vendor: bump netlink to latest master

backend/udp/network.go

@@ -155,7 +155,7 @@ func configureIface(ifname string, ipn ip.IP4Net, mtu int) error {
 		return fmt.Errorf("failed to lookup interface %v", ifname)
 	}
 
-	err = netlink.AddrAdd(iface, &netlink.Addr{ipn.ToIPNet(), ""})
+	err = netlink.AddrAdd(iface, &netlink.Addr{IPNet: ipn.ToIPNet(), Label: ""})
 	if err != nil {
 		return fmt.Errorf("failed to add IP address %v to %v: %v", ipn.String(), ifname, err)
 	}

backend/vxlan/device.go

@@ -302,7 +302,7 @@ func setAddr4(link *netlink.Vxlan, ipn *net.IPNet) error {
 		}
 	}
 
-	addr := netlink.Addr{ipn, ""}
+	addr := netlink.Addr{IPNet: ipn, Label: ""}
 	if err = netlink.AddrAdd(link, &addr); err != nil {
 		return fmt.Errorf("failed to add IP address %s to %s: %s", ipn.String(), link.Attrs().Name, err)
 	}

glide.lock

@@ -1,5 +1,5 @@
-hash: b37f2f98ef5d80a4d6a2c3338a8facd8d12e49b4ca82e8071ec8674a1ec4ba3a
-updated: 2016-06-15T11:14:13.332489872-07:00
+hash: d21f17e2078ce20f063091179f7688d67568e1508519fcfb7d9a39a65865eb0b
+updated: 2016-06-15T14:05:51.080462474-07:00
 imports:
 - name: github.com/aws/aws-sdk-go
   version: 5c0313fff8cd85670ae4fc996254dd7d66cb4cf7
@@ -58,9 +58,11 @@ imports:
 - name: github.com/vaughan0/go-ini
   version: a98ad7ee00ec53921f08832bc06ecf7fd600e6a1
 - name: github.com/vishvananda/netlink
-  version: a57a12c1b1d8aa37cf2f9da4fa7657b1f1fed88c
+  version: 293adec04146bfb07bf34f75921ebc9c1ee2cd74
   subpackages:
   - nl
+- name: github.com/vishvananda/netns
+  version: 8ba1072b58e0c2a240eb5f6120165c7776c3e7b8
 - name: golang.org/x/net
   version: b6d7b1396ec874c3b00f6c84cd4301a17c56c8ed
   subpackages:

glide.yaml

@@ -48,7 +48,7 @@ import:
 - package: github.com/vaughan0/go-ini
   version: a98ad7ee00ec53921f08832bc06ecf7fd600e6a1
 - package: github.com/vishvananda/netlink
-  version: a57a12c1b1d8aa37cf2f9da4fa7657b1f1fed88c
+  version: 293adec04146bfb07bf34f75921ebc9c1ee2cd74
   subpackages:
   - nl
 - package: golang.org/x/net

vendor/github.com/vishvananda/netlink/.travis.yml

@@ -1,3 +1,8 @@
 language: go
+before_script:
+  # make sure we keep path in tact when we sudo
+  - sudo sed -i -e 's/^Defaults\tsecure_path.*$//' /etc/sudoers
+  # modprobe ip_gre or else the first gre device can't be deleted
+  - sudo modprobe ip_gre
 install:
-      - go get github.com/vishvananda/netns
+  - go get github.com/vishvananda/netns

vendor/github.com/vishvananda/netlink/Makefile

@@ -11,14 +11,14 @@ goroot = $(addprefix ../../../,$(1))
 unroot = $(subst ../../../,,$(1))
 fmt = $(addprefix fmt-,$(1))
 
-all: fmt
+all: test
 
 $(call goroot,$(DEPS)):
 	go get $(call unroot,$@)
 
 .PHONY: $(call testdirs,$(DIRS))
 $(call testdirs,$(DIRS)):
-	sudo -E go test -v github.com/vishvananda/netlink/$@
+	sudo -E go test -test.parallel 4 -timeout 60s -v github.com/vishvananda/netlink/$@
 
 $(call fmt,$(call testdirs,$(DIRS))):
 	! gofmt -l $(subst fmt-,,$@)/*.go | grep ''

vendor/github.com/vishvananda/netlink/README.md

@@ -8,7 +8,7 @@ the kernel. It can be used to add and remove interfaces, set ip addresses
 and routes, and configure ipsec. Netlink communication requires elevated
 privileges, so in most cases this code needs to be run as root. Since
 low-level netlink messages are inscrutable at best, the library attempts
-to provide an api that is loosely modeled on the CLI provied by iproute2.
+to provide an api that is loosely modeled on the CLI provided by iproute2.
 Actions like `ip link add` will be accomplished via a similarly named
 function like AddLink(). This library began its life as a fork of the
 netlink functionality in

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

@@ -11,6 +11,8 @@ import (
 type Addr struct {
 	*net.IPNet
 	Label string
+	Flags int
+	Scope int
 }
 
 // String returns $ip/$netmask $label

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

@@ -2,6 +2,7 @@ package netlink
 
 import (
 	"fmt"
+	"log"
 	"net"
 	"strings"
 	"syscall"
@@ -9,32 +10,47 @@ import (
 	"github.com/vishvananda/netlink/nl"
 )
 
+// IFA_FLAGS is a u32 attribute.
+const IFA_FLAGS = 0x8
+
 // AddrAdd will add an IP address to a link device.
 // Equivalent to: `ip addr add $addr dev $link`
 func AddrAdd(link Link, addr *Addr) error {
+	return pkgHandle.AddrAdd(link, addr)
+}
 
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWADDR, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
-	return addrHandle(link, addr, req)
+// AddrAdd will add an IP address to a link device.
+// Equivalent to: `ip addr add $addr dev $link`
+func (h *Handle) AddrAdd(link Link, addr *Addr) error {
+	req := h.newNetlinkRequest(syscall.RTM_NEWADDR, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	return h.addrHandle(link, addr, req)
 }
 
 // AddrDel will delete an IP address from a link device.
 // Equivalent to: `ip addr del $addr dev $link`
 func AddrDel(link Link, addr *Addr) error {
-	req := nl.NewNetlinkRequest(syscall.RTM_DELADDR, syscall.NLM_F_ACK)
-	return addrHandle(link, addr, req)
+	return pkgHandle.AddrDel(link, addr)
+}
+
+// AddrDel will delete an IP address from a link device.
+// Equivalent to: `ip addr del $addr dev $link`
+func (h *Handle) AddrDel(link Link, addr *Addr) error {
+	req := h.newNetlinkRequest(syscall.RTM_DELADDR, syscall.NLM_F_ACK)
+	return h.addrHandle(link, addr, req)
 }
 
-func addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error {
+func (h *Handle) addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error {
 	base := link.Attrs()
 	if addr.Label != "" && !strings.HasPrefix(addr.Label, base.Name) {
 		return fmt.Errorf("label must begin with interface name")
 	}
-	ensureIndex(base)
+	h.ensureIndex(base)
 
 	family := nl.GetIPFamily(addr.IP)
 
 	msg := nl.NewIfAddrmsg(family)
 	msg.Index = uint32(base.Index)
+	msg.Scope = uint8(addr.Scope)
 	prefixlen, _ := addr.Mask.Size()
 	msg.Prefixlen = uint8(prefixlen)
 	req.AddData(msg)
@@ -52,6 +68,17 @@ func addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error {
 	addressData := nl.NewRtAttr(syscall.IFA_ADDRESS, addrData)
 	req.AddData(addressData)
 
+	if addr.Flags != 0 {
+		if addr.Flags <= 0xff {
+			msg.IfAddrmsg.Flags = uint8(addr.Flags)
+		} else {
+			b := make([]byte, 4)
+			native.PutUint32(b, uint32(addr.Flags))
+			flagsData := nl.NewRtAttr(IFA_FLAGS, b)
+			req.AddData(flagsData)
+		}
+	}
+
 	if addr.Label != "" {
 		labelData := nl.NewRtAttr(syscall.IFA_LABEL, nl.ZeroTerminated(addr.Label))
 		req.AddData(labelData)
@@ -65,7 +92,14 @@ func addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error {
 // Equivalent to: `ip addr show`.
 // The list can be filtered by link and ip family.
 func AddrList(link Link, family int) ([]Addr, error) {
-	req := nl.NewNetlinkRequest(syscall.RTM_GETADDR, syscall.NLM_F_DUMP)
+	return pkgHandle.AddrList(link, family)
+}
+
+// AddrList gets a list of IP addresses in the system.
+// Equivalent to: `ip addr show`.
+// The list can be filtered by link and ip family.
+func (h *Handle) AddrList(link Link, family int) ([]Addr, error) {
+	req := h.newNetlinkRequest(syscall.RTM_GETADDR, syscall.NLM_F_DUMP)
 	msg := nl.NewIfInfomsg(family)
 	req.AddData(msg)
 
@@ -74,55 +108,125 @@ func AddrList(link Link, family int) ([]Addr, error) {
 		return nil, err
 	}
 
-	index := 0
+	indexFilter := 0
 	if link != nil {
 		base := link.Attrs()
-		ensureIndex(base)
-		index = base.Index
+		h.ensureIndex(base)
+		indexFilter = base.Index
 	}
 
 	var res []Addr
 	for _, m := range msgs {
-		msg := nl.DeserializeIfAddrmsg(m)
+		addr, msgFamily, ifindex, err := parseAddr(m)
+		if err != nil {
+			return res, err
+		}
 
-		if link != nil && msg.Index != uint32(index) {
+		if link != nil && ifindex != indexFilter {
 			// Ignore messages from other interfaces
 			continue
 		}
 
-		attrs, err := nl.ParseRouteAttr(m[msg.Len():])
-		if err != nil {
-			return nil, err
+		if family != FAMILY_ALL && msgFamily != family {
+			continue
 		}
 
-		var local, dst *net.IPNet
-		var addr Addr
-		for _, attr := range attrs {
-			switch attr.Attr.Type {
-			case syscall.IFA_ADDRESS:
-				dst = &net.IPNet{
-					IP:   attr.Value,
-					Mask: net.CIDRMask(int(msg.Prefixlen), 8*len(attr.Value)),
-				}
-			case syscall.IFA_LOCAL:
-				local = &net.IPNet{
-					IP:   attr.Value,
-					Mask: net.CIDRMask(int(msg.Prefixlen), 8*len(attr.Value)),
-				}
-			case syscall.IFA_LABEL:
-				addr.Label = string(attr.Value[:len(attr.Value)-1])
+		res = append(res, addr)
+	}
+
+	return res, nil
+}
+
+func parseAddr(m []byte) (addr Addr, family, index int, err error) {
+	msg := nl.DeserializeIfAddrmsg(m)
+
+	family = -1
+	index = -1
+
+	attrs, err1 := nl.ParseRouteAttr(m[msg.Len():])
+	if err1 != nil {
+		err = err1
+		return
+	}
+
+	family = int(msg.Family)
+	index = int(msg.Index)
+
+	var local, dst *net.IPNet
+	for _, attr := range attrs {
+		switch attr.Attr.Type {
+		case syscall.IFA_ADDRESS:
+			dst = &net.IPNet{
+				IP:   attr.Value,
+				Mask: net.CIDRMask(int(msg.Prefixlen), 8*len(attr.Value)),
+			}
+		case syscall.IFA_LOCAL:
+			local = &net.IPNet{
+				IP:   attr.Value,
+				Mask: net.CIDRMask(int(msg.Prefixlen), 8*len(attr.Value)),
 			}
+		case syscall.IFA_LABEL:
+			addr.Label = string(attr.Value[:len(attr.Value)-1])
+		case IFA_FLAGS:
+			addr.Flags = int(native.Uint32(attr.Value[0:4]))
 		}
+	}
 
-		// IFA_LOCAL should be there but if not, fall back to IFA_ADDRESS
-		if local != nil {
-			addr.IPNet = local
-		} else {
-			addr.IPNet = dst
-		}
+	// IFA_LOCAL should be there but if not, fall back to IFA_ADDRESS
+	if local != nil {
+		addr.IPNet = local
+	} else {
+		addr.IPNet = dst
+	}
+	addr.Scope = int(msg.Scope)
 
-		res = append(res, addr)
+	return
+}
+
+type AddrUpdate struct {
+	LinkAddress net.IPNet
+	LinkIndex   int
+	NewAddr     bool // true=added false=deleted
+}
+
+// AddrSubscribe takes a chan down which notifications will be sent
+// when addresses change.  Close the 'done' chan to stop subscription.
+func AddrSubscribe(ch chan<- AddrUpdate, done <-chan struct{}) error {
+	s, err := nl.Subscribe(syscall.NETLINK_ROUTE, syscall.RTNLGRP_IPV4_IFADDR, syscall.RTNLGRP_IPV6_IFADDR)
+	if err != nil {
+		return err
 	}
+	if done != nil {
+		go func() {
+			<-done
+			s.Close()
+		}()
+	}
+	go func() {
+		defer close(ch)
+		for {
+			msgs, err := s.Receive()
+			if err != nil {
+				log.Printf("netlink.AddrSubscribe: Receive() error: %v", err)
+				return
+			}
+			for _, m := range msgs {
+				msgType := m.Header.Type
+				if msgType != syscall.RTM_NEWADDR && msgType != syscall.RTM_DELADDR {
+					log.Printf("netlink.AddrSubscribe: bad message type: %d", msgType)
+					continue
+				}
 
-	return res, nil
+				addr, _, ifindex, err := parseAddr(m.Data)
+				if err != nil {
+					log.Printf("netlink.AddrSubscribe: could not parse address: %v", err)
+					continue
+				}
+
+				ch <- AddrUpdate{LinkAddress: *addr.IPNet, LinkIndex: ifindex, NewAddr: msgType == syscall.RTM_NEWADDR}
+			}
+		}
+	}()
+
+	return nil
 }

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

@@ -1,10 +1,44 @@
 package netlink
 
 import (
+	"net"
+	"os"
+	"syscall"
 	"testing"
 )
 
-func TestAddrAddDel(t *testing.T) {
+func TestAddr(t *testing.T) {
+	if os.Getenv("TRAVIS_BUILD_DIR") != "" {
+		t.Skipf("Fails in travis with: addr_test.go:68: Address flags not set properly, got=0, expected=128")
+	}
+	// TODO: IFA_F_PERMANENT does not seem to be set by default on older kernels?
+	var address = &net.IPNet{IP: net.IPv4(127, 0, 0, 2), Mask: net.CIDRMask(24, 32)}
+	var addrTests = []struct {
+		addr     *Addr
+		expected *Addr
+	}{
+		{
+			&Addr{IPNet: address},
+			&Addr{IPNet: address, Label: "lo", Scope: syscall.RT_SCOPE_UNIVERSE, Flags: syscall.IFA_F_PERMANENT},
+		},
+		{
+			&Addr{IPNet: address, Label: "local"},
+			&Addr{IPNet: address, Label: "local", Scope: syscall.RT_SCOPE_UNIVERSE, Flags: syscall.IFA_F_PERMANENT},
+		},
+		{
+			&Addr{IPNet: address, Flags: syscall.IFA_F_OPTIMISTIC},
+			&Addr{IPNet: address, Label: "lo", Flags: syscall.IFA_F_OPTIMISTIC | syscall.IFA_F_PERMANENT, Scope: syscall.RT_SCOPE_UNIVERSE},
+		},
+		{
+			&Addr{IPNet: address, Flags: syscall.IFA_F_OPTIMISTIC | syscall.IFA_F_DADFAILED},
+			&Addr{IPNet: address, Label: "lo", Flags: syscall.IFA_F_OPTIMISTIC | syscall.IFA_F_DADFAILED | syscall.IFA_F_PERMANENT, Scope: syscall.RT_SCOPE_UNIVERSE},
+		},
+		{
+			&Addr{IPNet: address, Scope: syscall.RT_SCOPE_NOWHERE},
+			&Addr{IPNet: address, Label: "lo", Flags: syscall.IFA_F_PERMANENT, Scope: syscall.RT_SCOPE_NOWHERE},
+		},
+	}
+
 	tearDown := setUpNetlinkTest(t)
 	defer tearDown()
 
@@ -13,33 +47,66 @@ func TestAddrAddDel(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	addr, err := ParseAddr("127.1.1.1/24 local")
-	if err != nil {
-		t.Fatal(err)
-	}
+	for _, tt := range addrTests {
+		if err = AddrAdd(link, tt.addr); err != nil {
+			t.Fatal(err)
+		}
 
-	if err = AddrAdd(link, addr); err != nil {
-		t.Fatal(err)
-	}
+		addrs, err := AddrList(link, FAMILY_ALL)
+		if err != nil {
+			t.Fatal(err)
+		}
 
-	addrs, err := AddrList(link, FAMILY_ALL)
-	if err != nil {
-		t.Fatal(err)
-	}
+		if len(addrs) != 1 {
+			t.Fatal("Address not added properly")
+		}
 
-	if len(addrs) != 1 || !addr.Equal(addrs[0]) || addrs[0].Label != addr.Label {
-		t.Fatal("Address not added properly")
-	}
+		if !addrs[0].Equal(*tt.expected) {
+			t.Fatalf("Address ip no set properly, got=%s, expected=%s", addrs[0], tt.expected)
+		}
 
-	if err = AddrDel(link, addr); err != nil {
-		t.Fatal(err)
-	}
-	addrs, err = AddrList(link, FAMILY_ALL)
-	if err != nil {
-		t.Fatal(err)
-	}
+		if addrs[0].Label != tt.expected.Label {
+			t.Fatalf("Address label not set properly, got=%s, expected=%s", addrs[0].Label, tt.expected.Label)
+		}
+
+		if addrs[0].Flags != tt.expected.Flags {
+			t.Fatalf("Address flags not set properly, got=%d, expected=%d", addrs[0].Flags, tt.expected.Flags)
+		}
+
+		if addrs[0].Scope != tt.expected.Scope {
+			t.Fatalf("Address scope not set properly, got=%d, expected=%d", addrs[0].Scope, tt.expected.Scope)
+		}
+
+		// Pass FAMILY_V4, we should get the same results as FAMILY_ALL
+		addrs, err = AddrList(link, FAMILY_V4)
+		if err != nil {
+			t.Fatal(err)
+		}
+		if len(addrs) != 1 {
+			t.Fatal("Address not added properly")
+		}
+
+		// Pass a wrong family number, we should get nil list
+		addrs, err = AddrList(link, 0x8)
+		if err != nil {
+			t.Fatal(err)
+		}
+
+		if len(addrs) != 0 {
+			t.Fatal("Address not expected")
+		}
+
+		if err = AddrDel(link, tt.addr); err != nil {
+			t.Fatal(err)
+		}
+
+		addrs, err = AddrList(link, FAMILY_ALL)
+		if err != nil {
+			t.Fatal(err)
+		}
 
-	if len(addrs) != 0 {
-		t.Fatal("Address not removed properly")
+		if len(addrs) != 0 {
+			t.Fatal("Address not removed properly")
+		}
 	}
 }

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

@@ -0,0 +1,60 @@
+package netlink
+
+/*
+#include <asm/types.h>
+#include <asm/unistd.h>
+#include <errno.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <unistd.h>
+
+static int load_simple_bpf(int prog_type) {
+#ifdef __NR_bpf
+	// { return 1; }
+	__u64 __attribute__((aligned(8))) insns[] = {
+		0x00000001000000b7ull,
+		0x0000000000000095ull,
+	};
+	__u8 __attribute__((aligned(8))) license[] = "ASL2";
+	// Copied from a header file since libc is notoriously slow to update.
+	// The call will succeed or fail and that will be our indication on
+	// whether or not it is supported.
+	struct {
+		__u32 prog_type;
+		__u32 insn_cnt;
+		__u64 insns;
+		__u64 license;
+		__u32 log_level;
+		__u32 log_size;
+		__u64 log_buf;
+		__u32 kern_version;
+	} __attribute__((aligned(8))) attr = {
+		.prog_type = prog_type,
+		.insn_cnt = 2,
+		.insns = (uintptr_t)&insns,
+		.license = (uintptr_t)&license,
+	};
+	return syscall(__NR_bpf, 5, &attr, sizeof(attr));
+#else
+	errno = EINVAL;
+	return -1;
+#endif
+}
+*/
+import "C"
+
+type BpfProgType C.int
+
+const (
+	BPF_PROG_TYPE_UNSPEC BpfProgType = iota
+	BPF_PROG_TYPE_SOCKET_FILTER
+	BPF_PROG_TYPE_KPROBE
+	BPF_PROG_TYPE_SCHED_CLS
+	BPF_PROG_TYPE_SCHED_ACT
+)
+
+// loadSimpleBpf loads a trivial bpf program for testing purposes
+func loadSimpleBpf(progType BpfProgType) (int, error) {
+	fd, err := C.load_simple_bpf(C.int(progType))
+	return int(fd), err
+}

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

@@ -9,7 +9,7 @@ type Class interface {
 	Type() string
 }
 
-// Class represents a netlink class. A filter is associated with a link,
+// ClassAttrs represents a netlink class. A filter is associated with a link,
 // has a handle and a parent. The root filter of a device should have a
 // parent == HANDLE_ROOT.
 type ClassAttrs struct {
@@ -20,7 +20,7 @@ type ClassAttrs struct {
 }
 
 func (q ClassAttrs) String() string {
-	return fmt.Sprintf("{LinkIndex: %d, Handle: %s, Parent: %s, Leaf: %s}", q.LinkIndex, HandleStr(q.Handle), HandleStr(q.Parent), q.Leaf)
+	return fmt.Sprintf("{LinkIndex: %d, Handle: %s, Parent: %s, Leaf: %d}", q.LinkIndex, HandleStr(q.Handle), HandleStr(q.Parent), q.Leaf)
 }
 
 type HtbClassAttrs struct {
@@ -38,7 +38,7 @@ func (q HtbClassAttrs) String() string {
 	return fmt.Sprintf("{Rate: %d, Ceil: %d, Buffer: %d, Cbuffer: %d}", q.Rate, q.Ceil, q.Buffer, q.Cbuffer)
 }
 
-// Htb class
+// HtbClass represents an Htb class
 type HtbClass struct {
 	ClassAttrs
 	Rate    uint64
@@ -50,47 +50,15 @@ type HtbClass struct {
 	Prio    uint32
 }
 
-func NewHtbClass(attrs ClassAttrs, cattrs HtbClassAttrs) *HtbClass {
-	mtu := 1600
-	rate := cattrs.Rate / 8
-	ceil := cattrs.Ceil / 8
-	buffer := cattrs.Buffer
-	cbuffer := cattrs.Cbuffer
-	if ceil == 0 {
-		ceil = rate
-	}
-
-	if buffer == 0 {
-		buffer = uint32(float64(rate)/Hz() + float64(mtu))
-	}
-	buffer = uint32(Xmittime(rate, buffer))
-
-	if cbuffer == 0 {
-		cbuffer = uint32(float64(ceil)/Hz() + float64(mtu))
-	}
-	cbuffer = uint32(Xmittime(ceil, cbuffer))
-
-	return &HtbClass{
-		ClassAttrs: attrs,
-		Rate:       rate,
-		Ceil:       ceil,
-		Buffer:     buffer,
-		Cbuffer:    cbuffer,
-		Quantum:    10,
-		Level:      0,
-		Prio:       0,
-	}
-}
-
 func (q HtbClass) String() string {
 	return fmt.Sprintf("{Rate: %d, Ceil: %d, Buffer: %d, Cbuffer: %d}", q.Rate, q.Ceil, q.Buffer, q.Cbuffer)
 }
 
-func (class *HtbClass) Attrs() *ClassAttrs {
-	return &class.ClassAttrs
+func (q *HtbClass) Attrs() *ClassAttrs {
+	return &q.ClassAttrs
 }
 
-func (class *HtbClass) Type() string {
+func (q *HtbClass) Type() string {
 	return "htb"
 }
 

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

@@ -1,23 +1,70 @@
 package netlink
 
 import (
+	"errors"
 	"syscall"
 
 	"github.com/vishvananda/netlink/nl"
 )
 
+// NOTE: function is in here because it uses other linux functions
+func NewHtbClass(attrs ClassAttrs, cattrs HtbClassAttrs) *HtbClass {
+	mtu := 1600
+	rate := cattrs.Rate / 8
+	ceil := cattrs.Ceil / 8
+	buffer := cattrs.Buffer
+	cbuffer := cattrs.Cbuffer
+
+	if ceil == 0 {
+		ceil = rate
+	}
+
+	if buffer == 0 {
+		buffer = uint32(float64(rate)/Hz() + float64(mtu))
+	}
+	buffer = uint32(Xmittime(rate, buffer))
+
+	if cbuffer == 0 {
+		cbuffer = uint32(float64(ceil)/Hz() + float64(mtu))
+	}
+	cbuffer = uint32(Xmittime(ceil, cbuffer))
+
+	return &HtbClass{
+		ClassAttrs: attrs,
+		Rate:       rate,
+		Ceil:       ceil,
+		Buffer:     buffer,
+		Cbuffer:    cbuffer,
+		Quantum:    10,
+		Level:      0,
+		Prio:       0,
+	}
+}
+
 // ClassDel will delete a class from the system.
 // Equivalent to: `tc class del $class`
 func ClassDel(class Class) error {
-	return classModify(syscall.RTM_DELTCLASS, 0, class)
+	return pkgHandle.ClassDel(class)
+}
+
+// ClassDel will delete a class from the system.
+// Equivalent to: `tc class del $class`
+func (h *Handle) ClassDel(class Class) error {
+	return h.classModify(syscall.RTM_DELTCLASS, 0, class)
 }
 
 // ClassChange will change a class in place
 // Equivalent to: `tc class change $class`
 // The parent and handle MUST NOT be changed.
-
 func ClassChange(class Class) error {
-	return classModify(syscall.RTM_NEWTCLASS, 0, class)
+	return pkgHandle.ClassChange(class)
+}
+
+// ClassChange will change a class in place
+// Equivalent to: `tc class change $class`
+// The parent and handle MUST NOT be changed.
+func (h *Handle) ClassChange(class Class) error {
+	return h.classModify(syscall.RTM_NEWTCLASS, 0, class)
 }
 
 // ClassReplace will replace a class to the system.
@@ -26,21 +73,36 @@ func ClassChange(class Class) error {
 // If a class already exist with this parent/handle pair, the class is changed.
 // If a class does not already exist with this parent/handle, a new class is created.
 func ClassReplace(class Class) error {
-	return classModify(syscall.RTM_NEWTCLASS, syscall.NLM_F_CREATE, class)
+	return pkgHandle.ClassReplace(class)
+}
+
+// ClassReplace will replace a class to the system.
+// quivalent to: `tc class replace $class`
+// The handle MAY be changed.
+// If a class already exist with this parent/handle pair, the class is changed.
+// If a class does not already exist with this parent/handle, a new class is created.
+func (h *Handle) ClassReplace(class Class) error {
+	return h.classModify(syscall.RTM_NEWTCLASS, syscall.NLM_F_CREATE, class)
 }
 
 // ClassAdd will add a class to the system.
 // Equivalent to: `tc class add $class`
 func ClassAdd(class Class) error {
-	return classModify(
+	return pkgHandle.ClassAdd(class)
+}
+
+// ClassAdd will add a class to the system.
+// Equivalent to: `tc class add $class`
+func (h *Handle) ClassAdd(class Class) error {
+	return h.classModify(
 		syscall.RTM_NEWTCLASS,
 		syscall.NLM_F_CREATE|syscall.NLM_F_EXCL,
 		class,
 	)
 }
 
-func classModify(cmd, flags int, class Class) error {
-	req := nl.NewNetlinkRequest(cmd, flags|syscall.NLM_F_ACK)
+func (h *Handle) classModify(cmd, flags int, class Class) error {
+	req := h.newNetlinkRequest(cmd, flags|syscall.NLM_F_ACK)
 	base := class.Attrs()
 	msg := &nl.TcMsg{
 		Family:  nl.FAMILY_ALL,
@@ -65,15 +127,32 @@ func classPayload(req *nl.NetlinkRequest, class Class) error {
 	options := nl.NewRtAttr(nl.TCA_OPTIONS, nil)
 	if htb, ok := class.(*HtbClass); ok {
 		opt := nl.TcHtbCopt{}
-		opt.Rate.Rate = uint32(htb.Rate)
-		opt.Ceil.Rate = uint32(htb.Ceil)
 		opt.Buffer = htb.Buffer
 		opt.Cbuffer = htb.Cbuffer
 		opt.Quantum = htb.Quantum
 		opt.Level = htb.Level
 		opt.Prio = htb.Prio
 		// TODO: Handle Debug properly. For now default to 0
+		/* Calculate {R,C}Tab and set Rate and Ceil */
+		cellLog := -1
+		ccellLog := -1
+		linklayer := nl.LINKLAYER_ETHERNET
+		mtu := 1600
+		var rtab [256]uint32
+		var ctab [256]uint32
+		tcrate := nl.TcRateSpec{Rate: uint32(htb.Rate)}
+		if CalcRtable(&tcrate, rtab, cellLog, uint32(mtu), linklayer) < 0 {
+			return errors.New("HTB: failed to calculate rate table")
+		}
+		opt.Rate = tcrate
+		tcceil := nl.TcRateSpec{Rate: uint32(htb.Ceil)}
+		if CalcRtable(&tcceil, ctab, ccellLog, uint32(mtu), linklayer) < 0 {
+			return errors.New("HTB: failed to calculate ceil rate table")
+		}
+		opt.Ceil = tcceil
 		nl.NewRtAttrChild(options, nl.TCA_HTB_PARMS, opt.Serialize())
+		nl.NewRtAttrChild(options, nl.TCA_HTB_RTAB, SerializeRtab(rtab))
+		nl.NewRtAttrChild(options, nl.TCA_HTB_CTAB, SerializeRtab(ctab))
 	}
 	req.AddData(options)
 	return nil
@@ -83,14 +162,21 @@ func classPayload(req *nl.NetlinkRequest, class Class) error {
 // Equivalent to: `tc class show`.
 // Generally returns nothing if link and parent are not specified.
 func ClassList(link Link, parent uint32) ([]Class, error) {
-	req := nl.NewNetlinkRequest(syscall.RTM_GETTCLASS, syscall.NLM_F_DUMP)
+	return pkgHandle.ClassList(link, parent)
+}
+
+// ClassList gets a list of classes in the system.
+// Equivalent to: `tc class show`.
+// Generally returns nothing if link and parent are not specified.
+func (h *Handle) ClassList(link Link, parent uint32) ([]Class, error) {
+	req := h.newNetlinkRequest(syscall.RTM_GETTCLASS, syscall.NLM_F_DUMP)
 	msg := &nl.TcMsg{
 		Family: nl.FAMILY_ALL,
 		Parent: parent,
 	}
 	if link != nil {
 		base := link.Attrs()
-		ensureIndex(base)
+		h.ensureIndex(base)
 		msg.Ifindex = int32(base.Index)
 	}
 	req.AddData(msg)

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

@@ -269,14 +269,14 @@ func TestHtbClassAddHtbClassChangeDel(t *testing.T) {
 	// For change to work, the handle and parent cannot be changed.
 
