Merge pull request #324 from lstoll/lstoll-ipvlan-l3s

ipvlan: Support L3-Symmetric mode

Godeps/Godeps.json

@@ -1,7 +1,7 @@
 {
 	"ImportPath": "github.com/containernetworking/cni",
 	"GoVersion": "go1.6",
-	"GodepVersion": "v74",
+	"GodepVersion": "v75",
 	"Packages": [
 		"./..."
 	],
@@ -216,11 +216,11 @@
 		},
 		{
 			"ImportPath": "github.com/vishvananda/netlink",
-			"Rev": "9dee363ad4abbc3c9a4a24a9f1e33363e224b111"
+			"Rev": "a1f8555521646b5a9800f39c5fd477597697135c"
 		},
 		{
 			"ImportPath": "github.com/vishvananda/netlink/nl",
-			"Rev": "9dee363ad4abbc3c9a4a24a9f1e33363e224b111"
+			"Rev": "a1f8555521646b5a9800f39c5fd477597697135c"
 		},
 		{
 			"ImportPath": "golang.org/x/sys/unix",

plugins/main/ipvlan/ipvlan.go

@@ -60,6 +60,8 @@ func modeFromString(s string) (netlink.IPVlanMode, error) {
 		return netlink.IPVLAN_MODE_L2, nil
 	case "l3":
 		return netlink.IPVLAN_MODE_L3, nil
+	case "l3s":
+		return netlink.IPVLAN_MODE_L3S, nil
 	default:
 		return 0, fmt.Errorf("unknown ipvlan mode: %q", s)
 	}

vendor/github.com/vishvananda/netlink/bpf_linux.go

@@ -8,11 +8,11 @@ package netlink
 #include <stdint.h>
 #include <unistd.h>
 
-static int load_simple_bpf(int prog_type) {
+static int load_simple_bpf(int prog_type, int ret) {
 #ifdef __NR_bpf
-	// { return 1; }
+	// { return ret; }
 	__u64 __attribute__((aligned(8))) insns[] = {
-		0x00000001000000b7ull,
+		0x00000000000000b7ull | ((__u64)ret<<32),
 		0x0000000000000095ull,
 	};
 	__u8 __attribute__((aligned(8))) license[] = "ASL2";
@@ -51,10 +51,12 @@ const (
 	BPF_PROG_TYPE_KPROBE
 	BPF_PROG_TYPE_SCHED_CLS
 	BPF_PROG_TYPE_SCHED_ACT
+	BPF_PROG_TYPE_TRACEPOINT
+	BPF_PROG_TYPE_XDP
 )
 
 // loadSimpleBpf loads a trivial bpf program for testing purposes
-func loadSimpleBpf(progType BpfProgType) (int, error) {
-	fd, err := C.load_simple_bpf(C.int(progType))
+func loadSimpleBpf(progType BpfProgType, ret int) (int, error) {
+	fd, err := C.load_simple_bpf(C.int(progType), C.int(ret))
 	return int(fd), err
 }

vendor/github.com/vishvananda/netlink/filter.go

@@ -60,7 +60,7 @@ func (a TcAct) String() string {
 	case TC_ACT_JUMP:
 		return "jump"
 	}
-	return fmt.Sprintf("0x%x", a)
+	return fmt.Sprintf("0x%x", int32(a))
 }
 
 type TcPolAct int32
@@ -86,7 +86,7 @@ func (a TcPolAct) String() string {
 	case TC_POLICE_PIPE:
 		return "pipe"
 	}
-	return fmt.Sprintf("0x%x", a)
+	return fmt.Sprintf("0x%x", int32(a))
 }
 
 type ActionAttrs struct {

vendor/github.com/vishvananda/netlink/link.go

@@ -26,11 +26,15 @@ type LinkAttrs struct {
 	Name         string
 	HardwareAddr net.HardwareAddr
 	Flags        net.Flags
+	RawFlags     uint32
 	ParentIndex  int         // index of the parent link device
 	MasterIndex  int         // must be the index of a bridge
 	Namespace    interface{} // nil | NsPid | NsFd
 	Alias        string
 	Statistics   *LinkStatistics
+	Promisc      int
+	Xdp          *LinkXdp
+	EncapType    string
 }
 
