@aztec/p2p 0.0.1-commit.88c5703d4 → 0.0.1-commit.88e6f9396

package/src/services/libp2p/libp2p_service.ts

@@ -18,7 +18,6 @@ import {
   type CheckpointProposalCore,
   type Gossipable,
   P2PMessage,
-  type ValidationResult as P2PValidationResult,
   PeerErrorSeverity,
   PeerErrorSeverityByHarshness,
   TopicType,
@@ -226,7 +225,7 @@ export class LibP2PService extends WithTracer implements P2PService {
 
     const proposalValidatorOpts = {
       txsPermitted: !config.disableTransactions,
-      maxTxsPerBlock: config.validateMaxTxsPerBlock,
+      maxTxsPerBlock: config.validateMaxTxsPerBlock ?? config.validateMaxTxsPerCheckpoint,
     };
     this.blockProposalValidator = new BlockProposalValidator(epochCache, proposalValidatorOpts);
     this.checkpointProposalValidator = new CheckpointProposalValidator(epochCache, proposalValidatorOpts);
@@ -976,6 +975,11 @@ export class LibP2PService extends WithTracer implements P2PService {
       } else if (wasIgnored) {
         return { result: TopicValidatorResult.Ignore, obj: tx };
       } else {
+        this.logger.warn(`Gossiped tx ${txHash.toString()} unexpectedly rejected by pool`, {
+          source: source.toString(),
+          txHash: txHash.toString(),
+        });
+        this.peerManager.penalizePeer(source, PeerErrorSeverity.HighToleranceError);
         return { result: TopicValidatorResult.Reject };
       }
     };
@@ -1742,31 +1746,6 @@ export class LibP2PService extends WithTracer implements P2PService {
     return PeerErrorSeverity.HighToleranceError;
   }
 
-  /**
-   * Validate a checkpoint attestation.
-   *
-   * @param attestation - The checkpoint attestation to validate.
-   * @returns True if the checkpoint attestation is valid, false otherwise.
-   */
-  @trackSpan('Libp2pService.validateCheckpointAttestation', async (_, attestation) => ({
-    [Attributes.SLOT_NUMBER]: attestation.payload.header.slotNumber,
-    [Attributes.BLOCK_ARCHIVE]: attestation.archive.toString(),
-    [Attributes.P2P_ID]: await attestation.p2pMessageLoggingIdentifier().then(i => i.toString()),
-  }))
-  public async validateCheckpointAttestation(
-    peerId: PeerId,
-    attestation: CheckpointAttestation,
-  ): Promise<P2PValidationResult> {
-    const result = await this.checkpointAttestationValidator.validate(attestation);
-
-    if (result.result === 'reject') {
-      this.logger.warn(`Penalizing peer ${peerId} for checkpoint attestation validation failure`);
-      this.peerManager.penalizePeer(peerId, result.severity);
-    }
-
-    return result;
-  }
-
   public getPeerScore(peerId: PeerId): number {
     return this.node.services.pubsub.score.score(peerId.toString());
   }

package/src/services/peer-manager/peer_manager.ts

@@ -32,7 +32,7 @@ import { PeerScoreState, type PeerScoring } from './peer_scoring.js';
 const MAX_DIAL_ATTEMPTS = 3;
 const MAX_CACHED_PEERS = 100;
 const MAX_CACHED_PEER_AGE_MS = 5 * 60 * 1000; // 5 minutes
-const FAILED_PEER_BAN_TIME_MS = 5 * 60 * 1000; // 5 minutes timeout after failing MAX_DIAL_ATTEMPTS
+const DEFAULT_FAILED_PEER_BAN_TIME_MS = 5 * 60 * 1000; // 5 minutes timeout after failing MAX_DIAL_ATTEMPTS
 const GOODBYE_DIAL_TIMEOUT_MS = 1000;
 const FAILED_AUTH_HANDSHAKE_EXPIRY_MS = 60 * 60 * 1000; // 1 hour
 