 	// First, test it fails if we change the Handle.
-	old_handle := classattrs.Handle
+	oldHandle := classattrs.Handle
 	classattrs.Handle = MakeHandle(0xffff, 3)
 	class = NewHtbClass(classattrs, htbclassattrs)
 	if err := ClassChange(class); err == nil {
 		t.Fatal("ClassChange should not work when using a different handle.")
 	}
 	// It should work with the same handle
-	classattrs.Handle = old_handle
+	classattrs.Handle = oldHandle
 	htbclassattrs.Rate = 4321000
 	class = NewHtbClass(classattrs, htbclassattrs)
 	if err := ClassChange(class); err != nil {

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

@@ -1,17 +1,13 @@
 package netlink
 
-import (
-	"errors"
-	"fmt"
-	"github.com/vishvananda/netlink/nl"
-)
+import "fmt"
 
 type Filter interface {
 	Attrs() *FilterAttrs
 	Type() string
 }
 
-// Filter represents a netlink filter. A filter is associated with a link,
+// FilterAttrs represents a netlink filter. A filter is associated with a link,
 // has a handle and a parent. The root filter of a device should have a
 // parent == HANDLE_ROOT.
 type FilterAttrs struct {
@@ -26,11 +22,170 @@ func (q FilterAttrs) String() string {
 	return fmt.Sprintf("{LinkIndex: %d, Handle: %s, Parent: %s, Priority: %d, Protocol: %d}", q.LinkIndex, HandleStr(q.Handle), HandleStr(q.Parent), q.Priority, q.Protocol)
 }
 
+type TcAct int32
+
+const (
+	TC_ACT_UNSPEC     TcAct = -1
+	TC_ACT_OK         TcAct = 0
+	TC_ACT_RECLASSIFY TcAct = 1
+	TC_ACT_SHOT       TcAct = 2
+	TC_ACT_PIPE       TcAct = 3
+	TC_ACT_STOLEN     TcAct = 4
+	TC_ACT_QUEUED     TcAct = 5
+	TC_ACT_REPEAT     TcAct = 6
+	TC_ACT_REDIRECT   TcAct = 7
+	TC_ACT_JUMP       TcAct = 0x10000000
+)
+
+func (a TcAct) String() string {
+	switch a {
+	case TC_ACT_UNSPEC:
+		return "unspec"
+	case TC_ACT_OK:
+		return "ok"
+	case TC_ACT_RECLASSIFY:
+		return "reclassify"
+	case TC_ACT_SHOT:
+		return "shot"
+	case TC_ACT_PIPE:
+		return "pipe"
+	case TC_ACT_STOLEN:
+		return "stolen"
+	case TC_ACT_QUEUED:
+		return "queued"
+	case TC_ACT_REPEAT:
+		return "repeat"
+	case TC_ACT_REDIRECT:
+		return "redirect"
+	case TC_ACT_JUMP:
+		return "jump"
+	}
+	return fmt.Sprintf("0x%x", a)
+}
+
+type TcPolAct int32
+
+const (
+	TC_POLICE_UNSPEC     TcPolAct = TcPolAct(TC_ACT_UNSPEC)
+	TC_POLICE_OK         TcPolAct = TcPolAct(TC_ACT_OK)
+	TC_POLICE_RECLASSIFY TcPolAct = TcPolAct(TC_ACT_RECLASSIFY)
+	TC_POLICE_SHOT       TcPolAct = TcPolAct(TC_ACT_SHOT)
+	TC_POLICE_PIPE       TcPolAct = TcPolAct(TC_ACT_PIPE)
+)
+
+func (a TcPolAct) String() string {
+	switch a {
+	case TC_POLICE_UNSPEC:
+		return "unspec"
+	case TC_POLICE_OK:
+		return "ok"
+	case TC_POLICE_RECLASSIFY:
+		return "reclassify"
+	case TC_POLICE_SHOT:
+		return "shot"
+	case TC_POLICE_PIPE:
+		return "pipe"
+	}
+	return fmt.Sprintf("0x%x", a)
+}
+
+type ActionAttrs struct {
+	Index   int
+	Capab   int
+	Action  TcAct
+	Refcnt  int
+	Bindcnt int
+}
+
+func (q ActionAttrs) String() string {
+	return fmt.Sprintf("{Index: %d, Capab: %x, Action: %s, Refcnt: %d, Bindcnt: %d}", q.Index, q.Capab, q.Action.String(), q.Refcnt, q.Bindcnt)
+}
+
+// Action represents an action in any supported filter.
+type Action interface {
+	Attrs() *ActionAttrs
+	Type() string
+}
+
+type GenericAction struct {
+	ActionAttrs
+}
+
+func (action *GenericAction) Type() string {
+	return "generic"
+}
+
+func (action *GenericAction) Attrs() *ActionAttrs {
+	return &action.ActionAttrs
+}
+
+type BpfAction struct {
+	ActionAttrs
+	Fd   int
+	Name string
+}
+
+func (action *BpfAction) Type() string {
+	return "bpf"
+}
+
+func (action *BpfAction) Attrs() *ActionAttrs {
+	return &action.ActionAttrs
+}
+
+type MirredAct uint8
+
+func (a MirredAct) String() string {
+	switch a {
+	case TCA_EGRESS_REDIR:
+		return "egress redir"
+	case TCA_EGRESS_MIRROR:
+		return "egress mirror"
+	case TCA_INGRESS_REDIR:
+		return "ingress redir"
+	case TCA_INGRESS_MIRROR:
+		return "ingress mirror"
+	}
+	return "unknown"
+}
+
+const (
+	TCA_EGRESS_REDIR   MirredAct = 1 /* packet redirect to EGRESS*/
+	TCA_EGRESS_MIRROR  MirredAct = 2 /* mirror packet to EGRESS */
+	TCA_INGRESS_REDIR  MirredAct = 3 /* packet redirect to INGRESS*/
+	TCA_INGRESS_MIRROR MirredAct = 4 /* mirror packet to INGRESS */
+)
+
+type MirredAction struct {
+	ActionAttrs
+	MirredAction MirredAct
+	Ifindex      int
+}
+
+func (action *MirredAction) Type() string {
+	return "mirred"
+}
+
+func (action *MirredAction) Attrs() *ActionAttrs {
+	return &action.ActionAttrs
+}
+
+func NewMirredAction(redirIndex int) *MirredAction {
+	return &MirredAction{
+		ActionAttrs: ActionAttrs{
+			Action: TC_ACT_STOLEN,
+		},
+		MirredAction: TCA_EGRESS_REDIR,
+		Ifindex:      redirIndex,
+	}
+}
+
 // U32 filters on many packet related properties
 type U32 struct {
 	FilterAttrs
-	// Currently only supports redirecting to another interface
+	ClassId    uint32
 	RedirIndex int
+	Actions    []Action
 }
 
 func (filter *U32) Attrs() *FilterAttrs {
@@ -52,76 +207,25 @@ type FilterFwAttrs struct {
 	Rate      uint32
 	AvRate    uint32
 	PeakRate  uint32
-	Action    int
+	Action    TcPolAct
 	Overhead  uint16
 	LinkLayer int
 }
 
-// FwFilter filters on firewall marks
-type Fw struct {
+type BpfFilter struct {
 	FilterAttrs
-	ClassId uint32
-	Police  nl.TcPolice
-	InDev   string
-	// TODO Action
-	Mask   uint32
-	AvRate uint32
-	Rtab   [256]uint32
-	Ptab   [256]uint32
-}
-
-func NewFw(attrs FilterAttrs, fattrs FilterFwAttrs) (*Fw, error) {
-	var rtab [256]uint32
-	var ptab [256]uint32
-	rcell_log := -1
-	pcell_log := -1
-	avrate := fattrs.AvRate / 8
-	police := nl.TcPolice{}
-	police.Rate.Rate = fattrs.Rate / 8
-	police.PeakRate.Rate = fattrs.PeakRate / 8
-	buffer := fattrs.Buffer
-	linklayer := nl.LINKLAYER_ETHERNET
-
-	if fattrs.LinkLayer != nl.LINKLAYER_UNSPEC {
-		linklayer = fattrs.LinkLayer
-	}
-
-	police.Action = int32(fattrs.Action)
-	if police.Rate.Rate != 0 {
-		police.Rate.Mpu = fattrs.Mpu
-		police.Rate.Overhead = fattrs.Overhead
-		if CalcRtable(&police.Rate, rtab, rcell_log, fattrs.Mtu, linklayer) < 0 {
-			return nil, errors.New("TBF: failed to calculate rate table.")
-		}
-		police.Burst = uint32(Xmittime(uint64(police.Rate.Rate), uint32(buffer)))
-	}
-	police.Mtu = fattrs.Mtu
-	if police.PeakRate.Rate != 0 {
-		police.PeakRate.Mpu = fattrs.Mpu
-		police.PeakRate.Overhead = fattrs.Overhead
-		if CalcRtable(&police.PeakRate, ptab, pcell_log, fattrs.Mtu, linklayer) < 0 {
-			return nil, errors.New("POLICE: failed to calculate peak rate table.")
-		}
-	}
-
-	return &Fw{
-		FilterAttrs: attrs,
-		ClassId:     fattrs.ClassId,
-		InDev:       fattrs.InDev,
-		Mask:        fattrs.Mask,
-		Police:      police,
-		AvRate:      avrate,
-		Rtab:        rtab,
-		Ptab:        ptab,
-	}, nil
+	ClassId      uint32
+	Fd           int
+	Name         string
+	DirectAction bool
 }
 
-func (filter *Fw) Attrs() *FilterAttrs {
-	return &filter.FilterAttrs
+func (filter *BpfFilter) Type() string {
+	return "bpf"
 }
 
-func (filter *Fw) Type() string {
-	return "fw"
+func (filter *BpfFilter) Attrs() *FilterAttrs {
+	return &filter.FilterAttrs
 }
 
 // GenericFilter filters represent types that are not currently understood

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

@@ -3,16 +3,93 @@ package netlink
 import (
 	"bytes"
 	"encoding/binary"
+	"errors"
 	"fmt"
 	"syscall"
 
 	"github.com/vishvananda/netlink/nl"
 )
 
+// Fw filter filters on firewall marks
+// NOTE: this is in filter_linux because it refers to nl.TcPolice which
+//       is defined in nl/tc_linux.go
+type Fw struct {
+	FilterAttrs
+	ClassId uint32
+	// TODO remove nl type from interface
+	Police nl.TcPolice
+	InDev  string
+	// TODO Action
+	Mask   uint32
+	AvRate uint32
+	Rtab   [256]uint32
+	Ptab   [256]uint32
+}
+
+func NewFw(attrs FilterAttrs, fattrs FilterFwAttrs) (*Fw, error) {
+	var rtab [256]uint32
+	var ptab [256]uint32
+	rcellLog := -1
+	pcellLog := -1
+	avrate := fattrs.AvRate / 8
+	police := nl.TcPolice{}
+	police.Rate.Rate = fattrs.Rate / 8
+	police.PeakRate.Rate = fattrs.PeakRate / 8
+	buffer := fattrs.Buffer
+	linklayer := nl.LINKLAYER_ETHERNET
+
+	if fattrs.LinkLayer != nl.LINKLAYER_UNSPEC {
+		linklayer = fattrs.LinkLayer
+	}
+
+	police.Action = int32(fattrs.Action)
+	if police.Rate.Rate != 0 {
+		police.Rate.Mpu = fattrs.Mpu
+		police.Rate.Overhead = fattrs.Overhead
+		if CalcRtable(&police.Rate, rtab, rcellLog, fattrs.Mtu, linklayer) < 0 {
+			return nil, errors.New("TBF: failed to calculate rate table")
+		}
+		police.Burst = uint32(Xmittime(uint64(police.Rate.Rate), uint32(buffer)))
+	}
+	police.Mtu = fattrs.Mtu
+	if police.PeakRate.Rate != 0 {
+		police.PeakRate.Mpu = fattrs.Mpu
+		police.PeakRate.Overhead = fattrs.Overhead
+		if CalcRtable(&police.PeakRate, ptab, pcellLog, fattrs.Mtu, linklayer) < 0 {
+			return nil, errors.New("POLICE: failed to calculate peak rate table")
+		}
+	}
+
+	return &Fw{
+		FilterAttrs: attrs,
+		ClassId:     fattrs.ClassId,
+		InDev:       fattrs.InDev,
+		Mask:        fattrs.Mask,
+		Police:      police,
+		AvRate:      avrate,
+		Rtab:        rtab,
+		Ptab:        ptab,
+	}, nil
+}
+
+func (filter *Fw) Attrs() *FilterAttrs {
+	return &filter.FilterAttrs
+}
+
+func (filter *Fw) Type() string {
+	return "fw"
+}
+
 // FilterDel will delete a filter from the system.
 // Equivalent to: `tc filter del $filter`
 func FilterDel(filter Filter) error {
-	req := nl.NewNetlinkRequest(syscall.RTM_DELTFILTER, syscall.NLM_F_ACK)
+	return pkgHandle.FilterDel(filter)
+}
+
+// FilterDel will delete a filter from the system.
+// Equivalent to: `tc filter del $filter`
+func (h *Handle) FilterDel(filter Filter) error {
+	req := h.newNetlinkRequest(syscall.RTM_DELTFILTER, syscall.NLM_F_ACK)
 	base := filter.Attrs()
 	msg := &nl.TcMsg{
 		Family:  nl.FAMILY_ALL,
@@ -30,8 +107,14 @@ func FilterDel(filter Filter) error {
 // FilterAdd will add a filter to the system.
 // Equivalent to: `tc filter add $filter`
 func FilterAdd(filter Filter) error {
+	return pkgHandle.FilterAdd(filter)
+}
+
+// FilterAdd will add a filter to the system.
+// Equivalent to: `tc filter add $filter`
+func (h *Handle) FilterAdd(filter Filter) error {
 	native = nl.NativeEndian()
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWTFILTER, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	req := h.newNetlinkRequest(syscall.RTM_NEWTFILTER, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
 	base := filter.Attrs()
 	msg := &nl.TcMsg{
 		Family:  nl.FAMILY_ALL,
@@ -52,17 +135,17 @@ func FilterAdd(filter Filter) error {
 		}
 		sel.Keys = append(sel.Keys, nl.TcU32Key{})
 		nl.NewRtAttrChild(options, nl.TCA_U32_SEL, sel.Serialize())
-		actions := nl.NewRtAttrChild(options, nl.TCA_U32_ACT, nil)
-		table := nl.NewRtAttrChild(actions, nl.TCA_ACT_TAB, nil)
-		nl.NewRtAttrChild(table, nl.TCA_KIND, nl.ZeroTerminated("mirred"))
-		// redirect to other interface
-		mir := nl.TcMirred{
-			Action:  nl.TC_ACT_STOLEN,
-			Eaction: nl.TCA_EGRESS_REDIR,
-			Ifindex: uint32(u32.RedirIndex),
+		if u32.ClassId != 0 {
+			nl.NewRtAttrChild(options, nl.TCA_U32_CLASSID, nl.Uint32Attr(u32.ClassId))
+		}
+		actionsAttr := nl.NewRtAttrChild(options, nl.TCA_U32_ACT, nil)
+		// backwards compatibility
+		if u32.RedirIndex != 0 {
+			u32.Actions = append([]Action{NewMirredAction(u32.RedirIndex)}, u32.Actions...)
+		}
+		if err := encodeActions(actionsAttr, u32.Actions); err != nil {
+			return err
 		}
-		aopts := nl.NewRtAttrChild(table, nl.TCA_OPTIONS, nil)
-		nl.NewRtAttrChild(aopts, nl.TCA_MIRRED_PARMS, mir.Serialize())
 	} else if fw, ok := filter.(*Fw); ok {
 		if fw.Mask != 0 {
 			b := make([]byte, 4)
@@ -90,6 +173,21 @@ func FilterAdd(filter Filter) error {
 			native.PutUint32(b, fw.ClassId)
 			nl.NewRtAttrChild(options, nl.TCA_FW_CLASSID, b)
 		}
+	} else if bpf, ok := filter.(*BpfFilter); ok {
+		var bpfFlags uint32
+		if bpf.ClassId != 0 {
+			nl.NewRtAttrChild(options, nl.TCA_BPF_CLASSID, nl.Uint32Attr(bpf.ClassId))
+		}
+		if bpf.Fd >= 0 {
+			nl.NewRtAttrChild(options, nl.TCA_BPF_FD, nl.Uint32Attr((uint32(bpf.Fd))))
+		}
+		if bpf.Name != "" {
+			nl.NewRtAttrChild(options, nl.TCA_BPF_NAME, nl.ZeroTerminated(bpf.Name))
+		}
+		if bpf.DirectAction {
+			bpfFlags |= nl.TCA_BPF_FLAG_ACT_DIRECT
+		}
+		nl.NewRtAttrChild(options, nl.TCA_BPF_FLAGS, nl.Uint32Attr(bpfFlags))
 	}
 
 	req.AddData(options)
@@ -99,16 +197,23 @@ func FilterAdd(filter Filter) error {
 
 // FilterList gets a list of filters in the system.
 // Equivalent to: `tc filter show`.
-// Generally retunrs nothing if link and parent are not specified.
+// Generally returns nothing if link and parent are not specified.
 func FilterList(link Link, parent uint32) ([]Filter, error) {
-	req := nl.NewNetlinkRequest(syscall.RTM_GETTFILTER, syscall.NLM_F_DUMP)
+	return pkgHandle.FilterList(link, parent)
+}
+
+// FilterList gets a list of filters in the system.
+// Equivalent to: `tc filter show`.
+// Generally returns nothing if link and parent are not specified.
+func (h *Handle) FilterList(link Link, parent uint32) ([]Filter, error) {
+	req := h.newNetlinkRequest(syscall.RTM_GETTFILTER, syscall.NLM_F_DUMP)
 	msg := &nl.TcMsg{
 		Family: nl.FAMILY_ALL,
 		Parent: parent,
 	}
 	if link != nil {
 		base := link.Attrs()
-		ensureIndex(base)
+		h.ensureIndex(base)
 		msg.Ifindex = int32(base.Index)
 	}
 	req.AddData(msg)
@@ -147,29 +252,34 @@ func FilterList(link Link, parent uint32) ([]Filter, error) {
 					filter = &U32{}
 				case "fw":
 					filter = &Fw{}
+				case "bpf":
+					filter = &BpfFilter{}
 				default:
 					filter = &GenericFilter{FilterType: filterType}
 				}
 			case nl.TCA_OPTIONS:
+				data, err := nl.ParseRouteAttr(attr.Value)
+				if err != nil {
+					return nil, err
+				}
 				switch filterType {
 				case "u32":
-					data, err := nl.ParseRouteAttr(attr.Value)
-					if err != nil {
-						return nil, err
-					}
 					detailed, err = parseU32Data(filter, data)
 					if err != nil {
 						return nil, err
 					}
 				case "fw":
-					data, err := nl.ParseRouteAttr(attr.Value)
+					detailed, err = parseFwData(filter, data)
 					if err != nil {
 						return nil, err
 					}
-					detailed, err = parseFwData(filter, data)
+				case "bpf":
+					detailed, err = parseBpfData(filter, data)
 					if err != nil {
 						return nil, err
 					}
+				default:
+					detailed = true
 				}
 			}
 		}
@@ -183,6 +293,129 @@ func FilterList(link Link, parent uint32) ([]Filter, error) {
 	return res, nil
 }
 
+func toTcGen(attrs *ActionAttrs, tcgen *nl.TcGen) {
+	tcgen.Index = uint32(attrs.Index)
+	tcgen.Capab = uint32(attrs.Capab)
+	tcgen.Action = int32(attrs.Action)
+	tcgen.Refcnt = int32(attrs.Refcnt)
+	tcgen.Bindcnt = int32(attrs.Bindcnt)
+}
+
+func toAttrs(tcgen *nl.TcGen, attrs *ActionAttrs) {
+	attrs.Index = int(tcgen.Index)
+	attrs.Capab = int(tcgen.Capab)
+	attrs.Action = TcAct(tcgen.Action)
+	attrs.Refcnt = int(tcgen.Refcnt)
+	attrs.Bindcnt = int(tcgen.Bindcnt)
+}
+
+func encodeActions(attr *nl.RtAttr, actions []Action) error {
+	tabIndex := int(nl.TCA_ACT_TAB)
+
+	for _, action := range actions {
+		switch action := action.(type) {
+		default:
+			return fmt.Errorf("unknown action type %s", action.Type())
+		case *MirredAction:
+			table := nl.NewRtAttrChild(attr, tabIndex, nil)
+			tabIndex++
+			nl.NewRtAttrChild(table, nl.TCA_ACT_KIND, nl.ZeroTerminated("mirred"))
+			aopts := nl.NewRtAttrChild(table, nl.TCA_ACT_OPTIONS, nil)
+			mirred := nl.TcMirred{
+				Eaction: int32(action.MirredAction),
+				Ifindex: uint32(action.Ifindex),
+			}
+			toTcGen(action.Attrs(), &mirred.TcGen)
+			nl.NewRtAttrChild(aopts, nl.TCA_MIRRED_PARMS, mirred.Serialize())
+		case *BpfAction:
+			table := nl.NewRtAttrChild(attr, tabIndex, nil)
+			tabIndex++
+			nl.NewRtAttrChild(table, nl.TCA_ACT_KIND, nl.ZeroTerminated("bpf"))
+			aopts := nl.NewRtAttrChild(table, nl.TCA_ACT_OPTIONS, nil)
+			gen := nl.TcGen{}
+			toTcGen(action.Attrs(), &gen)
+			nl.NewRtAttrChild(aopts, nl.TCA_ACT_BPF_PARMS, gen.Serialize())
+			nl.NewRtAttrChild(aopts, nl.TCA_ACT_BPF_FD, nl.Uint32Attr(uint32(action.Fd)))
+			nl.NewRtAttrChild(aopts, nl.TCA_ACT_BPF_NAME, nl.ZeroTerminated(action.Name))
+		case *GenericAction:
+			table := nl.NewRtAttrChild(attr, tabIndex, nil)
+			tabIndex++
+			nl.NewRtAttrChild(table, nl.TCA_ACT_KIND, nl.ZeroTerminated("gact"))
+			aopts := nl.NewRtAttrChild(table, nl.TCA_ACT_OPTIONS, nil)
+			gen := nl.TcGen{}
+			toTcGen(action.Attrs(), &gen)
+			nl.NewRtAttrChild(aopts, nl.TCA_GACT_PARMS, gen.Serialize())
+		}
+	}
+	return nil
+}
+
+func parseActions(tables []syscall.NetlinkRouteAttr) ([]Action, error) {
+	var actions []Action
+	for _, table := range tables {
+		var action Action
+		var actionType string
+		aattrs, err := nl.ParseRouteAttr(table.Value)
+		if err != nil {
+			return nil, err
+		}
+	nextattr:
+		for _, aattr := range aattrs {
+			switch aattr.Attr.Type {
+			case nl.TCA_KIND:
+				actionType = string(aattr.Value[:len(aattr.Value)-1])
+				// only parse if the action is mirred or bpf
+				switch actionType {
+				case "mirred":
+					action = &MirredAction{}
+				case "bpf":
+					action = &BpfAction{}
+				case "gact":
+					action = &GenericAction{}
+				default:
+					break nextattr
+				}
+			case nl.TCA_OPTIONS:
+				adata, err := nl.ParseRouteAttr(aattr.Value)
+				if err != nil {
+					return nil, err
+				}
+				for _, adatum := range adata {
+					switch actionType {
+					case "mirred":
+						switch adatum.Attr.Type {
+						case nl.TCA_MIRRED_PARMS:
+							mirred := *nl.DeserializeTcMirred(adatum.Value)
+							toAttrs(&mirred.TcGen, action.Attrs())
+							action.(*MirredAction).ActionAttrs = ActionAttrs{}
+							action.(*MirredAction).Ifindex = int(mirred.Ifindex)
+							action.(*MirredAction).MirredAction = MirredAct(mirred.Eaction)
+						}
+					case "bpf":
+						switch adatum.Attr.Type {
+						case nl.TCA_ACT_BPF_PARMS:
+							gen := *nl.DeserializeTcGen(adatum.Value)
+							toAttrs(&gen, action.Attrs())
+						case nl.TCA_ACT_BPF_FD:
+							action.(*BpfAction).Fd = int(native.Uint32(adatum.Value[0:4]))
+						case nl.TCA_ACT_BPF_NAME:
+							action.(*BpfAction).Name = string(adatum.Value[:len(adatum.Value)-1])
+						}
+					case "gact":
+						switch adatum.Attr.Type {
+						case nl.TCA_GACT_PARMS:
+							gen := *nl.DeserializeTcGen(adatum.Value)
+							toAttrs(&gen, action.Attrs())
+						}
+					}
+				}
+			}
+		}
+		actions = append(actions, action)
+	}
+	return actions, nil
+}
+
 func parseU32Data(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error) {
 	native = nl.NativeEndian()
 	u32 := filter.(*U32)
@@ -197,34 +430,17 @@ func parseU32Data(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error)
 				return detailed, nil
 			}
 		case nl.TCA_U32_ACT:
-			table, err := nl.ParseRouteAttr(datum.Value)
+			tables, err := nl.ParseRouteAttr(datum.Value)
 			if err != nil {
 				return detailed, err
 			}
-			if len(table) != 1 || table[0].Attr.Type != nl.TCA_ACT_TAB {
-				return detailed, fmt.Errorf("Action table not formed properly")
+			u32.Actions, err = parseActions(tables)
+			if err != nil {
+				return detailed, err
 			}
-			aattrs, err := nl.ParseRouteAttr(table[0].Value)
-			for _, aattr := range aattrs {
-				switch aattr.Attr.Type {
-				case nl.TCA_KIND:
-					actionType := string(aattr.Value[:len(aattr.Value)-1])
-					// only parse if the action is mirred
-					if actionType != "mirred" {
-						return detailed, nil
-					}
-				case nl.TCA_OPTIONS:
-					adata, err := nl.ParseRouteAttr(aattr.Value)
-					if err != nil {
-						return detailed, err
-					}
-					for _, adatum := range adata {
-						switch adatum.Attr.Type {
-						case nl.TCA_MIRRED_PARMS:
-							mir := nl.DeserializeTcMirred(adatum.Value)
-							u32.RedirIndex = int(mir.Ifindex)
-						}
-					}
+			for _, action := range u32.Actions {
+				if action, ok := action.(*MirredAction); ok {
+					u32.RedirIndex = int(action.Ifindex)
 				}
 			}
 		}
@@ -261,6 +477,28 @@ func parseFwData(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error) {
 	return detailed, nil
 }
 
+func parseBpfData(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error) {
+	native = nl.NativeEndian()
+	bpf := filter.(*BpfFilter)
+	detailed := true
+	for _, datum := range data {
+		switch datum.Attr.Type {
+		case nl.TCA_BPF_FD:
+			bpf.Fd = int(native.Uint32(datum.Value[0:4]))
+		case nl.TCA_BPF_NAME:
+			bpf.Name = string(datum.Value[:len(datum.Value)-1])
+		case nl.TCA_BPF_CLASSID:
+			bpf.ClassId = native.Uint32(datum.Value[0:4])
+		case nl.TCA_BPF_FLAGS:
+			flags := native.Uint32(datum.Value[0:4])
+			if (flags & nl.TCA_BPF_FLAG_ACT_DIRECT) != 0 {
+				bpf.DirectAction = true
+			}
+		}
+	}
+	return detailed, nil
+}
+
 func AlignToAtm(size uint) uint {
 	var linksize, cells int
 	cells = int(size / nl.ATM_CELL_PAYLOAD)
@@ -283,27 +521,27 @@ func AdjustSize(sz uint, mpu uint, linklayer int) uint {
 	}
 }
 
-func CalcRtable(rate *nl.TcRateSpec, rtab [256]uint32, cell_log int, mtu uint32, linklayer int) int {
+func CalcRtable(rate *nl.TcRateSpec, rtab [256]uint32, cellLog int, mtu uint32, linklayer int) int {
 	bps := rate.Rate
 	mpu := rate.Mpu
 	var sz uint
 	if mtu == 0 {
 		mtu = 2047
 	}
-	if cell_log < 0 {
-		cell_log = 0
-		for (mtu >> uint(cell_log)) > 255 {
-			cell_log++
+	if cellLog < 0 {
+		cellLog = 0
+		for (mtu >> uint(cellLog)) > 255 {
+			cellLog++
 		}
 	}
 	for i := 0; i < 256; i++ {
-		sz = AdjustSize(uint((i+1)<<uint32(cell_log)), uint(mpu), linklayer)
+		sz = AdjustSize(uint((i+1)<<uint32(cellLog)), uint(mpu), linklayer)
 		rtab[i] = uint32(Xmittime(uint64(bps), uint32(sz)))
 	}
 	rate.CellAlign = -1
-	rate.CellLog = uint8(cell_log)
+	rate.CellLog = uint8(cellLog)
 	rate.Linklayer = uint8(linklayer & nl.TC_LINKLAYER_MASK)
-	return cell_log
+	return cellLog
 }
 
 func DeserializeRtab(b []byte) [256]uint32 {

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

@@ -3,8 +3,6 @@ package netlink
 import (
 	"syscall"
 	"testing"
-
-	"github.com/vishvananda/netlink/nl"
 )
 
 func TestFilterAddDel(t *testing.T) {
@@ -169,7 +167,7 @@ func TestFilterFwAddDel(t *testing.T) {
 		Buffer:   12345,
 		Rate:     1234,
 		PeakRate: 2345,
-		Action:   nl.TC_POLICE_SHOT,
+		Action:   TC_POLICE_SHOT,
 		ClassId:  MakeHandle(0xffff, 2),
 	}
 
