test(dgram): cover UDP control-plane starvation
- commit
- eac5808d11f8413c106eadf0a974800bf822fdb0
- author
- brice · · 2026-07-10 00:27 UTC
- parent
- 2479821
- signature
- GPG Verified
CD88F9E564E0102646FB6077A1A07C7C7C3E349D· arch-gpg Verified for the committer email
1 file · +122 -0
A internal/adapters/dgram/control_plane_test.go +122 -0
| @@ -0,0 +1,122 @@ | |||
| 1 | + | package dgram | |
| 2 | + | ||
| 3 | + | import ( | |
| 4 | + | "net" | |
| 5 | + | "testing" | |
| 6 | + | "time" | |
| 7 | + | ||
| 8 | + | "github.com/bnema/vev/internal/ports" | |
| 9 | + | pdgram "github.com/bnema/vev/pkg/dgram" | |
| 10 | + | ) | |
| 11 | + | ||
| 12 | + | func TestReadLoopDoesNotWriteACKSynchronously(t *testing.T) { | |
| 13 | + | clk := newManualClock(time.Date(2030, 1, 2, 3, 4, 5, 0, time.UTC)) | |
| 14 | + | aPC, bPC := newPair() | |
| 15 | + | a, err := NewTransportWithOptions(aPC, bPC.addr, key(), 1, 2, Options{ | |
| 16 | + | Clock: clk, ResendAfter: time.Hour, Heartbeat: time.Hour, | |
| 17 | + | }) | |
| 18 | + | if err != nil { | |
| 19 | + | t.Fatal(err) | |
| 20 | + | } | |
| 21 | + | b, err := NewTransportWithOptions(bPC, aPC.addr, key(), 2, 1, Options{ | |
| 22 | + | Clock: clk, ResendAfter: time.Hour, Heartbeat: time.Hour, | |
| 23 | + | }) | |
| 24 | + | if err != nil { | |
| 25 | + | _ = a.Close() | |
| 26 | + | t.Fatal(err) | |
| 27 | + | } | |
| 28 | + | releaseACK := make(chan struct{}) | |
| 29 | + | defer close(releaseACK) | |
| 30 | + | defer closeFloodTransports(a, b) | |
| 31 | + | ||
| 32 | + | ackStarted := make(chan struct{}, 1) | |
| 33 | + | aPC.drop = func(_ []byte, _ net.Addr) bool { | |
| 34 | + | ackStarted <- struct{}{} | |
| 35 | + | <-releaseACK | |
| 36 | + | return true | |
| 37 | + | } | |
| 38 | + | contacts := make(chan struct{}, 2) | |
| 39 | + | a.mu.Lock() | |
| 40 | + | a.afterAuthenticatedPacket = func() { contacts <- struct{}{} } | |
| 41 | + | // Keep diagnostics local to this test so each snapshot has an explicit | |
| 42 | + | // causative event instead of depending on observer scheduling. | |
| 43 | + | a.diagnosticCh = make(chan Diagnostic, 1) | |
| 44 | + | a.mu.Unlock() | |
| 45 | + | ||
| 46 | + | if err := b.Send(ports.Frame{Type: ports.MsgInput, Payload: []byte("first")}); err != nil { | |
| 47 | + | t.Fatal(err) | |
| 48 | + | } | |
| 49 | + | awaitSignal(t, contacts, "first authenticated record") | |
| 50 | + | awaitSignal(t, ackStarted, "blocked ACK") | |
| 51 | + | ||
| 52 | + | clk.advance(time.Second) | |
| 53 | + | if err := b.Send(ports.Frame{Type: ports.MsgInput, Payload: []byte("second")}); err != nil { | |
| 54 | + | t.Fatal(err) | |
| 55 | + | } | |
| 56 | + | awaitSignal(t, contacts, "second authenticated record while ACK write is blocked") | |
| 57 | + | a.emitDiagnostic() | |