@@ -776,7 +776,8 @@ export class PeerManager implements PeerManagerInterface {
         // Add to timed out peers
         this.timedOutPeers.set(id, {
           peerId: id,
-          timeoutUntilMs: this.dateProvider.now() + FAILED_PEER_BAN_TIME_MS,
+          timeoutUntilMs:
+            this.dateProvider.now() + (this.config.peerFailedBanTimeMs ?? DEFAULT_FAILED_PEER_BAN_TIME_MS),
         });
       }
     }
@@ -938,6 +939,8 @@ export class PeerManager implements PeerManagerInterface {
           `Received auth for validator ${sender.toString()} from peer ${peerIdString}, but this validator is already authenticated to peer ${peerForAddress.toString()}`,
           { ...logData, address: sender.toString() },
         );
+        this.markAuthHandshakeFailed(peerId);
+        this.markPeerForDisconnect(peerId);
         return;
       }
 

package/src/services/reqresp/batch-tx-requester/README.md

@@ -170,6 +170,37 @@ class BlockTxsResponse {
 
 The `BitVector` is a compact representation where each bit corresponds to a transaction index in the block proposal. This allows efficient capability advertisement without repeating full hashes.
 
+## Cancellation
+
+All cancellation is managed by a single `RequestTracker` instance, shared across the entire collection
+flow. The `RequestTracker` owns the deadline, tracks which txs are still missing, and exposes a
+`cancellationToken` promise that resolves when the request should stop (deadline hit, all txs fetched,
+or external `cancel()` call).
+
+Cancellation propagates from the deepest stack level upward:
+
+```
+RequestTracker.finish()
+  ├── resolves cancellationToken promise
+  │
+  ├── BatchTxRequester workers (deepest)
+  │     ├── shouldStop() checks requestTracker.cancelled → exit loop
+  │     ├── sleepClampedToDeadline races sleep vs cancellationToken → wakes
+  │     └── semaphore.acquire races vs cancellationToken → wakes
+  │           │
+  │           ▼ workers settle → txQueue.end() → generator returns
+  │
+  ├── Node collection loops
+  │     ├── notFinished() checks requestTracker.cancelled → exit loop
+  │     └── inter-retry sleep races vs cancellationToken → wakes
+  │           │
+  │           ▼ all node loops settle
+  │
+  └── collectFast (outermost)
+        awaits Promise.allSettled([reqresp, nodes]) → settles after inner tasks
+        finally: requestTracker.cancel() (idempotent), cleanup
+```
+
 ## Key Files
 
 | File | Description |
@@ -179,15 +210,16 @@ The `BitVector` is a compact representation where each bit corresponds to a tran
 | `peer_collection.ts` | Manages peer classification (dumb/smart/bad) and rate limiting |
 | `interface.ts` | Type definitions for dependencies |
 | `../protocols/block_txs/` | Wire protocol definitions (`BlockTxsRequest`, `BlockTxsResponse`, `BitVector`) |
+| `../../tx_collection/request_tracker.ts` | Centralized deadline, missing tx tracking, and cancellation signal |
 
 ## Stopping Conditions
 
-The `BatchTxRequester` stops when any of these conditions are met:
+The `BatchTxRequester` stops when any of these conditions are met, all managed by the `RequestTracker`:
 
-1. **All transactions fetched** - Success!
-2. **Deadline exceeded** - Timeout configured by caller
-3. **Abort signal** - External cancellation
-4. **No transactions to fetch** - Nothing was missing
+1. **All transactions fetched** - `markFetched()` removes the last missing tx, triggering `finish()`
+2. **Deadline exceeded** - `setTimeout` in `RequestTracker` fires, triggering `finish()`
+3. **External cancellation** - `RequestTracker.cancel()` called (e.g., from `stop()`, `stopCollectingForBlocksUpTo`)
+4. **No transactions to fetch** - Empty hash set at construction, `RequestTracker` finishes immediately
 
