1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
//! Outbound broadcast, interest-propagation, and state-sync helpers for
//! [`P2pConnManager`].
//!
//! Behavior-preserving extraction from `p2p_protoc.rs`.
use super::*;
impl P2pConnManager {
/// Broadcast a message to a list of peers off the event loop.
///
/// In production, the sends are spawned via `tokio::spawn` to prevent
/// self-deadlock: the bridge channel is drained by the event loop, so
/// inline fan-out of >capacity sends would block forever.
/// In simulation tests, sends run inline for determinism.
pub(super) async fn broadcast_to_peers(
bridge: &P2pBridge,
peers: Vec<SocketAddr>,
msg: crate::message::NetMessage,
) {
let bridge = bridge.clone();
let send_all = async move {
for peer_addr in peers {
// channel-safety: ok — in production this future is detached via
// `tokio::spawn(send_all)` below, so the blocking bridge.send cannot
// stall the event loop; it is only awaited inline under
// `simulation_tests` (single-threaded deterministic ordering).
if let Err(e) = bridge.send(peer_addr, msg.clone()).await {
tracing::warn!(
peer_addr = %peer_addr,
error = %e,
"Failed to send broadcast message to peer"
);
}
}
};
#[cfg(not(feature = "simulation_tests"))]
tokio::spawn(send_all);
#[cfg(feature = "simulation_tests")]
send_all.await;
}
/// Broadcast a hosting update message to all connected peers.
pub(super) async fn handle_hosting_broadcast(
&mut self,
message: crate::message::NeighborHostingMessage,
) {
tracing::debug!(
message = ?message,
peer_count = self.connections.len(),
phase = "broadcast",
"Broadcasting hosting update to connected peers"
);
let msg = crate::message::NetMessage::V1(crate::message::NetMessageV1::NeighborHosting {
message: message.clone(),
});
let peers: Vec<SocketAddr> = self.connections.keys().copied().collect();
Self::broadcast_to_peers(&self.bridge, peers, msg).await;
}
/// Broadcast ChangeInterests message to all connected peers.
pub(super) async fn handle_broadcast_change_interests(
&mut self,
added: Vec<u32>,
removed: Vec<u32>,
) {
tracing::debug!(
added_count = added.len(),
removed_count = removed.len(),
peer_count = self.connections.len(),
"Broadcasting ChangeInterests to connected peers"
);
let msg = crate::message::NetMessage::V1(crate::message::NetMessageV1::InterestSync {
message: crate::message::InterestMessage::ChangeInterests { added, removed },
});
let peers: Vec<SocketAddr> = self.connections.keys().copied().collect();
Self::broadcast_to_peers(&self.bridge, peers, msg).await;
}
/// Send an interest message to a specific peer.
pub(super) async fn handle_send_interest_message(
&mut self,
target: SocketAddr,
message: crate::message::InterestMessage,
) {
tracing::debug!(
target = %target,
"Sending interest message to peer"
);
let msg =
crate::message::NetMessage::V1(crate::message::NetMessageV1::InterestSync { message });
// Non-blocking (#4145): this handler is dispatched INLINE on the
// event-loop task from the NodeEvent::SendInterestMessage arm, so
// bridge.send()'s await on the cap-512 ev_listener_tx (drained by this
// same task) would self-stall. A dropped InterestSync is recovered by
// the next periodic InterestSync exchange / summary reconciliation.
if let Err(e) = self.bridge.try_send(target, msg) {
tracing::warn!(
peer_addr = %target,
error = %e,
"Failed to send interest message to peer (will re-sync periodically)"
);
}
}
/// Send an arbitrary `NetMessage` to a target peer without registering
/// a `pending_op_results` callback for the message's transaction.
///
/// Mirrors `handle_send_interest_message` but accepts a
/// fully-formed `NetMessage`. Used by the CONNECT joiner to
/// emit `ConnectFailed` upstream without disturbing its own
/// multi-reply receiver slot.
pub(super) async fn handle_send_net_message(
&mut self,
target: SocketAddr,
msg: crate::message::NetMessage,
) {
tracing::debug!(
target = %target,
tx = %msg.id(),
"Sending net message to peer"
);
// Non-blocking (#4145): dispatched INLINE on the event-loop task from
// the NodeEvent::SendNetMessage arm, so bridge.send()'s await on the
// cap-512 ev_listener_tx (drained by this same task) would self-stall.