@@ -246,3 +244,229 @@ func TestFilterFwAddDel(t *testing.T) {
 		t.Fatal("Failed to remove qdisc")
 	}
 }
+
+func TestFilterU32BpfAddDel(t *testing.T) {
+	tearDown := setUpNetlinkTest(t)
+	defer tearDown()
+	if err := LinkAdd(&Ifb{LinkAttrs{Name: "foo"}}); err != nil {
+		t.Fatal(err)
+	}
+	if err := LinkAdd(&Ifb{LinkAttrs{Name: "bar"}}); err != nil {
+		t.Fatal(err)
+	}
+	link, err := LinkByName("foo")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := LinkSetUp(link); err != nil {
+		t.Fatal(err)
+	}
+	redir, err := LinkByName("bar")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := LinkSetUp(redir); err != nil {
+		t.Fatal(err)
+	}
+	qdisc := &Ingress{
+		QdiscAttrs: QdiscAttrs{
+			LinkIndex: link.Attrs().Index,
+			Handle:    MakeHandle(0xffff, 0),
+			Parent:    HANDLE_INGRESS,
+		},
+	}
+	if err := QdiscAdd(qdisc); err != nil {
+		t.Fatal(err)
+	}
+	qdiscs, err := QdiscList(link)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(qdiscs) != 1 {
+		t.Fatal("Failed to add qdisc")
+	}
+	_, ok := qdiscs[0].(*Ingress)
+	if !ok {
+		t.Fatal("Qdisc is the wrong type")
+	}
+
+	fd, err := loadSimpleBpf(BPF_PROG_TYPE_SCHED_ACT)
+	if err != nil {
+		t.Skipf("Loading bpf program failed: %s", err)
+	}
+	filter := &U32{
+		FilterAttrs: FilterAttrs{
+			LinkIndex: link.Attrs().Index,
+			Parent:    MakeHandle(0xffff, 0),
+			Priority:  1,
+			Protocol:  syscall.ETH_P_ALL,
+		},
+		ClassId: MakeHandle(1, 1),
+		Actions: []Action{
+			&BpfAction{Fd: fd, Name: "simple"},
+			&MirredAction{
+				ActionAttrs: ActionAttrs{
+					Action: TC_ACT_STOLEN,
+				},
+				MirredAction: TCA_EGRESS_REDIR,
+				Ifindex:      redir.Attrs().Index,
+			},
+		},
+	}
+
+	if err := FilterAdd(filter); err != nil {
+		t.Fatal(err)
+	}
+
+	filters, err := FilterList(link, MakeHandle(0xffff, 0))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(filters) != 1 {
+		t.Fatal("Failed to add filter")
+	}
+	u32, ok := filters[0].(*U32)
+	if !ok {
+		t.Fatal("Filter is the wrong type")
+	}
+
+	if len(u32.Actions) != 2 {
+		t.Fatalf("Too few Actions in filter")
+	}
+	bpfAction, ok := u32.Actions[0].(*BpfAction)
+	if !ok {
+		t.Fatal("Action[0] is the wrong type")
+	}
+	if bpfAction.Fd != fd {
+		t.Fatal("Action Fd does not match")
+	}
+	if _, ok := u32.Actions[1].(*MirredAction); !ok {
+		t.Fatal("Action[1] is the wrong type")
+	}
+
+	if err := FilterDel(filter); err != nil {
+		t.Fatal(err)
+	}
+	filters, err = FilterList(link, MakeHandle(0xffff, 0))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(filters) != 0 {
+		t.Fatal("Failed to remove filter")
+	}
+
+	if err := QdiscDel(qdisc); err != nil {
+		t.Fatal(err)
+	}
+	qdiscs, err = QdiscList(link)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(qdiscs) != 0 {
+		t.Fatal("Failed to remove qdisc")
+	}
+}
+
+func TestFilterClsActBpfAddDel(t *testing.T) {
+	tearDown := setUpNetlinkTest(t)
+	defer tearDown()
+	if err := LinkAdd(&Ifb{LinkAttrs{Name: "foo"}}); err != nil {
+		t.Fatal(err)
+	}
+	link, err := LinkByName("foo")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := LinkSetUp(link); err != nil {
+		t.Fatal(err)
+	}
+	attrs := QdiscAttrs{
+		LinkIndex: link.Attrs().Index,
+		Handle:    MakeHandle(0xffff, 0),
+		Parent:    HANDLE_CLSACT,
+	}
+	qdisc := &GenericQdisc{
+		QdiscAttrs: attrs,
+		QdiscType:  "clsact",
+	}
+	// This feature was added in kernel 4.5
+	if err := QdiscAdd(qdisc); err != nil {
+		t.Skipf("Failed adding clsact qdisc, unsupported kernel")
+	}
+	qdiscs, err := QdiscList(link)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(qdiscs) != 1 {
+		t.Fatal("Failed to add qdisc")
+	}
+	if q, ok := qdiscs[0].(*GenericQdisc); !ok || q.Type() != "clsact" {
+		t.Fatal("qdisc is the wrong type")
+	}
+
+	filterattrs := FilterAttrs{
+		LinkIndex: link.Attrs().Index,
+		Parent:    HANDLE_MIN_EGRESS,
+		Handle:    MakeHandle(0, 1),
+		Protocol:  syscall.ETH_P_ALL,
+		Priority:  1,
+	}
+	fd, err := loadSimpleBpf(BPF_PROG_TYPE_SCHED_CLS)
+	if err != nil {
+		t.Skipf("Loading bpf program failed: %s", err)
+	}
+	filter := &BpfFilter{
+		FilterAttrs:  filterattrs,
+		Fd:           fd,
+		Name:         "simple",
+		DirectAction: true,
+	}
+	if filter.Fd < 0 {
+		t.Skipf("Failed to load bpf program")
+	}
+
+	if err := FilterAdd(filter); err != nil {
+		t.Fatal(err)
+	}
+
+	filters, err := FilterList(link, HANDLE_MIN_EGRESS)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(filters) != 1 {
+		t.Fatal("Failed to add filter")
+	}
+	bpf, ok := filters[0].(*BpfFilter)
+	if !ok {
+		t.Fatal("Filter is the wrong type")
+	}
+
+	if bpf.Fd != filter.Fd {
+		t.Fatal("Filter Fd does not match")
+	}
+	if bpf.DirectAction != filter.DirectAction {
+		t.Fatal("Filter DirectAction does not match")
+	}
+
+	if err := FilterDel(filter); err != nil {
+		t.Fatal(err)
+	}
+	filters, err = FilterList(link, HANDLE_MIN_EGRESS)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(filters) != 0 {
+		t.Fatal("Failed to remove filter")
+	}
+
+	if err := QdiscDel(qdisc); err != nil {
+		t.Fatal(err)
+	}
+	qdiscs, err = QdiscList(link)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(qdiscs) != 0 {
+		t.Fatal("Failed to remove qdisc")
+	}
+}

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

@@ -0,0 +1,86 @@
+package netlink
+
+import (
+	"sync/atomic"
+	"syscall"
+
+	"github.com/vishvananda/netlink/nl"
+	"github.com/vishvananda/netns"
+)
+
+// Empty handle used by the netlink package methods
+var pkgHandle = &Handle{}
+
+// Handle is an handle for the netlink requests
+// on a specific network namespace. All the requests
+// share the same netlink socket, which gets released
+// when the handle is deleted.
+type Handle struct {
+	seq          uint32
+	routeSocket  *nl.NetlinkSocket
+	xfrmSocket   *nl.NetlinkSocket
+	lookupByDump bool
+}
+
+// NewHandle returns a netlink handle on the current network namespace.
+func NewHandle() (*Handle, error) {
+	return newHandle(netns.None(), netns.None())
+}
+
+// NewHandle returns a netlink handle on the network namespace
+// specified by ns. If ns=netns.None(), current network namespace
+// will be assumed
+func NewHandleAt(ns netns.NsHandle) (*Handle, error) {
+	return newHandle(ns, netns.None())
+}
+
+// NewHandleAtFrom works as NewHandle but allows client to specify the
+// new and the origin netns Handle.
+func NewHandleAtFrom(newNs, curNs netns.NsHandle) (*Handle, error) {
+	return newHandle(newNs, curNs)
+}
+
+func newHandle(newNs, curNs netns.NsHandle) (*Handle, error) {
+	var (
+		err     error
+		rSocket *nl.NetlinkSocket
+		xSocket *nl.NetlinkSocket
+	)
+	rSocket, err = nl.GetNetlinkSocketAt(newNs, curNs, syscall.NETLINK_ROUTE)
+	if err != nil {
+		return nil, err
+	}
+	xSocket, err = nl.GetNetlinkSocketAt(newNs, curNs, syscall.NETLINK_XFRM)
+	if err != nil {
+		return nil, err
+	}
+	return &Handle{routeSocket: rSocket, xfrmSocket: xSocket}, nil
+}
+
+// Delete releases the resources allocated to this handle
+func (h *Handle) Delete() {
+	if h.routeSocket != nil {
+		h.routeSocket.Close()
+	}
+	if h.xfrmSocket != nil {
+		h.xfrmSocket.Close()
+	}
+	h.routeSocket, h.xfrmSocket = nil, nil
+}
+
+func (h *Handle) newNetlinkRequest(proto, flags int) *nl.NetlinkRequest {
+	// Do this so that package API still use nl package variable nextSeqNr
+	if h.routeSocket == nil {
+		return nl.NewNetlinkRequest(proto, flags)
+	}
+	return &nl.NetlinkRequest{
+		NlMsghdr: syscall.NlMsghdr{
+			Len:   uint32(syscall.SizeofNlMsghdr),
+			Type:  uint16(proto),
+			Flags: syscall.NLM_F_REQUEST | uint16(flags),
+			Seq:   atomic.AddUint32(&h.seq, 1),
+		},
+		RouteSocket: h.routeSocket,
+		XfmrSocket:  h.xfrmSocket,
+	}
+}

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

@@ -0,0 +1,286 @@
+package netlink
+
+import (
+	"crypto/rand"
+	"encoding/hex"
+	"fmt"
+	"io"
+	"net"
+	"sync"
+	"sync/atomic"
+	"testing"
+
+	"github.com/vishvananda/netns"
+)
+
+func TestHandleCreateDelete(t *testing.T) {
+	h, err := NewHandle()
+	if err != nil {
+		t.Fatal(err)
+	}
+	if h.routeSocket == nil || h.xfrmSocket == nil {
+		t.Fatalf("Handle socket(s) were not created")
+	}
+
+	h.Delete()
+	if h.routeSocket != nil || h.xfrmSocket != nil {
+		t.Fatalf("Handle socket(s) were not destroyed")
+	}
+}
+
+func TestHandleCreateNetns(t *testing.T) {
+	id := make([]byte, 4)
+	if _, err := io.ReadFull(rand.Reader, id); err != nil {
+		t.Fatal(err)
+	}
+	ifName := "dummy-" + hex.EncodeToString(id)
+
+	// Create an handle on the current netns
+	curNs, err := netns.Get()
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer curNs.Close()
+
+	ch, err := NewHandleAt(curNs)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer ch.Delete()
+
+	// Create an handle on a custom netns
+	newNs, err := netns.New()
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer newNs.Close()
+
+	nh, err := NewHandleAt(newNs)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer nh.Delete()
+
+	// Create an interface using the current handle
+	err = ch.LinkAdd(&Dummy{LinkAttrs{Name: ifName}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	l, err := ch.LinkByName(ifName)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if l.Type() != "dummy" {
+		t.Fatalf("Unexpected link type: %s", l.Type())
+	}
+
+	// Verify the new handle cannot find the interface
+	ll, err := nh.LinkByName(ifName)
+	if err == nil {
+		t.Fatalf("Unexpected link found on netns %s: %v", newNs, ll)
+	}
+
+	// Move the interface to the new netns
+	err = ch.LinkSetNsFd(l, int(newNs))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Verify new netns handle can find the interface while current cannot
+	ll, err = nh.LinkByName(ifName)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if ll.Type() != "dummy" {
+		t.Fatalf("Unexpected link type: %s", ll.Type())
+	}
+	ll, err = ch.LinkByName(ifName)
+	if err == nil {
+		t.Fatalf("Unexpected link found on netns %s: %v", curNs, ll)
+	}
+}
+
+var (
+	iter      = 10
+	numThread = uint32(4)
+	prefix    = "iface"
+	handle1   *Handle
+	handle2   *Handle
+	ns1       netns.NsHandle
+	ns2       netns.NsHandle
+	done      uint32
+	initError error
+	once      sync.Once
+)
+
+func getXfrmState(thread int) *XfrmState {
+	return &XfrmState{
+		Src:   net.IPv4(byte(192), byte(168), 1, byte(1+thread)),
+		Dst:   net.IPv4(byte(192), byte(168), 2, byte(1+thread)),
+		Proto: XFRM_PROTO_AH,
+		Mode:  XFRM_MODE_TUNNEL,
+		Spi:   thread,
+		Auth: &XfrmStateAlgo{
+			Name: "hmac(sha256)",
+			Key:  []byte("abcdefghijklmnopqrstuvwzyzABCDEF"),
+		},
+	}
+}
+
+func getXfrmPolicy(thread int) *XfrmPolicy {
+	return &XfrmPolicy{
+		Src:     &net.IPNet{IP: net.IPv4(byte(10), byte(10), byte(thread), 0), Mask: []byte{255, 255, 255, 0}},
+		Dst:     &net.IPNet{IP: net.IPv4(byte(10), byte(10), byte(thread), 0), Mask: []byte{255, 255, 255, 0}},
+		Proto:   17,
+		DstPort: 1234,
+		SrcPort: 5678,
+		Dir:     XFRM_DIR_OUT,
+		Tmpls: []XfrmPolicyTmpl{
+			{
+				Src:   net.IPv4(byte(192), byte(168), 1, byte(thread)),
+				Dst:   net.IPv4(byte(192), byte(168), 2, byte(thread)),
+				Proto: XFRM_PROTO_ESP,
+				Mode:  XFRM_MODE_TUNNEL,
+			},
+		},
+	}
+}
+func initParallel() {
+	ns1, initError = netns.New()
+	if initError != nil {
+		return
+	}
+	handle1, initError = NewHandleAt(ns1)
+	if initError != nil {
+		return
+	}
+	ns2, initError = netns.New()
+	if initError != nil {
+		return
+	}
+	handle2, initError = NewHandleAt(ns2)
+	if initError != nil {
+		return
+	}
+}
+
+func parallelDone() {
+	atomic.AddUint32(&done, 1)
+	if done == numThread {
+		if ns1.IsOpen() {
+			ns1.Close()
+		}
+		if ns2.IsOpen() {
+			ns2.Close()
+		}
+		if handle1 != nil {
+			handle1.Delete()
+		}
+		if handle2 != nil {
+			handle2.Delete()
+		}
+	}
+}
+
+// Do few route and xfrm operation on the two handles in parallel
+func runParallelTests(t *testing.T, thread int) {
+	defer parallelDone()
+
+	t.Parallel()
+
+	once.Do(initParallel)
+	if initError != nil {
+		t.Fatal(initError)
+	}
+
+	state := getXfrmState(thread)
+	policy := getXfrmPolicy(thread)
+	for i := 0; i < iter; i++ {
+		ifName := fmt.Sprintf("%s_%d_%d", prefix, thread, i)
+		link := &Dummy{LinkAttrs{Name: ifName}}
+		err := handle1.LinkAdd(link)
+		if err != nil {
+			t.Fatal(err)
+		}
+		l, err := handle1.LinkByName(ifName)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle1.LinkSetUp(l)
+		if err != nil {
+			t.Fatal(err)
+		}
+		handle1.LinkSetNsFd(l, int(ns2))
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle1.XfrmStateAdd(state)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle1.XfrmPolicyAdd(policy)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle2.LinkSetDown(l)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle2.XfrmStateAdd(state)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle2.XfrmPolicyAdd(policy)
+		if err != nil {
+			t.Fatal(err)
+		}
+		_, err = handle2.LinkByName(ifName)
+		if err != nil {
+			t.Fatal(err)
+		}
+		handle2.LinkSetNsFd(l, int(ns1))
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle1.LinkSetUp(l)
+		if err != nil {
+			t.Fatal(err)
+		}
+		l, err = handle1.LinkByName(ifName)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle1.XfrmPolicyDel(policy)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle2.XfrmPolicyDel(policy)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle1.XfrmStateDel(state)
+		if err != nil {
+			t.Fatal(err)
+		}
+		err = handle2.XfrmStateDel(state)
+		if err != nil {
+			t.Fatal(err)
+		}
+	}
+}
+
+func TestHandleParallel1(t *testing.T) {
+	runParallelTests(t, 1)
+}
+
+func TestHandleParallel2(t *testing.T) {
+	runParallelTests(t, 2)
+}
+
+func TestHandleParallel3(t *testing.T) {
+	runParallelTests(t, 3)
+}
+
+func TestHandleParallel4(t *testing.T) {
+	runParallelTests(t, 4)
+}

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

@@ -1,8 +1,8 @@
 package netlink
 
 import (
+	"fmt"
 	"net"
-	"syscall"
 )
 
 // Link represents a link device from netlink. Shared link attributes
@@ -29,6 +29,8 @@ type LinkAttrs struct {
 	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
 }
 
 // NewLinkAttrs returns LinkAttrs structure filled with default values
@@ -38,6 +40,35 @@ func NewLinkAttrs() LinkAttrs {
 	}
 }
 
+/*
+Ref: struct rtnl_link_stats {...}
+*/
+type LinkStatistics struct {
+	RxPackets         uint32
+	TxPackets         uint32
+	RxBytes           uint32
+	TxBytes           uint32
+	RxErrors          uint32
+	TxErrors          uint32
+	RxDropped         uint32
+	TxDropped         uint32
+	Multicast         uint32
+	Collisions        uint32
+	RxLengthErrors    uint32
+	RxOverErrors      uint32
+	RxCrcErrors       uint32
+	RxFrameErrors     uint32
+	RxFifoErrors      uint32
+	RxMissedErrors    uint32
+	TxAbortedErrors   uint32
+	TxCarrierErrors   uint32
+	TxFifoErrors      uint32
+	TxHeartbeatErrors uint32
+	TxWindowErrors    uint32
+	RxCompressed      uint32
+	TxCompressed      uint32
+}
+
 // Device links cannot be created via netlink. These links
 // are links created by udev like 'lo' and 'etho0'
 type Device struct {
@@ -141,11 +172,6 @@ func (macvtap Macvtap) Type() string {
 
 type TuntapMode uint16
 
-const (
-	TUNTAP_MODE_TUN TuntapMode = syscall.IFF_TUN
-	TUNTAP_MODE_TAP TuntapMode = syscall.IFF_TAP
-)
-
 // Tuntap links created via /dev/tun/tap, but can be destroyed via netlink
 type Tuntap struct {
 	LinkAttrs
@@ -202,6 +228,7 @@ type Vxlan struct {
 	RSC          bool
 	L2miss       bool
 	L3miss       bool
+	UDPCSum      bool
 	NoAge        bool
 	GBP          bool
 	Age          int
@@ -240,6 +267,316 @@ func (ipvlan *IPVlan) Type() string {
 	return "ipvlan"
 }
 
+// BondMode type
+type BondMode int
+
+func (b BondMode) String() string {
+	s, ok := bondModeToString[b]
+	if !ok {
+		return fmt.Sprintf("BondMode(%d)", b)
+	}
+	return s
+}
+
+// StringToBondMode returns bond mode, or uknonw is the s is invalid.
+func StringToBondMode(s string) BondMode {
+	mode, ok := StringToBondModeMap[s]
+	if !ok {
+		return BOND_MODE_UNKNOWN
+	}
+	return mode
+}
+
+// Possible BondMode
+const (
+	BOND_MODE_802_3AD BondMode = iota
+	BOND_MODE_BALANCE_RR
+	BOND_MODE_ACTIVE_BACKUP
+	BOND_MODE_BALANCE_XOR
+	BOND_MODE_BROADCAST
+	BOND_MODE_BALANCE_TLB
+	BOND_MODE_BALANCE_ALB
+	BOND_MODE_UNKNOWN
+)
+
+var bondModeToString = map[BondMode]string{
+	BOND_MODE_802_3AD:       "802.3ad",
+	BOND_MODE_BALANCE_RR:    "balance-rr",
+	BOND_MODE_ACTIVE_BACKUP: "active-backup",
+	BOND_MODE_BALANCE_XOR:   "balance-xor",
+	BOND_MODE_BROADCAST:     "broadcast",
+	BOND_MODE_BALANCE_TLB:   "balance-tlb",
+	BOND_MODE_BALANCE_ALB:   "balance-alb",
+}
+var StringToBondModeMap = map[string]BondMode{
+	"802.3ad":       BOND_MODE_802_3AD,
+	"balance-rr":    BOND_MODE_BALANCE_RR,
+	"active-backup": BOND_MODE_ACTIVE_BACKUP,
+	"balance-xor":   BOND_MODE_BALANCE_XOR,
+	"broadcast":     BOND_MODE_BROADCAST,
+	"balance-tlb":   BOND_MODE_BALANCE_TLB,
+	"balance-alb":   BOND_MODE_BALANCE_ALB,
+}
+
+// BondArpValidate type
+type BondArpValidate int
+
+// Possible BondArpValidate value
+const (
+	BOND_ARP_VALIDATE_NONE BondArpValidate = iota
+	BOND_ARP_VALIDATE_ACTIVE
+	BOND_ARP_VALIDATE_BACKUP
+	BOND_ARP_VALIDATE_ALL
+)
+
+// BondPrimaryReselect type
+type BondPrimaryReselect int
+
+// Possible BondPrimaryReselect value
+const (
+	BOND_PRIMARY_RESELECT_ALWAYS BondPrimaryReselect = iota
+	BOND_PRIMARY_RESELECT_BETTER
+	BOND_PRIMARY_RESELECT_FAILURE
+)
+
+// BondArpAllTargets type
+type BondArpAllTargets int
+
+// Possible BondArpAllTargets value
+const (
+	BOND_ARP_ALL_TARGETS_ANY BondArpAllTargets = iota
+	BOND_ARP_ALL_TARGETS_ALL
+)
+
+// BondFailOverMac type
+type BondFailOverMac int
+
+// Possible BondFailOverMac value
+const (
+	BOND_FAIL_OVER_MAC_NONE BondFailOverMac = iota
+	BOND_FAIL_OVER_MAC_ACTIVE
+	BOND_FAIL_OVER_MAC_FOLLOW
+)
+
+// BondXmitHashPolicy type
+type BondXmitHashPolicy int
+
+func (b BondXmitHashPolicy) String() string {
+	s, ok := bondXmitHashPolicyToString[b]
+	if !ok {
+		return fmt.Sprintf("XmitHashPolicy(%d)", b)
+	}
+	return s
+}
+
+// StringToBondXmitHashPolicy returns bond lacp arte, or uknonw is the s is invalid.
+func StringToBondXmitHashPolicy(s string) BondXmitHashPolicy {
+	lacp, ok := StringToBondXmitHashPolicyMap[s]
+	if !ok {
+		return BOND_XMIT_HASH_POLICY_UNKNOWN
+	}
+	return lacp
+}
+
+// Possible BondXmitHashPolicy value
+const (
+	BOND_XMIT_HASH_POLICY_LAYER2 BondXmitHashPolicy = iota
+	BOND_XMIT_HASH_POLICY_LAYER3_4
+	BOND_XMIT_HASH_POLICY_LAYER2_3
+	BOND_XMIT_HASH_POLICY_ENCAP2_3
+	BOND_XMIT_HASH_POLICY_ENCAP3_4
+	BOND_XMIT_HASH_POLICY_UNKNOWN
+)
+
+var bondXmitHashPolicyToString = map[BondXmitHashPolicy]string{
+	BOND_XMIT_HASH_POLICY_LAYER2:   "layer2",
+	BOND_XMIT_HASH_POLICY_LAYER3_4: "layer3+4",
+	BOND_XMIT_HASH_POLICY_LAYER2_3: "layer2+3",
+	BOND_XMIT_HASH_POLICY_ENCAP2_3: "encap2+3",
+	BOND_XMIT_HASH_POLICY_ENCAP3_4: "encap3+4",
+}
+var StringToBondXmitHashPolicyMap = map[string]BondXmitHashPolicy{
+	"layer2":   BOND_XMIT_HASH_POLICY_LAYER2,
+	"layer3+4": BOND_XMIT_HASH_POLICY_LAYER3_4,
+	"layer2+3": BOND_XMIT_HASH_POLICY_LAYER2_3,
+	"encap2+3": BOND_XMIT_HASH_POLICY_ENCAP2_3,
+	"encap3+4": BOND_XMIT_HASH_POLICY_ENCAP3_4,
+}
+
+// BondLacpRate type
+type BondLacpRate int
+
+func (b BondLacpRate) String() string {
+	s, ok := bondLacpRateToString[b]
+	if !ok {
+		return fmt.Sprintf("LacpRate(%d)", b)
+	}
+	return s
+}
+
+// StringToBondLacpRate returns bond lacp arte, or uknonw is the s is invalid.
+func StringToBondLacpRate(s string) BondLacpRate {
+	lacp, ok := StringToBondLacpRateMap[s]
+	if !ok {
+		return BOND_LACP_RATE_UNKNOWN
+	}
+	return lacp
+}
+
+// Possible BondLacpRate value
+const (
+	BOND_LACP_RATE_SLOW BondLacpRate = iota
+	BOND_LACP_RATE_FAST
+	BOND_LACP_RATE_UNKNOWN
+)
+
+var bondLacpRateToString = map[BondLacpRate]string{
+	BOND_LACP_RATE_SLOW: "slow",
+	BOND_LACP_RATE_FAST: "fast",
+}
+var StringToBondLacpRateMap = map[string]BondLacpRate{
+	"slow": BOND_LACP_RATE_SLOW,
+	"fast": BOND_LACP_RATE_FAST,
+}
+
+// BondAdSelect type
+type BondAdSelect int
+
+// Possible BondAdSelect value
+const (
+	BOND_AD_SELECT_STABLE BondAdSelect = iota
+	BOND_AD_SELECT_BANDWIDTH
+	BOND_AD_SELECT_COUNT
+)
+
+// BondAdInfo represents ad info for bond
+type BondAdInfo struct {
+	AggregatorId int
+	NumPorts     int
+	ActorKey     int
+	PartnerKey   int
+	PartnerMac   net.HardwareAddr
+}
+
+// Bond representation
+type Bond struct {
+	LinkAttrs
+	Mode            BondMode
+	ActiveSlave     int
+	Miimon          int
+	UpDelay         int
+	DownDelay       int
+	UseCarrier      int
+	ArpInterval     int
+	ArpIpTargets    []net.IP
+	ArpValidate     BondArpValidate
+	ArpAllTargets   BondArpAllTargets
+	Primary         int
+	PrimaryReselect BondPrimaryReselect
+	FailOverMac     BondFailOverMac
+	XmitHashPolicy  BondXmitHashPolicy
+	ResendIgmp      int
+	NumPeerNotif    int
+	AllSlavesActive int
+	MinLinks        int
+	LpInterval      int
+	PackersPerSlave int
+	LacpRate        BondLacpRate
+	AdSelect        BondAdSelect
+	// looking at iproute tool AdInfo can only be retrived. It can't be set.
+	AdInfo *BondAdInfo
+}
+
+func NewLinkBond(atr LinkAttrs) *Bond {
+	return &Bond{
+		LinkAttrs:       atr,
+		Mode:            -1,
+		ActiveSlave:     -1,
+		Miimon:          -1,
+		UpDelay:         -1,
+		DownDelay:       -1,
+		UseCarrier:      -1,
+		ArpInterval:     -1,
+		ArpIpTargets:    nil,
+		ArpValidate:     -1,
+		ArpAllTargets:   -1,
+		Primary:         -1,
+		PrimaryReselect: -1,
+		FailOverMac:     -1,
+		XmitHashPolicy:  -1,
+		ResendIgmp:      -1,
+		NumPeerNotif:    -1,
+		AllSlavesActive: -1,
+		MinLinks:        -1,
+		LpInterval:      -1,
+		PackersPerSlave: -1,
+		LacpRate:        -1,
+		AdSelect:        -1,
+	}
+}
+
+// Flag mask for bond options. Bond.Flagmask must be set to on for option to work.
+const (
+	BOND_MODE_MASK uint64 = 1 << (1 + iota)
+	BOND_ACTIVE_SLAVE_MASK
+	BOND_MIIMON_MASK
+	BOND_UPDELAY_MASK
+	BOND_DOWNDELAY_MASK
+	BOND_USE_CARRIER_MASK
+	BOND_ARP_INTERVAL_MASK
+	BOND_ARP_VALIDATE_MASK
+	BOND_ARP_ALL_TARGETS_MASK
+	BOND_PRIMARY_MASK
+	BOND_PRIMARY_RESELECT_MASK
+	BOND_FAIL_OVER_MAC_MASK
+	BOND_XMIT_HASH_POLICY_MASK
+	BOND_RESEND_IGMP_MASK
+	BOND_NUM_PEER_NOTIF_MASK
+	BOND_ALL_SLAVES_ACTIVE_MASK
+	BOND_MIN_LINKS_MASK
+	BOND_LP_INTERVAL_MASK
+	BOND_PACKETS_PER_SLAVE_MASK
+	BOND_LACP_RATE_MASK
+	BOND_AD_SELECT_MASK
+)
+
+// Attrs implementation.
+func (bond *Bond) Attrs() *LinkAttrs {
+	return &bond.LinkAttrs
+}
+
+// Type implementation fro Vxlan.
+func (bond *Bond) Type() string {
+	return "bond"
+}
+
+// Gretap devices must specify LocalIP and RemoteIP on create
+type Gretap struct {
+	LinkAttrs
+	IKey       uint32
+	OKey       uint32
+	EncapSport uint16
+	EncapDport uint16
+	Local      net.IP
+	Remote     net.IP
+	IFlags     uint16
+	OFlags     uint16
+	PMtuDisc   uint8
+	Ttl        uint8
+	Tos        uint8
+	EncapType  uint16
+	EncapFlags uint16
+	Link       uint32
+}
+
+func (gretap *Gretap) Attrs() *LinkAttrs {
+	return &gretap.LinkAttrs
+}
+
+func (gretap *Gretap) Type() string {
+	return "gretap"
+}
+
 // iproute2 supported devices;
 // vlan | veth | vcan | dummy | ifb | macvlan | macvtap |
 // bridge | bond | ipoib | ip6tnl | ipip | sit | vxlan |

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

@@ -12,6 +12,13 @@ import (
 	"github.com/vishvananda/netlink/nl"
 )
 
+const SizeofLinkStats = 0x5c
+
+const (
+	TUNTAP_MODE_TUN TuntapMode = syscall.IFF_TUN
+	TUNTAP_MODE_TAP TuntapMode = syscall.IFF_TAP
+)
+
 var native = nl.NativeEndian()
 var lookupByDump = false
 