| 58 | + | diagnostic := awaitResult(t, a.diagnosticCh, "second-record diagnostic") | |
| 59 | + | if diagnostic.SinceCompleteRecord != 0 { | |
| 60 | + | t.Fatalf("second record age=%v, want 0 before ACK release", diagnostic.SinceCompleteRecord) | |
| 61 | + | } | |
| 62 | + | } | |
| 63 | + | ||
| 64 | + | func TestHeartbeatContinuesDuringRetransmitStorm(t *testing.T) { | |
| 65 | + | clk := newManualClock(time.Date(2030, 1, 2, 3, 4, 5, 0, time.UTC)) | |
| 66 | + | aPC, bPC := newPair() | |
| 67 | + | const heartbeat = 11 * time.Millisecond | |
| 68 | + | a, err := NewTransportWithOptions(aPC, bPC.addr, key(), 1, 2, Options{ | |
| 69 | + | Clock: clk, MTU: 128, ResendAfter: 10 * time.Millisecond, Heartbeat: heartbeat, | |
| 70 | + | }) | |
| 71 | + | if err != nil { | |
| 72 | + | t.Fatal(err) | |
| 73 | + | } | |
| 74 | + | defer func() { _ = a.Close() }() | |
| 75 | + | for range 2 { | |
| 76 | + | awaitSignal(t, clk.timerCreated, "resend and heartbeat timer creation") | |
| 77 | + | } | |
| 78 | + | ||
| 79 | + | retransmitStarted := make(chan struct{}, 1) | |
| 80 | + | releaseRetransmit := make(chan struct{}) | |
| 81 | + | defer close(releaseRetransmit) | |
| 82 | + | heartbeats := make(chan struct{}, 2) | |
| 83 | + | aPC.drop = func(pkt []byte, _ net.Addr) bool { | |
| 84 | + | _, raw, err := a.codec.Open(pkt, a.sendDir, nil, nil) | |
| 85 | + | if err != nil { | |
| 86 | + | t.Errorf("open outbound packet: %v", err) | |
| 87 | + | return true | |
| 88 | + | } | |
| 89 | + | frag, err := pdgram.UnmarshalFragment(raw) | |
| 90 | + | if err != nil { | |
| 91 | + | t.Errorf("unmarshal outbound fragment: %v", err) | |
| 92 | + | return true | |
| 93 | + | } | |
| 94 | + | if frag.Count == 1 && len(frag.Data) == 9 && frag.Data[0] == recProbe { | |
| 95 | + | heartbeats <- struct{}{} | |
| 96 | + | return true | |
| 97 | + | } | |
| 98 | + | select { | |
| 99 | + | case retransmitStarted <- struct{}{}: | |
| 100 | + | <-releaseRetransmit | |
| 101 | + | default: | |
| 102 | + | } | |
| 103 | + | return true | |
| 104 | + | } | |
| 105 | + | ||
| 106 | + | now := clk.Now() | |
| 107 | + | a.mu.Lock() | |
| 108 | + | for seq := uint64(1); seq <= 64; seq++ { | |
| 109 | + | a.pending[seq] = &pending{ | |
| 110 | + | frame: ports.Frame{Type: ports.MsgInput, Payload: make([]byte, 24*a.mtu)}, | |
| 111 | + | enqueued: now.Add(-time.Second), first: now.Add(-time.Second), last: now.Add(-time.Second), | |
| 112 | + | } | |
| 113 | + | } | |
| 114 | + | a.mu.Unlock() | |
| 115 | + | ||
| 116 | + | clk.advance(10 * time.Millisecond) | |
| 117 | + | awaitSignal(t, retransmitStarted, "first retransmit in 64-record storm") | |
| 118 | + | for interval := 1; interval <= 2; interval++ { | |
| 119 | + | clk.advance(heartbeat) | |
| 120 | + | awaitSignal(t, heartbeats, "heartbeat during retransmit storm") | |