// This path carries the CONNECT joiner's upstream ConnectFailed; if the
// message is dropped here, the upstream connect operation still fails on
// its own OPERATION_TTL rather than hanging — so dropping is safe, and a
// wedged event loop (the alternative) would be strictly worse.
if let Err(e) = self.bridge.try_send(target, msg) {
tracing::warn!(
peer_addr = %target,
error = %e,
"Failed to send net message to peer (upstream op falls back to TTL)"
);
}
}
/// Broadcast ReadyState to all connected ring peers.
pub(super) async fn handle_broadcast_ready_state(&mut self, ready: bool) {
let msg =
crate::message::NetMessage::V1(crate::message::NetMessageV1::ReadyState { ready });
tracing::info!(
ready,
peer_count = self.connections.len(),
"Broadcasting ReadyState to connected peers"
);
let peers: Vec<SocketAddr> = self.connections.keys().copied().collect();
Self::broadcast_to_peers(&self.bridge, peers, msg).await;
}
/// Maximum retry attempts when a broadcast finds no targets.
const MAX_BROADCAST_RETRIES: u8 = 3;
/// Base delay between broadcast retries (scaled by attempt number for linear backoff).
const BROADCAST_RETRY_BASE_DELAY: Duration = Duration::from_secs(1);
/// Maximum entries in the no-target streak tracker. Prevents unbounded growth
/// from network-influenced contract keys.
const MAX_BROADCAST_STREAK_ENTRIES: usize = 256;
/// Notify interested network peers about a state change.
///
/// Echo-back is prevented by summary comparison: we skip peers whose cached
/// summary matches ours (they already have our state).
///
/// When no targets are found (race between contract updates and subscription
/// establishment), the broadcast is retried with linear backoff up to
/// [`Self::MAX_BROADCAST_RETRIES`] times.
pub(super) async fn handle_broadcast_state_change(
&mut self,
op_manager: &Arc<OpManager>,
key: freenet_stdlib::prelude::ContractKey,
new_state: freenet_stdlib::prelude::WrappedState,
// `false` for a fresh executor-emitted broadcast, `true` for a
// no-target retry re-emission this handler scheduled. Used only to
// count fresh logical broadcasts once for the #4281 summary.
is_retry: bool,
// `true` when this is a deferred re-emission of a stashed fresh-contract
// broadcast (issue #4359). Suppresses re-recording a `no_targets`
// propagation-summary event on give-up (the originating PUT already
// counted one when it first gave up) so a flush per interested peer
// does not inflate the #4281 stats.
is_reemit: bool,
) {
tracing::debug!(
contract = %key,
state_size = new_state.size(),
"BroadcastStateChange event received"
);
// Phase 7 egress self-block (#4300). A locally-applied UPDATE
// (e.g. delegate-driven) drives this fan-out to subscribers who
// don't know about our ban. If we have banned the contract, skip
// the fan-out entirely — and the no-target retry re-emission with
// it — rather than push state for a contract we have decided is
// harmful. There is no client to notify here (this is the
// fire-and-forget broadcast path), so we skip with a debug log
// instead of returning a typed error. Complements the
// client-originated UPDATE gate in `start_client_update` and the
// receive-side `UpdateMsg::*` drop in node.rs (PR #4299).
if op_manager.ring.contract_ban_list.is_banned(key.id()) {
tracing::debug!(
contract = %key,
phase = "broadcast_state_change_banned_skip",
"skipping state-change broadcast for banned contract"
);
// Drop any in-flight retry/streak bookkeeping for this
// contract: if it was banned mid-retry-cycle, a previously
// scheduled re-emission would otherwise leave a stale entry
// here that nothing clears until the contract is unbanned and
// broadcast again. Mirrors the cleanup the targets-found and
// retries-exhausted paths perform.