 ## Configuration
 
@@ -228,11 +260,15 @@ Request to peer fails
 ## Usage Example
 
 ```typescript
+const requestTracker = RequestTracker.create(
+  missingTxHashes,                          // TxHash[] - what we need
+  new Date(Date.now() + 5_000),             // deadline
+);
+
 const requester = new BatchTxRequester(
-  missingTxHashes,      // TxHash[] - what we need
+  requestTracker,       // IRequestTracker - tracks missing txs, deadline, and cancellation
   blockTxsSource,       // BlockTxsSource - the proposal or block we need txs for
   pinnedPeer,           // PeerId | undefined - peer expected to have the txs
-  timeoutMs,            // number - how long to try
   p2pService,           // BatchTxRequesterLibP2PService
 );
 
@@ -273,6 +309,8 @@ const txs = await BatchTxRequester.collectAllTxs(requester.run());
 │  1. Try RPC nodes first (fast)    │   │  Periodic polling of RPC nodes      │
 │  2. Fall back to BatchTxRequester │   │  and peers for missing txs          │
 │                                   │   │                                     │
+│  Creates RequestTracker per       │   │                                     │
+│  request with deadline            │   │                                     │
 └───────────────────┬───────────────┘   └─────────────────────────────────────┘
                     │
                     │ For 'proposal' and 'block' requests
@@ -281,6 +319,7 @@ const txs = await BatchTxRequester.collectAllTxs(requester.run());
 │                           BatchTxRequester                                  │
 │                                                                             │
 │  Aggressive parallel fetching from multiple peers                           │
+│  Shares RequestTracker with FastTxCollection for unified cancellation       │
 │  Uses BLOCK_TXS sub-protocol for efficient batching                         │
 └───────────────────┬─────────────────────────────────────────────────────────┘
                     │

package/src/services/reqresp/batch-tx-requester/batch_tx_requester.ts

@@ -1,15 +1,14 @@
 import { chunkWrapAround } from '@aztec/foundation/collection';
-import { TimeoutError } from '@aztec/foundation/error';
 import { type Logger, createLogger } from '@aztec/foundation/log';
 import { FifoMemoryQueue, type ISemaphore, Semaphore } from '@aztec/foundation/queue';
 import { sleep } from '@aztec/foundation/sleep';
-import { DateProvider, executeTimeout } from '@aztec/foundation/timer';
+import { DateProvider } from '@aztec/foundation/timer';
 import { PeerErrorSeverity } from '@aztec/stdlib/p2p';
 import { Tx, TxArray, TxHash } from '@aztec/stdlib/tx';
 
 import type { PeerId } from '@libp2p/interface';
 
-import type { IMissingTxsTracker } from '../../tx_collection/missing_txs_tracker.js';
+import type { IRequestTracker } from '../../tx_collection/request_tracker.js';
 import { ReqRespSubProtocol } from '.././interface.js';
 import { BlockTxsRequest, BlockTxsResponse, type BlockTxsSource } from '.././protocols/index.js';
 import { ReqRespStatus } from '.././status.js';
@@ -42,16 +41,14 @@ import { BatchRequestTxValidator, type IBatchRequestTxValidator } from './tx_val
  *    - Is the peer which was unable to send us successful response N times in a row
  * */
 export class BatchTxRequester {
+  private readonly requestTracker: IRequestTracker;
   private readonly blockTxsSource: BlockTxsSource;
   private readonly pinnedPeer: PeerId | undefined;
-  private readonly timeoutMs: number;
   private readonly p2pService: BatchTxRequesterLibP2PService;
   private readonly logger: Logger;
-  private readonly dateProvider: DateProvider;
   private readonly opts: BatchTxRequesterOptions;
   private readonly peers: IPeerCollection;
   private readonly txsMetadata: ITxMetadataCollection;
-  private readonly deadline: number;
   private readonly smartRequesterSemaphore: ISemaphore;
   private readonly txQueue: FifoMemoryQueue<Tx>;