 // NewLinkAttrs returns LinkAttrs structure filled with default values
@@ -69,6 +73,11 @@ type LinkStatistics struct {
 	TxCompressed      uint32
 }
 
+type LinkXdp struct {
+	Fd       int
+	Attached bool
+}
+
 // Device links cannot be created via netlink. These links
 // are links created by udev like 'lo' and 'etho0'
 type Device struct {
@@ -253,6 +262,7 @@ type IPVlanMode uint16
 const (
 	IPVLAN_MODE_L2 IPVlanMode = iota
 	IPVLAN_MODE_L3
+	IPVLAN_MODE_L3S
 	IPVLAN_MODE_MAX
 )
 

vendor/github.com/vishvananda/netlink/link_linux.go

@@ -55,6 +55,44 @@ func (h *Handle) ensureIndex(link *LinkAttrs) {
 	}
 }
 
+func (h *Handle) SetPromiscOn(link Link) error {
+	base := link.Attrs()
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+
+	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
+	msg.Change = syscall.IFF_PROMISC
+	msg.Flags = syscall.IFF_UP
+	msg.Index = int32(base.Index)
+	req.AddData(msg)
+
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
+func SetPromiscOn(link Link) error {
+	return pkgHandle.SetPromiscOn(link)
+}
+
+func (h *Handle) SetPromiscOff(link Link) error {
+	base := link.Attrs()
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+
+	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
+	msg.Change = syscall.IFF_PROMISC
+	msg.Flags = 0 & ^syscall.IFF_UP
+	msg.Index = int32(base.Index)
+	req.AddData(msg)
+
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
+func SetPromiscOff(link Link) error {
+	return pkgHandle.SetPromiscOff(link)
+}
+
 // LinkSetUp enables the link device.
 // Equivalent to: `ip link set $link up`
 func LinkSetUp(link Link) error {
@@ -378,6 +416,23 @@ func (h *Handle) LinkSetNsFd(link Link, fd int) error {
 	return err
 }
 
+// LinkSetXdpFd adds a bpf function to the driver. The fd must be a bpf
+// program loaded with bpf(type=BPF_PROG_TYPE_XDP)
+func LinkSetXdpFd(link Link, fd int) error {
+	base := link.Attrs()
+	ensureIndex(base)
+	req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+
+	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
+	msg.Index = int32(base.Index)
+	req.AddData(msg)
+
+	addXdpAttrs(&LinkXdp{Fd: fd}, req)
+
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
 func boolAttr(val bool) []byte {
 	var v uint8
 	if val {
@@ -655,6 +710,10 @@ func (h *Handle) LinkAdd(link Link) error {
 		req.AddData(attr)
 	}
 
+	if base.Xdp != nil {
+		addXdpAttrs(base.Xdp, req)
+	}
+
 	linkInfo := nl.NewRtAttr(syscall.IFLA_LINKINFO, nil)
 	nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_KIND, nl.NonZeroTerminated(link.Type()))
 
@@ -877,7 +936,10 @@ func linkDeserialize(m []byte) (Link, error) {
 		return nil, err
 	}
 