@@ -33,12 +40,27 @@ func ensureIndex(link *LinkAttrs) {
 	}
 }
 
+func (h *Handle) ensureIndex(link *LinkAttrs) {
+	if link != nil && link.Index == 0 {
+		newlink, _ := h.LinkByName(link.Name)
+		if newlink != nil {
+			link.Index = newlink.Attrs().Index
+		}
+	}
+}
+
 // LinkSetUp enables the link device.
 // Equivalent to: `ip link set $link up`
 func LinkSetUp(link Link) error {
+	return pkgHandle.LinkSetUp(link)
+}
+
+// LinkSetUp enables the link device.
+// Equivalent to: `ip link set $link up`
+func (h *Handle) LinkSetUp(link Link) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_NEWLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Change = syscall.IFF_UP
@@ -53,9 +75,15 @@ func LinkSetUp(link Link) error {
 // LinkSetDown disables link device.
 // Equivalent to: `ip link set $link down`
 func LinkSetDown(link Link) error {
+	return pkgHandle.LinkSetDown(link)
+}
+
+// LinkSetDown disables link device.
+// Equivalent to: `ip link set $link down`
+func (h *Handle) LinkSetDown(link Link) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_NEWLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Change = syscall.IFF_UP
@@ -70,9 +98,15 @@ func LinkSetDown(link Link) error {
 // LinkSetMTU sets the mtu of the link device.
 // Equivalent to: `ip link set $link mtu $mtu`
 func LinkSetMTU(link Link, mtu int) error {
+	return pkgHandle.LinkSetMTU(link, mtu)
+}
+
+// LinkSetMTU sets the mtu of the link device.
+// Equivalent to: `ip link set $link mtu $mtu`
+func (h *Handle) LinkSetMTU(link Link, mtu int) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Index = int32(base.Index)
@@ -91,9 +125,15 @@ func LinkSetMTU(link Link, mtu int) error {
 // LinkSetName sets the name of the link device.
 // Equivalent to: `ip link set $link name $name`
 func LinkSetName(link Link, name string) error {
+	return pkgHandle.LinkSetName(link, name)
+}
+
+// LinkSetName sets the name of the link device.
+// Equivalent to: `ip link set $link name $name`
+func (h *Handle) LinkSetName(link Link, name string) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Index = int32(base.Index)
@@ -106,12 +146,42 @@ func LinkSetName(link Link, name string) error {
 	return err
 }
 
+// LinkSetAlias sets the alias of the link device.
+// Equivalent to: `ip link set dev $link alias $name`
+func LinkSetAlias(link Link, name string) error {
+	return pkgHandle.LinkSetAlias(link, name)
+}
+
+// LinkSetAlias sets the alias of the link device.
+// Equivalent to: `ip link set dev $link alias $name`
+func (h *Handle) LinkSetAlias(link Link, name string) error {
+	base := link.Attrs()
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+
+	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
+	msg.Index = int32(base.Index)
+	req.AddData(msg)
+
+	data := nl.NewRtAttr(syscall.IFLA_IFALIAS, []byte(name))
+	req.AddData(data)
+
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
 // LinkSetHardwareAddr sets the hardware address of the link device.
 // Equivalent to: `ip link set $link address $hwaddr`
 func LinkSetHardwareAddr(link Link, hwaddr net.HardwareAddr) error {
+	return pkgHandle.LinkSetHardwareAddr(link, hwaddr)
+}
+
+// LinkSetHardwareAddr sets the hardware address of the link device.
+// Equivalent to: `ip link set $link address $hwaddr`
+func (h *Handle) LinkSetHardwareAddr(link Link, hwaddr net.HardwareAddr) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Index = int32(base.Index)
@@ -124,24 +194,111 @@ func LinkSetHardwareAddr(link Link, hwaddr net.HardwareAddr) error {
 	return err
 }
 
+// LinkSetVfHardwareAddr sets the hardware address of a vf for the link.
+// Equivalent to: `ip link set $link vf $vf mac $hwaddr`
+func LinkSetVfHardwareAddr(link Link, vf int, hwaddr net.HardwareAddr) error {
+	return pkgHandle.LinkSetVfHardwareAddr(link, vf, hwaddr)
+}
+
+// LinkSetVfHardwareAddr sets the hardware address of a vf for the link.
+// Equivalent to: `ip link set $link vf $vf mac $hwaddr`
+func (h *Handle) LinkSetVfHardwareAddr(link Link, vf int, hwaddr net.HardwareAddr) error {
+	base := link.Attrs()
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+
+	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
+	msg.Index = int32(base.Index)
+	req.AddData(msg)
+
+	data := nl.NewRtAttr(nl.IFLA_VFINFO_LIST, nil)
+	info := nl.NewRtAttrChild(data, nl.IFLA_VF_INFO, nil)
+	vfmsg := nl.VfMac{
+		Vf: uint32(vf),
+	}
+	copy(vfmsg.Mac[:], []byte(hwaddr))
+	nl.NewRtAttrChild(info, nl.IFLA_VF_MAC, vfmsg.Serialize())
+	req.AddData(data)
+
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
+// LinkSetVfVlan sets the vlan of a vf for the link.
+// Equivalent to: `ip link set $link vf $vf vlan $vlan`
+func LinkSetVfVlan(link Link, vf, vlan int) error {
+	return pkgHandle.LinkSetVfVlan(link, vf, vlan)
+}
+
+// LinkSetVfVlan sets the vlan of a vf for the link.
+// Equivalent to: `ip link set $link vf $vf vlan $vlan`
+func (h *Handle) LinkSetVfVlan(link Link, vf, vlan int) error {
+	base := link.Attrs()
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+
+	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
+	msg.Index = int32(base.Index)
+	req.AddData(msg)
+
+	data := nl.NewRtAttr(nl.IFLA_VFINFO_LIST, nil)
+	info := nl.NewRtAttrChild(data, nl.IFLA_VF_INFO, nil)
+	vfmsg := nl.VfVlan{
+		Vf:   uint32(vf),
+		Vlan: uint32(vlan),
+	}
+	nl.NewRtAttrChild(info, nl.IFLA_VF_VLAN, vfmsg.Serialize())
+	req.AddData(data)
+
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
 // LinkSetMaster sets the master of the link device.
 // Equivalent to: `ip link set $link master $master`
 func LinkSetMaster(link Link, master *Bridge) error {
+	return pkgHandle.LinkSetMaster(link, master)
+}
+
+// LinkSetMaster sets the master of the link device.
+// Equivalent to: `ip link set $link master $master`
+func (h *Handle) LinkSetMaster(link Link, master *Bridge) error {
 	index := 0
 	if master != nil {
 		masterBase := master.Attrs()
-		ensureIndex(masterBase)
+		h.ensureIndex(masterBase)
 		index = masterBase.Index
 	}
-	return LinkSetMasterByIndex(link, index)
+	if index <= 0 {
+		return fmt.Errorf("Device does not exist")
+	}
+	return h.LinkSetMasterByIndex(link, index)
+}
+
+// LinkSetNoMaster removes the master of the link device.
+// Equivalent to: `ip link set $link nomaster`
+func LinkSetNoMaster(link Link) error {
+	return pkgHandle.LinkSetNoMaster(link)
+}
+
+// LinkSetNoMaster removes the master of the link device.
+// Equivalent to: `ip link set $link nomaster`
+func (h *Handle) LinkSetNoMaster(link Link) error {
+	return h.LinkSetMasterByIndex(link, 0)
 }
 
 // LinkSetMasterByIndex sets the master of the link device.
 // Equivalent to: `ip link set $link master $master`
 func LinkSetMasterByIndex(link Link, masterIndex int) error {
+	return pkgHandle.LinkSetMasterByIndex(link, masterIndex)
+}
+
+// LinkSetMasterByIndex sets the master of the link device.
+// Equivalent to: `ip link set $link master $master`
+func (h *Handle) LinkSetMasterByIndex(link Link, masterIndex int) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Index = int32(base.Index)
@@ -161,9 +318,16 @@ func LinkSetMasterByIndex(link Link, masterIndex int) error {
 // pid must be a pid of a running process.
 // Equivalent to: `ip link set $link netns $pid`
 func LinkSetNsPid(link Link, nspid int) error {
+	return pkgHandle.LinkSetNsPid(link, nspid)
+}
+
+// LinkSetNsPid puts the device into a new network namespace. The
+// pid must be a pid of a running process.
+// Equivalent to: `ip link set $link netns $pid`
+func (h *Handle) LinkSetNsPid(link Link, nspid int) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Index = int32(base.Index)
@@ -183,9 +347,16 @@ func LinkSetNsPid(link Link, nspid int) error {
 // fd must be an open file descriptor to a network namespace.
 // Similar to: `ip link set $link netns $ns`
 func LinkSetNsFd(link Link, fd int) error {
+	return pkgHandle.LinkSetNsFd(link, fd)
+}
+
+// LinkSetNsFd puts the device into a new network namespace. The
+// fd must be an open file descriptor to a network namespace.
+// Similar to: `ip link set $link netns $ns`
+func (h *Handle) LinkSetNsFd(link Link, fd int) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Index = int32(base.Index)
@@ -250,10 +421,12 @@ func addVxlanAttrs(vxlan *Vxlan, linkInfo *nl.RtAttr) {
 	nl.NewRtAttrChild(data, nl.IFLA_VXLAN_L2MISS, boolAttr(vxlan.L2miss))
 	nl.NewRtAttrChild(data, nl.IFLA_VXLAN_L3MISS, boolAttr(vxlan.L3miss))
 
+	if vxlan.UDPCSum {
+		nl.NewRtAttrChild(data, nl.IFLA_VXLAN_UDP_CSUM, boolAttr(vxlan.UDPCSum))
+	}
 	if vxlan.GBP {
 		nl.NewRtAttrChild(data, nl.IFLA_VXLAN_GBP, boolAttr(vxlan.GBP))
 	}
-
 	if vxlan.NoAge {
 		nl.NewRtAttrChild(data, nl.IFLA_VXLAN_AGEING, nl.Uint32Attr(0))
 	} else if vxlan.Age > 0 {
@@ -263,7 +436,7 @@ func addVxlanAttrs(vxlan *Vxlan, linkInfo *nl.RtAttr) {
 		nl.NewRtAttrChild(data, nl.IFLA_VXLAN_LIMIT, nl.Uint32Attr(uint32(vxlan.Limit)))
 	}
 	if vxlan.Port > 0 {
-		nl.NewRtAttrChild(data, nl.IFLA_VXLAN_PORT, nl.Uint16Attr(uint16(vxlan.Port)))
+		nl.NewRtAttrChild(data, nl.IFLA_VXLAN_PORT, htons(uint16(vxlan.Port)))
 	}
 	if vxlan.PortLow > 0 || vxlan.PortHigh > 0 {
 		pr := vxlanPortRange{uint16(vxlan.PortLow), uint16(vxlan.PortHigh)}
@@ -275,10 +448,98 @@ func addVxlanAttrs(vxlan *Vxlan, linkInfo *nl.RtAttr) {
 	}
 }
 
+func addBondAttrs(bond *Bond, linkInfo *nl.RtAttr) {
+	data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil)
+	if bond.Mode >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_MODE, nl.Uint8Attr(uint8(bond.Mode)))
+	}
+	if bond.ActiveSlave >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_ACTIVE_SLAVE, nl.Uint32Attr(uint32(bond.ActiveSlave)))
+	}
+	if bond.Miimon >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_MIIMON, nl.Uint32Attr(uint32(bond.Miimon)))
+	}
+	if bond.UpDelay >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_UPDELAY, nl.Uint32Attr(uint32(bond.UpDelay)))
+	}
+	if bond.DownDelay >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_DOWNDELAY, nl.Uint32Attr(uint32(bond.DownDelay)))
+	}
+	if bond.UseCarrier >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_USE_CARRIER, nl.Uint8Attr(uint8(bond.UseCarrier)))
+	}
+	if bond.ArpInterval >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_ARP_INTERVAL, nl.Uint32Attr(uint32(bond.ArpInterval)))
+	}
+	if bond.ArpIpTargets != nil {
+		msg := nl.NewRtAttrChild(data, nl.IFLA_BOND_ARP_IP_TARGET, nil)
+		for i := range bond.ArpIpTargets {
+			ip := bond.ArpIpTargets[i].To4()
+			if ip != nil {
+				nl.NewRtAttrChild(msg, i, []byte(ip))
+				continue
+			}
+			ip = bond.ArpIpTargets[i].To16()
+			if ip != nil {
+				nl.NewRtAttrChild(msg, i, []byte(ip))
+			}
+		}
+	}
+	if bond.ArpValidate >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_ARP_VALIDATE, nl.Uint32Attr(uint32(bond.ArpValidate)))
+	}
+	if bond.ArpAllTargets >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_ARP_ALL_TARGETS, nl.Uint32Attr(uint32(bond.ArpAllTargets)))
+	}
+	if bond.Primary >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_PRIMARY, nl.Uint32Attr(uint32(bond.Primary)))
+	}
+	if bond.PrimaryReselect >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_PRIMARY_RESELECT, nl.Uint8Attr(uint8(bond.PrimaryReselect)))
+	}
+	if bond.FailOverMac >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_FAIL_OVER_MAC, nl.Uint8Attr(uint8(bond.FailOverMac)))
+	}
+	if bond.XmitHashPolicy >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_XMIT_HASH_POLICY, nl.Uint8Attr(uint8(bond.XmitHashPolicy)))
+	}
+	if bond.ResendIgmp >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_RESEND_IGMP, nl.Uint32Attr(uint32(bond.ResendIgmp)))
+	}
+	if bond.NumPeerNotif >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_NUM_PEER_NOTIF, nl.Uint8Attr(uint8(bond.NumPeerNotif)))
+	}
+	if bond.AllSlavesActive >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_ALL_SLAVES_ACTIVE, nl.Uint8Attr(uint8(bond.AllSlavesActive)))
+	}
+	if bond.MinLinks >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_MIN_LINKS, nl.Uint32Attr(uint32(bond.MinLinks)))
+	}
+	if bond.LpInterval >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_LP_INTERVAL, nl.Uint32Attr(uint32(bond.LpInterval)))
+	}
+	if bond.PackersPerSlave >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_PACKETS_PER_SLAVE, nl.Uint32Attr(uint32(bond.PackersPerSlave)))
+	}
+	if bond.LacpRate >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_AD_LACP_RATE, nl.Uint8Attr(uint8(bond.LacpRate)))
+	}
+	if bond.AdSelect >= 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_BOND_AD_SELECT, nl.Uint8Attr(uint8(bond.AdSelect)))
+	}
+}
+
 // LinkAdd adds a new link device. The type and features of the device
-// are taken fromt the parameters in the link object.
+// are taken from the parameters in the link object.
 // Equivalent to: `ip link add $link`
 func LinkAdd(link Link) error {
+	return pkgHandle.LinkAdd(link)
+}
+
+// LinkAdd adds a new link device. The type and features of the device
+// are taken fromt the parameters in the link object.
+// Equivalent to: `ip link add $link`
+func (h *Handle) LinkAdd(link Link) error {
 	// TODO: set mtu and hardware address
 	// TODO: support extra data for macvlan
 	base := link.Attrs()
@@ -315,19 +576,40 @@ func LinkAdd(link Link) error {
 		if errno != 0 {
 			return fmt.Errorf("Tuntap IOCTL TUNSETPERSIST failed, errno %v", errno)
 		}
-		ensureIndex(base)
+		h.ensureIndex(base)
 
 		// can't set master during create, so set it afterwards
 		if base.MasterIndex != 0 {
 			// TODO: verify MasterIndex is actually a bridge?
-			return LinkSetMasterByIndex(link, base.MasterIndex)
+			return h.LinkSetMasterByIndex(link, base.MasterIndex)
 		}
 		return nil
 	}
 
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWLINK, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	req := h.newNetlinkRequest(syscall.RTM_NEWLINK, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
+	// TODO: make it shorter
+	if base.Flags&net.FlagUp != 0 {
+		msg.Change = syscall.IFF_UP
+		msg.Flags = syscall.IFF_UP
+	}
+	if base.Flags&net.FlagBroadcast != 0 {
+		msg.Change |= syscall.IFF_BROADCAST
+		msg.Flags |= syscall.IFF_BROADCAST
+	}
+	if base.Flags&net.FlagLoopback != 0 {
+		msg.Change |= syscall.IFF_LOOPBACK
+		msg.Flags |= syscall.IFF_LOOPBACK
+	}
+	if base.Flags&net.FlagPointToPoint != 0 {
+		msg.Change |= syscall.IFF_POINTOPOINT
+		msg.Flags |= syscall.IFF_POINTOPOINT
+	}
+	if base.Flags&net.FlagMulticast != 0 {
+		msg.Change |= syscall.IFF_MULTICAST
+		msg.Flags |= syscall.IFF_MULTICAST
+	}
 	req.AddData(msg)
 
 	if base.ParentIndex != 0 {
@@ -388,6 +670,8 @@ func LinkAdd(link Link) error {
 
 	} else if vxlan, ok := link.(*Vxlan); ok {
 		addVxlanAttrs(vxlan, linkInfo)
+	} else if bond, ok := link.(*Bond); ok {
+		addBondAttrs(bond, linkInfo)
 	} else if ipv, ok := link.(*IPVlan); ok {
 		data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil)
 		nl.NewRtAttrChild(data, nl.IFLA_IPVLAN_MODE, nl.Uint16Attr(uint16(ipv.Mode)))
@@ -396,6 +680,13 @@ func LinkAdd(link Link) error {
 			data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil)
 			nl.NewRtAttrChild(data, nl.IFLA_MACVLAN_MODE, nl.Uint32Attr(macvlanModes[macv.Mode]))
 		}
+	} else if macv, ok := link.(*Macvtap); ok {
+		if macv.Mode != MACVLAN_MODE_DEFAULT {
+			data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil)
+			nl.NewRtAttrChild(data, nl.IFLA_MACVLAN_MODE, nl.Uint32Attr(macvlanModes[macv.Mode]))
+		}
+	} else if gretap, ok := link.(*Gretap); ok {
+		addGretapAttrs(gretap, linkInfo)
 	}
 
 	req.AddData(linkInfo)
@@ -405,12 +696,12 @@ func LinkAdd(link Link) error {
 		return err
 	}
 
-	ensureIndex(base)
+	h.ensureIndex(base)
 
 	// can't set master during create, so set it afterwards
 	if base.MasterIndex != 0 {
 		// TODO: verify MasterIndex is actually a bridge?
-		return LinkSetMasterByIndex(link, base.MasterIndex)
+		return h.LinkSetMasterByIndex(link, base.MasterIndex)
 	}
 	return nil
 }
@@ -419,11 +710,18 @@ func LinkAdd(link Link) error {
 // the link object for it to be deleted. The other values are ignored.
 // Equivalent to: `ip link del $link`
 func LinkDel(link Link) error {
+	return pkgHandle.LinkDel(link)
+}
+
+// LinkDel deletes link device. Either Index or Name must be set in
+// the link object for it to be deleted. The other values are ignored.
+// Equivalent to: `ip link del $link`
+func (h *Handle) LinkDel(link Link) error {
 	base := link.Attrs()
 
-	ensureIndex(base)
+	h.ensureIndex(base)
 
-	req := nl.NewNetlinkRequest(syscall.RTM_DELLINK, syscall.NLM_F_ACK)
+	req := h.newNetlinkRequest(syscall.RTM_DELLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Index = int32(base.Index)
@@ -433,8 +731,8 @@ func LinkDel(link Link) error {
 	return err
 }
 
-func linkByNameDump(name string) (Link, error) {
-	links, err := LinkList()
+func (h *Handle) linkByNameDump(name string) (Link, error) {
+	links, err := h.LinkList()
 	if err != nil {
 		return nil, err
 	}
@@ -447,13 +745,32 @@ func linkByNameDump(name string) (Link, error) {
 	return nil, fmt.Errorf("Link %s not found", name)
 }
 
+func (h *Handle) linkByAliasDump(alias string) (Link, error) {
+	links, err := h.LinkList()
+	if err != nil {
+		return nil, err
+	}
+
+	for _, link := range links {
+		if link.Attrs().Alias == alias {
+			return link, nil
+		}
+	}
+	return nil, fmt.Errorf("Link alias %s not found", alias)
+}
+
 // LinkByName finds a link by name and returns a pointer to the object.
 func LinkByName(name string) (Link, error) {
-	if lookupByDump {
-		return linkByNameDump(name)
+	return pkgHandle.LinkByName(name)
+}
+
+// LinkByName finds a link by name and returns a pointer to the object.
+func (h *Handle) LinkByName(name string) (Link, error) {
+	if h.lookupByDump {
+		return h.linkByNameDump(name)
 	}
 
-	req := nl.NewNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_ACK)
+	req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	req.AddData(msg)
@@ -465,8 +782,40 @@ func LinkByName(name string) (Link, error) {
 	if err == syscall.EINVAL {
 		// older kernels don't support looking up via IFLA_IFNAME
 		// so fall back to dumping all links
-		lookupByDump = true
-		return linkByNameDump(name)
+		h.lookupByDump = true
+		return h.linkByNameDump(name)
+	}
+
+	return link, err
+}
+
+// LinkByAlias finds a link by its alias and returns a pointer to the object.
+// If there are multiple links with the alias it returns the first one
+func LinkByAlias(alias string) (Link, error) {
+	return pkgHandle.LinkByAlias(alias)
+}
+
+// LinkByAlias finds a link by its alias and returns a pointer to the object.
+// If there are multiple links with the alias it returns the first one
+func (h *Handle) LinkByAlias(alias string) (Link, error) {
+	if h.lookupByDump {
+		return h.linkByAliasDump(alias)
+	}
+
+	req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_ACK)
+
+	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
+	req.AddData(msg)
+
+	nameData := nl.NewRtAttr(syscall.IFLA_IFALIAS, nl.ZeroTerminated(alias))
+	req.AddData(nameData)
+
+	link, err := execGetLink(req)
+	if err == syscall.EINVAL {
+		// older kernels don't support looking up via IFLA_IFALIAS
+		// so fall back to dumping all links
+		h.lookupByDump = true
+		return h.linkByAliasDump(alias)
 	}
 
 	return link, err
@@ -474,7 +823,12 @@ func LinkByName(name string) (Link, error) {
 
 // LinkByIndex finds a link by index and returns a pointer to the object.
 func LinkByIndex(index int) (Link, error) {
-	req := nl.NewNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_ACK)
+	return pkgHandle.LinkByIndex(index)
+}
+
+// LinkByIndex finds a link by index and returns a pointer to the object.
+func (h *Handle) LinkByIndex(index int) (Link, error) {
+	req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	msg.Index = int32(index)
@@ -543,12 +897,16 @@ func linkDeserialize(m []byte) (Link, error) {
 						link = &Veth{}
 					case "vxlan":
 						link = &Vxlan{}
+					case "bond":
+						link = &Bond{}
 					case "ipvlan":
 						link = &IPVlan{}
 					case "macvlan":
 						link = &Macvlan{}
 					case "macvtap":
 						link = &Macvtap{}
+					case "gretap":
+						link = &Gretap{}
 					default:
 						link = &GenericLink{LinkType: linkType}
 					}
@@ -562,12 +920,16 @@ func linkDeserialize(m []byte) (Link, error) {
 						parseVlanData(link, data)
 					case "vxlan":
 						parseVxlanData(link, data)
+					case "bond":
+						parseBondData(link, data)
 					case "ipvlan":
 						parseIPVlanData(link, data)
 					case "macvlan":
 						parseMacvlanData(link, data)
 					case "macvtap":
 						parseMacvtapData(link, data)
+					case "gretap":
+						parseGretapData(link, data)
 					}
 				}
 			}
@@ -591,6 +953,10 @@ func linkDeserialize(m []byte) (Link, error) {
 			base.MasterIndex = int(native.Uint32(attr.Value[0:4]))
 		case syscall.IFLA_TXQLEN:
 			base.TxQLen = int(native.Uint32(attr.Value[0:4]))
+		case syscall.IFLA_IFALIAS:
+			base.Alias = string(attr.Value[:len(attr.Value)-1])
+		case syscall.IFLA_STATS:
+			base.Statistics = parseLinkStats(attr.Value[:])
 		}
 	}
 	// Links that don't have IFLA_INFO_KIND are hardware devices
@@ -605,9 +971,15 @@ func linkDeserialize(m []byte) (Link, error) {
 // LinkList gets a list of link devices.
 // Equivalent to: `ip link show`
 func LinkList() ([]Link, error) {
+	return pkgHandle.LinkList()
+}
+
+// LinkList gets a list of link devices.
+// Equivalent to: `ip link show`
+func (h *Handle) LinkList() ([]Link, error) {
 	// NOTE(vish): This duplicates functionality in net/iface_linux.go, but we need
 	//             to get the message ourselves to parse link type.
-	req := nl.NewNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_DUMP)
+	req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_DUMP)
 
 	msg := nl.NewIfInfomsg(syscall.AF_UNSPEC)
 	req.AddData(msg)
@@ -632,6 +1004,7 @@ func LinkList() ([]Link, error) {
 // LinkUpdate is used to pass information back from LinkSubscribe()
 type LinkUpdate struct {
 	nl.IfInfomsg
+	Header syscall.NlMsghdr
 	Link
 }
 
@@ -661,7 +1034,7 @@ func LinkSubscribe(ch chan<- LinkUpdate, done <-chan struct{}) error {
 				if err != nil {
 					return
 				}
-				ch <- LinkUpdate{IfInfomsg: *ifmsg, Link: link}
+				ch <- LinkUpdate{IfInfomsg: *ifmsg, Header: m.Header, Link: link}
 			}
 		}
 	}()
@@ -670,33 +1043,57 @@ func LinkSubscribe(ch chan<- LinkUpdate, done <-chan struct{}) error {
 }
 
 func LinkSetHairpin(link Link, mode bool) error {
-	return setProtinfoAttr(link, mode, nl.IFLA_BRPORT_MODE)
+	return pkgHandle.LinkSetHairpin(link, mode)
+}
+
+func (h *Handle) LinkSetHairpin(link Link, mode bool) error {
+	return h.setProtinfoAttr(link, mode, nl.IFLA_BRPORT_MODE)
 }
 
 func LinkSetGuard(link Link, mode bool) error {
-	return setProtinfoAttr(link, mode, nl.IFLA_BRPORT_GUARD)
+	return pkgHandle.LinkSetGuard(link, mode)
+}
+
+func (h *Handle) LinkSetGuard(link Link, mode bool) error {
+	return h.setProtinfoAttr(link, mode, nl.IFLA_BRPORT_GUARD)
 }
 
 func LinkSetFastLeave(link Link, mode bool) error {
-	return setProtinfoAttr(link, mode, nl.IFLA_BRPORT_FAST_LEAVE)
+	return pkgHandle.LinkSetFastLeave(link, mode)
+}
+
+func (h *Handle) LinkSetFastLeave(link Link, mode bool) error {
+	return h.setProtinfoAttr(link, mode, nl.IFLA_BRPORT_FAST_LEAVE)
 }
 
 func LinkSetLearning(link Link, mode bool) error {
-	return setProtinfoAttr(link, mode, nl.IFLA_BRPORT_LEARNING)
+	return pkgHandle.LinkSetLearning(link, mode)
+}
+
+func (h *Handle) LinkSetLearning(link Link, mode bool) error {
+	return h.setProtinfoAttr(link, mode, nl.IFLA_BRPORT_LEARNING)
 }
 
 func LinkSetRootBlock(link Link, mode bool) error {
-	return setProtinfoAttr(link, mode, nl.IFLA_BRPORT_PROTECT)
+	return pkgHandle.LinkSetRootBlock(link, mode)
+}
+
+func (h *Handle) LinkSetRootBlock(link Link, mode bool) error {
+	return h.setProtinfoAttr(link, mode, nl.IFLA_BRPORT_PROTECT)
 }
 
 func LinkSetFlood(link Link, mode bool) error {
-	return setProtinfoAttr(link, mode, nl.IFLA_BRPORT_UNICAST_FLOOD)
+	return pkgHandle.LinkSetFlood(link, mode)
 }
 
-func setProtinfoAttr(link Link, mode bool, attr int) error {
+func (h *Handle) LinkSetFlood(link Link, mode bool) error {
+	return h.setProtinfoAttr(link, mode, nl.IFLA_BRPORT_UNICAST_FLOOD)
+}
+
+func (h *Handle) setProtinfoAttr(link Link, mode bool, attr int) error {
 	base := link.Attrs()
-	ensureIndex(base)
-	req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
+	h.ensureIndex(base)
+	req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK)
 
 	msg := nl.NewIfInfomsg(syscall.AF_BRIDGE)
 	msg.Index = int32(base.Index)
@@ -752,6 +1149,8 @@ func parseVxlanData(link Link, data []syscall.NetlinkRouteAttr) {
 			vxlan.L2miss = int8(datum.Value[0]) != 0
 		case nl.IFLA_VXLAN_L3MISS:
 			vxlan.L3miss = int8(datum.Value[0]) != 0
+		case nl.IFLA_VXLAN_UDP_CSUM:
+			vxlan.UDPCSum = int8(datum.Value[0]) != 0
 		case nl.IFLA_VXLAN_GBP:
 			vxlan.GBP = int8(datum.Value[0]) != 0
 		case nl.IFLA_VXLAN_AGEING:
@@ -760,7 +1159,7 @@ func parseVxlanData(link Link, data []syscall.NetlinkRouteAttr) {
 		case nl.IFLA_VXLAN_LIMIT:
 			vxlan.Limit = int(native.Uint32(datum.Value[0:4]))
 		case nl.IFLA_VXLAN_PORT:
-			vxlan.Port = int(native.Uint16(datum.Value[0:2]))
+			vxlan.Port = int(ntohs(datum.Value[0:2]))
 		case nl.IFLA_VXLAN_PORT_RANGE:
 			buf := bytes.NewBuffer(datum.Value[0:4])
 			var pr vxlanPortRange
@@ -772,6 +1171,60 @@ func parseVxlanData(link Link, data []syscall.NetlinkRouteAttr) {
 	}
 }
 
