p2p/discover: document BFS choice, add RandomWorkers split

Two related changes to CrawlIterator:

(1) Add a file-level commentary block explaining why the iterator uses a
FIFO queue (BFS over the FINDNODE-response graph) and what it is *not*
suitable for (target-directed lookup -- use RandomNodes() / the alpha=3
lookup iterator for that). The choice was inherited from dcrawl.nim
without explicit reasoning; making it visible avoids future readers
re-deriving the survey-vs-lookup distinction.

The BFS rationale is two-fold:

 - Coverage: BFS reaches every peer within N hops of the seeds in
   order, so a time-bounded run produces a representative sample of the
   reachable graph rather than a deep tendril through one sub-region.
 - Adversarial resilience: a peer returning malicious "neighbour"
   claims, dead-end peers, or eclipse-style sub-graphs cannot
   monopolise the worker pool, because pending work from other branches
   sits ahead of the attacker's responses in the queue. DFS would
   amplify each of these attacks.

(2) Add a RandomWorkers field to CrawlOptions. Of the Workers-sized
worker pool, the first (Workers - RandomWorkers) workers pop the FIFO
front (BFS), while RandomWorkers workers pop a uniform-random queue
index via swap-and-pop (O(1)). Total worker count is unchanged.

Default RandomWorkers = Workers / 4 (4 of 16 with the default
parallelism). At this ratio:

 - Cold-start cost is negligible: 12 of 16 workers still drain FIFO,
   so the first ~1s of a fresh crawl behaves like pure BFS.
 - 25% of pops break strict FIFO ordering, providing a mild
   anti-fingerprint defence against an attacker who could otherwise
   predict our processing order from the contents of their own
   FINDNODE responses.

Operators can override per-run via the new --random-workers CLI flag
on `devp2p discv4 crawl` and `discv5 crawl`. Negative value forces
pure BFS; positive value selects an explicit count.

The new TestCrawlIteratorRandomWorkers covers four pop-policy
configurations (all-fifo, all-random, half-half, default) and
asserts the iterator still terminates and emits each node exactly
once in each.

cmd/devp2p/discv4cmd.go

@@ -91,7 +91,7 @@ var (
 		Name:   "crawl",
 		Usage:  "Updates a nodes.json file with random nodes found in the DHT",
 		Action: discv4Crawl,
-		Flags:  slices.Concat(discoveryNodeFlags, []cli.Flag{crawlTimeoutFlag, crawlParallelismFlag, crawlModeFlag}),
+		Flags:  slices.Concat(discoveryNodeFlags, []cli.Flag{crawlTimeoutFlag, crawlParallelismFlag, crawlModeFlag, crawlRandomWorkersFlag}),
 	}
 	discv4TestCommand = &cli.Command{
 		Name:   "test",
@@ -143,6 +143,11 @@ var (
 		Usage: "Crawl iterator mode: 'lookup' (alpha-bounded Kademlia lookup) or 'fast' (one FINDNODE per peer with rotating prefix; sized by -parallel).",
 		Value: "lookup",
 	}
+	crawlRandomWorkersFlag = &cli.IntFlag{
+		Name:  "random-workers",
+		Usage: "Of the -parallel workers in -mode=fast, how many pop a random queue item rather than the FIFO front. 0 = library default (parallel/4); negative = pure BFS.",
+		Value: 0,
+	}
 	remoteEnodeFlag = &cli.StringFlag{
 		Name:    "remote",
 		Usage:   "Enode of the remote node under test",
@@ -264,7 +269,7 @@ func discv4Crawl(ctx *cli.Context) error {
 	disc, config := startV4(ctx)
 	defer disc.Close()
 
-	iter, err := newDiscv4CrawlIterator(disc, config.Bootnodes, ctx.String(crawlModeFlag.Name), ctx.Int(crawlParallelismFlag.Name))
+	iter, err := newDiscv4CrawlIterator(disc, config.Bootnodes, ctx.String(crawlModeFlag.Name), ctx.Int(crawlParallelismFlag.Name), ctx.Int(crawlRandomWorkersFlag.Name))
 	if err != nil {
 		return err
 	}
@@ -278,14 +283,15 @@ func discv4Crawl(ctx *cli.Context) error {
 	return nil
 }
 