   private readonly txValidator: IBatchRequestTxValidator;
@@ -60,21 +57,19 @@ export class BatchTxRequester {
   private readonly txBatchSize: number;
 
   constructor(
-    missingTxsTracker: IMissingTxsTracker,
+    requestTracker: IRequestTracker,
     blockTxsSource: BlockTxsSource,
     pinnedPeer: PeerId | undefined,
-    timeoutMs: number,
     p2pService: BatchTxRequesterLibP2PService,
     logger?: Logger,
     dateProvider?: DateProvider,
     opts?: BatchTxRequesterOptions,
   ) {
+    this.requestTracker = requestTracker;
     this.blockTxsSource = blockTxsSource;
     this.pinnedPeer = pinnedPeer;
-    this.timeoutMs = timeoutMs;
     this.p2pService = p2pService;
     this.logger = logger ?? createLogger('p2p:reqresp_batch');
-    this.dateProvider = dateProvider ?? new DateProvider();
     this.opts = opts ?? {};
 
     this.smartParallelWorkerCount =
@@ -82,7 +77,6 @@ export class BatchTxRequester {
     this.dumbParallelWorkerCount =
       this.opts.dumbParallelWorkerCount ?? DEFAULT_BATCH_TX_REQUESTER_DUMB_PARALLEL_WORKER_COUNT;
     this.txBatchSize = this.opts.txBatchSize ?? DEFAULT_BATCH_TX_REQUESTER_TX_BATCH_SIZE;
-    this.deadline = this.dateProvider.now() + this.timeoutMs;
     this.txQueue = new FifoMemoryQueue(this.logger);
     this.txValidator = this.opts.txValidator ?? new BatchRequestTxValidator(this.p2pService.txValidatorConfig);
 
@@ -93,12 +87,12 @@ export class BatchTxRequester {
       this.peers = new PeerCollection(
         this.p2pService.connectionSampler,
         this.pinnedPeer,
-        this.dateProvider,
+        dateProvider ?? new DateProvider(),
         badPeerThreshold,
         this.p2pService.peerScoring,
       );
     }
-    this.txsMetadata = new MissingTxMetadataCollection(missingTxsTracker, this.txBatchSize);
+    this.txsMetadata = new MissingTxMetadataCollection(requestTracker, this.txBatchSize);
     this.smartRequesterSemaphore = this.opts.semaphore ?? new Semaphore(0);
   }
 
@@ -106,40 +100,30 @@ export class BatchTxRequester {
    * Fetches all missing transactions and yields them  one by one
    * */
   public async *run(): AsyncGenerator<Tx, Tx | undefined, unknown> {
-    // Our timeout is represented in milliseconds but queue expects seconds
-    // We also want to make sure we wait at least 1 second in case of very low timeouts
-    const timeoutQueueAfter = Math.max(Math.ceil(this.timeoutMs / 1_000), 1);
     try {
       if (this.txsMetadata.getMissingTxHashes().size === 0) {
         return undefined;
       }
 
-      // Start workers in background
-      const workersPromise = executeTimeout(
-        () => Promise.allSettled([this.smartRequester(), this.dumbRequester(), this.pinnedPeerRequester()]),
-        this.timeoutMs,
-      ).finally(() => {
+      // Start workers in background. Workers stop themselves via requestTracker.checkCancelled().
+      const workersPromise = Promise.allSettled([
+        this.smartRequester(),
+        this.dumbRequester(),
+        this.pinnedPeerRequester(),
+      ]).finally(() => {
         this.txQueue.end();
       });
 
+      // Yield txs as workers put them on the queue. The queue's end() drains remaining items
+      // before returning null, so we don't lose any txs.
       while (true) {
-        const tx = await this.txQueue.get(timeoutQueueAfter);
+        const tx = await this.txQueue.get();
 
-        // null indicates that the queue has ended
         if (tx === null) {
           break;
         }
 
         yield tx;
-
-        if (this.shouldStop()) {
-          // Drain queue before ending
-          let remaining;
-          while ((remaining = this.txQueue.getImmediate()) !== undefined) {
-            yield remaining;
-          }
-          break;
-        }
       }
 
       this.unlockSmartRequesterSemaphores();
@@ -360,7 +344,10 @@ export class BatchTxRequester {
   ) {
     try {
