. See #2146 (comment)

It should, but I complained because in general we have a strong guarantee that if something exists in-memory, it can be serialized. We don't have a reasonable way to provide/handle errors serializing objects, so we generally panic when writing to a sink that cannot fail (eg &mut Vec::new()). If we're particularly worried about it, we should make the fields private and enforce the preconditions when updating/creating a route, but I'm not sure we need to be - while somewhat nonsensical, it is possible to pay such a route just fine.

jkczyz · 2023-04-20T20:58:12Z

lightning/src/routing/router.rs

+				BlindedHop { blinded_node_id: ln_test_utils::pubkey(49), encrypted_payload: Vec::new() }
+			],
+		};
+		// (De)serialize a Route with 1 blinded path out of two total paths.


Likewise, would this ever happen?

Probably not, but it was broken in a prior iteration so thought I'd test it. Fine to remove.

lightning/src/routing/scoring.rs

jkczyz · 2023-04-20T21:12:13Z

lightning/src/routing/router.rs

@@ -2111,6 +2111,8 @@ fn add_random_cltv_offset(route: &mut Route, payment_params: &PaymentParameters,
 	let network_nodes = network_graph.nodes();

 	for path in route.paths.iter_mut() {
+		if path.blinded_tail.as_ref().map_or(false, |tail| tail.hops.len() > 1) { continue }


Could probably use a comment as to why this will skip.

I'm also confused as to why we're doing this? I guess we just assume there's enough privacy if we're paying a blinded path? Maybe let's make the constant 2 instead?

I think I was concerned we might blow up the cltv expiry if blinded path recipients add their own shadow offsets, but I think it makes more sense to just always add a shadow offset for blinded paths, rather than only adding it for 1-hop blinded paths. Updated.

TheBlueMatt

Given we're refusing to pay blinded paths and we moved the value calculation to happen before the addition of the blinded paths object, I don't really care if this is squashed anymore or not. Because our value calculations are right I'm okay with not, even if its a bit weird our serialization is lossy in the intermediate commits.

lightning/src/routing/router.rs

TheBlueMatt · 2023-04-20T21:46:53Z

lightning/src/routing/router.rs

 		let mut max_path_offset = payment_params.max_total_cltv_expiry_delta - path_total_cltv_expiry_delta;
 		max_path_offset = cmp::max(
 			max_path_offset - (max_path_offset % MEDIAN_HOP_CLTV_EXPIRY_DELTA),
 			max_path_offset % MEDIAN_HOP_CLTV_EXPIRY_DELTA);
 		shadow_ctlv_expiry_delta_offset = cmp::min(shadow_ctlv_expiry_delta_offset, max_path_offset);

 		// Add 'shadow' CLTV offset to the final hop
-		if let Some(last_hop) = path.last_mut() {
+		if let Some(tail) = path.blinded_tail.as_mut() {
+			tail.final_cltv_expiry_delta = tail.final_cltv_expiry_delta


Okay, wait, I'm confused on this again - the way I read the current docs the blinded path's final_cltv_expiry_delta is the final cltv expiry delta the recipient wants to receive, and the cltv_expiry_delta of the last hop is the difference between that and the introduction point's cltv expiry. Thus, if we only want to change the last hop's cltv (and have it bubble up through the route) we should just change the blinded path part.

TheBlueMatt · 2023-04-20T21:47:12Z

lightning/src/routing/router.rs

@@ -2111,6 +2111,8 @@ fn add_random_cltv_offset(route: &mut Route, payment_params: &PaymentParameters,
 	let network_nodes = network_graph.nodes();

 	for path in route.paths.iter_mut() {
+		if path.blinded_tail.as_ref().map_or(false, |tail| tail.hops.len() > 1) { continue }


I'm also confused as to why we're doing this? I guess we just assume there's enough privacy if we're paying a blinded path? Maybe let's make the constant 2 instead?

TheBlueMatt · 2023-04-20T21:48:57Z

lightning/src/ln/outbound_payment.rs

@@ -721,8 +718,8 @@ impl OutboundPayments {
 			}
 		};
 		for path in route.paths.iter() {
-			if path.len() == 0 {
-				log_error!(logger, "length-0 path in route");
+			if path.hops.len() == 0 {


Wait, isn't this actually okay? If we're the introduction node for a blinded path away from us the hops should be empty?

In the case that we're the intro node, we'll want to advance the blinded path before pathfinding (otherwise how can we pathfind to ourselves?). So I think we-are-the-intro-node should be an error in payment parameters initialization.

Therefore, I think hops.len() will always be > 0? If that's not the case, I have a bunch of path.hops.first().unwrap()s to update..

Ah, okay, I wasn't sure what the approach was, that makes sense.

TheBlueMatt · 2023-04-21T15:44:51Z

LGTM, feel free to squash.

