Merge branch 'fix/iso-public-ip-boot-race' of github.com:mulgadc/spinifex into fix/iso-public-ip-boot-race

go.mod

@@ -12,6 +12,7 @@ require (
 	github.com/insomniacslk/dhcp v0.0.0-20260407060928-11b94ed970f2
 	github.com/kdomanski/iso9660 v0.4.0
 	github.com/klauspost/cpuid/v2 v2.3.0
+	github.com/mdlayher/packet v1.1.2
 	github.com/mulgadc/predastore v1.1.6
 	github.com/mulgadc/viperblock v1.2.1
 	github.com/nats-io/nats-server/v2 v2.12.7
@@ -23,6 +24,7 @@ require (
 	github.com/spf13/viper v1.21.0
 	github.com/stretchr/testify v1.11.1
 	golang.org/x/crypto v0.50.0
+	golang.org/x/sys v0.43.0
 	gopkg.in/ini.v1 v1.67.1
 )
 
@@ -86,7 +88,6 @@ require (
 	github.com/mattn/go-isatty v0.0.20 // indirect
 	github.com/mattn/go-localereader v0.0.1 // indirect
 	github.com/mattn/go-runewidth v0.0.21 // indirect
-	github.com/mdlayher/packet v1.1.2 // indirect
 	github.com/mdlayher/socket v0.4.1 // indirect
 	github.com/minio/crc64nvme v1.1.1 // indirect
 	github.com/minio/highwayhash v1.0.4 // indirect
@@ -123,7 +124,6 @@ require (
 	golang.org/x/mod v0.34.0 // indirect
 	golang.org/x/net v0.53.0 // indirect
 	golang.org/x/sync v0.20.0 // indirect
-	golang.org/x/sys v0.43.0 // indirect
 	golang.org/x/telemetry v0.0.0-20260311193753-579e4da9a98c // indirect
 	golang.org/x/term v0.42.0 // indirect
 	golang.org/x/text v0.36.0 // indirect

spinifex/services/vpcd/dhcp/nclient4.go

@@ -3,12 +3,46 @@ package dhcp
 import (
 	"context"
 	"fmt"
+	"net"
 	"time"
 
 	"github.com/insomniacslk/dhcp/dhcpv4"
 	"github.com/insomniacslk/dhcp/dhcpv4/nclient4"
+	"github.com/mdlayher/packet"
+	"golang.org/x/sys/unix"
 )
 
+// dhcpClientPort is the IANA-assigned UDP port for DHCP clients.
+const dhcpClientPort = 68
+
+// openPromiscBridgeConn opens an AF_PACKET DGRAM socket on iface filtered
+// to IPv4 and enables PACKET_MR_PROMISC on it. Equivalent to nclient4's
+// own NewRawUDPConn but with promisc on.
+//
+// Why promisc: vpcd uses a synthetic per-tenant chaddr (e.g. 02:f3:...)
+// that never matches the bridge's MAC. Many real-world DHCP servers
+// ignore the BOOTP BROADCAST flag and reply with a unicast OFFER addressed
+// to that chaddr. A Linux bridge floods unknown unicast to slave ports
+// but does NOT pass it up to the bridge netdev unless IFF_PROMISC is set,
+// so the AF_PACKET socket misses the OFFER and DORA times out. Promisc
+// flips IFF_PROMISC for the lifetime of the socket (per-DORA), and the
+// kernel refcounts it so it auto-clears on Close.
+func openPromiscBridgeConn(iface string) (net.PacketConn, net.HardwareAddr, error) {
+	ifc, err := net.InterfaceByName(iface)
+	if err != nil {
+		return nil, nil, fmt.Errorf("lookup iface %s: %w", iface, err)
+	}
+	raw, err := packet.Listen(ifc, packet.Datagram, unix.ETH_P_IP, nil)
+	if err != nil {
+		return nil, nil, fmt.Errorf("packet.Listen on %s: %w", iface, err)
+	}
+	if err := raw.SetPromiscuous(true); err != nil {
+		_ = raw.Close()
+		return nil, nil, fmt.Errorf("set promisc on %s: %w", iface, err)
+	}
+	return nclient4.NewBroadcastUDPConn(raw, &net.UDPAddr{Port: dhcpClientPort}), ifc.HardwareAddr, nil
+}
+
 // NClient4Client is the production DHCP client backed by
 // github.com/insomniacslk/dhcp/dhcpv4/nclient4. Each Acquire/Renew/Release
 // opens an AF_PACKET socket on the target bridge for the duration of the
@@ -39,12 +73,17 @@ func (c *NClient4Client) Acquire(ctx context.Context, req AcquireRequest) (*Leas
 		return nil, fmt.Errorf("dhcp acquire: hw_addr is required")
 	}
 
-	client, err := nclient4.New(req.Bridge,
+	conn, ifaceMAC, err := openPromiscBridgeConn(req.Bridge)
+	if err != nil {
+		return nil, fmt.Errorf("open promisc conn on %s: %w", req.Bridge, err)
+	}
+	client, err := nclient4.NewWithConn(conn, ifaceMAC,
 		nclient4.WithHWAddr(req.HWAddr),
 		nclient4.WithTimeout(c.timeout),
 		nclient4.WithRetry(c.retry),
 	)
 	if err != nil {
+		_ = conn.Close()
 		return nil, fmt.Errorf("open nclient4 on %s: %w", req.Bridge, err)
 	}
 	defer func() { _ = client.Close() }()
@@ -71,12 +110,17 @@ func (c *NClient4Client) Renew(ctx context.Context, lease *Lease) (*Lease, error
 		return nil, fmt.Errorf("dhcp renew: %w", err)
 	}
 
-	client, err := nclient4.New(lease.Bridge,
+	conn, ifaceMAC, err := openPromiscBridgeConn(lease.Bridge)
+	if err != nil {
+		return nil, fmt.Errorf("open promisc conn on %s for renew: %w", lease.Bridge, err)
+	}
+	client, err := nclient4.NewWithConn(conn, ifaceMAC,
 		nclient4.WithHWAddr(lease.HWAddr),
 		nclient4.WithTimeout(c.timeout),
 		nclient4.WithRetry(c.retry),
 	)
 	if err != nil {
+		_ = conn.Close()
 		return nil, fmt.Errorf("open nclient4 on %s for renew: %w", lease.Bridge, err)
 	}
 	defer func() { _ = client.Close() }()
@@ -105,11 +149,16 @@ func (c *NClient4Client) Release(ctx context.Context, lease *Lease) error {
 		return fmt.Errorf("dhcp release: %w", err)
 	}
 
-	client, err := nclient4.New(lease.Bridge,
+	conn, ifaceMAC, err := openPromiscBridgeConn(lease.Bridge)
+	if err != nil {
+		return fmt.Errorf("open promisc conn on %s for release: %w", lease.Bridge, err)
+	}
+	client, err := nclient4.NewWithConn(conn, ifaceMAC,
 		nclient4.WithHWAddr(lease.HWAddr),
 		nclient4.WithTimeout(c.timeout),
 	)
 	if err != nil {
+		_ = conn.Close()
 		return fmt.Errorf("open nclient4 on %s for release: %w", lease.Bridge, err)
 	}
 	defer func() { _ = client.Close() }()