       this.logger.trace(`Smart worker ${workerIndex} started`);
-      await executeTimeout((_: AbortSignal) => this.smartRequesterSemaphore.acquire(), this.timeoutMs);
+      await Promise.race([this.smartRequesterSemaphore.acquire(), this.requestTracker.cancellationToken]);
+      if (this.requestTracker.checkCancelled()) {
+        return;
+      }
       this.logger.trace(`Smart worker ${workerIndex} acquired semaphore`);
 
       while (!this.shouldStop()) {
@@ -384,7 +371,10 @@ export class BatchTxRequester {
           //
           // When a dumb peer responds with valid txIndices, it gets
           // promoted to smart and releases the semaphore, waking this worker.
-          await executeTimeout((_: AbortSignal) => this.smartRequesterSemaphore.acquire(), this.timeoutMs);
+          await Promise.race([this.smartRequesterSemaphore.acquire(), this.requestTracker.cancellationToken]);
+          if (this.requestTracker.checkCancelled()) {
+            break;
+          }
           this.logger.debug(`Worker loop smart: acquired next smart peer`);
           continue;
         }
@@ -411,11 +401,7 @@ export class BatchTxRequester {
         });
       }
     } catch (err: any) {
-      if (err instanceof TimeoutError) {
-        this.logger.debug(`Smart worker ${workerIndex} timed out waiting for semaphore`);
-      } else {
-        this.logger.error(`Smart worker ${workerIndex} encountered an error: ${err}`);
-      }
+      this.logger.error(`Smart worker ${workerIndex} encountered an error: ${err}`);
     } finally {
       this.logger.debug(`Smart worker ${workerIndex} finished`);
     }
@@ -651,27 +637,14 @@ export class BatchTxRequester {
   }
 
   /*
-   * @returns true if all missing txs have been fetched */
-  private fetchedAllTxs() {
-    return this.txsMetadata.getMissingTxHashes().size == 0;
-  }
-
-  /*
-   * Checks if the BatchTxRequester should stop fetching missing txs
-   * Conditions for stopping are:
-   * - There have been no missing transactions to start with
-   * - All transactions have been fetched
-   * - The deadline has been hit (no more time to fetch)
-   * - This process has been cancelled via abortSignal
-   *
-   * @returns true if BatchTxRequester should stop, otherwise false*/
+   * Checks if the BatchTxRequester should stop fetching missing txs.
+   * Delegates to requestTracker which covers: deadline hit, all txs fetched, or external cancellation. */
   private shouldStop() {
-    const aborted = this.opts.abortSignal?.aborted ?? false;
-    if (aborted) {
+    if (this.requestTracker.checkCancelled()) {
       this.unlockSmartRequesterSemaphores();
     }
 
-    return aborted || this.fetchedAllTxs() || this.dateProvider.now() > this.deadline;
+    return this.requestTracker.checkCancelled();
   }
 
   /*
@@ -689,10 +662,9 @@ export class BatchTxRequester {
    * This ensures we don't sleep past the deadline.
    * */
   private async sleepClampedToDeadline(durationMs: number) {
-    const remaining = this.deadline - this.dateProvider.now();
-    const thereIsTimeRemaining = remaining > 0;
-    if (thereIsTimeRemaining) {
-      await sleep(Math.min(durationMs, remaining));
+    if (this.requestTracker.checkCancelled()) {
+      return;
     }
+    await Promise.race([sleep(durationMs), this.requestTracker.cancellationToken]);
   }
 }

package/src/services/reqresp/batch-tx-requester/interface.ts

@@ -49,7 +49,6 @@ export interface BatchTxRequesterOptions {
   //Injectable for testing purposes
   semaphore?: ISemaphore;
   peerCollection?: IPeerCollection;
-  abortSignal?: AbortSignal;
   /** Optional tx validator for testing - if not provided, one is created from p2pService.txValidatorConfig */
   txValidator?: IBatchRequestTxValidator;
 }

package/src/services/reqresp/batch-tx-requester/missing_txs.ts

@@ -2,7 +2,7 @@ import { type Tx, TxHash } from '@aztec/stdlib/tx';
 
 import type { PeerId } from '@libp2p/interface';
 