-	base := LinkAttrs{Index: int(msg.Index), Flags: linkFlags(msg.Flags)}
+	base := LinkAttrs{Index: int(msg.Index), RawFlags: msg.Flags, Flags: linkFlags(msg.Flags), EncapType: msg.EncapType()}
+	if msg.Flags&syscall.IFF_PROMISC != 0 {
+		base.Promisc = 1
+	}
 	var link Link
 	linkType := ""
 	for _, attr := range attrs {
@@ -964,6 +1026,12 @@ func linkDeserialize(m []byte) (Link, error) {
 			base.Alias = string(attr.Value[:len(attr.Value)-1])
 		case syscall.IFLA_STATS:
 			base.Statistics = parseLinkStats(attr.Value[:])
+		case nl.IFLA_XDP:
+			xdp, err := parseLinkXdp(attr.Value[:])
+			if err != nil {
+				return nil, err
+			}
+			base.Xdp = xdp
 		}
 	}
 	// Links that don't have IFLA_INFO_KIND are hardware devices
@@ -1395,3 +1463,28 @@ func parseGretapData(link Link, data []syscall.NetlinkRouteAttr) {
 func parseLinkStats(data []byte) *LinkStatistics {
 	return (*LinkStatistics)(unsafe.Pointer(&data[0:SizeofLinkStats][0]))
 }
+
+func addXdpAttrs(xdp *LinkXdp, req *nl.NetlinkRequest) {
+	attrs := nl.NewRtAttr(nl.IFLA_XDP|syscall.NLA_F_NESTED, nil)
+	b := make([]byte, 4)
+	native.PutUint32(b, uint32(xdp.Fd))
+	nl.NewRtAttrChild(attrs, nl.IFLA_XDP_FD, b)
+	req.AddData(attrs)
+}
+
+func parseLinkXdp(data []byte) (*LinkXdp, error) {
+	attrs, err := nl.ParseRouteAttr(data)
+	if err != nil {
+		return nil, err
+	}
+	xdp := &LinkXdp{}
+	for _, attr := range attrs {
+		switch attr.Attr.Type {
+		case nl.IFLA_XDP_FD:
+			xdp.Fd = int(native.Uint32(attr.Value[0:4]))
+		case nl.IFLA_XDP_ATTACHED:
+			xdp.Attached = attr.Value[0] != 0
+		}
+	}
+	return xdp, nil
+}

vendor/github.com/vishvananda/netlink/neigh_linux.go

@@ -151,8 +151,10 @@ func neighHandle(neigh *Neigh, req *nl.NetlinkRequest) error {
 	dstData := nl.NewRtAttr(NDA_DST, ipData)
 	req.AddData(dstData)
 
-	hwData := nl.NewRtAttr(NDA_LLADDR, []byte(neigh.HardwareAddr))
-	req.AddData(hwData)
+	if neigh.Flags != NTF_PROXY || neigh.HardwareAddr != nil {
+		hwData := nl.NewRtAttr(NDA_LLADDR, []byte(neigh.HardwareAddr))
+		req.AddData(hwData)
+	}
 
 	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
 	return err
@@ -165,14 +167,33 @@ func NeighList(linkIndex, family int) ([]Neigh, error) {
 	return pkgHandle.NeighList(linkIndex, family)
 }
 
+// NeighProxyList gets a list of neighbor proxies in the system.
+// Equivalent to: `ip neighbor show proxy`.
+// The list can be filtered by link and ip family.
+func NeighProxyList(linkIndex, family int) ([]Neigh, error) {
+	return pkgHandle.NeighProxyList(linkIndex, family)
+}
+
 // NeighList gets a list of IP-MAC mappings in the system (ARP table).
 // Equivalent to: `ip neighbor show`.
 // The list can be filtered by link and ip family.
 func (h *Handle) NeighList(linkIndex, family int) ([]Neigh, error) {
+	return h.neighList(linkIndex, family, 0)
+}
+
+// NeighProxyList gets a list of neighbor proxies in the system.
+// Equivalent to: `ip neighbor show proxy`.
+// The list can be filtered by link, ip family.
+func (h *Handle) NeighProxyList(linkIndex, family int) ([]Neigh, error) {
+	return h.neighList(linkIndex, family, NTF_PROXY)
+}
+
+func (h *Handle) neighList(linkIndex, family, flags int) ([]Neigh, error) {
 	req := h.newNetlinkRequest(syscall.RTM_GETNEIGH, syscall.NLM_F_DUMP)
 	msg := Ndmsg{
 		Family: uint8(family),
 		Index:  uint32(linkIndex),
+		Flags:  uint8(flags),
 	}
 	req.AddData(&msg)
 