self.broadcast_retries.remove(&key);
self.broadcast_no_target_streak.remove(&key);
return;
}
let self_addr = op_manager.ring.connection_manager.get_own_addr();
let Some(self_addr) = self_addr else {
tracing::warn!(
contract = %key,
"Cannot broadcast state change - no own address"
);
return;
};
let target_result = op_manager.get_broadcast_targets_update(&key, &self_addr);
tracing::debug!(
contract = %key,
target_count = target_result.targets.len(),
self_addr = %self_addr,
"BroadcastStateChange: found targets"
);
if target_result.targets.is_empty() {
// Record the NO_TARGETS outcome once per *fresh* no-target broadcast
// for the #4281 summary, but NOT for retry re-emissions. Gating on
// `is_retry` (carried on the event) rather than the per-contract
// `broadcast_retries` counter is what makes the count accurate:
// retries and fresh broadcasts share that counter, so a fresh
// broadcast arriving mid-retry-cycle would otherwise be absorbed
// and uncounted (Codex P2). With the explicit flag, every fresh
// no-target broadcast is counted exactly once and retries never
// double-count. If a later retry heals (finds targets), the
// targets-found path below additionally records that success — both
// the initial miss and the eventual propagation are surfaced.
//
// `is_reemit` (issue #4359) is also excluded: a deferred re-emission
// of a stashed fresh-contract broadcast that STILL finds no targets
// must not re-record a `no_targets` event — the originating PUT
// already counted one when it first gave up, and counting again per
// interested-peer flush would inflate the #4281 no_targets/targets_avg
// stats.
if !is_retry && !is_reemit {
op_manager.update_propagation_stats.record_broadcast(
*key.id(),
0,
target_result.interest_resolve_failed,
);
}
let retry_count = self.broadcast_retries.entry(key).or_insert(0);
if *retry_count < Self::MAX_BROADCAST_RETRIES {
*retry_count += 1;
let attempt = *retry_count;
tracing::info!(
contract = %key,
self_addr = %self_addr,
attempt,
max_retries = Self::MAX_BROADCAST_RETRIES,
"BROADCAST_NO_TARGETS: scheduling retry - subscriptions may still be in-flight"
);
// Schedule a delayed re-emission of BroadcastStateChange.
// DELIBERATELY blocking — this whole block runs inside a
// `tokio::spawn`, so the blocking await cannot wedge the
// event loop, only its own detached task. Switching to
// `try_notify_node_event` here would silently drop the
// retry in the exact recovery path the executor-side
// try_notify sites depend on for healing.
//
// Spawn-task accumulation is bounded per-contract by
// `MAX_BROADCAST_RETRIES = 3` and
// `BROADCAST_RETRY_BASE_DELAY = 1s`. Worst-case under
// simultaneous wedges is
// `MAX_BROADCAST_RETRIES × concurrent_contracts_in_flight`
// spawn tasks blocked up to 30 s each on the same
// saturated channel — bounded by the small per-tick
// contract-touch count, so soft memory pressure only.
// (Per PR #4231 re-review.)
let op_mgr = op_manager.clone();
let delay = Self::BROADCAST_RETRY_BASE_DELAY * u32::from(attempt);
tokio::spawn(async move {
tokio::time::sleep(delay).await;
if let Err(e) = op_mgr
.notify_node_event(crate::message::NodeEvent::BroadcastStateChange {
key,
new_state,
is_retry: true,
is_reemit: false,
})
.await
{
tracing::warn!(
contract = %key,
error = %e,
"Failed to re-emit BroadcastStateChange for retry"
);
}
});
} else {
// Retries exhausted. Track consecutive no-target cycles to
// suppress repetitive WARN logging after the first occurrence.
self.broadcast_retries.remove(&key);
// Issue #4359: a fresh contract id loses the race between this
// broadcast give-up (~6 s) and the much slower interest/
// subscription resolve for a never-before-seen key. Instead of
// permanently abandoning the state — which silently leaves it
// locally-hosted only while every other node GETs NotFound —
// stash it. When the first interested peer/subscriber for this
// contract appears later, the interest path drains the stash
// and re-emits a single BroadcastStateChange, which then finds
// the freshly-registered target and propagates. The store is
// bounded by size and TTL so a churn of fresh ids that never
// gain a subscriber cannot pin memory.
op_manager
.pending_broadcasts
.stash(*key.id(), new_state.clone());
// Lost-wakeup guard (#4359 re-review): stash() runs here on the
// event-loop task while the interest-resolve flush runs on the
// subscribe/op driver task — they are concurrent with no
// happens-before. An interested peer could have registered (and
// already flushed → found nothing to take) in the window between
// `get_broadcast_targets_update` above and this stash. That would
// leave the state stashed with the only interested peer already
// past its flush, draining nothing until the NEXT subscriber or
// the TTL. Re-resolve targets now that the stash is in place: if a
// target has appeared, take the stash back and re-emit
// immediately rather than waiting on a future flush.