This lays groundwork for adding blinded path info to Path

valentinewallace · 2023-04-21T15:49:07Z

Squashed without diff

valentinewallace · 2023-04-21T18:37:12Z

Squashed the commit disallowing paying blinded payment paths into the commit that adds blinded path info to Path. So we shouldn't need to squash the last 10 commits into one anymore

…ed paths for now

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch from 9ffccf0 to 510db9f Compare April 3, 2023 22:28

valentinewallace added this to the 0.0.115 milestone Apr 3, 2023

valentinewallace marked this pull request as draft April 4, 2023 19:34

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch 3 times, most recently from 6e98fa5 to b5f4349 Compare April 5, 2023 17:25

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch from b5f4349 to f2a9bb2 Compare April 5, 2023 18:09

valentinewallace marked this pull request as ready for review April 5, 2023 21:26

dunxen reviewed Apr 6, 2023

View reviewed changes

lightning/src/util/ser.rs Outdated Show resolved Hide resolved

lightning/src/ln/channelmanager.rs Show resolved Hide resolved

valentinewallace commented Apr 6, 2023

View reviewed changes

lightning/src/routing/router.rs Show resolved Hide resolved

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch from f2a9bb2 to 86056fc Compare April 6, 2023 20:05

TheBlueMatt reviewed Apr 7, 2023

View reviewed changes

jkczyz reviewed Apr 7, 2023

View reviewed changes

lightning/src/routing/router.rs Outdated Show resolved Hide resolved

lightning/src/routing/router.rs Outdated Show resolved Hide resolved

lightning/src/routing/router.rs Outdated Show resolved Hide resolved

lightning/src/routing/scoring.rs Outdated Show resolved Hide resolved

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch 5 times, most recently from 4d1c9a0 to 96e3ff0 Compare April 12, 2023 19:36

TheBlueMatt reviewed Apr 13, 2023

View reviewed changes

lightning/src/routing/router.rs Outdated Show resolved Hide resolved

lightning/src/routing/router.rs Outdated Show resolved Hide resolved

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch 3 times, most recently from d74e4ed to 35e6e58 Compare April 18, 2023 18:13

TheBlueMatt reviewed Apr 18, 2023

View reviewed changes

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch 2 times, most recently from a526988 to 9861527 Compare April 19, 2023 17:15

jkczyz reviewed Apr 20, 2023

View reviewed changes

TheBlueMatt reviewed Apr 20, 2023

View reviewed changes

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch 2 times, most recently from dd8e87b to d974016 Compare April 21, 2023 15:33

valentinewallace added 3 commits April 21, 2023 11:48

Minor Route/RouteHop doc updates

b5827f7

Add utilities for getting a path's final value and cltv delta

93afed5

Replace Vec<RouteHop> with new Path struct

d5b05e5

This lays groundwork for adding blinded path info to Path

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch from d974016 to 6383245 Compare April 21, 2023 15:48

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch 2 times, most recently from 871a609 to 7631b6a Compare April 21, 2023 18:35

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch from 7631b6a to dc426a4 Compare April 21, 2023 19:34

valentinewallace added 9 commits April 21, 2023 15:35

Add blinded path {metadata} fields to Path, but disallow paying blind…

64c26c8

…ed paths for now

Support (de)ser for Path::blinded_tails in events

97a8c91

Support (de)serializing Path::blinded_tails in Routes

c8fd77d

Support (de)serializing Path::blinded_tails in HTLCSource

2e51a1c

Account for Path::blinded_tail in InflightHtlcs::process_path

dac4a1c

Test scoring paths with blinded tails

976411d

Account for Path::blinded_tail when adding a shadow cltv offset

5a6f9b0

Update changelog for backwards compat

5c2cf77

Fix outbound_payment for new Path::blinded_tail

b131634

valentinewallace force-pushed the 2023-03-blinded-pathfinding-groundwork branch from dc426a4 to b131634 Compare April 21, 2023 19:35

TheBlueMatt approved these changes Apr 21, 2023

View reviewed changes

jkczyz approved these changes Apr 24, 2023

View reviewed changes

TheBlueMatt merged commit 607727f into lightningdevkit:main Apr 24, 2023
14 checks passed

valentinewallace mentioned this pull request May 10, 2023

Routing to blinded payment paths #2120

Merged

1 task

jkczyz mentioned this pull request May 10, 2023

BOLT12 Offers Tracking Issue #1970

Open

57 tasks

Blinded pathfinding groundwork #2146

Blinded pathfinding groundwork #2146

Conversation

valentinewallace commented Apr 3, 2023

valentinewallace commented Apr 3, 2023

codecov-commenter commented Apr 4, 2023 • edited

Codecov Report

valentinewallace commented Apr 5, 2023

TheBlueMatt left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

TheBlueMatt commented Apr 7, 2023

jkczyz left a comment

Choose a reason for hiding this comment

valentinewallace commented Apr 12, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

valentinewallace commented Apr 20, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

valentinewallace Apr 20, 2023 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

TheBlueMatt left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

TheBlueMatt commented Apr 21, 2023

valentinewallace commented Apr 21, 2023

valentinewallace commented Apr 21, 2023

codecov-commenter commented Apr 4, 2023 •

edited

valentinewallace Apr 20, 2023 •

edited