+func parseBondData(link Link, data []syscall.NetlinkRouteAttr) {
+	bond := NewLinkBond(NewLinkAttrs())
+	for i := range data {
+		switch data[i].Attr.Type {
+		case nl.IFLA_BOND_MODE:
+			bond.Mode = BondMode(data[i].Value[0])
+		case nl.IFLA_BOND_ACTIVE_SLAVE:
+			bond.ActiveSlave = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_MIIMON:
+			bond.Miimon = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_UPDELAY:
+			bond.UpDelay = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_DOWNDELAY:
+			bond.DownDelay = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_USE_CARRIER:
+			bond.UseCarrier = int(data[i].Value[0])
+		case nl.IFLA_BOND_ARP_INTERVAL:
+			bond.ArpInterval = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_ARP_IP_TARGET:
+			// TODO: implement
+		case nl.IFLA_BOND_ARP_VALIDATE:
+			bond.ArpValidate = BondArpValidate(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_ARP_ALL_TARGETS:
+			bond.ArpAllTargets = BondArpAllTargets(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_PRIMARY:
+			bond.Primary = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_PRIMARY_RESELECT:
+			bond.PrimaryReselect = BondPrimaryReselect(data[i].Value[0])
+		case nl.IFLA_BOND_FAIL_OVER_MAC:
+			bond.FailOverMac = BondFailOverMac(data[i].Value[0])
+		case nl.IFLA_BOND_XMIT_HASH_POLICY:
+			bond.XmitHashPolicy = BondXmitHashPolicy(data[i].Value[0])
+		case nl.IFLA_BOND_RESEND_IGMP:
+			bond.ResendIgmp = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_NUM_PEER_NOTIF:
+			bond.NumPeerNotif = int(data[i].Value[0])
+		case nl.IFLA_BOND_ALL_SLAVES_ACTIVE:
+			bond.AllSlavesActive = int(data[i].Value[0])
+		case nl.IFLA_BOND_MIN_LINKS:
+			bond.MinLinks = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_LP_INTERVAL:
+			bond.LpInterval = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_PACKETS_PER_SLAVE:
+			bond.PackersPerSlave = int(native.Uint32(data[i].Value[0:4]))
+		case nl.IFLA_BOND_AD_LACP_RATE:
+			bond.LacpRate = BondLacpRate(data[i].Value[0])
+		case nl.IFLA_BOND_AD_SELECT:
+			bond.AdSelect = BondAdSelect(data[i].Value[0])
+		case nl.IFLA_BOND_AD_INFO:
+			// TODO: implement
+		}
+	}
+}
+
 func parseIPVlanData(link Link, data []syscall.NetlinkRouteAttr) {
 	ipv := link.(*IPVlan)
 	for _, datum := range data {
@@ -828,3 +1281,100 @@ func linkFlags(rawFlags uint32) net.Flags {
 	}
 	return f
 }
+
+func htonl(val uint32) []byte {
+	bytes := make([]byte, 4)
+	binary.BigEndian.PutUint32(bytes, val)
+	return bytes
+}
+
+func htons(val uint16) []byte {
+	bytes := make([]byte, 2)
+	binary.BigEndian.PutUint16(bytes, val)
+	return bytes
+}
+
+func ntohl(buf []byte) uint32 {
+	return binary.BigEndian.Uint32(buf)
+}
+
+func ntohs(buf []byte) uint16 {
+	return binary.BigEndian.Uint16(buf)
+}
+
+func addGretapAttrs(gretap *Gretap, linkInfo *nl.RtAttr) {
+	data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil)
+
+	ip := gretap.Local.To4()
+	if ip != nil {
+		nl.NewRtAttrChild(data, nl.IFLA_GRE_LOCAL, []byte(ip))
+	}
+	ip = gretap.Remote.To4()
+	if ip != nil {
+		nl.NewRtAttrChild(data, nl.IFLA_GRE_REMOTE, []byte(ip))
+	}
+
+	if gretap.IKey != 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_GRE_IKEY, htonl(gretap.IKey))
+		gretap.IFlags |= uint16(nl.GRE_KEY)
+	}
+
+	if gretap.OKey != 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_GRE_OKEY, htonl(gretap.OKey))
+		gretap.OFlags |= uint16(nl.GRE_KEY)
+	}
+
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_IFLAGS, htons(gretap.IFlags))
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_OFLAGS, htons(gretap.OFlags))
+
+	if gretap.Link != 0 {
+		nl.NewRtAttrChild(data, nl.IFLA_GRE_LINK, nl.Uint32Attr(gretap.Link))
+	}
+
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_PMTUDISC, nl.Uint8Attr(gretap.PMtuDisc))
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_TTL, nl.Uint8Attr(gretap.Ttl))
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_TOS, nl.Uint8Attr(gretap.Tos))
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_ENCAP_TYPE, nl.Uint16Attr(gretap.EncapType))
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_ENCAP_FLAGS, nl.Uint16Attr(gretap.EncapFlags))
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_ENCAP_SPORT, htons(gretap.EncapSport))
+	nl.NewRtAttrChild(data, nl.IFLA_GRE_ENCAP_DPORT, htons(gretap.EncapDport))
+}
+
+func parseGretapData(link Link, data []syscall.NetlinkRouteAttr) {
+	gre := link.(*Gretap)
+	for _, datum := range data {
+		switch datum.Attr.Type {
+		case nl.IFLA_GRE_OKEY:
+			gre.IKey = ntohl(datum.Value[0:4])
+		case nl.IFLA_GRE_IKEY:
+			gre.OKey = ntohl(datum.Value[0:4])
+		case nl.IFLA_GRE_LOCAL:
+			gre.Local = net.IP(datum.Value[0:4])
+		case nl.IFLA_GRE_REMOTE:
+			gre.Remote = net.IP(datum.Value[0:4])
+		case nl.IFLA_GRE_ENCAP_SPORT:
+			gre.EncapSport = ntohs(datum.Value[0:2])
+		case nl.IFLA_GRE_ENCAP_DPORT:
+			gre.EncapDport = ntohs(datum.Value[0:2])
+		case nl.IFLA_GRE_IFLAGS:
+			gre.IFlags = ntohs(datum.Value[0:2])
+		case nl.IFLA_GRE_OFLAGS:
+			gre.OFlags = ntohs(datum.Value[0:2])
+
+		case nl.IFLA_GRE_TTL:
+			gre.Ttl = uint8(datum.Value[0])
+		case nl.IFLA_GRE_TOS:
+			gre.Tos = uint8(datum.Value[0])
+		case nl.IFLA_GRE_PMTUDISC:
+			gre.PMtuDisc = uint8(datum.Value[0])
+		case nl.IFLA_GRE_ENCAP_TYPE:
+			gre.EncapType = native.Uint16(datum.Value[0:2])
+		case nl.IFLA_GRE_ENCAP_FLAGS:
+			gre.EncapFlags = native.Uint16(datum.Value[0:2])
+		}
+	}
+}
+
+func parseLinkStats(data []byte) *LinkStatistics {
+	return (*LinkStatistics)(unsafe.Pointer(&data[0:SizeofLinkStats][0]))
+}

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

@@ -3,6 +3,7 @@ package netlink
 import (
 	"bytes"
 	"net"
+	"os"
 	"syscall"
 	"testing"
 	"time"
@@ -45,16 +46,6 @@ func testLinkAddDel(t *testing.T, link Link) {
 		}
 	}
 
-	if rBase.ParentIndex == 0 && base.ParentIndex != 0 {
-		t.Fatal("Created link doesn't have a Parent but it should")
-	} else if rBase.ParentIndex != 0 && base.ParentIndex == 0 {
-		t.Fatal("Created link has a Parent but it shouldn't")
-	} else if rBase.ParentIndex != 0 && base.ParentIndex != 0 {
-		if rBase.ParentIndex != base.ParentIndex {
-			t.Fatal("Link.ParentIndex doesn't match")
-		}
-	}
-
 	if veth, ok := result.(*Veth); ok {
 		if rBase.TxQLen != base.TxQLen {
 			t.Fatalf("qlen is %d, should be %d", rBase.TxQLen, base.TxQLen)
@@ -63,17 +54,30 @@ func testLinkAddDel(t *testing.T, link Link) {
 			t.Fatalf("MTU is %d, should be %d", rBase.MTU, base.MTU)
 		}
 
-		if veth.PeerName != "" {
-			var peer *Veth
-			other, err := LinkByName(veth.PeerName)
-			if err != nil {
-				t.Fatalf("Peer %s not created", veth.PeerName)
+		if original, ok := link.(*Veth); ok {
+			if original.PeerName != "" {
+				var peer *Veth
+				other, err := LinkByName(original.PeerName)
+				if err != nil {
+					t.Fatalf("Peer %s not created", veth.PeerName)
+				}
+				if peer, ok = other.(*Veth); !ok {
+					t.Fatalf("Peer %s is incorrect type", veth.PeerName)
+				}
+				if peer.TxQLen != testTxQLen {
+					t.Fatalf("TxQLen of peer is %d, should be %d", peer.TxQLen, testTxQLen)
+				}
 			}
-			if peer, ok = other.(*Veth); !ok {
-				t.Fatalf("Peer %s is incorrect type", veth.PeerName)
-			}
-			if peer.TxQLen != testTxQLen {
-				t.Fatalf("TxQLen of peer is %d, should be %d", peer.TxQLen, testTxQLen)
+		}
+	} else {
+		// recent kernels set the parent index for veths in the response
+		if rBase.ParentIndex == 0 && base.ParentIndex != 0 {
+			t.Fatalf("Created link doesn't have parent %d but it should", base.ParentIndex)
+		} else if rBase.ParentIndex != 0 && base.ParentIndex == 0 {
+			t.Fatalf("Created link has parent %d but it shouldn't", rBase.ParentIndex)
+		} else if rBase.ParentIndex != 0 && base.ParentIndex != 0 {
+			if rBase.ParentIndex != base.ParentIndex {
+				t.Fatalf("Link.ParentIndex doesn't match %d != %d", rBase.ParentIndex, base.ParentIndex)
 			}
 		}
 	}
@@ -106,6 +110,16 @@ func testLinkAddDel(t *testing.T, link Link) {
 		}
 	}
 
+	if macv, ok := link.(*Macvtap); ok {
+		other, ok := result.(*Macvtap)
+		if !ok {
+			t.Fatal("Result of create is not a macvtap")
+		}
+		if macv.Mode != other.Mode {
+			t.Fatalf("Got unexpected mode: %d, expected: %d", other.Mode, macv.Mode)
+		}
+	}
+
 	if err = LinkDel(link); err != nil {
 		t.Fatal(err)
 	}
@@ -117,6 +131,7 @@ func testLinkAddDel(t *testing.T, link Link) {
 
 	if len(links) != num {
 		t.Fatal("Link not removed properly")
+		return
 	}
 }
 
@@ -199,6 +214,19 @@ func TestLinkAddDelBridge(t *testing.T) {
 	testLinkAddDel(t, &Bridge{LinkAttrs{Name: "foo", MTU: 1400}})
 }
 
+func TestLinkAddDelGretap(t *testing.T) {
+	tearDown := setUpNetlinkTest(t)
+	defer tearDown()
+
+	testLinkAddDel(t, &Gretap{
+		LinkAttrs: LinkAttrs{Name: "foo"},
+		IKey:      0x101,
+		OKey:      0x101,
+		PMtuDisc:  1,
+		Local:     net.IPv4(127, 0, 0, 1),
+		Remote:    net.IPv4(127, 0, 0, 1)})
+}
+
 func TestLinkAddDelVlan(t *testing.T) {
 	tearDown := setUpNetlinkTest(t)
 	defer tearDown()
@@ -229,6 +257,16 @@ func TestLinkAddDelMacvlan(t *testing.T) {
 		Mode:      MACVLAN_MODE_PRIVATE,
 	})
 
+	testLinkAddDel(t, &Macvlan{
+		LinkAttrs: LinkAttrs{Name: "bar", ParentIndex: parent.Attrs().Index},
+		Mode:      MACVLAN_MODE_BRIDGE,
+	})
+
+	testLinkAddDel(t, &Macvlan{
+		LinkAttrs: LinkAttrs{Name: "bar", ParentIndex: parent.Attrs().Index},
+		Mode:      MACVLAN_MODE_VEPA,
+	})
+
 	if err := LinkDel(parent); err != nil {
 		t.Fatal(err)
 	}
@@ -250,6 +288,20 @@ func TestLinkAddDelMacvtap(t *testing.T) {
 		},
 	})
 
+	testLinkAddDel(t, &Macvtap{
+		Macvlan: Macvlan{
+			LinkAttrs: LinkAttrs{Name: "bar", ParentIndex: parent.Attrs().Index},
+			Mode:      MACVLAN_MODE_BRIDGE,
+		},
+	})
+
+	testLinkAddDel(t, &Macvtap{
+		Macvlan: Macvlan{
+			LinkAttrs: LinkAttrs{Name: "bar", ParentIndex: parent.Attrs().Index},
+			Mode:      MACVLAN_MODE_VEPA,
+		},
+	})
+
 	if err := LinkDel(parent); err != nil {
 		t.Fatal(err)
 	}
@@ -259,7 +311,15 @@ func TestLinkAddDelVeth(t *testing.T) {
 	tearDown := setUpNetlinkTest(t)
 	defer tearDown()
 
-	testLinkAddDel(t, &Veth{LinkAttrs{Name: "foo", TxQLen: testTxQLen, MTU: 1400}, "bar"})
+	veth := &Veth{LinkAttrs: LinkAttrs{Name: "foo", TxQLen: testTxQLen, MTU: 1400}, PeerName: "bar"}
+	testLinkAddDel(t, veth)
+}
+
+func TestLinkAddDelBond(t *testing.T) {
+	tearDown := setUpNetlinkTest(t)
+	defer tearDown()
+
+	testLinkAddDel(t, NewLinkBond(LinkAttrs{Name: "foo"}))
 }
 
 func TestLinkAddVethWithDefaultTxQLen(t *testing.T) {
@@ -389,6 +449,12 @@ func TestLinkSetUnsetResetMaster(t *testing.T) {
 		t.Fatal(err)
 	}
 
+	nonexistsmaster := &Bridge{LinkAttrs{Name: "foobar"}}
+
+	if err := LinkSetMaster(slave, nonexistsmaster); err == nil {
+		t.Fatal("error expected")
+	}
+
 	if err := LinkSetMaster(slave, master); err != nil {
 		t.Fatal(err)
 	}
@@ -415,7 +481,7 @@ func TestLinkSetUnsetResetMaster(t *testing.T) {
 		t.Fatal("Master not reset properly")
 	}
 
-	if err := LinkSetMaster(slave, nil); err != nil {
+	if err := LinkSetNoMaster(slave); err != nil {
 		t.Fatal(err)
 	}
 
@@ -531,6 +597,9 @@ func TestLinkAddDelVxlan(t *testing.T) {
 }
 
 func TestLinkAddDelIPVlanL2(t *testing.T) {
+	if os.Getenv("TRAVIS_BUILD_DIR") != "" {
+		t.Skipf("Kernel in travis is too old for this test")
+	}
 	tearDown := setUpNetlinkTest(t)
 	defer tearDown()
 	parent := &Dummy{LinkAttrs{Name: "foo"}}
@@ -550,6 +619,9 @@ func TestLinkAddDelIPVlanL2(t *testing.T) {
 }
 
 func TestLinkAddDelIPVlanL3(t *testing.T) {
+	if os.Getenv("TRAVIS_BUILD_DIR") != "" {
+		t.Skipf("Kernel in travis is too old for this test")
+	}
 	tearDown := setUpNetlinkTest(t)
 	defer tearDown()
 	parent := &Dummy{LinkAttrs{Name: "foo"}}
@@ -670,6 +742,25 @@ func TestLinkSet(t *testing.T) {
 	if !bytes.Equal(link.Attrs().HardwareAddr, addr) {
 		t.Fatalf("hardware address not changed!")
 	}
+
+	err = LinkSetAlias(link, "barAlias")
+	if err != nil {
+		t.Fatalf("Could not set alias: %v", err)
+	}
+
+	link, err = LinkByName("bar")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if link.Attrs().Alias != "barAlias" {
+		t.Fatalf("alias not changed!")
+	}
+
+	link, err = LinkByAlias("barAlias")
+	if err != nil {
+		t.Fatal(err)
+	}
 }
 
 func expectLinkUpdate(ch <-chan LinkUpdate, ifaceName string, up bool) bool {
@@ -722,3 +813,51 @@ func TestLinkSubscribe(t *testing.T) {
 		t.Fatal("Del update not received as expected")
 	}
 }
+
+func TestLinkStats(t *testing.T) {
+	defer setUpNetlinkTest(t)()
+
+	// Create a veth pair and verify the cross-stats once both
+	// ends are brought up and some ICMPv6 packets are exchanged
+	v0 := "v0"
+	v1 := "v1"
+
+	vethLink := &Veth{LinkAttrs: LinkAttrs{Name: v0}, PeerName: v1}
+	if err := LinkAdd(vethLink); err != nil {
+		t.Fatal(err)
+	}
+
+	veth0, err := LinkByName(v0)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := LinkSetUp(veth0); err != nil {
+		t.Fatal(err)
+	}
+
+	veth1, err := LinkByName(v1)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := LinkSetUp(veth1); err != nil {
+		t.Fatal(err)
+	}
+
+	time.Sleep(2 * time.Second)
+
+	// verify statistics
+	veth0, err = LinkByName(v0)
+	if err != nil {
+		t.Fatal(err)
+	}
+	veth1, err = LinkByName(v1)
+	if err != nil {
+		t.Fatal(err)
+	}
+	v0Stats := veth0.Attrs().Statistics
+	v1Stats := veth1.Attrs().Statistics
+	if v0Stats.RxPackets != v1Stats.TxPackets || v0Stats.TxPackets != v1Stats.RxPackets ||
+		v0Stats.RxBytes != v1Stats.TxBytes || v0Stats.TxBytes != v1Stats.RxBytes {
+		t.Fatalf("veth ends counters differ:\n%v\n%v", v0Stats, v1Stats)
+	}
+}

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

@@ -67,30 +67,62 @@ func (msg *Ndmsg) Len() int {
 // NeighAdd will add an IP to MAC mapping to the ARP table
 // Equivalent to: `ip neigh add ....`
 func NeighAdd(neigh *Neigh) error {
-	return neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL)
+	return pkgHandle.NeighAdd(neigh)
+}
+
+// NeighAdd will add an IP to MAC mapping to the ARP table
+// Equivalent to: `ip neigh add ....`
+func (h *Handle) NeighAdd(neigh *Neigh) error {
+	return h.neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL)
 }
 
 // NeighSet will add or replace an IP to MAC mapping to the ARP table
 // Equivalent to: `ip neigh replace....`
 func NeighSet(neigh *Neigh) error {
-	return neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_REPLACE)
+	return pkgHandle.NeighSet(neigh)
+}
+
+// NeighSet will add or replace an IP to MAC mapping to the ARP table
+// Equivalent to: `ip neigh replace....`
+func (h *Handle) NeighSet(neigh *Neigh) error {
+	return h.neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_REPLACE)
 }
 
 // NeighAppend will append an entry to FDB
 // Equivalent to: `bridge fdb append...`
 func NeighAppend(neigh *Neigh) error {
-	return neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_APPEND)
+	return pkgHandle.NeighAppend(neigh)
 }
 
+// NeighAppend will append an entry to FDB
+// Equivalent to: `bridge fdb append...`
+func (h *Handle) NeighAppend(neigh *Neigh) error {
+	return h.neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_APPEND)
+}
+
+// NeighAppend will append an entry to FDB
+// Equivalent to: `bridge fdb append...`
 func neighAdd(neigh *Neigh, mode int) error {
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWNEIGH, mode|syscall.NLM_F_ACK)
+	return pkgHandle.neighAdd(neigh, mode)
+}
+
+// NeighAppend will append an entry to FDB
+// Equivalent to: `bridge fdb append...`
+func (h *Handle) neighAdd(neigh *Neigh, mode int) error {
+	req := h.newNetlinkRequest(syscall.RTM_NEWNEIGH, mode|syscall.NLM_F_ACK)
 	return neighHandle(neigh, req)
 }
 
 // NeighDel will delete an IP address from a link device.
 // Equivalent to: `ip addr del $addr dev $link`
 func NeighDel(neigh *Neigh) error {
-	req := nl.NewNetlinkRequest(syscall.RTM_DELNEIGH, syscall.NLM_F_ACK)
+	return pkgHandle.NeighDel(neigh)
+}
+
+// NeighDel will delete an IP address from a link device.
+// Equivalent to: `ip addr del $addr dev $link`
+func (h *Handle) NeighDel(neigh *Neigh) error {
+	req := h.newNetlinkRequest(syscall.RTM_DELNEIGH, syscall.NLM_F_ACK)
 	return neighHandle(neigh, req)
 }
 
@@ -130,7 +162,14 @@ func neighHandle(neigh *Neigh, req *nl.NetlinkRequest) error {
 // Equivalent to: `ip neighbor show`.
 // The list can be filtered by link and ip family.
 func NeighList(linkIndex, family int) ([]Neigh, error) {
-	req := nl.NewNetlinkRequest(syscall.RTM_GETNEIGH, syscall.NLM_F_DUMP)
+	return pkgHandle.NeighList(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) {
+	req := h.newNetlinkRequest(syscall.RTM_GETNEIGH, syscall.NLM_F_DUMP)
 	msg := Ndmsg{
 		Family: uint8(family),
 		Index:  uint32(linkIndex),

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

@@ -8,18 +8,7 @@
 // interface that is loosly modeled on the iproute2 cli.
 package netlink
 
-import (
-	"net"
-
-	"github.com/vishvananda/netlink/nl"
-)
-
-const (
-	// Family type definitions
-	FAMILY_ALL = nl.FAMILY_ALL
-	FAMILY_V4  = nl.FAMILY_V4
-	FAMILY_V6  = nl.FAMILY_V6
-)
+import "net"
 
 // ParseIPNet parses a string in ip/net format and returns a net.IPNet.
 // This is valuable because addresses in netlink are often IPNets and
@@ -33,7 +22,10 @@ func ParseIPNet(s string) (*net.IPNet, error) {
 	return &net.IPNet{IP: ip, Mask: ipNet.Mask}, nil
 }
 
-// NewIPNet generates an IPNet from an ip address using a netmask of 32.
+// NewIPNet generates an IPNet from an ip address using a netmask of 32 or 128.
 func NewIPNet(ip net.IP) *net.IPNet {
-	return &net.IPNet{IP: ip, Mask: net.CIDRMask(32, 32)}
+	if ip.To4() != nil {
+		return &net.IPNet{IP: ip, Mask: net.CIDRMask(32, 32)}
+	}
+	return &net.IPNet{IP: ip, Mask: net.CIDRMask(128, 128)}
 }

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

@@ -0,0 +1,10 @@
+package netlink
+
+import "github.com/vishvananda/netlink/nl"
+
+// Family type definitions
+const (
+	FAMILY_ALL = nl.FAMILY_ALL
+	FAMILY_V4  = nl.FAMILY_V4
+	FAMILY_V6  = nl.FAMILY_V6
+)

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

@@ -138,6 +138,6 @@ func NeighList(linkIndex, family int) ([]Neigh, error) {
 	return nil, ErrNotImplemented
 }
 
-func NeighDeserialize(m []byte) (*Ndmsg, *Neigh, error) {
-	return nil, nil, ErrNotImplemented
+func NeighDeserialize(m []byte) (*Neigh, error) {
+	return nil, ErrNotImplemented
 }

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

@@ -1,7 +1,13 @@
 package nl
 
+import (
+	"unsafe"
+)
+
 const (
 	DEFAULT_CHANGE = 0xFFFFFFFF
+	// doesn't exist in syscall
+	IFLA_VFINFO_LIST = 0x16
 )
 
 const (
@@ -102,3 +108,289 @@ const (
 	MACVLAN_MODE_PASSTHRU = 8
 	MACVLAN_MODE_SOURCE   = 16
 )
+
+const (
+	IFLA_BOND_UNSPEC = iota
+	IFLA_BOND_MODE
+	IFLA_BOND_ACTIVE_SLAVE
+	IFLA_BOND_MIIMON
+	IFLA_BOND_UPDELAY
+	IFLA_BOND_DOWNDELAY
+	IFLA_BOND_USE_CARRIER
+	IFLA_BOND_ARP_INTERVAL
+	IFLA_BOND_ARP_IP_TARGET
+	IFLA_BOND_ARP_VALIDATE
+	IFLA_BOND_ARP_ALL_TARGETS
+	IFLA_BOND_PRIMARY
+	IFLA_BOND_PRIMARY_RESELECT
+	IFLA_BOND_FAIL_OVER_MAC
+	IFLA_BOND_XMIT_HASH_POLICY
+	IFLA_BOND_RESEND_IGMP
+	IFLA_BOND_NUM_PEER_NOTIF
+	IFLA_BOND_ALL_SLAVES_ACTIVE
+	IFLA_BOND_MIN_LINKS
+	IFLA_BOND_LP_INTERVAL
+	IFLA_BOND_PACKETS_PER_SLAVE
+	IFLA_BOND_AD_LACP_RATE
+	IFLA_BOND_AD_SELECT
+	IFLA_BOND_AD_INFO
+)
+
+const (
+	IFLA_BOND_AD_INFO_UNSPEC = iota
+	IFLA_BOND_AD_INFO_AGGREGATOR
+	IFLA_BOND_AD_INFO_NUM_PORTS
+	IFLA_BOND_AD_INFO_ACTOR_KEY
+	IFLA_BOND_AD_INFO_PARTNER_KEY
+	IFLA_BOND_AD_INFO_PARTNER_MAC
+)
+
+const (
+	IFLA_BOND_SLAVE_UNSPEC = iota
+	IFLA_BOND_SLAVE_STATE
+	IFLA_BOND_SLAVE_MII_STATUS
+	IFLA_BOND_SLAVE_LINK_FAILURE_COUNT
+	IFLA_BOND_SLAVE_PERM_HWADDR
+	IFLA_BOND_SLAVE_QUEUE_ID
+	IFLA_BOND_SLAVE_AD_AGGREGATOR_ID
+)
+
+const (
+	IFLA_GRE_UNSPEC = iota
+	IFLA_GRE_LINK
+	IFLA_GRE_IFLAGS
+	IFLA_GRE_OFLAGS
+	IFLA_GRE_IKEY
+	IFLA_GRE_OKEY
+	IFLA_GRE_LOCAL
+	IFLA_GRE_REMOTE
+	IFLA_GRE_TTL
+	IFLA_GRE_TOS
+	IFLA_GRE_PMTUDISC
+	IFLA_GRE_ENCAP_LIMIT
+	IFLA_GRE_FLOWINFO
+	IFLA_GRE_FLAGS
+	IFLA_GRE_ENCAP_TYPE
+	IFLA_GRE_ENCAP_FLAGS
+	IFLA_GRE_ENCAP_SPORT
+	IFLA_GRE_ENCAP_DPORT
+	IFLA_GRE_COLLECT_METADATA
+	IFLA_GRE_MAX = IFLA_GRE_COLLECT_METADATA
+)
+
+const (
+	GRE_CSUM    = 0x8000
+	GRE_ROUTING = 0x4000
+	GRE_KEY     = 0x2000
+	GRE_SEQ     = 0x1000
+	GRE_STRICT  = 0x0800
+	GRE_REC     = 0x0700
+	GRE_FLAGS   = 0x00F8
+	GRE_VERSION = 0x0007
+)
+
+const (
+	IFLA_VF_INFO_UNSPEC = iota
+	IFLA_VF_INFO
+	IFLA_VF_INFO_MAX = IFLA_VF_INFO
+)
+
+const (
+	IFLA_VF_UNSPEC = iota
+	IFLA_VF_MAC    /* Hardware queue specific attributes */
+	IFLA_VF_VLAN
+	IFLA_VF_TX_RATE      /* Max TX Bandwidth Allocation */
+	IFLA_VF_SPOOFCHK     /* Spoof Checking on/off switch */
+	IFLA_VF_LINK_STATE   /* link state enable/disable/auto switch */
+	IFLA_VF_RATE         /* Min and Max TX Bandwidth Allocation */
+	IFLA_VF_RSS_QUERY_EN /* RSS Redirection Table and Hash Key query
+	 * on/off switch
+	 */
+	IFLA_VF_STATS /* network device statistics */
+	IFLA_VF_MAX   = IFLA_VF_STATS
+)
+
+const (
+	IFLA_VF_LINK_STATE_AUTO    = iota /* link state of the uplink */
+	IFLA_VF_LINK_STATE_ENABLE         /* link always up */
+	IFLA_VF_LINK_STATE_DISABLE        /* link always down */
+	IFLA_VF_LINK_STATE_MAX     = IFLA_VF_LINK_STATE_DISABLE
+)
+
+const (
+	IFLA_VF_STATS_RX_PACKETS = iota
+	IFLA_VF_STATS_TX_PACKETS
+	IFLA_VF_STATS_RX_BYTES
+	IFLA_VF_STATS_TX_BYTES
+	IFLA_VF_STATS_BROADCAST
+	IFLA_VF_STATS_MULTICAST
+	IFLA_VF_STATS_MAX = IFLA_VF_STATS_MULTICAST
+)
+
+const (
+	SizeofVfMac        = 0x24
+	SizeofVfVlan       = 0x0c
+	SizeofVfTxRate     = 0x08
+	SizeofVfRate       = 0x0c
+	SizeofVfSpoofchk   = 0x08
+	SizeofVfLinkState  = 0x08
+	SizeofVfRssQueryEn = 0x08
+)
+
+// struct ifla_vf_mac {
+//   __u32 vf;
+//   __u8 mac[32]; /* MAX_ADDR_LEN */
+// };
+
+type VfMac struct {
+	Vf  uint32
+	Mac [32]byte
+}
+
+func (msg *VfMac) Len() int {
+	return SizeofVfMac
+}
+
+func DeserializeVfMac(b []byte) *VfMac {
+	return (*VfMac)(unsafe.Pointer(&b[0:SizeofVfMac][0]))
+}
+
+func (msg *VfMac) Serialize() []byte {
+	return (*(*[SizeofVfMac]byte)(unsafe.Pointer(msg)))[:]
+}
+
+// struct ifla_vf_vlan {
+//   __u32 vf;
+//   __u32 vlan; /* 0 - 4095, 0 disables VLAN filter */
+//   __u32 qos;
+// };
+
+type VfVlan struct {
+	Vf   uint32
+	Vlan uint32
+	Qos  uint32
+}
+
+func (msg *VfVlan) Len() int {
+	return SizeofVfVlan
+}
+
+func DeserializeVfVlan(b []byte) *VfVlan {
+	return (*VfVlan)(unsafe.Pointer(&b[0:SizeofVfVlan][0]))
+}
+
+func (msg *VfVlan) Serialize() []byte {
+	return (*(*[SizeofVfVlan]byte)(unsafe.Pointer(msg)))[:]
+}
+
+// struct ifla_vf_tx_rate {
+//   __u32 vf;
+//   __u32 rate; /* Max TX bandwidth in Mbps, 0 disables throttling */
+// };
+
+type VfTxRate struct {
+	Vf   uint32
+	Rate uint32
+}
+
+func (msg *VfTxRate) Len() int {
+	return SizeofVfTxRate
+}
+
+func DeserializeVfTxRate(b []byte) *VfTxRate {
+	return (*VfTxRate)(unsafe.Pointer(&b[0:SizeofVfTxRate][0]))
+}
+
+func (msg *VfTxRate) Serialize() []byte {
+	return (*(*[SizeofVfTxRate]byte)(unsafe.Pointer(msg)))[:]
+}
+
+// struct ifla_vf_rate {
+//   __u32 vf;
+//   __u32 min_tx_rate; /* Min Bandwidth in Mbps */
+//   __u32 max_tx_rate; /* Max Bandwidth in Mbps */
+// };
+
+type VfRate struct {
+	Vf        uint32
+	MinTxRate uint32
+	MaxTxRate uint32
+}
+
+func (msg *VfRate) Len() int {
+	return SizeofVfRate
+}
+
+func DeserializeVfRate(b []byte) *VfRate {
+	return (*VfRate)(unsafe.Pointer(&b[0:SizeofVfRate][0]))
+}
+
+func (msg *VfRate) Serialize() []byte {
+	return (*(*[SizeofVfRate]byte)(unsafe.Pointer(msg)))[:]
+}
+
+// struct ifla_vf_spoofchk {
+//   __u32 vf;
+//   __u32 setting;
+// };
+
+type VfSpoofchk struct {
+	Vf      uint32
+	Setting uint32
+}
+
+func (msg *VfSpoofchk) Len() int {
+	return SizeofVfSpoofchk
+}
+
+func DeserializeVfSpoofchk(b []byte) *VfSpoofchk {
+	return (*VfSpoofchk)(unsafe.Pointer(&b[0:SizeofVfSpoofchk][0]))
+}
+
+func (msg *VfSpoofchk) Serialize() []byte {
+	return (*(*[SizeofVfSpoofchk]byte)(unsafe.Pointer(msg)))[:]
+}
+
+// struct ifla_vf_link_state {
+//   __u32 vf;
+//   __u32 link_state;
+// };
+
+type VfLinkState struct {
+	Vf        uint32
+	LinkState uint32
+}
+
+func (msg *VfLinkState) Len() int {
+	return SizeofVfLinkState
+}
+
+func DeserializeVfLinkState(b []byte) *VfLinkState {
+	return (*VfLinkState)(unsafe.Pointer(&b[0:SizeofVfLinkState][0]))
+}
+
+func (msg *VfLinkState) Serialize() []byte {
+	return (*(*[SizeofVfLinkState]byte)(unsafe.Pointer(msg)))[:]
+}
+
+// struct ifla_vf_rss_query_en {
+//   __u32 vf;
+//   __u32 setting;
+// };
+
+type VfRssQueryEn struct {
+	Vf      uint32
+	Setting uint32
+}
+
+func (msg *VfRssQueryEn) Len() int {
+	return SizeofVfRssQueryEn
+}
+
+func DeserializeVfRssQueryEn(b []byte) *VfRssQueryEn {
+	return (*VfRssQueryEn)(unsafe.Pointer(&b[0:SizeofVfRssQueryEn][0]))
+}
+
+func (msg *VfRssQueryEn) Serialize() []byte {
+	return (*(*[SizeofVfRssQueryEn]byte)(unsafe.Pointer(msg)))[:]
+}