-func newDiscv4CrawlIterator(disc *discover.UDPv4, bootnodes []*enode.Node, mode string, parallel int) (enode.Iterator, error) {
+func newDiscv4CrawlIterator(disc *discover.UDPv4, bootnodes []*enode.Node, mode string, parallel, randomWorkers int) (enode.Iterator, error) {
 	switch mode {
 	case "", "lookup":
 		return disc.RandomNodes(), nil
 	case "fast":
 		return disc.CrawlIterator(discover.CrawlOptions{
-			Workers: parallel,
+			Workers:       parallel,
-			Seeds:   bootnodes,
+			RandomWorkers: randomWorkers,
+			Seeds:         bootnodes,
 		}), nil
 	default:
 		return nil, fmt.Errorf("unknown -%s value %q (want 'lookup' or 'fast')", crawlModeFlag.Name, mode)

cmd/devp2p/discv5cmd.go

@@ -61,6 +61,7 @@ var (
 			crawlTimeoutFlag,
 			crawlParallelismFlag,
 			crawlModeFlag,
+			crawlRandomWorkersFlag,
 		}),
 	}
 	discv5TestCommand = &cli.Command{
@@ -114,7 +115,7 @@ func discv5Crawl(ctx *cli.Context) error {
 	disc, config := startV5(ctx)
 	defer disc.Close()
 
-	iter, err := newDiscv5CrawlIterator(disc, config.Bootnodes, ctx.String(crawlModeFlag.Name), ctx.Int(crawlParallelismFlag.Name))
+	iter, err := newDiscv5CrawlIterator(disc, config.Bootnodes, ctx.String(crawlModeFlag.Name), ctx.Int(crawlParallelismFlag.Name), ctx.Int(crawlRandomWorkersFlag.Name))
 	if err != nil {
 		return err
 	}
@@ -128,14 +129,15 @@ func discv5Crawl(ctx *cli.Context) error {
 	return nil
 }
 
-func newDiscv5CrawlIterator(disc *discover.UDPv5, bootnodes []*enode.Node, mode string, parallel int) (enode.Iterator, error) {
+func newDiscv5CrawlIterator(disc *discover.UDPv5, bootnodes []*enode.Node, mode string, parallel, randomWorkers int) (enode.Iterator, error) {
 	switch mode {
 	case "", "lookup":
 		return disc.RandomNodes(), nil
 	case "fast":
 		return disc.CrawlIterator(discover.CrawlOptions{
-			Workers: parallel,
+			Workers:       parallel,
-			Seeds:   bootnodes,
+			RandomWorkers: randomWorkers,
+			Seeds:         bootnodes,
 		}), nil
 	default:
 		return nil, fmt.Errorf("unknown -%s value %q (want 'lookup' or 'fast')", crawlModeFlag.Name, mode)

p2p/discover/crawliter.go

@@ -18,6 +18,7 @@ package discover
 
 import (
 	crand "crypto/rand"
+	"math/rand/v2"
 	"sync"
 	"sync/atomic"
 
@@ -25,6 +26,37 @@ import (
 	"github.com/ethereum/go-ethereum/p2p/enode"
 )
 
