Merge pull request #17839 from karalabe/downloader-invalid-hash-chain-fix

eth/downloader: fix invalid hash chain error due to head mini reorg

eth/downloader/downloader.go

@@ -60,6 +60,9 @@ var (
 	maxHeadersProcess = 2048      // Number of header download results to import at once into the chain
 	maxResultsProcess = 2048      // Number of content download results to import at once into the chain
 
+	reorgProtThreshold   = 48 // Threshold number of recent blocks to disable mini reorg protection
+	reorgProtHeaderDelay = 2  // Number of headers to delay delivering to cover mini reorgs
+
 	fsHeaderCheckFrequency = 100             // Verification frequency of the downloaded headers during fast sync
 	fsHeaderSafetyNet      = 2048            // Number of headers to discard in case a chain violation is detected
 	fsHeaderForceVerify    = 24              // Number of headers to verify before and after the pivot to accept it
@@ -847,6 +850,30 @@ func (d *Downloader) fetchHeaders(p *peerConnection, from uint64, pivot uint64)
 				}
 				headers = filled[proced:]
 				from += uint64(proced)
+			} else {
+				// If we're closing in on the chain head, but haven't yet reached it, delay
+				// the last few headers so mini reorgs on the head don't cause invalid hash
+				// chain errors.
+				if n := len(headers); n > 0 {
+					// Retrieve the current head we're at
+					head := uint64(0)
+					if d.mode == LightSync {
+						head = d.lightchain.CurrentHeader().Number.Uint64()
+					} else {
+						head = d.blockchain.CurrentFastBlock().NumberU64()
+						if full := d.blockchain.CurrentBlock().NumberU64(); head < full {
+							head = full
+						}
+					}
+					// If the head is way older than this batch, delay the last few headers
+					if head+uint64(reorgProtThreshold) < headers[n-1].Number.Uint64() {
+						delay := reorgProtHeaderDelay
+						if delay > n {
+							delay = n
+						}
+						headers = headers[:n-delay]
+					}
+				}
 			}
 			// Insert all the new headers and fetch the next batch
 			if len(headers) > 0 {
@@ -857,8 +884,18 @@ func (d *Downloader) fetchHeaders(p *peerConnection, from uint64, pivot uint64)
 					return errCancelHeaderFetch
 				}
 				from += uint64(len(headers))
-			}
 				getHeaders(from)
+			} else {
+				// No headers delivered, or all of them being delayed, sleep a bit and retry
+				p.log.Trace("All headers delayed, waiting")
+				select {
+				case <-time.After(fsHeaderContCheck):
+					getHeaders(from)
+					continue
+				case <-d.cancelCh:
+					return errCancelHeaderFetch
+				}
+			}
 
 		case <-timeout.C:
 			if d.dropPeer == nil {

eth/downloader/downloader_test.go

@@ -744,7 +744,7 @@ func testThrottling(t *testing.T, protocol int, mode SyncMode) {
 			tester.downloader.queue.lock.Unlock()
 			tester.lock.Unlock()
 
-			if cached == blockCacheItems || retrieved+cached+frozen == targetBlocks+1 {
+			if cached == blockCacheItems || cached == blockCacheItems-reorgProtHeaderDelay || retrieved+cached+frozen == targetBlocks+1 || retrieved+cached+frozen == targetBlocks+1-reorgProtHeaderDelay {
 				break
 			}
 		}
@@ -754,7 +754,7 @@ func testThrottling(t *testing.T, protocol int, mode SyncMode) {
 		tester.lock.RLock()
 		retrieved = len(tester.ownBlocks)
 		tester.lock.RUnlock()
-		if cached != blockCacheItems && retrieved+cached+frozen != targetBlocks+1 {
+		if cached != blockCacheItems && cached != blockCacheItems-reorgProtHeaderDelay && retrieved+cached+frozen != targetBlocks+1 && retrieved+cached+frozen != targetBlocks+1-reorgProtHeaderDelay {
 			t.Fatalf("block count mismatch: have %v, want %v (owned %v, blocked %v, target %v)", cached, blockCacheItems, retrieved, frozen, targetBlocks+1)
 		}
 		// Permit the blocked blocks to import