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

@@ -0,0 +1,199 @@
+package nl
+
+import (
+	"bytes"
+	"crypto/rand"
+	"encoding/binary"
+	"testing"
+)
+
+func (msg *VfMac) write(b []byte) {
+	native := NativeEndian()
+	native.PutUint32(b[0:4], uint32(msg.Vf))
+	copy(b[4:36], msg.Mac[:])
+}
+
+func (msg *VfMac) serializeSafe() []byte {
+	length := SizeofVfMac
+	b := make([]byte, length)
+	msg.write(b)
+	return b
+}
+
+func deserializeVfMacSafe(b []byte) *VfMac {
+	var msg = VfMac{}
+	binary.Read(bytes.NewReader(b[0:SizeofVfMac]), NativeEndian(), &msg)
+	return &msg
+}
+
+func TestVfMacDeserializeSerialize(t *testing.T) {
+	var orig = make([]byte, SizeofVfMac)
+	rand.Read(orig)
+	safemsg := deserializeVfMacSafe(orig)
+	msg := DeserializeVfMac(orig)
+	testDeserializeSerialize(t, orig, safemsg, msg)
+}
+
+func (msg *VfVlan) write(b []byte) {
+	native := NativeEndian()
+	native.PutUint32(b[0:4], uint32(msg.Vf))
+	native.PutUint32(b[4:8], uint32(msg.Vlan))
+	native.PutUint32(b[8:12], uint32(msg.Qos))
+}
+
+func (msg *VfVlan) serializeSafe() []byte {
+	length := SizeofVfVlan
+	b := make([]byte, length)
+	msg.write(b)
+	return b
+}
+
+func deserializeVfVlanSafe(b []byte) *VfVlan {
+	var msg = VfVlan{}
+	binary.Read(bytes.NewReader(b[0:SizeofVfVlan]), NativeEndian(), &msg)
+	return &msg
+}
+
+func TestVfVlanDeserializeSerialize(t *testing.T) {
+	var orig = make([]byte, SizeofVfVlan)
+	rand.Read(orig)
+	safemsg := deserializeVfVlanSafe(orig)
+	msg := DeserializeVfVlan(orig)
+	testDeserializeSerialize(t, orig, safemsg, msg)
+}
+
+func (msg *VfTxRate) write(b []byte) {
+	native := NativeEndian()
+	native.PutUint32(b[0:4], uint32(msg.Vf))
+	native.PutUint32(b[4:8], uint32(msg.Rate))
+}
+
+func (msg *VfTxRate) serializeSafe() []byte {
+	length := SizeofVfTxRate
+	b := make([]byte, length)
+	msg.write(b)
+	return b
+}
+
+func deserializeVfTxRateSafe(b []byte) *VfTxRate {
+	var msg = VfTxRate{}
+	binary.Read(bytes.NewReader(b[0:SizeofVfTxRate]), NativeEndian(), &msg)
+	return &msg
+}
+
+func TestVfTxRateDeserializeSerialize(t *testing.T) {
+	var orig = make([]byte, SizeofVfTxRate)
+	rand.Read(orig)
+	safemsg := deserializeVfTxRateSafe(orig)
+	msg := DeserializeVfTxRate(orig)
+	testDeserializeSerialize(t, orig, safemsg, msg)
+}
+
+func (msg *VfRate) write(b []byte) {
+	native := NativeEndian()
+	native.PutUint32(b[0:4], uint32(msg.Vf))
+	native.PutUint32(b[4:8], uint32(msg.MinTxRate))
+	native.PutUint32(b[8:12], uint32(msg.MaxTxRate))
+}
+
+func (msg *VfRate) serializeSafe() []byte {
+	length := SizeofVfRate
+	b := make([]byte, length)
+	msg.write(b)
+	return b
+}
+
+func deserializeVfRateSafe(b []byte) *VfRate {
+	var msg = VfRate{}
+	binary.Read(bytes.NewReader(b[0:SizeofVfRate]), NativeEndian(), &msg)
+	return &msg
+}
+
+func TestVfRateDeserializeSerialize(t *testing.T) {
+	var orig = make([]byte, SizeofVfRate)
+	rand.Read(orig)
+	safemsg := deserializeVfRateSafe(orig)
+	msg := DeserializeVfRate(orig)
+	testDeserializeSerialize(t, orig, safemsg, msg)
+}
+
+func (msg *VfSpoofchk) write(b []byte) {
+	native := NativeEndian()
+	native.PutUint32(b[0:4], uint32(msg.Vf))
+	native.PutUint32(b[4:8], uint32(msg.Setting))
+}
+
+func (msg *VfSpoofchk) serializeSafe() []byte {
+	length := SizeofVfSpoofchk
+	b := make([]byte, length)
+	msg.write(b)
+	return b
+}
+
+func deserializeVfSpoofchkSafe(b []byte) *VfSpoofchk {
+	var msg = VfSpoofchk{}
+	binary.Read(bytes.NewReader(b[0:SizeofVfSpoofchk]), NativeEndian(), &msg)
+	return &msg
+}
+
+func TestVfSpoofchkDeserializeSerialize(t *testing.T) {
+	var orig = make([]byte, SizeofVfSpoofchk)
+	rand.Read(orig)
+	safemsg := deserializeVfSpoofchkSafe(orig)
+	msg := DeserializeVfSpoofchk(orig)
+	testDeserializeSerialize(t, orig, safemsg, msg)
+}
+
+func (msg *VfLinkState) write(b []byte) {
+	native := NativeEndian()
+	native.PutUint32(b[0:4], uint32(msg.Vf))
+	native.PutUint32(b[4:8], uint32(msg.LinkState))
+}
+
+func (msg *VfLinkState) serializeSafe() []byte {
+	length := SizeofVfLinkState
+	b := make([]byte, length)
+	msg.write(b)
+	return b
+}
+
+func deserializeVfLinkStateSafe(b []byte) *VfLinkState {
+	var msg = VfLinkState{}
+	binary.Read(bytes.NewReader(b[0:SizeofVfLinkState]), NativeEndian(), &msg)
+	return &msg
+}
+
+func TestVfLinkStateDeserializeSerialize(t *testing.T) {
+	var orig = make([]byte, SizeofVfLinkState)
+	rand.Read(orig)
+	safemsg := deserializeVfLinkStateSafe(orig)
+	msg := DeserializeVfLinkState(orig)
+	testDeserializeSerialize(t, orig, safemsg, msg)
+}
+
+func (msg *VfRssQueryEn) write(b []byte) {
+	native := NativeEndian()
+	native.PutUint32(b[0:4], uint32(msg.Vf))
+	native.PutUint32(b[4:8], uint32(msg.Setting))
+}
+
+func (msg *VfRssQueryEn) serializeSafe() []byte {
+	length := SizeofVfRssQueryEn
+	b := make([]byte, length)
+	msg.write(b)
+	return b
+}
+
+func deserializeVfRssQueryEnSafe(b []byte) *VfRssQueryEn {
+	var msg = VfRssQueryEn{}
+	binary.Read(bytes.NewReader(b[0:SizeofVfRssQueryEn]), NativeEndian(), &msg)
+	return &msg
+}
+
+func TestVfRssQueryEnDeserializeSerialize(t *testing.T) {
+	var orig = make([]byte, SizeofVfRssQueryEn)
+	rand.Read(orig)
+	safemsg := deserializeVfRssQueryEnSafe(orig)
+	msg := DeserializeVfRssQueryEn(orig)
+	testDeserializeSerialize(t, orig, safemsg, msg)
+}

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

@@ -6,9 +6,13 @@ import (
 	"encoding/binary"
 	"fmt"
 	"net"
+	"runtime"
+	"sync"
 	"sync/atomic"
 	"syscall"
 	"unsafe"
+
+	"github.com/vishvananda/netns"
 )
 
 const (
@@ -171,7 +175,9 @@ func (a *RtAttr) Serialize() []byte {
 
 type NetlinkRequest struct {
 	syscall.NlMsghdr
-	Data []NetlinkRequestData
+	Data        []NetlinkRequestData
+	RouteSocket *NetlinkSocket
+	XfmrSocket  *NetlinkSocket
 }
 
 // Serialize the Netlink Request into a byte array
@@ -203,14 +209,35 @@ func (req *NetlinkRequest) AddData(data NetlinkRequestData) {
 }
 
 // Execute the request against a the given sockType.
-// Returns a list of netlink messages in seriaized format, optionally filtered
+// Returns a list of netlink messages in serialized format, optionally filtered
 // by resType.
 func (req *NetlinkRequest) Execute(sockType int, resType uint16) ([][]byte, error) {
-	s, err := getNetlinkSocket(sockType)
-	if err != nil {
-		return nil, err
+	var (
+		s   *NetlinkSocket
+		err error
+	)
+
+	switch sockType {
+	case syscall.NETLINK_XFRM:
+		s = req.XfmrSocket
+	case syscall.NETLINK_ROUTE:
+		s = req.RouteSocket
+	default:
+		return nil, fmt.Errorf("Socket type %d is not handled", sockType)
+	}
+
+	sharedSocket := s != nil
+
+	if s == nil {
+		s, err = getNetlinkSocket(sockType)
+		if err != nil {
+			return nil, err
+		}
+		defer s.Close()
+	} else {
+		s.Lock()
+		defer s.Unlock()
 	}
-	defer s.Close()
 
 	if err := s.Send(req); err != nil {
 		return nil, err
@@ -231,7 +258,10 @@ done:
 		}
 		for _, m := range msgs {
 			if m.Header.Seq != req.Seq {
-				return nil, fmt.Errorf("Wrong Seq nr %d, expected 1", m.Header.Seq)
+				if sharedSocket {
+					continue
+				}
+				return nil, fmt.Errorf("Wrong Seq nr %d, expected %d", m.Header.Seq, req.Seq)
 			}
 			if m.Header.Pid != pid {
 				return nil, fmt.Errorf("Wrong pid %d, expected %d", m.Header.Pid, pid)
@@ -276,6 +306,7 @@ func NewNetlinkRequest(proto, flags int) *NetlinkRequest {
 type NetlinkSocket struct {
 	fd  int
 	lsa syscall.SockaddrNetlink
+	sync.Mutex
 }
 
 func getNetlinkSocket(protocol int) (*NetlinkSocket, error) {
@@ -295,6 +326,32 @@ func getNetlinkSocket(protocol int) (*NetlinkSocket, error) {
 	return s, nil
 }
 
+// GetNetlinkSocketAt opens a netlink socket in the network namespace newNs
+// and positions the thread back into the network namespace specified by curNs,
+// when done. If curNs is close, the function derives the current namespace and
+// moves back into it when done. If newNs is close, the socket will be opened
+// in the current network namespace.
+func GetNetlinkSocketAt(newNs, curNs netns.NsHandle, protocol int) (*NetlinkSocket, error) {
+	var err error
+
+	if newNs.IsOpen() {
+		runtime.LockOSThread()
+		defer runtime.UnlockOSThread()
+		if !curNs.IsOpen() {
+			if curNs, err = netns.Get(); err != nil {
+				return nil, fmt.Errorf("could not get current namespace while creating netlink socket: %v", err)
+			}
+			defer curNs.Close()
+		}
+		if err := netns.Set(newNs); err != nil {
+			return nil, fmt.Errorf("failed to set into network namespace %d while creating netlink socket: %v", newNs, err)
+		}
+		defer netns.Set(curNs)
+	}
+
+	return getNetlinkSocket(protocol)
+}
+
 // Create a netlink socket with a given protocol (e.g. NETLINK_ROUTE)
 // and subscribe it to multicast groups passed in variable argument list.
 // Returns the netlink socket on which Receive() method can be called
@@ -323,6 +380,7 @@ func Subscribe(protocol int, groups ...uint) (*NetlinkSocket, error) {
 
 func (s *NetlinkSocket) Close() {
 	syscall.Close(s.fd)
+	s.fd = -1
 }
 
 func (s *NetlinkSocket) GetFd() int {
@@ -330,6 +388,9 @@ func (s *NetlinkSocket) GetFd() int {
 }
 
 func (s *NetlinkSocket) Send(request *NetlinkRequest) error {
+	if s.fd < 0 {
+		return fmt.Errorf("Send called on a closed socket")
+	}
 	if err := syscall.Sendto(s.fd, request.Serialize(), 0, &s.lsa); err != nil {
 		return err
 	}
@@ -337,6 +398,9 @@ func (s *NetlinkSocket) Send(request *NetlinkRequest) error {
 }
 
 func (s *NetlinkSocket) Receive() ([]syscall.NetlinkMessage, error) {
+	if s.fd < 0 {
+		return nil, fmt.Errorf("Receive called on a closed socket")
+	}
 	rb := make([]byte, syscall.Getpagesize())
 	nr, _, err := syscall.Recvfrom(s.fd, rb, 0)
 	if err != nil {

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

@@ -40,3 +40,15 @@ func DeserializeRtMsg(b []byte) *RtMsg {
 func (msg *RtMsg) Serialize() []byte {
 	return (*(*[syscall.SizeofRtMsg]byte)(unsafe.Pointer(msg)))[:]
 }
+
+type RtNexthop struct {
+	syscall.RtNexthop
+}
+
+func DeserializeRtNexthop(b []byte) *RtNexthop {
+	return (*RtNexthop)(unsafe.Pointer(&b[0:syscall.SizeofRtNexthop][0]))
+}
+
+func (msg *RtNexthop) Serialize() []byte {
+	return (*(*[syscall.SizeofRtNexthop]byte)(unsafe.Pointer(msg)))[:]
+}

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

@@ -0,0 +1,37 @@
+package nl
+
+// syscall package lack of rule atributes type.
+// Thus there are defined below
+const (
+	FRA_UNSPEC  = iota
+	FRA_DST     /* destination address */
+	FRA_SRC     /* source address */
+	FRA_IIFNAME /* interface name */
+	FRA_GOTO    /* target to jump to (FR_ACT_GOTO) */
+	FRA_UNUSED2
+	FRA_PRIORITY /* priority/preference */
+	FRA_UNUSED3
+	FRA_UNUSED4
+	FRA_UNUSED5
+	FRA_FWMARK /* mark */
+	FRA_FLOW   /* flow/class id */
+	FRA_TUN_ID
+	FRA_SUPPRESS_IFGROUP
+	FRA_SUPPRESS_PREFIXLEN
+	FRA_TABLE  /* Extended table id */
+	FRA_FWMASK /* mask for netfilter mark */
+	FRA_OIFNAME
+)
+
+// ip rule netlink request types
+const (
+	FR_ACT_UNSPEC = iota
+	FR_ACT_TO_TBL /* Pass to fixed table */
+	FR_ACT_GOTO   /* Jump to another rule */
+	FR_ACT_NOP    /* No operation */
+	FR_ACT_RES3
+	FR_ACT_RES4
+	FR_ACT_BLACKHOLE   /* Drop without notification */
+	FR_ACT_UNREACHABLE /* Drop with ENETUNREACH */
+	FR_ACT_PROHIBIT    /* Drop with EACCES */
+)

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

@@ -49,6 +49,15 @@ const (
 	TCAA_MAX    = 1
 )
 
+const (
+	TCA_ACT_UNSPEC = iota
+	TCA_ACT_KIND
+	TCA_ACT_OPTIONS
+	TCA_ACT_INDEX
+	TCA_ACT_STATS
+	TCA_ACT_MAX
+)
+
 const (
 	TCA_PRIO_UNSPEC = iota
 	TCA_PRIO_MQ
@@ -69,7 +78,8 @@ const (
 	SizeofTcHtbGlob      = 0x14
 	SizeofTcU32Key       = 0x10
 	SizeofTcU32Sel       = 0x10 // without keys
-	SizeofTcMirred       = 0x1c
+	SizeofTcGen          = 0x14
+	SizeofTcMirred       = SizeofTcGen + 0x08
 	SizeofTcPolice       = 2*SizeofTcRateSpec + 0x20
 )
 
@@ -506,42 +516,92 @@ func (x *TcU32Sel) Serialize() []byte {
 	return buf
 }
 
+type TcGen struct {
+	Index   uint32
+	Capab   uint32
+	Action  int32
+	Refcnt  int32
+	Bindcnt int32
+}
+
+func (msg *TcGen) Len() int {
+	return SizeofTcGen
+}
+
+func DeserializeTcGen(b []byte) *TcGen {
+	return (*TcGen)(unsafe.Pointer(&b[0:SizeofTcGen][0]))
+}
+
+func (x *TcGen) Serialize() []byte {
+	return (*(*[SizeofTcGen]byte)(unsafe.Pointer(x)))[:]
+}
+
+// #define tc_gen \
+//   __u32                 index; \
+//   __u32                 capab; \
+//   int                   action; \
+//   int                   refcnt; \
+//   int                   bindcnt
+
 const (
-	TCA_ACT_MIRRED = 8
+	TCA_ACT_GACT = 5
 )
 
 const (
-	TCA_MIRRED_UNSPEC = iota
-	TCA_MIRRED_TM
-	TCA_MIRRED_PARMS
-	TCA_MIRRED_MAX = TCA_MIRRED_PARMS
+	TCA_GACT_UNSPEC = iota
+	TCA_GACT_TM
+	TCA_GACT_PARMS
+	TCA_GACT_PROB
+	TCA_GACT_MAX = TCA_GACT_PROB
 )
 
+type TcGact TcGen
+
 const (
-	TCA_EGRESS_REDIR   = 1 /* packet redirect to EGRESS*/
-	TCA_EGRESS_MIRROR  = 2 /* mirror packet to EGRESS */
-	TCA_INGRESS_REDIR  = 3 /* packet redirect to INGRESS*/
-	TCA_INGRESS_MIRROR = 4 /* mirror packet to INGRESS */
+	TCA_ACT_BPF = 13
 )
 
 const (
-	TC_ACT_UNSPEC     = int32(-1)
-	TC_ACT_OK         = 0
-	TC_ACT_RECLASSIFY = 1
-	TC_ACT_SHOT       = 2
-	TC_ACT_PIPE       = 3
-	TC_ACT_STOLEN     = 4
-	TC_ACT_QUEUED     = 5
-	TC_ACT_REPEAT     = 6
-	TC_ACT_JUMP       = 0x10000000
+	TCA_ACT_BPF_UNSPEC = iota
+	TCA_ACT_BPF_TM
+	TCA_ACT_BPF_PARMS
+	TCA_ACT_BPF_OPS_LEN
+	TCA_ACT_BPF_OPS
+	TCA_ACT_BPF_FD
+	TCA_ACT_BPF_NAME
+	TCA_ACT_BPF_MAX = TCA_ACT_BPF_NAME
+)
+
+const (
+	TCA_BPF_FLAG_ACT_DIRECT uint32 = 1 << iota
+)
+
+const (
+	TCA_BPF_UNSPEC = iota
+	TCA_BPF_ACT
+	TCA_BPF_POLICE
+	TCA_BPF_CLASSID
+	TCA_BPF_OPS_LEN
+	TCA_BPF_OPS
+	TCA_BPF_FD
+	TCA_BPF_NAME
+	TCA_BPF_FLAGS
+	TCA_BPF_MAX = TCA_BPF_FLAGS
+)
+
+type TcBpf TcGen
+
+const (
+	TCA_ACT_MIRRED = 8
+)
+
+const (
+	TCA_MIRRED_UNSPEC = iota
+	TCA_MIRRED_TM
+	TCA_MIRRED_PARMS
+	TCA_MIRRED_MAX = TCA_MIRRED_PARMS
 )
 
-// #define tc_gen \
-//   __u32                 index; \
-//   __u32                 capab; \
-//   int                   action; \
-//   int                   refcnt; \
-//   int                   bindcnt
 // struct tc_mirred {
 // 	tc_gen;
 // 	int                     eaction;   /* one of IN/EGRESS_MIRROR/REDIR */
@@ -549,11 +609,7 @@ const (
 // };
 
 type TcMirred struct {
-	Index   uint32
-	Capab   uint32
-	Action  int32
-	Refcnt  int32
-	Bindcnt int32
+	TcGen
 	Eaction int32
 	Ifindex uint32
 }
@@ -570,14 +626,6 @@ func (x *TcMirred) Serialize() []byte {
 	return (*(*[SizeofTcMirred]byte)(unsafe.Pointer(x)))[:]
 }
 
-const (
-	TC_POLICE_UNSPEC     = TC_ACT_UNSPEC
-	TC_POLICE_OK         = TC_ACT_OK
-	TC_POLICE_RECLASSIFY = TC_ACT_RECLASSIFY
-	TC_POLICE_SHOT       = TC_ACT_SHOT
-	TC_POLICE_PIPE       = TC_ACT_PIPE
-)
-
 // struct tc_police {
 // 	__u32			index;
 // 	int			action;

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

@@ -78,6 +78,7 @@ const (
 	SizeofXfrmLifetimeCfg = 0x40
 	SizeofXfrmLifetimeCur = 0x20
 	SizeofXfrmId          = 0x18
+	SizeofXfrmMark        = 0x08
 )
 
 // typedef union {
@@ -256,3 +257,20 @@ func DeserializeXfrmId(b []byte) *XfrmId {
 func (msg *XfrmId) Serialize() []byte {
 	return (*(*[SizeofXfrmId]byte)(unsafe.Pointer(msg)))[:]
 }
+
+type XfrmMark struct {
+	Value uint32
+	Mask  uint32
+}
+
+func (msg *XfrmMark) Len() int {
+	return SizeofXfrmMark
+}
+
+func DeserializeXfrmMark(b []byte) *XfrmMark {
+	return (*XfrmMark)(unsafe.Pointer(&b[0:SizeofXfrmMark][0]))
+}
+
+func (msg *XfrmMark) Serialize() []byte {
+	return (*(*[SizeofXfrmMark]byte)(unsafe.Pointer(msg)))[:]
+}

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

@@ -5,12 +5,13 @@ import (
 )
 
 const (
-	SizeofXfrmUsersaId   = 0x18
-	SizeofXfrmStats      = 0x0c
-	SizeofXfrmUsersaInfo = 0xe0
-	SizeofXfrmAlgo       = 0x44
-	SizeofXfrmAlgoAuth   = 0x48
-	SizeofXfrmEncapTmpl  = 0x18
+	SizeofXfrmUsersaId    = 0x18
+	SizeofXfrmStats       = 0x0c
+	SizeofXfrmUsersaInfo  = 0xe0
+	SizeofXfrmAlgo        = 0x44
+	SizeofXfrmAlgoAuth    = 0x48
+	SizeofXfrmEncapTmpl   = 0x18
+	SizeofXfrmUsersaFlush = 0x8
 )
 
 // struct xfrm_usersa_id {
@@ -219,3 +220,23 @@ func DeserializeXfrmEncapTmpl(b []byte) *XfrmEncapTmpl {
 func (msg *XfrmEncapTmpl) Serialize() []byte {
 	return (*(*[SizeofXfrmEncapTmpl]byte)(unsafe.Pointer(msg)))[:]
 }
+
+// struct xfrm_usersa_flush {
+//    __u8 proto;
+// };
+
+type XfrmUsersaFlush struct {
+	Proto uint8
+}
+
+func (msg *XfrmUsersaFlush) Len() int {
+	return SizeofXfrmUsersaFlush
+}
+
+func DeserializeXfrmUsersaFlush(b []byte) *XfrmUsersaFlush {
+	return (*XfrmUsersaFlush)(unsafe.Pointer(&b[0:SizeofXfrmUsersaFlush][0]))
+}
+
+func (msg *XfrmUsersaFlush) Serialize() []byte {
+	return (*(*[SizeofXfrmUsersaFlush]byte)(unsafe.Pointer(msg)))[:]
+}

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

@@ -127,14 +127,18 @@ func deserializeXfrmAlgoSafe(b []byte) *XfrmAlgo {
 }
 
 func TestXfrmAlgoDeserializeSerialize(t *testing.T) {
+	native := NativeEndian()
 	// use a 32 byte key len
 	var orig = make([]byte, SizeofXfrmAlgo+32)
 	rand.Read(orig)
 	// set the key len to 256 bits
-	orig[64] = 0
-	orig[65] = 1
-	orig[66] = 0
-	orig[67] = 0
+	var KeyLen uint32 = 0x00000100
+	// Little Endian    Big Endian
+	// orig[64] = 0     orig[64] = 0
+	// orig[65] = 1     orig[65] = 0
+	// orig[66] = 0     orig[66] = 1
+	// orig[67] = 0     orig[67] = 0
+	native.PutUint32(orig[64:68], KeyLen)
 	safemsg := deserializeXfrmAlgoSafe(orig)
 	msg := DeserializeXfrmAlgo(orig)
 	testDeserializeSerialize(t, orig, safemsg, msg)
@@ -164,14 +168,18 @@ func deserializeXfrmAlgoAuthSafe(b []byte) *XfrmAlgoAuth {
 }
 
 func TestXfrmAlgoAuthDeserializeSerialize(t *testing.T) {
+	native := NativeEndian()
 	// use a 32 byte key len
 	var orig = make([]byte, SizeofXfrmAlgoAuth+32)
 	rand.Read(orig)
 	// set the key len to 256 bits
-	orig[64] = 0
-	orig[65] = 1
-	orig[66] = 0
-	orig[67] = 0
+	var KeyLen uint32 = 0x00000100
+	// Little Endian    Big Endian
+	// orig[64] = 0     orig[64] = 0
+	// orig[65] = 1     orig[65] = 0
+	// orig[66] = 0     orig[66] = 1
+	// orig[67] = 0     orig[67] = 0
+	native.PutUint32(orig[64:68], KeyLen)
 	safemsg := deserializeXfrmAlgoAuthSafe(orig)
 	msg := DeserializeXfrmAlgoAuth(orig)
 	testDeserializeSerialize(t, orig, safemsg, msg)
@@ -205,3 +213,29 @@ func TestXfrmEncapTmplDeserializeSerialize(t *testing.T) {
 	msg := DeserializeXfrmEncapTmpl(orig)
 	testDeserializeSerialize(t, orig, safemsg, msg)
 }
+
+func (msg *XfrmMark) write(b []byte) {
+	native := NativeEndian()
+	native.PutUint32(b[0:4], msg.Value)
+	native.PutUint32(b[4:8], msg.Mask)
+}
+
+func (msg *XfrmMark) serializeSafe() []byte {
+	b := make([]byte, SizeofXfrmMark)
+	msg.write(b)
+	return b
+}
+
+func deserializeXfrmMarkSafe(b []byte) *XfrmMark {
+	var msg = XfrmMark{}
+	binary.Read(bytes.NewReader(b[0:SizeofXfrmMark]), NativeEndian(), &msg)
+	return &msg
+}
+
+func TestXfrmMarkDeserializeSerialize(t *testing.T) {
+	var orig = make([]byte, SizeofXfrmMark)
+	rand.Read(orig)
+	safemsg := deserializeXfrmMarkSafe(orig)
+	msg := DeserializeXfrmMark(orig)
+	testDeserializeSerialize(t, orig, safemsg, msg)
+}

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

@@ -8,10 +8,14 @@ import (
 )
 
 func LinkGetProtinfo(link Link) (Protinfo, error) {
+	return pkgHandle.LinkGetProtinfo(link)
+}
+
+func (h *Handle) LinkGetProtinfo(link Link) (Protinfo, error) {
 	base := link.Attrs()
-	ensureIndex(base)
+	h.ensureIndex(base)
 	var pi Protinfo
-	req := nl.NewNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_DUMP)
+	req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_DUMP)
 	msg := nl.NewIfInfomsg(syscall.AF_BRIDGE)
 	req.AddData(msg)
 	msgs, err := req.Execute(syscall.NETLINK_ROUTE, 0)