-import type { IMissingTxsTracker } from '../../tx_collection/missing_txs_tracker.js';
+import type { IRequestTracker } from '../../tx_collection/request_tracker.js';
 import { DEFAULT_BATCH_TX_REQUESTER_TX_BATCH_SIZE } from './config.js';
 import type { ITxMetadataCollection } from './interface.js';
 
@@ -41,10 +41,10 @@ export class MissingTxMetadataCollection implements ITxMetadataCollection {
   private txMetadata = new Map<string, MissingTxMetadata>();
 
   constructor(
-    private missingTxsTracker: IMissingTxsTracker,
+    private requestTracker: IRequestTracker,
     private readonly txBatchSize: number = DEFAULT_BATCH_TX_REQUESTER_TX_BATCH_SIZE,
   ) {
-    missingTxsTracker.missingTxHashes.forEach(hash => this.txMetadata.set(hash, new MissingTxMetadata(hash)));
+    requestTracker.missingTxHashes.forEach(hash => this.txMetadata.set(hash, new MissingTxMetadata(hash)));
   }
 
   public getPrioritizingNotInFlightAndLowerRequestCount(txs: string[]): MissingTxMetadata[] {
@@ -65,7 +65,7 @@ export class MissingTxMetadataCollection implements ITxMetadataCollection {
   }
 
   public getMissingTxHashes(): Set<string> {
-    return this.missingTxsTracker.missingTxHashes;
+    return this.requestTracker.missingTxHashes;
   }
 
   public getTxsPeerHas(peer: PeerId): Set<string> {
@@ -128,7 +128,7 @@ export class MissingTxMetadataCollection implements ITxMetadataCollection {
   }
 
   public alreadyFetched(txHash: TxHash): boolean {
-    return !this.missingTxsTracker.isMissing(txHash.toString());
+    return !this.requestTracker.isMissing(txHash.toString());
   }
 
   public markFetched(peerId: PeerId, tx: Tx): boolean {
@@ -144,7 +144,7 @@ export class MissingTxMetadataCollection implements ITxMetadataCollection {
     }
 
     txMeta.peers.add(peerId.toString());
-    return this.missingTxsTracker.markFetched(tx);
+    return this.requestTracker.markFetched(tx);
   }
 
   public markPeerHas(peerId: PeerId, txHash: TxHash[]) {

package/src/services/reqresp/reqresp.ts

@@ -16,7 +16,7 @@ import {
   IndividualReqRespTimeoutError,
   InvalidResponseError,
 } from '../../errors/reqresp.error.js';
-import { SnappyTransform } from '../encoding.js';
+import { OversizedSnappyResponseError, SnappyTransform } from '../encoding.js';
 import type { PeerScoring } from '../peer-manager/peer_scoring.js';
 import {
   DEFAULT_INDIVIDUAL_REQUEST_TIMEOUT_MS,
@@ -553,16 +553,10 @@ export class ReqResp implements ReqRespInterface {
         data: message,
       };
     } catch (e: any) {
+      // All errors (invalid status bytes, oversized snappy responses, corrupt data, etc.)
+      // are re-thrown so the caller can penalize the peer via handleResponseError.
       this.logger.debug(`Reading message failed: ${e.message}`);
-
-      let status = ReqRespStatus.UNKNOWN;
-      if (e instanceof ReqRespStatusError) {
-        status = e.status;
-      }
-
-      return {
-        status,
-      };
+      throw e;
     }
   }
 
@@ -780,6 +774,20 @@ export class ReqResp implements ReqRespInterface {
       return undefined;
     }
 
+    // Invalid status byte: the peer sent a status byte that doesn't match any known status code.
+    // This is a protocol violation, penalize harshly.
+    if (e instanceof ReqRespStatusError) {
+      this.logger.warn(`Invalid status byte from peer ${peerId.toString()} in ${subProtocol}: ${e.message}`, logTags);
+      return PeerErrorSeverity.LowToleranceError;
+    }
+
+    // Oversized snappy response: the peer is sending data that exceeds the allowed size.
+    // This is a protocol violation that wastes bandwidth, so penalize harshly.
+    if (e instanceof OversizedSnappyResponseError) {
+      this.logger.warn(`Oversized response from peer ${peerId.toString()} in ${subProtocol}: ${e.message}`, logTags);
+      return PeerErrorSeverity.LowToleranceError;
+    }
+
     return this.categorizeConnectionErrors(e, peerId, subProtocol);
   }