let recheck = op_manager.get_broadcast_targets_update(&key, &self_addr);
if !recheck.targets.is_empty() {
if let Some(stashed) = op_manager.pending_broadcasts.take(key.id()) {
tracing::debug!(
contract = %key,
target_count = recheck.targets.len(),
phase = "pending_broadcast_stash_recheck",
"A target appeared while giving up; re-emitting the deferred \
fresh-contract broadcast immediately instead of leaving it \
stashed for a future flush (#4359)"
);
if let Err(e) = op_manager.try_notify_node_event(
crate::message::NodeEvent::BroadcastStateChange {
key,
new_state: stashed.clone(),
is_retry: false,
is_reemit: true,
},
) {
// Channel full/closed: re-stash so the next interest
// flush still recovers it. Losing it would re-open the
// locally-hosted-only failure this fix closes.
op_manager.pending_broadcasts.stash(*key.id(), stashed);
tracing::debug!(
contract = %key,
error = %e,
"pending_broadcast_stash_recheck: immediate re-emit dropped; \
re-stashed for the next interest signal"
);
}
}
}
// Evict oldest entry if at capacity to prevent unbounded growth.
if !self.broadcast_no_target_streak.contains_key(&key)
&& self.broadcast_no_target_streak.len() >= Self::MAX_BROADCAST_STREAK_ENTRIES
{
if let Some(evict_key) = self.broadcast_no_target_streak.keys().next().cloned()
{
self.broadcast_no_target_streak.remove(&evict_key);
}
}
let streak = self.broadcast_no_target_streak.entry(key).or_insert(0);
*streak = streak.saturating_add(1);
let current_streak = *streak;
if current_streak <= 1 {
tracing::warn!(
contract = %key,
self_addr = %self_addr,
"BROADCAST_NO_TARGETS: no targets found after {} retries, giving up",
Self::MAX_BROADCAST_RETRIES
);
} else {
tracing::debug!(
contract = %key,
self_addr = %self_addr,
consecutive_misses = current_streak,
"BROADCAST_NO_TARGETS: still no targets (suppressing repeated warns)"
);
}
// Emit delivery summary for diagnostics (only on first miss per streak)
if current_streak <= 1 {
let update_tx =
crate::message::Transaction::new::<crate::operations::update::UpdateMsg>();
if let Some(log) = NetEventLog::broadcast_delivery_summary(
&update_tx,
&op_manager.ring,
key,
&target_result,
0,
0,
0,
) {
self.bridge
.log_register
.register_events(Either::Left(log))
.await;
}
}
// The NO_TARGETS outcome was already recorded for the #4281
// summary at the start of this retry cycle (retry_count == 0
// above); retry exhaustion does not record again, so a stuck
// contract counts as one no_targets per fresh broadcast, not
// once per retry.
}
return;
}
// Targets found - clear any pending retry and streak state for this contract
self.broadcast_retries.remove(&key);
self.broadcast_no_target_streak.remove(&key);
// Also drop any deferred re-broadcast stash (#4359): this fan-out is
// reaching targets now, so a previously-abandoned state for this
// contract is superseded and must not be re-emitted later as stale.
let _ = op_manager.pending_broadcasts.take(key.id());
// Record a successful fan-out for the #4281 propagation summary. Fires
// whether targets were found on the initial attempt or on a healing
// retry, so a recovered in-flight-subscription race is surfaced as a
// propagated update (with its real target count). A fresh broadcast
// that found targets on the first try has no `broadcast_retries` entry,
// so it is recorded exactly once here.
op_manager.update_propagation_stats.record_broadcast(
*key.id(),
target_result.targets.len(),
target_result.interest_resolve_failed,
);
// In production, enqueue each target into the broadcast queue.
// The queue worker handles delta computation and streaming with bounded
// concurrency (default: 2 concurrent streams) to prevent uplink saturation.
//
// In simulation tests, call inline to preserve deterministic message
// ordering — tokio::spawn in start_paused(true) changes broadcast
// delivery order and breaks convergence.