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

@@ -8,16 +8,21 @@ import (
 const (
 	HANDLE_NONE      = 0
 	HANDLE_INGRESS   = 0xFFFFFFF1
+	HANDLE_CLSACT    = HANDLE_INGRESS
 	HANDLE_ROOT      = 0xFFFFFFFF
 	PRIORITY_MAP_LEN = 16
 )
+const (
+	HANDLE_MIN_INGRESS = 0xFFFFFFF2
+	HANDLE_MIN_EGRESS  = 0xFFFFFFF3
+)
 
 type Qdisc interface {
 	Attrs() *QdiscAttrs
 	Type() string
 }
 
-// Qdisc represents a netlink qdisc. A qdisc is associated with a link,
+// QdiscAttrs represents a netlink qdisc. A qdisc is associated with a link,
 // has a handle, a parent and a refcnt. The root qdisc of a device should
 // have parent == HANDLE_ROOT.
 type QdiscAttrs struct {
@@ -28,7 +33,7 @@ type QdiscAttrs struct {
 }
 
 func (q QdiscAttrs) String() string {
-	return fmt.Sprintf("{LinkIndex: %d, Handle: %s, Parent: %s, Refcnt: %s}", q.LinkIndex, HandleStr(q.Handle), HandleStr(q.Parent), q.Refcnt)
+	return fmt.Sprintf("{LinkIndex: %d, Handle: %s, Parent: %s, Refcnt: %d}", q.LinkIndex, HandleStr(q.Handle), HandleStr(q.Parent), q.Refcnt)
 }
 
 func MakeHandle(major, minor uint16) uint32 {
@@ -149,7 +154,7 @@ type NetemQdiscAttrs struct {
 
 func (q NetemQdiscAttrs) String() string {
 	return fmt.Sprintf(
-		"{Latency: %d, Limit: %d, Loss: %d, Gap: %d, Duplicate: %d, Jitter: %d}",
+		"{Latency: %d, Limit: %d, Loss: %f, Gap: %d, Duplicate: %f, Jitter: %d}",
 		q.Latency, q.Limit, q.Loss, q.Gap, q.Duplicate, q.Jitter,
 	)
 }
@@ -171,70 +176,6 @@ type Netem struct {
 	CorruptCorr   uint32
 }
 
-func NewNetem(attrs QdiscAttrs, nattrs NetemQdiscAttrs) *Netem {
-	var limit uint32 = 1000
-	var loss_corr, delay_corr, duplicate_corr uint32
-	var reorder_prob, reorder_corr uint32
-	var corrupt_prob, corrupt_corr uint32
-
-	latency := nattrs.Latency
-	loss := Percentage2u32(nattrs.Loss)
-	gap := nattrs.Gap
-	duplicate := Percentage2u32(nattrs.Duplicate)
-	jitter := nattrs.Jitter
-
-	// Correlation
-	if latency > 0 && jitter > 0 {
-		delay_corr = Percentage2u32(nattrs.DelayCorr)
-	}
-	if loss > 0 {
-		loss_corr = Percentage2u32(nattrs.LossCorr)
-	}
-	if duplicate > 0 {
-		duplicate_corr = Percentage2u32(nattrs.DuplicateCorr)
-	}
-	// FIXME should validate values(like loss/duplicate are percentages...)
-	latency = time2Tick(latency)
-
-	if nattrs.Limit != 0 {
-		limit = nattrs.Limit
-	}
-	// Jitter is only value if latency is > 0
-	if latency > 0 {
-		jitter = time2Tick(jitter)
-	}
-
-	reorder_prob = Percentage2u32(nattrs.ReorderProb)
-	reorder_corr = Percentage2u32(nattrs.ReorderCorr)
-
-	if reorder_prob > 0 {
-		// ERROR if lantency == 0
-		if gap == 0 {
-			gap = 1
-		}
-	}
-
-	corrupt_prob = Percentage2u32(nattrs.CorruptProb)
-	corrupt_corr = Percentage2u32(nattrs.CorruptCorr)
-
-	return &Netem{
-		QdiscAttrs:    attrs,
-		Latency:       latency,
-		DelayCorr:     delay_corr,
-		Limit:         limit,
-		Loss:          loss,
-		LossCorr:      loss_corr,
-		Gap:           gap,
-		Duplicate:     duplicate,
-		DuplicateCorr: duplicate_corr,
-		Jitter:        jitter,
-		ReorderProb:   reorder_prob,
-		ReorderCorr:   reorder_corr,
-		CorruptProb:   corrupt_prob,
-		CorruptCorr:   corrupt_corr,
-	}
-}
-
 func (qdisc *Netem) Attrs() *QdiscAttrs {
 	return &qdisc.QdiscAttrs
 }

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

@@ -10,24 +10,109 @@ import (
 	"github.com/vishvananda/netlink/nl"
 )
 
+// NOTE function is here because it uses other linux functions
+func NewNetem(attrs QdiscAttrs, nattrs NetemQdiscAttrs) *Netem {
+	var limit uint32 = 1000
+	var lossCorr, delayCorr, duplicateCorr uint32
+	var reorderProb, reorderCorr uint32
+	var corruptProb, corruptCorr uint32
+
+	latency := nattrs.Latency
+	loss := Percentage2u32(nattrs.Loss)
+	gap := nattrs.Gap
+	duplicate := Percentage2u32(nattrs.Duplicate)
+	jitter := nattrs.Jitter
+
+	// Correlation
+	if latency > 0 && jitter > 0 {
+		delayCorr = Percentage2u32(nattrs.DelayCorr)
+	}
+	if loss > 0 {
+		lossCorr = Percentage2u32(nattrs.LossCorr)
+	}
+	if duplicate > 0 {
+		duplicateCorr = Percentage2u32(nattrs.DuplicateCorr)
+	}
+	// FIXME should validate values(like loss/duplicate are percentages...)
+	latency = time2Tick(latency)
+
+	if nattrs.Limit != 0 {
+		limit = nattrs.Limit
+	}
+	// Jitter is only value if latency is > 0
+	if latency > 0 {
+		jitter = time2Tick(jitter)
+	}
+
+	reorderProb = Percentage2u32(nattrs.ReorderProb)
+	reorderCorr = Percentage2u32(nattrs.ReorderCorr)
+
+	if reorderProb > 0 {
+		// ERROR if lantency == 0
+		if gap == 0 {
+			gap = 1
+		}
+	}
+
+	corruptProb = Percentage2u32(nattrs.CorruptProb)
+	corruptCorr = Percentage2u32(nattrs.CorruptCorr)
+
+	return &Netem{
+		QdiscAttrs:    attrs,
+		Latency:       latency,
+		DelayCorr:     delayCorr,
+		Limit:         limit,
+		Loss:          loss,
+		LossCorr:      lossCorr,
+		Gap:           gap,
+		Duplicate:     duplicate,
+		DuplicateCorr: duplicateCorr,
+		Jitter:        jitter,
+		ReorderProb:   reorderProb,
+		ReorderCorr:   reorderCorr,
+		CorruptProb:   corruptProb,
+		CorruptCorr:   corruptCorr,
+	}
+}
+
 // QdiscDel will delete a qdisc from the system.
 // Equivalent to: `tc qdisc del $qdisc`
 func QdiscDel(qdisc Qdisc) error {
-	return qdiscModify(syscall.RTM_DELQDISC, 0, qdisc)
+	return pkgHandle.QdiscDel(qdisc)
+}
+
+// QdiscDel will delete a qdisc from the system.
+// Equivalent to: `tc qdisc del $qdisc`
+func (h *Handle) QdiscDel(qdisc Qdisc) error {
+	return h.qdiscModify(syscall.RTM_DELQDISC, 0, qdisc)
 }
 
 // QdiscChange will change a qdisc in place
 // Equivalent to: `tc qdisc change $qdisc`
 // The parent and handle MUST NOT be changed.
 func QdiscChange(qdisc Qdisc) error {
-	return qdiscModify(syscall.RTM_NEWQDISC, 0, qdisc)
+	return pkgHandle.QdiscChange(qdisc)
+}
+
+// QdiscChange will change a qdisc in place
+// Equivalent to: `tc qdisc change $qdisc`
+// The parent and handle MUST NOT be changed.
+func (h *Handle) QdiscChange(qdisc Qdisc) error {
+	return h.qdiscModify(syscall.RTM_NEWQDISC, 0, qdisc)
 }
 
 // QdiscReplace will replace a qdisc to the system.
 // Equivalent to: `tc qdisc replace $qdisc`
 // The handle MUST change.
 func QdiscReplace(qdisc Qdisc) error {
-	return qdiscModify(
+	return pkgHandle.QdiscReplace(qdisc)
+}
+
+// QdiscReplace will replace a qdisc to the system.
+// Equivalent to: `tc qdisc replace $qdisc`
+// The handle MUST change.
+func (h *Handle) QdiscReplace(qdisc Qdisc) error {
+	return h.qdiscModify(
 		syscall.RTM_NEWQDISC,
 		syscall.NLM_F_CREATE|syscall.NLM_F_REPLACE,
 		qdisc)
@@ -36,14 +121,20 @@ func QdiscReplace(qdisc Qdisc) error {
 // QdiscAdd will add a qdisc to the system.
 // Equivalent to: `tc qdisc add $qdisc`
 func QdiscAdd(qdisc Qdisc) error {
-	return qdiscModify(
+	return pkgHandle.QdiscAdd(qdisc)
+}
+
+// QdiscAdd will add a qdisc to the system.
+// Equivalent to: `tc qdisc add $qdisc`
+func (h *Handle) QdiscAdd(qdisc Qdisc) error {
+	return h.qdiscModify(
 		syscall.RTM_NEWQDISC,
 		syscall.NLM_F_CREATE|syscall.NLM_F_EXCL,
 		qdisc)
 }
 
-func qdiscModify(cmd, flags int, qdisc Qdisc) error {
-	req := nl.NewNetlinkRequest(cmd, flags|syscall.NLM_F_ACK)
+func (h *Handle) qdiscModify(cmd, flags int, qdisc Qdisc) error {
+	req := h.newNetlinkRequest(cmd, flags|syscall.NLM_F_ACK)
 	base := qdisc.Attrs()
 	msg := &nl.TcMsg{
 		Family:  nl.FAMILY_ALL,
@@ -139,11 +230,18 @@ func qdiscPayload(req *nl.NetlinkRequest, qdisc Qdisc) error {
 // Equivalent to: `tc qdisc show`.
 // The list can be filtered by link.
 func QdiscList(link Link) ([]Qdisc, error) {
-	req := nl.NewNetlinkRequest(syscall.RTM_GETQDISC, syscall.NLM_F_DUMP)
+	return pkgHandle.QdiscList(link)
+}
+
+// QdiscList gets a list of qdiscs in the system.
+// Equivalent to: `tc qdisc show`.
+// The list can be filtered by link.
+func (h *Handle) QdiscList(link Link) ([]Qdisc, error) {
+	req := h.newNetlinkRequest(syscall.RTM_GETQDISC, syscall.NLM_F_DUMP)
 	index := int32(0)
 	if link != nil {
 		base := link.Attrs()
-		ensureIndex(base)
+		h.ensureIndex(base)
 		index = int32(base.Index)
 	}
 	msg := &nl.TcMsg{
@@ -334,9 +432,9 @@ const (
 )
 
 var (
-	tickInUsec  float64 = 0.0
-	clockFactor float64 = 0.0
-	hz          float64 = 0.0
+	tickInUsec  float64
+	clockFactor float64
+	hz          float64
 )
 
 func initClock() {

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

@@ -3,43 +3,37 @@ package netlink
 import (
 	"fmt"
 	"net"
-	"syscall"
 )
 
 // Scope is an enum representing a route scope.
 type Scope uint8
 
-const (
-	SCOPE_UNIVERSE Scope = syscall.RT_SCOPE_UNIVERSE
-	SCOPE_SITE     Scope = syscall.RT_SCOPE_SITE
-	SCOPE_LINK     Scope = syscall.RT_SCOPE_LINK
-	SCOPE_HOST     Scope = syscall.RT_SCOPE_HOST
-	SCOPE_NOWHERE  Scope = syscall.RT_SCOPE_NOWHERE
-)
-
 type NextHopFlag int
 
-const (
-	FLAG_ONLINK    NextHopFlag = syscall.RTNH_F_ONLINK
-	FLAG_PERVASIVE NextHopFlag = syscall.RTNH_F_PERVASIVE
-)
-
-// Route represents a netlink route. A route is associated with a link,
-// has a destination network, an optional source ip, and optional
-// gateway. Advanced route parameters and non-main routing tables are
-// currently not supported.
+// Route represents a netlink route.
 type Route struct {
-	LinkIndex int
-	Scope     Scope
-	Dst       *net.IPNet
-	Src       net.IP
-	Gw        net.IP
-	Flags     int
+	LinkIndex  int
+	ILinkIndex int
+	Scope      Scope
+	Dst        *net.IPNet
+	Src        net.IP
+	Gw         net.IP
+	MultiPath  []*NexthopInfo
+	Protocol   int
+	Priority   int
+	Table      int
+	Type       int
+	Tos        int
+	Flags      int
 }
 
 func (r Route) String() string {
-	return fmt.Sprintf("{Ifindex: %d Dst: %s Src: %s Gw: %s Flags: %s}", r.LinkIndex, r.Dst,
-		r.Src, r.Gw, r.ListFlags())
+	if len(r.MultiPath) > 0 {
+		return fmt.Sprintf("{Dst: %s Src: %s Gw: %s Flags: %s Table: %d}", r.Dst,
+			r.Src, r.MultiPath, r.ListFlags(), r.Table)
+	}
+	return fmt.Sprintf("{Ifindex: %d Dst: %s Src: %s Gw: %s Flags: %s Table: %d}", r.LinkIndex, r.Dst,
+		r.Src, r.Gw, r.ListFlags(), r.Table)
 }
 
 func (r *Route) SetFlag(flag NextHopFlag) {
@@ -55,23 +49,18 @@ type flagString struct {
 	s string
 }
 
-var testFlags = []flagString{
-	flagString{f: FLAG_ONLINK, s: "onlink"},
-	flagString{f: FLAG_PERVASIVE, s: "pervasive"},
-}
-
-func (r *Route) ListFlags() []string {
-	var flags []string
-	for _, tf := range testFlags {
-		if r.Flags&int(tf.f) != 0 {
-			flags = append(flags, tf.s)
-		}
-	}
-	return flags
-}
-
 // RouteUpdate is sent when a route changes - type is RTM_NEWROUTE or RTM_DELROUTE
 type RouteUpdate struct {
 	Type uint16
 	Route
 }
+
+type NexthopInfo struct {
+	LinkIndex int
+	Hops      int
+	Gw        net.IP
+}
+
+func (n *NexthopInfo) String() string {
+	return fmt.Sprintf("{Ifindex: %d Weight: %d, Gw: %s}", n.LinkIndex, n.Hops+1, n.Gw)
+}

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

@@ -10,27 +10,78 @@ import (
 
 // RtAttr is shared so it is in netlink_linux.go
 
+const (
+	SCOPE_UNIVERSE Scope = syscall.RT_SCOPE_UNIVERSE
+	SCOPE_SITE     Scope = syscall.RT_SCOPE_SITE
+	SCOPE_LINK     Scope = syscall.RT_SCOPE_LINK
+	SCOPE_HOST     Scope = syscall.RT_SCOPE_HOST
+	SCOPE_NOWHERE  Scope = syscall.RT_SCOPE_NOWHERE
+)
+
+const (
+	RT_FILTER_PROTOCOL uint64 = 1 << (1 + iota)
+	RT_FILTER_SCOPE
+	RT_FILTER_TYPE
+	RT_FILTER_TOS
+	RT_FILTER_IIF
+	RT_FILTER_OIF
+	RT_FILTER_DST
+	RT_FILTER_SRC
+	RT_FILTER_GW
+	RT_FILTER_TABLE
+)
+
+const (
+	FLAG_ONLINK    NextHopFlag = syscall.RTNH_F_ONLINK
+	FLAG_PERVASIVE NextHopFlag = syscall.RTNH_F_PERVASIVE
+)
+
+var testFlags = []flagString{
+	{f: FLAG_ONLINK, s: "onlink"},
+	{f: FLAG_PERVASIVE, s: "pervasive"},
+}
+
+func (r *Route) ListFlags() []string {
+	var flags []string
+	for _, tf := range testFlags {
+		if r.Flags&int(tf.f) != 0 {
+			flags = append(flags, tf.s)
+		}
+	}
+	return flags
+}
+
 // RouteAdd will add a route to the system.
 // Equivalent to: `ip route add $route`
 func RouteAdd(route *Route) error {
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWROUTE, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
-	return routeHandle(route, req, nl.NewRtMsg())
+	return pkgHandle.RouteAdd(route)
+}
+
+// RouteAdd will add a route to the system.
+// Equivalent to: `ip route add $route`
+func (h *Handle) RouteAdd(route *Route) error {
+	req := h.newNetlinkRequest(syscall.RTM_NEWROUTE, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	return h.routeHandle(route, req, nl.NewRtMsg())
 }
 
 // RouteDel will delete a route from the system.
 // Equivalent to: `ip route del $route`
 func RouteDel(route *Route) error {
-	req := nl.NewNetlinkRequest(syscall.RTM_DELROUTE, syscall.NLM_F_ACK)
-	return routeHandle(route, req, nl.NewRtDelMsg())
+	return pkgHandle.RouteDel(route)
 }
 
-func routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg) error {
+// RouteDel will delete a route from the system.
+// Equivalent to: `ip route del $route`
+func (h *Handle) RouteDel(route *Route) error {
+	req := h.newNetlinkRequest(syscall.RTM_DELROUTE, syscall.NLM_F_ACK)
+	return h.routeHandle(route, req, nl.NewRtDelMsg())
+}
+
+func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg) error {
 	if (route.Dst == nil || route.Dst.IP == nil) && route.Src == nil && route.Gw == nil {
 		return fmt.Errorf("one of Dst.IP, Src, or Gw must not be nil")
 	}
 
-	msg.Scope = uint8(route.Scope)
-	msg.Flags = uint32(route.Flags)
 	family := -1
 	var rtAttrs []*nl.RtAttr
 
@@ -79,8 +130,66 @@ func routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg) error {
 		rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_GATEWAY, gwData))
 	}
 
-	msg.Family = uint8(family)
+	if len(route.MultiPath) > 0 {
+		buf := []byte{}
+		for _, nh := range route.MultiPath {
+			rtnh := &nl.RtNexthop{
+				RtNexthop: syscall.RtNexthop{
+					Hops:    uint8(nh.Hops),
+					Ifindex: int32(nh.LinkIndex),
+					Len:     uint16(syscall.SizeofRtNexthop),
+				},
+			}
+			var gwData []byte
+			if nh.Gw != nil {
+				gwFamily := nl.GetIPFamily(nh.Gw)
+				if family != -1 && family != gwFamily {
+					return fmt.Errorf("gateway, source, and destination ip are not the same IP family")
+				}
+				var gw *nl.RtAttr
+				if gwFamily == FAMILY_V4 {
+					gw = nl.NewRtAttr(syscall.RTA_GATEWAY, []byte(nh.Gw.To4()))
+				} else {
+					gw = nl.NewRtAttr(syscall.RTA_GATEWAY, []byte(nh.Gw.To16()))
+				}
+				gwData := gw.Serialize()
+				rtnh.Len += uint16(len(gwData))
+			}
+			buf = append(buf, rtnh.Serialize()...)
+			buf = append(buf, gwData...)
+		}
+		rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_MULTIPATH, buf))
+	}
 
+	if route.Table > 0 {
+		if route.Table >= 256 {
+			msg.Table = syscall.RT_TABLE_UNSPEC
+			b := make([]byte, 4)
+			native.PutUint32(b, uint32(route.Table))
+			rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_TABLE, b))
+		} else {
+			msg.Table = uint8(route.Table)
+		}
+	}
+
+	if route.Priority > 0 {
+		b := make([]byte, 4)
+		native.PutUint32(b, uint32(route.Priority))
+		rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_PRIORITY, b))
+	}
+	if route.Tos > 0 {
+		msg.Tos = uint8(route.Tos)
+	}
+	if route.Protocol > 0 {
+		msg.Protocol = uint8(route.Protocol)
+	}
+	if route.Type > 0 {
+		msg.Type = uint8(route.Type)
+	}
+
+	msg.Flags = uint32(route.Flags)
+	msg.Scope = uint8(route.Scope)
+	msg.Family = uint8(family)
 	req.AddData(msg)
 	for _, attr := range rtAttrs {
 		req.AddData(attr)
@@ -102,61 +211,108 @@ func routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg) error {
 // Equivalent to: `ip route show`.
 // The list can be filtered by link and ip family.
 func RouteList(link Link, family int) ([]Route, error) {
-	req := nl.NewNetlinkRequest(syscall.RTM_GETROUTE, syscall.NLM_F_DUMP)
-	msg := nl.NewIfInfomsg(family)
-	req.AddData(msg)
+	return pkgHandle.RouteList(link, family)
+}
+
+// RouteList gets a list of routes in the system.
+// Equivalent to: `ip route show`.
+// The list can be filtered by link and ip family.
+func (h *Handle) RouteList(link Link, family int) ([]Route, error) {
+	var routeFilter *Route
+	if link != nil {
+		routeFilter = &Route{
+			LinkIndex: link.Attrs().Index,
+		}
+	}
+	return h.RouteListFiltered(family, routeFilter, RT_FILTER_OIF)
+}
+
+// RouteListFiltered gets a list of routes in the system filtered with specified rules.
+// All rules must be defined in RouteFilter struct
+func RouteListFiltered(family int, filter *Route, filterMask uint64) ([]Route, error) {
+	return pkgHandle.RouteListFiltered(family, filter, filterMask)
+}
+
+// RouteListFiltered gets a list of routes in the system filtered with specified rules.
+// All rules must be defined in RouteFilter struct
+func (h *Handle) RouteListFiltered(family int, filter *Route, filterMask uint64) ([]Route, error) {
+	req := h.newNetlinkRequest(syscall.RTM_GETROUTE, syscall.NLM_F_DUMP)
+	infmsg := nl.NewIfInfomsg(family)
+	req.AddData(infmsg)
 
 	msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWROUTE)
 	if err != nil {
 		return nil, err
 	}
 
-	index := 0
-	if link != nil {
-		base := link.Attrs()
-		ensureIndex(base)
-		index = base.Index
-	}
-
 	var res []Route
 	for _, m := range msgs {
 		msg := nl.DeserializeRtMsg(m)
-
 		if msg.Flags&syscall.RTM_F_CLONED != 0 {
 			// Ignore cloned routes
 			continue
 		}
-
 		if msg.Table != syscall.RT_TABLE_MAIN {
-			// Ignore non-main tables
-			continue
+			if filter == nil || filter != nil && filterMask&RT_FILTER_TABLE == 0 {
+				// Ignore non-main tables
+				continue
+			}
 		}
-
 		route, err := deserializeRoute(m)
 		if err != nil {
 			return nil, err
 		}
-
-		if link != nil && route.LinkIndex != index {
-			// Ignore routes from other interfaces
-			continue
+		if filter != nil {
+			switch {
+			case filterMask&RT_FILTER_TABLE != 0 && route.Table != filter.Table:
+				continue
+			case filterMask&RT_FILTER_PROTOCOL != 0 && route.Protocol != filter.Protocol:
+				continue
+			case filterMask&RT_FILTER_SCOPE != 0 && route.Scope != filter.Scope:
+				continue
+			case filterMask&RT_FILTER_TYPE != 0 && route.Type != filter.Type:
+				continue
+			case filterMask&RT_FILTER_TOS != 0 && route.Tos != filter.Tos:
+				continue
+			case filterMask&RT_FILTER_OIF != 0 && route.LinkIndex != filter.LinkIndex:
+				continue
+			case filterMask&RT_FILTER_IIF != 0 && route.ILinkIndex != filter.ILinkIndex:
+				continue
+			case filterMask&RT_FILTER_GW != 0 && !route.Gw.Equal(filter.Gw):
+				continue
+			case filterMask&RT_FILTER_SRC != 0 && !route.Src.Equal(filter.Src):
+				continue
+			case filterMask&RT_FILTER_DST != 0 && filter.Dst != nil:
+				if route.Dst == nil {
+					continue
+				}
+				aMaskLen, aMaskBits := route.Dst.Mask.Size()
+				bMaskLen, bMaskBits := filter.Dst.Mask.Size()
+				if !(route.Dst.IP.Equal(filter.Dst.IP) && aMaskLen == bMaskLen && aMaskBits == bMaskBits) {
+					continue
+				}
+			}
 		}
 		res = append(res, route)
 	}
-
 	return res, nil
 }
 
 // deserializeRoute decodes a binary netlink message into a Route struct
 func deserializeRoute(m []byte) (Route, error) {
-	route := Route{}
 	msg := nl.DeserializeRtMsg(m)
 	attrs, err := nl.ParseRouteAttr(m[msg.Len():])
 	if err != nil {
-		return route, err
+		return Route{}, err
+	}
+	route := Route{
+		Scope:    Scope(msg.Scope),
+		Protocol: int(msg.Protocol),
+		Table:    int(msg.Table),
+		Type:     int(msg.Type),
+		Tos:      int(msg.Tos),
+		Flags:    int(msg.Flags),
 	}
-	route.Scope = Scope(msg.Scope)
-	route.Flags = int(msg.Flags)
 
 	native := nl.NativeEndian()
 	for _, attr := range attrs {
@@ -171,8 +327,47 @@ func deserializeRoute(m []byte) (Route, error) {
 				Mask: net.CIDRMask(int(msg.Dst_len), 8*len(attr.Value)),
 			}
 		case syscall.RTA_OIF:
-			routeIndex := int(native.Uint32(attr.Value[0:4]))
-			route.LinkIndex = routeIndex
+			route.LinkIndex = int(native.Uint32(attr.Value[0:4]))
+		case syscall.RTA_IIF:
+			route.ILinkIndex = int(native.Uint32(attr.Value[0:4]))
+		case syscall.RTA_PRIORITY:
+			route.Priority = int(native.Uint32(attr.Value[0:4]))
+		case syscall.RTA_TABLE:
+			route.Table = int(native.Uint32(attr.Value[0:4]))
+		case syscall.RTA_MULTIPATH:
+			parseRtNexthop := func(value []byte) (*NexthopInfo, []byte, error) {
+				if len(value) < syscall.SizeofRtNexthop {
+					return nil, nil, fmt.Errorf("Lack of bytes")
+				}
+				nh := nl.DeserializeRtNexthop(value)
+				if len(value) < int(nh.RtNexthop.Len) {
+					return nil, nil, fmt.Errorf("Lack of bytes")
+				}
+				info := &NexthopInfo{
+					LinkIndex: int(nh.RtNexthop.Ifindex),
+					Hops:      int(nh.RtNexthop.Hops),
+				}
+				attrs, err := nl.ParseRouteAttr(value[syscall.SizeofRtNexthop:int(nh.RtNexthop.Len)])
+				if err != nil {
+					return nil, nil, err
+				}
+				for _, attr := range attrs {
+					switch attr.Attr.Type {
+					case syscall.RTA_GATEWAY:
+						info.Gw = net.IP(attr.Value)
+					}
+				}
+				return info, value[int(nh.RtNexthop.Len):], nil
+			}
+			rest := attr.Value
+			for len(rest) > 0 {
+				info, buf, err := parseRtNexthop(rest)
+				if err != nil {
+					return route, err
+				}
+				route.MultiPath = append(route.MultiPath, info)
+				rest = buf
+			}
 		}
 	}
 	return route, nil
@@ -181,7 +376,13 @@ func deserializeRoute(m []byte) (Route, error) {
 // RouteGet gets a route to a specific destination from the host system.
 // Equivalent to: 'ip route get'.
 func RouteGet(destination net.IP) ([]Route, error) {
-	req := nl.NewNetlinkRequest(syscall.RTM_GETROUTE, syscall.NLM_F_REQUEST)
+	return pkgHandle.RouteGet(destination)
+}
+
+// RouteGet gets a route to a specific destination from the host system.
+// Equivalent to: 'ip route get'.
+func (h *Handle) RouteGet(destination net.IP) ([]Route, error) {
+	req := h.newNetlinkRequest(syscall.RTM_GETROUTE, syscall.NLM_F_REQUEST)
 	family := nl.GetIPFamily(destination)
 	var destinationData []byte
 	var bitlen uint8

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

@@ -18,17 +18,19 @@ func TestRouteAddDel(t *testing.T) {
 	}
 
 	// bring the interface up
-	if err = LinkSetUp(link); err != nil {
+	if err := LinkSetUp(link); err != nil {
 		t.Fatal(err)
 	}
 
 	// add a gateway route
-	_, dst, err := net.ParseCIDR("192.168.0.0/24")
+	dst := &net.IPNet{
+		IP:   net.IPv4(192, 168, 0, 0),
+		Mask: net.CIDRMask(24, 32),
+	}
 
-	ip := net.ParseIP("127.1.1.1")
+	ip := net.IPv4(127, 1, 1, 1)
 	route := Route{LinkIndex: link.Attrs().Index, Dst: dst, Src: ip}
-	err = RouteAdd(&route)
-	if err != nil {
+	if err := RouteAdd(&route); err != nil {
 		t.Fatal(err)
 	}
 	routes, err := RouteList(link, FAMILY_V4)
@@ -36,10 +38,10 @@ func TestRouteAddDel(t *testing.T) {
 		t.Fatal(err)
 	}
 	if len(routes) != 1 {
-		t.Fatal("Link not added properly")
+		t.Fatal("Route not added properly")
 	}
 
-	dstIP := net.ParseIP("192.168.0.42")
+	dstIP := net.IPv4(192, 168, 0, 42)
 	routeToDstIP, err := RouteGet(dstIP)
 	if err != nil {
 		t.Fatal(err)
@@ -48,12 +50,9 @@ func TestRouteAddDel(t *testing.T) {
 	if len(routeToDstIP) == 0 {
 		t.Fatal("Default route not present")
 	}
-
-	err = RouteDel(&route)
-	if err != nil {
+	if err := RouteDel(&route); err != nil {
 		t.Fatal(err)
 	}
-
 	routes, err = RouteList(link, FAMILY_V4)
 	if err != nil {
 		t.Fatal(err)
@@ -122,25 +121,127 @@ func TestRouteSubscribe(t *testing.T) {
 	}
 
 	// add a gateway route
-	_, dst, err := net.ParseCIDR("192.168.0.0/24")
+	dst := &net.IPNet{
+		IP:   net.IPv4(192, 168, 0, 0),
+		Mask: net.CIDRMask(24, 32),
+	}
 
-	ip := net.ParseIP("127.1.1.1")
+	ip := net.IPv4(127, 1, 1, 1)
 	route := Route{LinkIndex: link.Attrs().Index, Dst: dst, Src: ip}
-	err = RouteAdd(&route)
-	if err != nil {
+	if err := RouteAdd(&route); err != nil {
 		t.Fatal(err)
 	}
 
 	if !expectRouteUpdate(ch, syscall.RTM_NEWROUTE, dst.IP) {
 		t.Fatal("Add update not received as expected")
 	}
+	if err := RouteDel(&route); err != nil {
+		t.Fatal(err)
+	}
+	if !expectRouteUpdate(ch, syscall.RTM_DELROUTE, dst.IP) {
+		t.Fatal("Del update not received as expected")
+	}
+}
+
+func TestRouteExtraFields(t *testing.T) {
+	tearDown := setUpNetlinkTest(t)
+	defer tearDown()
 