+// CrawlIterator performs a breadth-first crawl of the discv4/discv5
+// FINDNODE-response graph. Each worker pops a peer from the work queue,
+// issues one FINDNODE call against it, and feeds any newly-seen peers
+// back into the queue (and the iterator's output buffer).
+//
+// BFS (FIFO queue) was chosen over DFS or target-directed lookup for two
+// reasons that align with the iterator's intended use case (survey-style
+// crawls — devp2p crawl, geth dial-candidate discovery):
+//
+//   - Coverage: BFS reaches every peer within N hops of the seeds in
+//     order, so a time-bounded run produces a representative sample of
+//     the reachable graph rather than a deep tendril through one
+//     sub-region.
+//   - Adversarial resilience: a peer returning malicious "neighbour"
+//     claims, dead-end peers, or eclipse-style sub-graphs cannot
+//     monopolise the worker pool, because pending work from other
+//     branches sits ahead of the attacker's responses in the queue.
+//     DFS would amplify each of these attacks.
+//
+// CrawlIterator is NOT a target-directed search. To find a specific node
+// (or the K closest to a known target), use the lookup-based iterator
+// returned by [UDPv4.RandomNodes] / [UDPv5.RandomNodes] instead — those
+// run an alpha-bounded Kademlia lookup, which is the right shape for
+// "find peer X" but the wrong shape for "survey the network".
+//
+// Pop policy can be tuned via [CrawlOptions.RandomWorkers]: any subset
+// of the worker pool can be configured to pop a uniform-random queue
+// item instead of the FIFO front, breaking strict ordering as a mild
+// anti-fingerprint defence. The remaining workers pop FIFO and preserve
+// the BFS character. See the field doc for details.
+
 // CrawlOptions configures a CrawlIterator.
 type CrawlOptions struct {
 	// Workers is the number of concurrent FINDNODE calls in flight.
@@ -58,6 +90,23 @@ type CrawlOptions struct {
 	// once across a long crawl. Realistic bound is the reachable DHT size
 	// (~1M peers, ~50 MB).
 	OutputCap int
+
+	// RandomWorkers is the number of workers in the pool that pop a uniform-
+	// random queue item instead of the FIFO front. The remaining
+	// Workers - RandomWorkers workers behave as today, popping front. Total
+	// worker count is unchanged at Workers; only the BFS-vs-random split
+	// differs.
+	//
+	// Zero (the struct zero-value) selects the library default of
+	// Workers/4 rounded down. Pass a negative value (e.g. -1) to force
+	// pure BFS with zero random workers. A positive value > Workers is
+	// clamped to Workers (i.e. fully-random pop with no FIFO workers).
+	//
+	// The default Workers/4 mix preserves the BFS coverage character (cold
+	// start is dominated by the FIFO majority) while breaking strict FIFO
+	// ordering, a mild anti-fingerprint defence. See the file-level comment
+	// for details.
+	RandomWorkers int
 }
 
 func (o *CrawlOptions) withDefaults() {
@@ -73,6 +122,17 @@ func (o *CrawlOptions) withDefaults() {
 	if o.OutputCap <= 0 {
 		o.OutputCap = 16 * o.Workers
 	}
+	switch {
+	case o.RandomWorkers < 0:
+		// Negative means: explicit pure BFS, no random workers.
+		o.RandomWorkers = 0
+	case o.RandomWorkers == 0:
+		// Zero (struct zero-value) means: library default.
+		o.RandomWorkers = o.Workers / 4
+	case o.RandomWorkers > o.Workers:
+		// Clamp; can't have more random workers than total workers.
+		o.RandomWorkers = o.Workers
+	}
 }
 
 // CrawlIterator returns an enode.Iterator that performs a breadth-first
@@ -166,10 +226,17 @@ func newCrawlIterator(opts CrawlOptions, queryFn func(*enode.Node, int) ([]*enod
 		it.inflight++
 	}
 
-	// Workers.
+	// Workers. The first (Workers - RandomWorkers) workers pop FIFO; the
-	for i := 0; i < opts.Workers; i++ {
+	// remaining RandomWorkers pop a uniform-random queue index. Both share
+	// the same queue, mutex, cond, and termination semantics.
+	fifoCount := opts.Workers - opts.RandomWorkers
+	for i := 0; i < fifoCount; i++ {
 		it.wg.Add(1)
-		go it.worker()
+		go it.worker(false)
+	}
+	for i := 0; i < opts.RandomWorkers; i++ {
+		it.wg.Add(1)
+		go it.worker(true)
 	}
 	return it
 }
@@ -207,8 +274,10 @@ func (it *crawlIterator) discover(n *enode.Node) {
 }
 