| 121 | + | } | |
| 122 | + | } | |
| @@ -0,0 +1,122 @@ | |||||
| 1 | + | package dgram | |||
| 2 | + | ||||
| 3 | + | import ( | |||
| 4 | + | "net" | |||
| 5 | + | "testing" | |||
| 6 | + | "time" | |||
| 7 | + | ||||
| 8 | + | "github.com/bnema/vev/internal/ports" | |||
| 9 | + | pdgram "github.com/bnema/vev/pkg/dgram" | |||
| 10 | + | ) | |||
| 11 | + | ||||
| 12 | + | func TestReadLoopDoesNotWriteACKSynchronously(t *testing.T) { | |||
| 13 | + | clk := newManualClock(time.Date(2030, 1, 2, 3, 4, 5, 0, time.UTC)) | |||
| 14 | + | aPC, bPC := newPair() | |||
| 15 | + | a, err := NewTransportWithOptions(aPC, bPC.addr, key(), 1, 2, Options{ | |||
| 16 | + | Clock: clk, ResendAfter: time.Hour, Heartbeat: time.Hour, | |||
| 17 | + | }) | |||
| 18 | + | if err != nil { | |||
| 19 | + | t.Fatal(err) | |||
| 20 | + | } | |||
| 21 | + | b, err := NewTransportWithOptions(bPC, aPC.addr, key(), 2, 1, Options{ | |||
| 22 | + | Clock: clk, ResendAfter: time.Hour, Heartbeat: time.Hour, | |||
| 23 | + | }) | |||
| 24 | + | if err != nil { | |||
| 25 | + | _ = a.Close() | |||
| 26 | + | t.Fatal(err) | |||
| 27 | + | } | |||
| 28 | + | releaseACK := make(chan struct{}) | |||
| 29 | + | defer close(releaseACK) | |||
| 30 | + | defer closeFloodTransports(a, b) | |||
| 31 | + | ||||
| 32 | + | ackStarted := make(chan struct{}, 1) | |||
| 33 | + | aPC.drop = func(_ []byte, _ net.Addr) bool { | |||
| 34 | + | ackStarted <- struct{}{} | |||
| 35 | + | <-releaseACK | |||
| 36 | + | return true | |||
| 37 | + | } | |||
| 38 | + | contacts := make(chan struct{}, 2) | |||
| 39 | + | a.mu.Lock() | |||
| 40 | + | a.afterAuthenticatedPacket = func() { contacts <- struct{}{} } | |||
| 41 | + | // Keep diagnostics local to this test so each snapshot has an explicit | |||
| 42 | + | // causative event instead of depending on observer scheduling. | |||
| 43 | + | a.diagnosticCh = make(chan Diagnostic, 1) | |||
| 44 | + | a.mu.Unlock() | |||
| 45 | + | ||||
| 46 | + | if err := b.Send(ports.Frame{Type: ports.MsgInput, Payload: []byte("first")}); err != nil { | |||
| 47 | + | t.Fatal(err) | |||
| 48 | + | } | |||
| 49 | + | awaitSignal(t, contacts, "first authenticated record") | |||
| 50 | + | awaitSignal(t, ackStarted, "blocked ACK") | |||
| 51 | + | ||||
| 52 | + | clk.advance(time.Second) | |||
| 53 | + | if err := b.Send(ports.Frame{Type: ports.MsgInput, Payload: []byte("second")}); err != nil { | |||
| 54 | + | t.Fatal(err) | |||
| 55 | + | } | |||
| 56 | + | awaitSignal(t, contacts, "second authenticated record while ACK write is blocked") | |||
| 57 | + | a.emitDiagnostic() | |||
| 58 | + | diagnostic := awaitResult(t, a.diagnosticCh, "second-record diagnostic") | |||
| 59 | + | if diagnostic.SinceCompleteRecord != 0 { | |||