#[cfg(not(feature = "simulation_tests"))]
{
for target in &target_result.targets {
self.broadcast_queue
.enqueue(key, target.clone(), new_state.clone())
.await;
}
// Emit broadcast emitted telemetry (issue #3622)
//
// Production path limitations:
// - Transaction ID is synthetic (not from the original update operation).
// BroadcastQueue creates its own TX per target, so this ID cannot be
// correlated with downstream BroadcastReceived/BroadcastApplied events.
// - `broadcasted_to` = number of targets enqueued, not actual delivery
// count. Actual sends are async via BroadcastQueue.
// - `broadcast_to` is empty to avoid cloning up to 512 PeerKeyLocations
// purely for telemetry. The peer list can be reconstructed from
// BroadcastReceived events on the receiving end.
let update_tx =
crate::message::Transaction::new::<crate::operations::update::UpdateMsg>();
let enqueued_count = target_result.targets.len();
if let Some(log) = NetEventLog::update_broadcast_emitted(
&update_tx,
&op_manager.ring,
key,
Vec::new(),
enqueued_count,
new_state,
) {
self.bridge
.log_register
.register_events(Either::Left(log))
.await;
}
}
#[cfg(feature = "simulation_tests")]
{
Self::broadcast_state_to_peers(
self.bridge.clone(),
op_manager,
key,
new_state,
target_result,
)
.await;
}
}
/// Send state to a specific peer that reported a stale summary.
///
/// Unlike `handle_broadcast_state_change` which fans out to ALL subscribers,
/// this targets only the peer that needs catching up. Used by the interest
/// sync summary-mismatch handler to avoid O(peers^2) broadcast storms.
pub(super) async fn handle_sync_state_to_peer(
&mut self,
op_manager: &Arc<OpManager>,
key: freenet_stdlib::prelude::ContractKey,
new_state: freenet_stdlib::prelude::WrappedState,
target_addr: std::net::SocketAddr,
) {
let target = op_manager
.ring
.connection_manager
.get_peer_by_addr(target_addr);
let Some(target) = target else {
tracing::debug!(
contract = %key,
peer = %target_addr,
"SyncStateToPeer: peer not found in connection manager, skipping"
);
return;
};
tracing::debug!(
contract = %key,
peer = %target_addr,
"SyncStateToPeer: sending state to stale peer"
);
#[cfg(not(feature = "simulation_tests"))]
{
self.broadcast_queue.enqueue(key, target, new_state).await;
}
#[cfg(feature = "simulation_tests")]
{
super::super::broadcast_queue::broadcast_to_single_peer(
&self.bridge,
op_manager,
key,
new_state,
target,
)
.await;
}
}
/// Send state change broadcasts to interested peers.
///
/// Used only in simulation tests for inline (deterministic) broadcast ordering.
/// In production, the `BroadcastQueue` handles per-target sends with bounded
/// concurrency.
#[cfg(feature = "simulation_tests")]
async fn broadcast_state_to_peers(
bridge: P2pBridge,
op_manager: &Arc<OpManager>,
key: freenet_stdlib::prelude::ContractKey,
new_state: freenet_stdlib::prelude::WrappedState,
target_result: crate::operations::update::BroadcastTargetResult,
) {
// Skip the whole fan-out when we hold no local state for `key`: there is
// nothing to broadcast, and the per-target `get_contract_summary` calls
// below are the residual #4473 summarize storm on this path. Mirrors the
// production `broadcast_to_single_peer` gate. See
// `broadcast_queue::should_broadcast_contract`.