vendor/github.com/vishvananda/netlink/nl/link_linux.go

@@ -1,13 +1,35 @@
 package nl
 
 import (
+	"syscall"
 	"unsafe"
 )
 
 const (
 	DEFAULT_CHANGE = 0xFFFFFFFF
 	// doesn't exist in syscall
-	IFLA_VFINFO_LIST = 0x16
+	IFLA_VFINFO_LIST = syscall.IFLA_IFALIAS + 1 + iota
+	IFLA_STATS64
+	IFLA_VF_PORTS
+	IFLA_PORT_SELF
+	IFLA_AF_SPEC
+	IFLA_GROUP
+	IFLA_NET_NS_FD
+	IFLA_EXT_MASK
+	IFLA_PROMISCUITY
+	IFLA_NUM_TX_QUEUES
+	IFLA_NUM_RX_QUEUES
+	IFLA_CARRIER
+	IFLA_PHYS_PORT_ID
+	IFLA_CARRIER_CHANGES
+	IFLA_PHYS_SWITCH_ID
+	IFLA_LINK_NETNSID
+	IFLA_PHYS_PORT_NAME
+	IFLA_PROTO_DOWN
+	IFLA_GSO_MAX_SEGS
+	IFLA_GSO_MAX_SIZE
+	IFLA_PAD
+	IFLA_XDP
 )
 
 const (
@@ -89,11 +111,6 @@ const (
 	IFLA_IPVLAN_MAX = IFLA_IPVLAN_MODE
 )
 
-const (
-	// not defined in syscall
-	IFLA_NET_NS_FD = 28
-)
-
 const (
 	IFLA_MACVLAN_UNSPEC = iota
 	IFLA_MACVLAN_MODE
@@ -394,3 +411,10 @@ func DeserializeVfRssQueryEn(b []byte) *VfRssQueryEn {
 func (msg *VfRssQueryEn) Serialize() []byte {
 	return (*(*[SizeofVfRssQueryEn]byte)(unsafe.Pointer(msg)))[:]
 }
+
+const (
+	IFLA_XDP_UNSPEC   = iota
+	IFLA_XDP_FD       /* fd of xdp program to attach, or -1 to remove */
+	IFLA_XDP_ATTACHED /* read-only bool indicating if prog is attached */
+	IFLA_XDP_MAX      = IFLA_XDP_ATTACHED
+)

vendor/github.com/vishvananda/netlink/nl/nl_linux.go

@@ -100,6 +100,147 @@ func (msg *IfInfomsg) Len() int {
 	return syscall.SizeofIfInfomsg
 }
 
