Correct + update scoring attempts right at the maximum amounts
What the developer wrote
Authored by Matt Corallo
Correct + update scoring attempts right at the maximum amounts
While sending a payment at exactly a channel's max or the liquidity
upper bound we've calculated is unlikely to work, we shouldn't
consider it impossible. Here we fix a few cases where we previously
did so.
This isn't really fixing anything, as we only shift the boundary by
1 msat, but it makes the next commit more logical.
What changed, and why it matters
This commit tweaks how the Lightning payment router estimates whether a payment can fit through a channel. Previously, sending an amount exactly equal to a channel's maximum capacity or estimated liquidity upper bound was treated as impossible (maximum penalty). Now it is treated as merely very unlikely (a large but finite penalty), shifting the boundary by 1 millisatoshi. The commit message explicitly says this is not really fixing anything, just making the next commit more logical. There is no indication this is a security fix or that it can be exploited.
No security action required. Treat as a normal routing-quality improvement. If reviewing a series, evaluate the follow-up commit for any actual security relevance.
Security signals we found
No security-relevant keywords in commit title or message
No unsafe code, no input parsing, no serialization format changes
No bounds-checking or integer-overflow fixes
Change is explicitly described by the author as not fixing anything meaningful
Evidence from the diff
The change is in lightning/src/routing/scoring.rs. It changes two boundary checks from >= to > so that amounts exactly equal to max_liquidity_msat or available_capacity no longer trigger the considered_impossible_penalty_msat or max historical penalty. Test expectations are updated accordingly, shifting the ‘impossible’ threshold up by 1 msat. This is a scoring/routing behavior adjustment, not a memory-safety, cryptographic, or consensus bug.
Changed components
lightning/src/routing/scoring.rsProbabilisticScorer channel penalty calculationInspect captured patch +19 / −13
diff --git a/lightning/src/routing/scoring.rs b/lightning/src/routing/scoring.rs
index c2d3e51..7130c92 100644
--- a/lightning/src/routing/scoring.rs
+++ b/lightning/src/routing/scoring.rs
@@ -1464,8 +1464,7 @@ impl<
// liquidity penalty at all (as the success probability is 100%).
} else if total_inflight_amount_msat >= max_liquidity_msat {
// Equivalent to hitting the else clause below with the amount equal to the effective
- // capacity and without any certainty on the liquidity upper bound, plus the
- // impossibility penalty.
+ // capacity and without any certainty on the liquidity upper bound.
let negative_log10_times_2048 = NEGATIVE_LOG10_UPPER_BOUND * 2048;
res = Self::combined_penalty_msat(amount_msat, negative_log10_times_2048,
score_params.liquidity_penalty_multiplier_msat,
@@ -1489,12 +1488,11 @@ impl<
}
}
- if total_inflight_amount_msat >= max_liquidity_msat {
+ if total_inflight_amount_msat > max_liquidity_msat {
res = res.saturating_add(score_params.considered_impossible_penalty_msat);
}
if total_inflight_amount_msat >= available_capacity {
- // We're trying to send more than the capacity, use a max penalty.
res = res.saturating_add(Self::combined_penalty_msat(amount_msat,
NEGATIVE_LOG10_UPPER_BOUND * 2048,
score_params.historical_liquidity_penalty_multiplier_msat,
@@ -3214,6 +3212,8 @@ mod tests {
let usage = ChannelUsage { amount_msat: 250, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 300);
let usage = ChannelUsage { amount_msat: 500, ..usage };
+ assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 2000);
+ let usage = ChannelUsage { amount_msat: 501, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), u64::max_value());
let usage = ChannelUsage { amount_msat: 750, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), u64::max_value());
@@ -3431,22 +3431,22 @@ mod tests {
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 0);
let usage = ChannelUsage { amount_msat: 1, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 0);
- let usage = ChannelUsage { amount_msat: 1_023, ..usage };
- assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 2_000);
let usage = ChannelUsage { amount_msat: 1_024, ..usage };
+ assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 2_000);
+ let usage = ChannelUsage { amount_msat: 1_025, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), u64::max_value());
// Fully decay liquidity upper bound.
scorer.time_passed(Duration::from_secs(10 * 9));
let usage = ChannelUsage { amount_msat: 0, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 0);
- let usage = ChannelUsage { amount_msat: 1_024, ..usage };
+ let usage = ChannelUsage { amount_msat: 1_025, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), u64::max_value());
scorer.time_passed(Duration::from_secs(10 * 10));
let usage = ChannelUsage { amount_msat: 0, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 0);
- let usage = ChannelUsage { amount_msat: 1_024, ..usage };
+ let usage = ChannelUsage { amount_msat: 1_025, ..usage };
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), u64::max_value());
}
@@ -3520,7 +3520,7 @@ mod tests {
let mut scorer = ProbabilisticScorer::new(decay_params, &network_graph, &logger);
let source = source_node_id();
let usage = ChannelUsage {
- amount_msat: 500,
+ amount_msat: 501,
inflight_htlc_msat: 0,
effective_capacity: EffectiveCapacity::Total { capacity_msat: 1_000, htlc_maximum_msat: 1_000 },
};
@@ -3535,10 +3535,10 @@ mod tests {
assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), u64::max_value());
scorer.time_passed(Duration::from_secs(10));
- assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 473);
+ assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 477);
scorer.payment_path_failed(&payment_path_for_amount(250), 43, Duration::from_secs(10));
- assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 300);
+ assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 304);
let mut serialized_scorer = Vec::new();
scorer.write(&mut serialized_scorer).unwrap();
@@ -3546,7 +3546,7 @@ mod tests {
let mut serialized_scorer = io::Cursor::new(&serialized_scorer);
let deserialized_scorer =
<ProbabilisticScorer<_, _>>::read(&mut serialized_scorer, (decay_params, &network_graph, &logger)).unwrap();
- assert_eq!(deserialized_scorer.channel_penalty_msat(&candidate, usage, ¶ms), 300);
+ assert_eq!(deserialized_scorer.channel_penalty_msat(&candidate, usage, ¶ms), 304);
}
#[rustfmt::skip]
@@ -3577,7 +3577,13 @@ mod tests {
info,
short_channel_id: 42,
});
- assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), u64::max_value());
+ assert_eq!(scorer.channel_penalty_msat(&candidate, usage, ¶ms), 2000);
+
+ let over_usage = ChannelUsage {
+ amount_msat: 501,
+ ..usage
+ };
+ assert_eq!(scorer.channel_penalty_msat(&candidate, over_usage, ¶ms), u64::max_value());
if decay_before_reload {
scorer.time_passed(Duration::from_secs(10));
Why this scored 16/100
Community notes
Notes can correct, qualify, or add evidence to the AI analysis. Every note shown here has been validated by a human moderator.
The AI analysis stands alone for now. Submit a note if you can add evidence or important context.