 // popWork blocks until either a peer is available to query, or the iterator
-// has nothing left to do. Returns (nil, false) on termination.
+// has nothing left to do. Returns (nil, false) on termination. If random is
-func (it *crawlIterator) popWork() (*enode.Node, bool) {
+// true, pops a uniform-random index via swap-and-pop; otherwise pops the
+// FIFO front. Both modes share the same termination semantics.
+func (it *crawlIterator) popWork(random bool) (*enode.Node, bool) {
 	it.mu.Lock()
 	defer it.mu.Unlock()
 	for {
@@ -216,6 +285,14 @@ func (it *crawlIterator) popWork() (*enode.Node, bool) {
 			return nil, false
 		}
 		if len(it.queue) > 0 {
+			if random {
+				i := rand.IntN(len(it.queue))
+				n := it.queue[i]
+				// Swap-and-pop: O(1) removal from middle.
+				it.queue[i] = it.queue[len(it.queue)-1]
+				it.queue = it.queue[:len(it.queue)-1]
+				return n, true
+			}
 			n := it.queue[0]
 			it.queue = it.queue[1:]
 			return n, true
@@ -243,10 +320,10 @@ func (it *crawlIterator) finishWork() {
 	}
 }
 
-func (it *crawlIterator) worker() {
+func (it *crawlIterator) worker(random bool) {
 	defer it.wg.Done()
 	for {
-		n, ok := it.popWork()
+		n, ok := it.popWork(random)
 		if !ok {
 			return
 		}

p2p/discover/crawliter_test.go

@@ -223,6 +223,69 @@ func TestCrawlIteratorOutputCap(t *testing.T) {
 	}
 }
 
+// TestCrawlIteratorRandomWorkers verifies that with a mixed FIFO + random
+// worker pool, the iterator still terminates and emits each node exactly
+// once on the synthetic graph from TestCrawlIteratorTerminates.
+func TestCrawlIteratorRandomWorkers(t *testing.T) {
+	for _, tc := range []struct {
+		name          string
+		workers       int
+		randomWorkers int
+	}{
+		{"all-fifo", 4, -1},   // -1 → pure BFS
+		{"all-random", 4, 4},  // all workers random-pop
+		{"half-half", 8, 4},   // 4 FIFO + 4 random
+		{"default", 16, 0},    // 0 → library default = 4 random of 16
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			nodes := makeTestNodes(t, 50)
+			neighbours := make(map[enode.ID][]*enode.Node, len(nodes))
+			for i, n := range nodes {
+				var ns []*enode.Node
+				for k := 1; k <= 5; k++ {
+					ns = append(ns, nodes[(i+k)%len(nodes)])
+				}
+				neighbours[n.ID()] = ns
+			}
+			var calls atomic.Int64
+			queryFn := func(dst *enode.Node, _ int) ([]*enode.Node, error) {
+				calls.Add(1)
+				return neighbours[dst.ID()], nil
+			}
+			it := newCrawlIterator(CrawlOptions{
+				Workers:       tc.workers,
+				RandomWorkers: tc.randomWorkers,
+				Seeds:         []*enode.Node{nodes[0]},
+				Drange:        16,
+			}, queryFn)
+			seen := make(map[enode.ID]int)
+			done := make(chan struct{})
+			go func() {
+				for it.Next() {
+					seen[it.Node().ID()]++
+				}
+				close(done)
+			}()
+			select {
+			case <-done:
+			case <-time.After(5 * time.Second):
+				t.Fatal("iterator did not terminate within 5s")
+			}
+			if got := len(seen); got != len(nodes) {
+				t.Fatalf("emitted %d distinct nodes, want %d", got, len(nodes))
+			}
+			for id, c := range seen {
+				if c != 1 {
+					t.Errorf("node %x emitted %d times, want 1", id[:4], c)
+				}
+			}
+			if got, want := calls.Load(), int64(len(nodes)); got != want {
+				t.Errorf("queryFn invoked %d times, want %d", got, want)
+			}
+		})
+	}
+}
+
 // TestCrawlIteratorRotation verifies that the d argument passed to queryFn
 // rotates through 0..Drange-1.
 func TestCrawlIteratorRotation(t *testing.T) {