+func (msg *IfInfomsg) EncapType() string {
+	switch msg.Type {
+	case 0:
+		return "generic"
+	case syscall.ARPHRD_ETHER:
+		return "ether"
+	case syscall.ARPHRD_EETHER:
+		return "eether"
+	case syscall.ARPHRD_AX25:
+		return "ax25"
+	case syscall.ARPHRD_PRONET:
+		return "pronet"
+	case syscall.ARPHRD_CHAOS:
+		return "chaos"
+	case syscall.ARPHRD_IEEE802:
+		return "ieee802"
+	case syscall.ARPHRD_ARCNET:
+		return "arcnet"
+	case syscall.ARPHRD_APPLETLK:
+		return "atalk"
+	case syscall.ARPHRD_DLCI:
+		return "dlci"
+	case syscall.ARPHRD_ATM:
+		return "atm"
+	case syscall.ARPHRD_METRICOM:
+		return "metricom"
+	case syscall.ARPHRD_IEEE1394:
+		return "ieee1394"
+	case syscall.ARPHRD_INFINIBAND:
+		return "infiniband"
+	case syscall.ARPHRD_SLIP:
+		return "slip"
+	case syscall.ARPHRD_CSLIP:
+		return "cslip"
+	case syscall.ARPHRD_SLIP6:
+		return "slip6"
+	case syscall.ARPHRD_CSLIP6:
+		return "cslip6"
+	case syscall.ARPHRD_RSRVD:
+		return "rsrvd"
+	case syscall.ARPHRD_ADAPT:
+		return "adapt"
+	case syscall.ARPHRD_ROSE:
+		return "rose"
+	case syscall.ARPHRD_X25:
+		return "x25"
+	case syscall.ARPHRD_HWX25:
+		return "hwx25"
+	case syscall.ARPHRD_PPP:
+		return "ppp"
+	case syscall.ARPHRD_HDLC:
+		return "hdlc"
+	case syscall.ARPHRD_LAPB:
+		return "lapb"
+	case syscall.ARPHRD_DDCMP:
+		return "ddcmp"
+	case syscall.ARPHRD_RAWHDLC:
+		return "rawhdlc"
+	case syscall.ARPHRD_TUNNEL:
+		return "ipip"
+	case syscall.ARPHRD_TUNNEL6:
+		return "tunnel6"
+	case syscall.ARPHRD_FRAD:
+		return "frad"
+	case syscall.ARPHRD_SKIP:
+		return "skip"
+	case syscall.ARPHRD_LOOPBACK:
+		return "loopback"
+	case syscall.ARPHRD_LOCALTLK:
+		return "ltalk"
+	case syscall.ARPHRD_FDDI:
+		return "fddi"
+	case syscall.ARPHRD_BIF:
+		return "bif"
+	case syscall.ARPHRD_SIT:
+		return "sit"
+	case syscall.ARPHRD_IPDDP:
+		return "ip/ddp"
+	case syscall.ARPHRD_IPGRE:
+		return "gre"
+	case syscall.ARPHRD_PIMREG:
+		return "pimreg"
+	case syscall.ARPHRD_HIPPI:
+		return "hippi"
+	case syscall.ARPHRD_ASH:
+		return "ash"
+	case syscall.ARPHRD_ECONET:
+		return "econet"
+	case syscall.ARPHRD_IRDA:
+		return "irda"
+	case syscall.ARPHRD_FCPP:
+		return "fcpp"
+	case syscall.ARPHRD_FCAL:
+		return "fcal"
+	case syscall.ARPHRD_FCPL:
+		return "fcpl"
+	case syscall.ARPHRD_FCFABRIC:
+		return "fcfb0"
+	case syscall.ARPHRD_FCFABRIC + 1:
+		return "fcfb1"
+	case syscall.ARPHRD_FCFABRIC + 2:
+		return "fcfb2"
+	case syscall.ARPHRD_FCFABRIC + 3:
+		return "fcfb3"
+	case syscall.ARPHRD_FCFABRIC + 4:
+		return "fcfb4"
+	case syscall.ARPHRD_FCFABRIC + 5:
+		return "fcfb5"
+	case syscall.ARPHRD_FCFABRIC + 6:
+		return "fcfb6"
+	case syscall.ARPHRD_FCFABRIC + 7:
+		return "fcfb7"
+	case syscall.ARPHRD_FCFABRIC + 8:
+		return "fcfb8"
+	case syscall.ARPHRD_FCFABRIC + 9:
+		return "fcfb9"
+	case syscall.ARPHRD_FCFABRIC + 10:
+		return "fcfb10"
+	case syscall.ARPHRD_FCFABRIC + 11:
+		return "fcfb11"
+	case syscall.ARPHRD_FCFABRIC + 12:
+		return "fcfb12"
+	case syscall.ARPHRD_IEEE802_TR:
+		return "tr"
+	case syscall.ARPHRD_IEEE80211:
+		return "ieee802.11"
+	case syscall.ARPHRD_IEEE80211_PRISM:
+		return "ieee802.11/prism"
+	case syscall.ARPHRD_IEEE80211_RADIOTAP:
+		return "ieee802.11/radiotap"
+	case syscall.ARPHRD_IEEE802154:
+		return "ieee802.15.4"
+
+	case 65534:
+		return "none"
+	case 65535:
+		return "void"
+	}
+	return fmt.Sprintf("unknown%d", msg.Type)
+}
+
 func rtaAlignOf(attrlen int) int {
 	return (attrlen + syscall.RTA_ALIGNTO - 1) & ^(syscall.RTA_ALIGNTO - 1)
 }