if !super::super::broadcast_queue::should_broadcast_contract(op_manager, &key) {
tracing::trace!(
contract = %key,
"Skipping broadcast fan-out - contract not hosted or in use"
);
return;
}
// Get our summary once for all targets
let our_summary = op_manager
.interest_manager
.get_contract_summary(op_manager, &key)
.await;
let update_tx = crate::message::Transaction::new::<crate::operations::update::UpdateMsg>();
tracing::debug!(
tx = %update_tx,
contract = %key,
target_count = target_result.targets.len(),
"Broadcasting state change to network peers"
);
let mut skipped_summary_match: usize = 0;
let mut send_success: usize = 0;
let mut send_failed: usize = 0;
for target in &target_result.targets {
let Some(peer_addr) = target.socket_addr() else {
continue;
};
let peer_key = PeerKey::from(target.pub_key().clone());
// Get peer's cached summary
let their_summary = op_manager
.interest_manager
.get_peer_summary(&key, &peer_key);
// Skip if summaries are equal (no change to send)
if let (Some(ours), Some(theirs)) = (&our_summary, &their_summary) {
if ours.as_ref() == theirs.as_ref() {
tracing::trace!(
contract = %key,
peer = %peer_addr,
"Skipping broadcast - peer already has our state"
);
skipped_summary_match += 1;
continue;
}
}
// Compute delta if we have their summary (uses memoization cache)
// Track whether we successfully computed a delta vs sent full state
let (payload, sent_delta) = match (&our_summary, &their_summary) {
(Some(ours), Some(theirs)) => {
match op_manager
.interest_manager
.compute_delta(op_manager, &key, theirs, ours, new_state.size())
.await
{
Ok(Some(delta)) => (
crate::message::DeltaOrFullState::Delta(delta.as_ref().to_vec()),
true,
),
Ok(None) => {
tracing::debug!(
contract = %key,
"Delta computation returned no change, sending full state"
);
(
crate::message::DeltaOrFullState::FullState(
new_state.as_ref().to_vec(),
),
false,
)
}
Err(err) => {
tracing::debug!(
contract = %key,
error = %err,
"Delta computation failed, falling back to full state"
);
(
crate::message::DeltaOrFullState::FullState(
new_state.as_ref().to_vec(),
),
false,
)
}
}
}
_ => (
crate::message::DeltaOrFullState::FullState(new_state.as_ref().to_vec()),
false,
),
};
// Save payload size before moving it into the message
let payload_size = payload.size();
// Check if we should use streaming for full state broadcasts
let use_streaming = matches!(&payload, crate::message::DeltaOrFullState::FullState(_))
&& crate::operations::should_use_streaming(
op_manager.streaming_threshold,
payload_size,
);
let send_result = if use_streaming {
let sender_summary_bytes = our_summary
.as_ref()
.map(|s| s.as_ref().to_vec())
.unwrap_or_default();
let state_bytes = match payload {
crate::message::DeltaOrFullState::FullState(data) => data,
_ => unreachable!("checked above"),
};
let streaming_payload = crate::operations::update::BroadcastStreamingPayload {
state_bytes,
sender_summary_bytes,
};
let payload_bytes = match bincode::serialize(&streaming_payload) {
Ok(b) => b,
Err(e) => {
tracing::warn!(
tx = %update_tx,
error = %e,
"Failed to serialize BroadcastStreamingPayload, skipping"
);
continue;
}
};
let sid = StreamId::next_operations();
tracing::debug!(
tx = %update_tx,
contract = %key,
peer = %peer_addr,
stream_id = %sid,
payload_size,
"Using streaming for BroadcastTo"
);
let msg = crate::operations::update::UpdateMsg::BroadcastToStreaming {
id: update_tx,
stream_id: sid,
key,
total_size: payload_bytes.len() as u64,
};
let net_msg: NetMessage = msg.into();
// Serialize metadata for embedding in fragment #1 (fix #2757)
let metadata = match bincode::serialize(&net_msg) {
Ok(bytes) => Some(bytes::Bytes::from(bytes)),
Err(e) => {
tracing::warn!(
?peer_addr,
error = %e,
"Failed to serialize BroadcastTo metadata for embedding"
);
None
}
};
// channel-safety: ok — broadcast_state_to_peers is
// `#[cfg(feature = "simulation_tests")]` only and never compiled into
// the production binary; the sim runner drives inline deterministic
// broadcast ordering. Production uses the BroadcastQueue path.
let send_res = bridge.send(peer_addr, net_msg).await;
match send_res {
Ok(()) => {
// The metadata message went out; the streamed full state is
// the actual payload. The success arm below caches
// `our_summary` for this peer (FIX 1 of #4145), which is only
// correct if the STATE was sent — so this branch's result must
// reflect the stream send, not just the metadata send. If
// `send_stream` fails, a streamed full state that never landed
// would otherwise be marked as delivered, recreating the
// #2763/#4235 divergence hazard (a wrongly-cached summary makes
// the next delta unappliable). Propagate the stream-send error.