| 60 | + | t.Fatalf("second record age=%v, want 0 before ACK release", diagnostic.SinceCompleteRecord) | |||
| 61 | + | } | |||
| 62 | + | } | |||
| 63 | + | ||||
| 64 | + | func TestHeartbeatContinuesDuringRetransmitStorm(t *testing.T) { | |||
| 65 | + | clk := newManualClock(time.Date(2030, 1, 2, 3, 4, 5, 0, time.UTC)) | |||
| 66 | + | aPC, bPC := newPair() | |||
| 67 | + | const heartbeat = 11 * time.Millisecond | |||
| 68 | + | a, err := NewTransportWithOptions(aPC, bPC.addr, key(), 1, 2, Options{ | |||
| 69 | + | Clock: clk, MTU: 128, ResendAfter: 10 * time.Millisecond, Heartbeat: heartbeat, | |||
| 70 | + | }) | |||
| 71 | + | if err != nil { | |||
| 72 | + | t.Fatal(err) | |||
| 73 | + | } | |||
| 74 | + | defer func() { _ = a.Close() }() | |||
| 75 | + | for range 2 { | |||
| 76 | + | awaitSignal(t, clk.timerCreated, "resend and heartbeat timer creation") | |||
| 77 | + | } | |||
| 78 | + | ||||
| 79 | + | retransmitStarted := make(chan struct{}, 1) | |||
| 80 | + | releaseRetransmit := make(chan struct{}) | |||
| 81 | + | defer close(releaseRetransmit) | |||
| 82 | + | heartbeats := make(chan struct{}, 2) | |||
| 83 | + | aPC.drop = func(pkt []byte, _ net.Addr) bool { | |||
| 84 | + | _, raw, err := a.codec.Open(pkt, a.sendDir, nil, nil) | |||
| 85 | + | if err != nil { | |||
| 86 | + | t.Errorf("open outbound packet: %v", err) | |||
| 87 | + | return true | |||
| 88 | + | } | |||
| 89 | + | frag, err := pdgram.UnmarshalFragment(raw) | |||
| 90 | + | if err != nil { | |||
| 91 | + | t.Errorf("unmarshal outbound fragment: %v", err) | |||
| 92 | + | return true | |||
| 93 | + | } | |||
| 94 | + | if frag.Count == 1 && len(frag.Data) == 9 && frag.Data[0] == recProbe { | |||
| 95 | + | heartbeats <- struct{}{} | |||
| 96 | + | return true | |||
| 97 | + | } | |||
| 98 | + | select { | |||
| 99 | + | case retransmitStarted <- struct{}{}: | |||
| 100 | + | <-releaseRetransmit | |||
| 101 | + | default: | |||
| 102 | + | } | |||
| 103 | + | return true | |||
| 104 | + | } | |||
| 105 | + | ||||
| 106 | + | now := clk.Now() | |||
| 107 | + | a.mu.Lock() | |||
| 108 | + | for seq := uint64(1); seq <= 64; seq++ { | |||
| 109 | + | a.pending[seq] = &pending{ | |||
| 110 | + | frame: ports.Frame{Type: ports.MsgInput, Payload: make([]byte, 24*a.mtu)}, | |||
| 111 | + | enqueued: now.Add(-time.Second), first: now.Add(-time.Second), last: now.Add(-time.Second), | |||
| 112 | + | } | |||
| 113 | + | } | |||
| 114 | + | a.mu.Unlock() | |||
| 115 | + | ||||
| 116 | + | clk.advance(10 * time.Millisecond) | |||
| 117 | + | awaitSignal(t, retransmitStarted, "first retransmit in 64-record storm") | |||
| 118 | + | for interval := 1; interval <= 2; interval++ { | |||
| 119 | + | clk.advance(heartbeat) | |||
| 120 | + | awaitSignal(t, heartbeats, "heartbeat during retransmit storm") | |||
| 121 | + | } | |||
| 122 | + | } | |||