vendor/github.com/vishvananda/netlink/route_linux.go

@@ -153,7 +153,7 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg
 				} else {
 					gw = nl.NewRtAttr(syscall.RTA_GATEWAY, []byte(nh.Gw.To16()))
 				}
-				gwData := gw.Serialize()
+				gwData = gw.Serialize()
 				rtnh.Len += uint16(len(gwData))
 			}
 			buf = append(buf, rtnh.Serialize()...)

vendor/github.com/vishvananda/netlink/rule_linux.go

@@ -82,41 +82,46 @@ func ruleHandle(rule *Rule, req *nl.NetlinkRequest) error {
 		req.AddData(rtAttrs[i])
 	}
 
-	var (
-		b      = make([]byte, 4)
-		native = nl.NativeEndian()
-	)
+	native := nl.NativeEndian()
 
 	if rule.Priority >= 0 {
+		b := make([]byte, 4)
 		native.PutUint32(b, uint32(rule.Priority))
 		req.AddData(nl.NewRtAttr(nl.FRA_PRIORITY, b))
 	}
 	if rule.Mark >= 0 {
+		b := make([]byte, 4)
 		native.PutUint32(b, uint32(rule.Mark))
 		req.AddData(nl.NewRtAttr(nl.FRA_FWMARK, b))
 	}
 	if rule.Mask >= 0 {
+		b := make([]byte, 4)
 		native.PutUint32(b, uint32(rule.Mask))
 		req.AddData(nl.NewRtAttr(nl.FRA_FWMASK, b))
 	}
 	if rule.Flow >= 0 {
+		b := make([]byte, 4)
 		native.PutUint32(b, uint32(rule.Flow))
 		req.AddData(nl.NewRtAttr(nl.FRA_FLOW, b))
 	}
 	if rule.TunID > 0 {
+		b := make([]byte, 4)
 		native.PutUint32(b, uint32(rule.TunID))
 		req.AddData(nl.NewRtAttr(nl.FRA_TUN_ID, b))
 	}
 	if rule.Table >= 256 {
+		b := make([]byte, 4)
 		native.PutUint32(b, uint32(rule.Table))
 		req.AddData(nl.NewRtAttr(nl.FRA_TABLE, b))
 	}
 	if msg.Table > 0 {
 		if rule.SuppressPrefixlen >= 0 {
+			b := make([]byte, 4)
 			native.PutUint32(b, uint32(rule.SuppressPrefixlen))
 			req.AddData(nl.NewRtAttr(nl.FRA_SUPPRESS_PREFIXLEN, b))
 		}
 		if rule.SuppressIfgroup >= 0 {
+			b := make([]byte, 4)
 			native.PutUint32(b, uint32(rule.SuppressIfgroup))
 			req.AddData(nl.NewRtAttr(nl.FRA_SUPPRESS_IFGROUP, b))
 		}
@@ -129,6 +134,7 @@ func ruleHandle(rule *Rule, req *nl.NetlinkRequest) error {
 	}
 	if rule.Goto >= 0 {
 		msg.Type = nl.FR_ACT_NOP
+		b := make([]byte, 4)
 		native.PutUint32(b, uint32(rule.Goto))
 		req.AddData(nl.NewRtAttr(nl.FRA_GOTO, b))
 	}