// channel-safety: ok — sim-only path (see the
// `#[cfg(feature = "simulation_tests")]` gate on this fn);
// never compiled into the production binary.
bridge
.send_stream(
peer_addr,
sid,
bytes::Bytes::from(payload_bytes),
metadata,
)
.await
.inspect_err(|err| {
tracing::warn!(
tx = %update_tx,
peer = %peer_addr,
error = %err,
"Failed to send broadcast stream data"
);
})
}
// Metadata send failed; nothing was streamed. Propagate the error.
Err(err) => Err(err),
}
} else {
let msg = crate::operations::update::UpdateMsg::BroadcastTo {
id: update_tx,
key,
payload,
// Include our summary so peer doesn't echo back
sender_summary_bytes: our_summary
.as_ref()
.map(|s| s.as_ref().to_vec())
.unwrap_or_default(),
};
// channel-safety: ok — sim-only path (see the
// `#[cfg(feature = "simulation_tests")]` gate on this fn); never
// compiled into the production binary.
bridge.send(peer_addr, msg.into()).await
};
if let Err(err) = send_result {
send_failed += 1;
tracing::warn!(
tx = %update_tx,
peer = %peer_addr,
error = %err,
"Failed to send state change broadcast"
);
} else {
send_success += 1;
// Track delta vs full state sends for testing (PR #2763)
if sent_delta {
op_manager
.interest_manager
.record_delta_send(new_state.size(), payload_size);
crate::config::GlobalTestMetrics::record_delta_send();
} else {
op_manager.interest_manager.record_full_state_send();
crate::config::GlobalTestMetrics::record_full_state_send();
}
// Issue #3046: Refresh the peer's interest TTL on every successful
// broadcast send. Without this, interest entries for peers who only
// receive full-state broadcasts (no delta → no update_peer_summary
// call) expire after INTEREST_TTL, even though broadcasts
// are being successfully delivered. This caused ~49% of River room
// subscribers to miss updates.
op_manager
.interest_manager
.refresh_peer_interest(&key, &peer_key);
// Issue #4145: Cache the peer summary on ANY successful broadcast
// send — delta OR full state — not just deltas. Separate from the
// TTL refresh above: update_peer_summary sets the cached summary
// (used for delta computation), refresh_peer_interest only extends
// the TTL.
//
// PR #2763 originally gated this on `sent_delta`, which created a
// chicken-and-egg: a delta needs the peer's cached summary, but the
// summary was only cached after a delta — so every NEW subscriber
// (or a peer whose summary was cleared) starts on full state and is
// trapped sending full state forever. Under sustained fan-out that
// is the #4233 full-state broadcast storm.
//
// Safe now because this runs only inside the send-success (`else`)
// arm. This is the inline non-streaming `BroadcastTo` path: a
// successful `bridge.send` is the terminal state we can observe
// (the same assumption the delta path already made). Caching on a
// delivered broadcast is safe even when the cache is momentarily
// wrong: a wrongly-cached summary is corrected by the periodic
// InterestSync summary exchange (~5 min, node.rs) and by the
// delta-apply-failure → ResyncRequest path that clears the
// sender's cached summary (node.rs ~2119). Caching here lets the
// NEXT broadcast to this peer be a small delta. The streaming
// (full-state) path is gated more strictly on a `Delivered`
// completion (#4235) in `broadcast_queue.rs` — note that signal is
// sender-side completion, not a receiver ack, so a lost stream
// tail is covered by the same two backstops.
// (Telemetry above still records delta-vs-full-state separately.)
if let Some(summary) = &our_summary {
op_manager.interest_manager.update_peer_summary(
&key,
&peer_key,
Some(summary.clone()),
);
}
}
}
// Emit broadcast emitted telemetry (issue #3622)
if let Some(log) = NetEventLog::update_broadcast_emitted(
&update_tx,
&op_manager.ring,
key,
target_result.targets.clone(),
send_success,
new_state,
) {
bridge.log_register.register_events(Either::Left(log)).await;
}
// Emit broadcast delivery summary telemetry (issue #3046)
if let Some(log) = NetEventLog::broadcast_delivery_summary(
&update_tx,
&op_manager.ring,
key,
&target_result,
skipped_summary_match,
send_success,
send_failed,
) {
bridge.log_register.register_events(Either::Left(log)).await;
}
}
}