-	err = RouteDel(&route)
+	// get loopback interface
+	link, err := LinkByName("lo")
 	if err != nil {
 		t.Fatal(err)
 	}
+	// bring the interface up
+	if err = LinkSetUp(link); err != nil {
+		t.Fatal(err)
+	}
 
-	if !expectRouteUpdate(ch, syscall.RTM_DELROUTE, dst.IP) {
-		t.Fatal("Del update not received as expected")
+	// add a gateway route
+	dst := &net.IPNet{
+		IP:   net.IPv4(1, 1, 1, 1),
+		Mask: net.CIDRMask(32, 32),
+	}
+
+	src := net.IPv4(127, 3, 3, 3)
+	route := Route{
+		LinkIndex: link.Attrs().Index,
+		Dst:       dst,
+		Src:       src,
+		Scope:     syscall.RT_SCOPE_LINK,
+		Priority:  13,
+		Table:     syscall.RT_TABLE_MAIN,
+		Type:      syscall.RTN_UNICAST,
+		Tos:       14,
+	}
+	if err := RouteAdd(&route); err != nil {
+		t.Fatal(err)
+	}
+	routes, err := RouteListFiltered(FAMILY_V4, &Route{
+		Dst:   dst,
+		Src:   src,
+		Scope: syscall.RT_SCOPE_LINK,
+		Table: syscall.RT_TABLE_MAIN,
+		Type:  syscall.RTN_UNICAST,
+		Tos:   14,
+	}, RT_FILTER_DST|RT_FILTER_SRC|RT_FILTER_SCOPE|RT_FILTER_TABLE|RT_FILTER_TYPE|RT_FILTER_TOS)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(routes) != 1 {
+		t.Fatal("Route not added properly")
+	}
+
+	if routes[0].Scope != syscall.RT_SCOPE_LINK {
+		t.Fatal("Invalid Scope. Route not added properly")
+	}
+	if routes[0].Priority != 13 {
+		t.Fatal("Invalid Priority. Route not added properly")
+	}
+	if routes[0].Table != syscall.RT_TABLE_MAIN {
+		t.Fatal("Invalid Scope. Route not added properly")
+	}
+	if routes[0].Type != syscall.RTN_UNICAST {
+		t.Fatal("Invalid Type. Route not added properly")
+	}
+	if routes[0].Tos != 14 {
+		t.Fatal("Invalid Tos. Route not added properly")
+	}
+}
+
+func TestRouteMultiPath(t *testing.T) {
+	tearDown := setUpNetlinkTest(t)
+	defer tearDown()
+
+	// get loopback interface
+	link, err := LinkByName("lo")
+	if err != nil {
+		t.Fatal(err)
+	}
+	// bring the interface up
+	if err = LinkSetUp(link); err != nil {
+		t.Fatal(err)
+	}
+
+	// add a gateway route
+	dst := &net.IPNet{
+		IP:   net.IPv4(192, 168, 0, 0),
+		Mask: net.CIDRMask(24, 32),
+	}
+
+	idx := link.Attrs().Index
+	route := Route{Dst: dst, MultiPath: []*NexthopInfo{&NexthopInfo{LinkIndex: idx}, &NexthopInfo{LinkIndex: idx}}}
+	if err := RouteAdd(&route); err != nil {
+		t.Fatal(err)
+	}
+	routes, err := RouteList(nil, FAMILY_V4)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(routes) != 1 {
+		t.Fatal("MultiPath Route not added properly")
+	}
+	if len(routes[0].MultiPath) != 2 {
+		t.Fatal("MultiPath Route not added properly")
 	}
 }

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

@@ -0,0 +1,7 @@
+// +build !linux
+
+package netlink
+
+func (r *Route) ListFlags() []string {
+	return []string{}
+}

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

@@ -0,0 +1,40 @@
+package netlink
+
+import (
+	"fmt"
+	"net"
+)
+
+// Rule represents a netlink rule.
+type Rule struct {
+	Priority          int
+	Table             int
+	Mark              int
+	Mask              int
+	TunID             uint
+	Goto              int
+	Src               *net.IPNet
+	Dst               *net.IPNet
+	Flow              int
+	IifName           string
+	OifName           string
+	SuppressIfgroup   int
+	SuppressPrefixlen int
+}
+
+func (r Rule) String() string {
+	return fmt.Sprintf("ip rule %d: from %s table %d", r.Priority, r.Src, r.Table)
+}
+
+// NewRule return empty rules.
+func NewRule() *Rule {
+	return &Rule{
+		SuppressIfgroup:   -1,
+		SuppressPrefixlen: -1,
+		Priority:          -1,
+		Mark:              -1,
+		Mask:              -1,
+		Goto:              -1,
+		Flow:              -1,
+	}
+}

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

@@ -0,0 +1,215 @@
+package netlink
+
+import (
+	"fmt"
+	"net"
+	"syscall"
+
+	"github.com/vishvananda/netlink/nl"
+)
+
+// RuleAdd adds a rule to the system.
+// Equivalent to: ip rule add
+func RuleAdd(rule *Rule) error {
+	return pkgHandle.RuleAdd(rule)
+}
+
+// RuleAdd adds a rule to the system.
+// Equivalent to: ip rule add
+func (h *Handle) RuleAdd(rule *Rule) error {
+	req := h.newNetlinkRequest(syscall.RTM_NEWRULE, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	return ruleHandle(rule, req)
+}
+
+// RuleDel deletes a rule from the system.
+// Equivalent to: ip rule del
+func RuleDel(rule *Rule) error {
+	return pkgHandle.RuleDel(rule)
+}
+
+// RuleDel deletes a rule from the system.
+// Equivalent to: ip rule del
+func (h *Handle) RuleDel(rule *Rule) error {
+	req := h.newNetlinkRequest(syscall.RTM_DELRULE, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	return ruleHandle(rule, req)
+}
+
+func ruleHandle(rule *Rule, req *nl.NetlinkRequest) error {
+	msg := nl.NewRtMsg()
+	msg.Family = syscall.AF_INET
+	var dstFamily uint8
+
+	var rtAttrs []*nl.RtAttr
+	if rule.Dst != nil && rule.Dst.IP != nil {
+		dstLen, _ := rule.Dst.Mask.Size()
+		msg.Dst_len = uint8(dstLen)
+		msg.Family = uint8(nl.GetIPFamily(rule.Dst.IP))
+		dstFamily = msg.Family
+		var dstData []byte
+		if msg.Family == syscall.AF_INET {
+			dstData = rule.Dst.IP.To4()
+		} else {
+			dstData = rule.Dst.IP.To16()
+		}
+		rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_DST, dstData))
+	}
+
+	if rule.Src != nil && rule.Src.IP != nil {
+		msg.Family = uint8(nl.GetIPFamily(rule.Src.IP))
+		if dstFamily != 0 && dstFamily != msg.Family {
+			return fmt.Errorf("source and destination ip are not the same IP family")
+		}
+		srcLen, _ := rule.Src.Mask.Size()
+		msg.Src_len = uint8(srcLen)
+		var srcData []byte
+		if msg.Family == syscall.AF_INET {
+			srcData = rule.Src.IP.To4()
+		} else {
+			srcData = rule.Src.IP.To16()
+		}
+		rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_SRC, srcData))
+	}
+
+	if rule.Table >= 0 {
+		msg.Table = uint8(rule.Table)
+		if rule.Table >= 256 {
+			msg.Table = syscall.RT_TABLE_UNSPEC
+		}
+	}
+
+	req.AddData(msg)
+	for i := range rtAttrs {
+		req.AddData(rtAttrs[i])
+	}
+
+	var (
+		b      = make([]byte, 4)
+		native = nl.NativeEndian()
+	)
+
+	if rule.Priority >= 0 {
+		native.PutUint32(b, uint32(rule.Priority))
+		req.AddData(nl.NewRtAttr(nl.FRA_PRIORITY, b))
+	}
+	if rule.Mark >= 0 {
+		native.PutUint32(b, uint32(rule.Mark))
+		req.AddData(nl.NewRtAttr(nl.FRA_FWMARK, b))
+	}
+	if rule.Mask >= 0 {
+		native.PutUint32(b, uint32(rule.Mask))
+		req.AddData(nl.NewRtAttr(nl.FRA_FWMASK, b))
+	}
+	if rule.Flow >= 0 {
+		native.PutUint32(b, uint32(rule.Flow))
+		req.AddData(nl.NewRtAttr(nl.FRA_FLOW, b))
+	}
+	if rule.TunID > 0 {
+		native.PutUint32(b, uint32(rule.TunID))
+		req.AddData(nl.NewRtAttr(nl.FRA_TUN_ID, b))
+	}
+	if rule.Table >= 256 {
+		native.PutUint32(b, uint32(rule.Table))
+		req.AddData(nl.NewRtAttr(nl.FRA_TABLE, b))
+	}
+	if msg.Table > 0 {
+		if rule.SuppressPrefixlen >= 0 {
+			native.PutUint32(b, uint32(rule.SuppressPrefixlen))
+			req.AddData(nl.NewRtAttr(nl.FRA_SUPPRESS_PREFIXLEN, b))
+		}
+		if rule.SuppressIfgroup >= 0 {
+			native.PutUint32(b, uint32(rule.SuppressIfgroup))
+			req.AddData(nl.NewRtAttr(nl.FRA_SUPPRESS_IFGROUP, b))
+		}
+	}
+	if rule.IifName != "" {
+		req.AddData(nl.NewRtAttr(nl.FRA_IIFNAME, []byte(rule.IifName)))
+	}
+	if rule.OifName != "" {
+		req.AddData(nl.NewRtAttr(nl.FRA_OIFNAME, []byte(rule.OifName)))
+	}
+	if rule.Goto >= 0 {
+		msg.Type = nl.FR_ACT_NOP
+		native.PutUint32(b, uint32(rule.Goto))
+		req.AddData(nl.NewRtAttr(nl.FRA_GOTO, b))
+	}
+
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
+// RuleList lists rules in the system.
+// Equivalent to: ip rule list
+func RuleList(family int) ([]Rule, error) {
+	return pkgHandle.RuleList(family)
+}
+
+// RuleList lists rules in the system.
+// Equivalent to: ip rule list
+func (h *Handle) RuleList(family int) ([]Rule, error) {
+	req := h.newNetlinkRequest(syscall.RTM_GETRULE, syscall.NLM_F_DUMP|syscall.NLM_F_REQUEST)
+	msg := nl.NewIfInfomsg(family)
+	req.AddData(msg)
+
+	msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWRULE)
+	if err != nil {
+		return nil, err
+	}
+
+	native := nl.NativeEndian()
+	var res = make([]Rule, 0)
+	for i := range msgs {
+		msg := nl.DeserializeRtMsg(msgs[i])
+		attrs, err := nl.ParseRouteAttr(msgs[i][msg.Len():])
+		if err != nil {
+			return nil, err
+		}
+
+		rule := NewRule()
+
+		for j := range attrs {
+			switch attrs[j].Attr.Type {
+			case syscall.RTA_TABLE:
+				rule.Table = int(native.Uint32(attrs[j].Value[0:4]))
+			case nl.FRA_SRC:
+				rule.Src = &net.IPNet{
+					IP:   attrs[j].Value,
+					Mask: net.CIDRMask(int(msg.Src_len), 8*len(attrs[j].Value)),
+				}
+			case nl.FRA_DST:
+				rule.Dst = &net.IPNet{
+					IP:   attrs[j].Value,
+					Mask: net.CIDRMask(int(msg.Dst_len), 8*len(attrs[j].Value)),
+				}
+			case nl.FRA_FWMARK:
+				rule.Mark = int(native.Uint32(attrs[j].Value[0:4]))
+			case nl.FRA_FWMASK:
+				rule.Mask = int(native.Uint32(attrs[j].Value[0:4]))
+			case nl.FRA_TUN_ID:
+				rule.TunID = uint(native.Uint64(attrs[j].Value[0:4]))
+			case nl.FRA_IIFNAME:
+				rule.IifName = string(attrs[j].Value[:len(attrs[j].Value)-1])
+			case nl.FRA_OIFNAME:
+				rule.OifName = string(attrs[j].Value[:len(attrs[j].Value)-1])
+			case nl.FRA_SUPPRESS_PREFIXLEN:
+				i := native.Uint32(attrs[j].Value[0:4])
+				if i != 0xffffffff {
+					rule.SuppressPrefixlen = int(i)
+				}
+			case nl.FRA_SUPPRESS_IFGROUP:
+				i := native.Uint32(attrs[j].Value[0:4])
+				if i != 0xffffffff {
+					rule.SuppressIfgroup = int(i)
+				}
+			case nl.FRA_FLOW:
+				rule.Flow = int(native.Uint32(attrs[j].Value[0:4]))
+			case nl.FRA_GOTO:
+				rule.Goto = int(native.Uint32(attrs[j].Value[0:4]))
+			case nl.FRA_PRIORITY:
+				rule.Priority = int(native.Uint32(attrs[j].Value[0:4]))
+			}
+		}
+		res = append(res, *rule)
+	}
+
+	return res, nil
+}

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

@@ -0,0 +1,66 @@
+package netlink
+
+import (
+	"net"
+	"syscall"
+	"testing"
+)
+
+func TestRuleAddDel(t *testing.T) {
+	srcNet := &net.IPNet{IP: net.IPv4(172, 16, 0, 1), Mask: net.CIDRMask(16, 32)}
+	dstNet := &net.IPNet{IP: net.IPv4(172, 16, 1, 1), Mask: net.CIDRMask(24, 32)}
+
+	rulesBegin, err := RuleList(syscall.AF_INET)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	rule := NewRule()
+	rule.Table = syscall.RT_TABLE_MAIN
+	rule.Src = srcNet
+	rule.Dst = dstNet
+	rule.Priority = 5
+	rule.OifName = "lo"
+	rule.IifName = "lo"
+	if err := RuleAdd(rule); err != nil {
+		t.Fatal(err)
+	}
+
+	rules, err := RuleList(syscall.AF_INET)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if len(rules) != len(rulesBegin)+1 {
+		t.Fatal("Rule not added properly")
+	}
+
+	// find this rule
+	var found bool
+	for i := range rules {
+		if rules[i].Table == rule.Table &&
+			rules[i].Src != nil && rules[i].Src.String() == srcNet.String() &&
+			rules[i].Dst != nil && rules[i].Dst.String() == dstNet.String() &&
+			rules[i].OifName == rule.OifName &&
+			rules[i].Priority == rule.Priority &&
+			rules[i].IifName == rule.IifName {
+			found = true
+		}
+	}
+	if !found {
+		t.Fatal("Rule has diffrent options than one added")
+	}
+
+	if err := RuleDel(rule); err != nil {
+		t.Fatal(err)
+	}
+
+	rulesEnd, err := RuleList(syscall.AF_INET)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if len(rulesEnd) != len(rulesBegin) {
+		t.Fatal("Rule not removed properly")
+	}
+}

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

@@ -13,7 +13,7 @@ const (
 	XFRM_PROTO_ESP       Proto = syscall.IPPROTO_ESP
 	XFRM_PROTO_AH        Proto = syscall.IPPROTO_AH
 	XFRM_PROTO_HAO       Proto = syscall.IPPROTO_DSTOPTS
-	XFRM_PROTO_COMP      Proto = syscall.IPPROTO_COMP
+	XFRM_PROTO_COMP      Proto = 0x6c // NOTE not defined on darwin
 	XFRM_PROTO_IPSEC_ANY Proto = syscall.IPPROTO_RAW
 )
 
@@ -62,3 +62,13 @@ func (m Mode) String() string {
 	}
 	return fmt.Sprintf("%d", m)
 }
+
+// XfrmMark represents the mark associated to the state or policy
+type XfrmMark struct {
+	Value uint32
+	Mask  uint32
+}
+
+func (m *XfrmMark) String() string {
+	return fmt.Sprintf("(0x%x,0x%x)", m.Value, m.Mask)
+}

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

@@ -43,17 +43,32 @@ type XfrmPolicyTmpl struct {
 	Src   net.IP
 	Proto Proto
 	Mode  Mode
+	Spi   int
 	Reqid int
 }
 
+func (t XfrmPolicyTmpl) String() string {
+	return fmt.Sprintf("{Dst: %v, Src: %v, Proto: %s, Mode: %s, Spi: 0x%x, Reqid: 0x%x}",
+		t.Dst, t.Src, t.Proto, t.Mode, t.Spi, t.Reqid)
+}
+
 // XfrmPolicy represents an ipsec policy. It represents the overlay network
 // and has a list of XfrmPolicyTmpls representing the base addresses of
 // the policy.
 type XfrmPolicy struct {
 	Dst      *net.IPNet
 	Src      *net.IPNet
+	Proto    Proto
+	DstPort  int
+	SrcPort  int
 	Dir      Dir
 	Priority int
 	Index    int
+	Mark     *XfrmMark
 	Tmpls    []XfrmPolicyTmpl
 }
+
+func (p XfrmPolicy) String() string {
+	return fmt.Sprintf("{Dst: %v, Src: %v, Proto: %s, DstPort: %d, SrcPort: %d, Dir: %s, Priority: %d, Index: %d, Mark: %s, Tmpls: %s}",
+		p.Dst, p.Src, p.Proto, p.DstPort, p.SrcPort, p.Dir, p.Priority, p.Index, p.Mark, p.Tmpls)
+}

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

@@ -7,19 +7,55 @@ import (
 )
 
 func selFromPolicy(sel *nl.XfrmSelector, policy *XfrmPolicy) {
-	sel.Family = uint16(nl.GetIPFamily(policy.Dst.IP))
-	sel.Daddr.FromIP(policy.Dst.IP)
-	sel.Saddr.FromIP(policy.Src.IP)
-	prefixlenD, _ := policy.Dst.Mask.Size()
-	sel.PrefixlenD = uint8(prefixlenD)
-	prefixlenS, _ := policy.Src.Mask.Size()
-	sel.PrefixlenS = uint8(prefixlenS)
+	sel.Family = uint16(nl.FAMILY_V4)
+	if policy.Dst != nil {
+		sel.Family = uint16(nl.GetIPFamily(policy.Dst.IP))
+		sel.Daddr.FromIP(policy.Dst.IP)
+		prefixlenD, _ := policy.Dst.Mask.Size()
+		sel.PrefixlenD = uint8(prefixlenD)
+	}
+	if policy.Src != nil {
+		sel.Saddr.FromIP(policy.Src.IP)
+		prefixlenS, _ := policy.Src.Mask.Size()
+		sel.PrefixlenS = uint8(prefixlenS)
+	}
+	sel.Proto = uint8(policy.Proto)
+	sel.Dport = nl.Swap16(uint16(policy.DstPort))
+	sel.Sport = nl.Swap16(uint16(policy.SrcPort))
+	if sel.Dport != 0 {
+		sel.DportMask = ^uint16(0)
+	}
+	if sel.Sport != 0 {
+		sel.SportMask = ^uint16(0)
+	}
 }
 
 // XfrmPolicyAdd will add an xfrm policy to the system.
 // Equivalent to: `ip xfrm policy add $policy`
 func XfrmPolicyAdd(policy *XfrmPolicy) error {
-	req := nl.NewNetlinkRequest(nl.XFRM_MSG_NEWPOLICY, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	return pkgHandle.XfrmPolicyAdd(policy)
+}
+
+// XfrmPolicyAdd will add an xfrm policy to the system.
+// Equivalent to: `ip xfrm policy add $policy`
+func (h *Handle) XfrmPolicyAdd(policy *XfrmPolicy) error {
+	return h.xfrmPolicyAddOrUpdate(policy, nl.XFRM_MSG_NEWPOLICY)
+}
+
+// XfrmPolicyUpdate will update an xfrm policy to the system.
+// Equivalent to: `ip xfrm policy update $policy`
+func XfrmPolicyUpdate(policy *XfrmPolicy) error {
+	return pkgHandle.XfrmPolicyUpdate(policy)
+}
+
+// XfrmPolicyUpdate will update an xfrm policy to the system.
+// Equivalent to: `ip xfrm policy update $policy`
+func (h *Handle) XfrmPolicyUpdate(policy *XfrmPolicy) error {
+	return h.xfrmPolicyAddOrUpdate(policy, nl.XFRM_MSG_UPDPOLICY)
+}
+
+func (h *Handle) xfrmPolicyAddOrUpdate(policy *XfrmPolicy, nlProto int) error {
+	req := h.newNetlinkRequest(nlProto, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
 
 	msg := &nl.XfrmUserpolicyInfo{}
 	selFromPolicy(&msg.Sel, policy)
@@ -39,6 +75,7 @@ func XfrmPolicyAdd(policy *XfrmPolicy) error {
 		userTmpl.XfrmId.Daddr.FromIP(tmpl.Dst)
 		userTmpl.Saddr.FromIP(tmpl.Src)
 		userTmpl.XfrmId.Proto = uint8(tmpl.Proto)
+		userTmpl.XfrmId.Spi = nl.Swap32(uint32(tmpl.Spi))
 		userTmpl.Mode = uint8(tmpl.Mode)
 		userTmpl.Reqid = uint32(tmpl.Reqid)
 		userTmpl.Aalgos = ^uint32(0)
@@ -49,6 +86,10 @@ func XfrmPolicyAdd(policy *XfrmPolicy) error {
 		tmpls := nl.NewRtAttr(nl.XFRMA_TMPL, tmplData)
 		req.AddData(tmpls)
 	}
+	if policy.Mark != nil {
+		out := nl.NewRtAttr(nl.XFRMA_MARK, writeMark(policy.Mark))
+		req.AddData(out)
+	}
 
 	_, err := req.Execute(syscall.NETLINK_XFRM, 0)
 	return err
@@ -58,15 +99,14 @@ func XfrmPolicyAdd(policy *XfrmPolicy) error {
 // the Tmpls are ignored when matching the policy to delete.
 // Equivalent to: `ip xfrm policy del $policy`
 func XfrmPolicyDel(policy *XfrmPolicy) error {
-	req := nl.NewNetlinkRequest(nl.XFRM_MSG_DELPOLICY, syscall.NLM_F_ACK)
-
-	msg := &nl.XfrmUserpolicyId{}
-	selFromPolicy(&msg.Sel, policy)
-	msg.Index = uint32(policy.Index)
-	msg.Dir = uint8(policy.Dir)
-	req.AddData(msg)
+	return pkgHandle.XfrmPolicyDel(policy)
+}
 
-	_, err := req.Execute(syscall.NETLINK_XFRM, 0)
+// XfrmPolicyDel will delete an xfrm policy from the system. Note that
+// the Tmpls are ignored when matching the policy to delete.
+// Equivalent to: `ip xfrm policy del $policy`
+func (h *Handle) XfrmPolicyDel(policy *XfrmPolicy) error {
+	_, err := h.xfrmPolicyGetOrDelete(policy, nl.XFRM_MSG_DELPOLICY)
 	return err
 }
 
@@ -74,7 +114,14 @@ func XfrmPolicyDel(policy *XfrmPolicy) error {
 // Equivalent to: `ip xfrm policy show`.
 // The list can be filtered by ip family.
 func XfrmPolicyList(family int) ([]XfrmPolicy, error) {
-	req := nl.NewNetlinkRequest(nl.XFRM_MSG_GETPOLICY, syscall.NLM_F_DUMP)
+	return pkgHandle.XfrmPolicyList(family)
+}
+
+// XfrmPolicyList gets a list of xfrm policies in the system.
+// Equivalent to: `ip xfrm policy show`.
+// The list can be filtered by ip family.
+func (h *Handle) XfrmPolicyList(family int) ([]XfrmPolicy, error) {
+	req := h.newNetlinkRequest(nl.XFRM_MSG_GETPOLICY, syscall.NLM_F_DUMP)
 
 	msg := nl.NewIfInfomsg(family)
 	req.AddData(msg)
@@ -86,42 +133,125 @@ func XfrmPolicyList(family int) ([]XfrmPolicy, error) {
 
 	var res []XfrmPolicy
 	for _, m := range msgs {
-		msg := nl.DeserializeXfrmUserpolicyInfo(m)
-
-		if family != FAMILY_ALL && family != int(msg.Sel.Family) {
+		if policy, err := parseXfrmPolicy(m, family); err == nil {
+			res = append(res, *policy)
+		} else if err == familyError {
 			continue
+		} else {
+			return nil, err
 		}
+	}
+	return res, nil
+}
 
-		var policy XfrmPolicy
+// XfrmPolicyGet gets a the policy described by the index or selector, if found.
+// Equivalent to: `ip xfrm policy get { SELECTOR | index INDEX } dir DIR [ctx CTX ] [ mark MARK [ mask MASK ] ] [ ptype PTYPE ]`.
+func XfrmPolicyGet(policy *XfrmPolicy) (*XfrmPolicy, error) {
+	return pkgHandle.XfrmPolicyGet(policy)
+}
 
-		policy.Dst = msg.Sel.Daddr.ToIPNet(msg.Sel.PrefixlenD)
-		policy.Src = msg.Sel.Saddr.ToIPNet(msg.Sel.PrefixlenS)
-		policy.Priority = int(msg.Priority)
-		policy.Index = int(msg.Index)
-		policy.Dir = Dir(msg.Dir)
+// XfrmPolicyGet gets a the policy described by the index or selector, if found.
+// Equivalent to: `ip xfrm policy get { SELECTOR | index INDEX } dir DIR [ctx CTX ] [ mark MARK [ mask MASK ] ] [ ptype PTYPE ]`.
+func (h *Handle) XfrmPolicyGet(policy *XfrmPolicy) (*XfrmPolicy, error) {
+	return h.xfrmPolicyGetOrDelete(policy, nl.XFRM_MSG_GETPOLICY)
+}
 
-		attrs, err := nl.ParseRouteAttr(m[msg.Len():])
-		if err != nil {
-			return nil, err
-		}
+// XfrmPolicyFlush will flush the policies on the system.
+// Equivalent to: `ip xfrm policy flush`
+func XfrmPolicyFlush() error {
+	return pkgHandle.XfrmPolicyFlush()
+}
+
+// XfrmPolicyFlush will flush the policies on the system.
+// Equivalent to: `ip xfrm policy flush`
+func (h *Handle) XfrmPolicyFlush() error {
+	req := h.newNetlinkRequest(nl.XFRM_MSG_FLUSHPOLICY, syscall.NLM_F_ACK)
+	_, err := req.Execute(syscall.NETLINK_XFRM, 0)
+	return err
+}
+
+func (h *Handle) xfrmPolicyGetOrDelete(policy *XfrmPolicy, nlProto int) (*XfrmPolicy, error) {
+	req := h.newNetlinkRequest(nlProto, syscall.NLM_F_ACK)
+
+	msg := &nl.XfrmUserpolicyId{}
+	selFromPolicy(&msg.Sel, policy)
+	msg.Index = uint32(policy.Index)
+	msg.Dir = uint8(policy.Dir)
+	req.AddData(msg)
+
+	if policy.Mark != nil {
+		out := nl.NewRtAttr(nl.XFRMA_MARK, writeMark(policy.Mark))
+		req.AddData(out)
+	}
+
+	resType := nl.XFRM_MSG_NEWPOLICY
+	if nlProto == nl.XFRM_MSG_DELPOLICY {
+		resType = 0
+	}
+
+	msgs, err := req.Execute(syscall.NETLINK_XFRM, uint16(resType))
+	if err != nil {
+		return nil, err
+	}
 
-		for _, attr := range attrs {
-			switch attr.Attr.Type {
-			case nl.XFRMA_TMPL:
-				max := len(attr.Value)
-				for i := 0; i < max; i += nl.SizeofXfrmUserTmpl {
-					var resTmpl XfrmPolicyTmpl
-					tmpl := nl.DeserializeXfrmUserTmpl(attr.Value[i : i+nl.SizeofXfrmUserTmpl])
-					resTmpl.Dst = tmpl.XfrmId.Daddr.ToIP()
-					resTmpl.Src = tmpl.Saddr.ToIP()
-					resTmpl.Proto = Proto(tmpl.XfrmId.Proto)
-					resTmpl.Mode = Mode(tmpl.Mode)
-					resTmpl.Reqid = int(tmpl.Reqid)
-					policy.Tmpls = append(policy.Tmpls, resTmpl)
-				}
+	if nlProto == nl.XFRM_MSG_DELPOLICY {
+		return nil, err
+	}
+
+	p, err := parseXfrmPolicy(msgs[0], FAMILY_ALL)
+	if err != nil {
+		return nil, err
+	}
+
+	return p, nil
+}
+
+func parseXfrmPolicy(m []byte, family int) (*XfrmPolicy, error) {
+	msg := nl.DeserializeXfrmUserpolicyInfo(m)
+
+	// This is mainly for the policy dump
+	if family != FAMILY_ALL && family != int(msg.Sel.Family) {
+		return nil, familyError
+	}
+
+	var policy XfrmPolicy
+
+	policy.Dst = msg.Sel.Daddr.ToIPNet(msg.Sel.PrefixlenD)
+	policy.Src = msg.Sel.Saddr.ToIPNet(msg.Sel.PrefixlenS)
+	policy.Proto = Proto(msg.Sel.Proto)
+	policy.DstPort = int(nl.Swap16(msg.Sel.Dport))
+	policy.SrcPort = int(nl.Swap16(msg.Sel.Sport))
+	policy.Priority = int(msg.Priority)
+	policy.Index = int(msg.Index)
+	policy.Dir = Dir(msg.Dir)
+
+	attrs, err := nl.ParseRouteAttr(m[msg.Len():])
+	if err != nil {
+		return nil, err
+	}
+
+	for _, attr := range attrs {
+		switch attr.Attr.Type {
+		case nl.XFRMA_TMPL:
+			max := len(attr.Value)
+			for i := 0; i < max; i += nl.SizeofXfrmUserTmpl {
+				var resTmpl XfrmPolicyTmpl
+				tmpl := nl.DeserializeXfrmUserTmpl(attr.Value[i : i+nl.SizeofXfrmUserTmpl])
+				resTmpl.Dst = tmpl.XfrmId.Daddr.ToIP()
+				resTmpl.Src = tmpl.Saddr.ToIP()
+				resTmpl.Proto = Proto(tmpl.XfrmId.Proto)
+				resTmpl.Mode = Mode(tmpl.Mode)
+				resTmpl.Spi = int(nl.Swap32(tmpl.XfrmId.Spi))
+				resTmpl.Reqid = int(tmpl.Reqid)
+				policy.Tmpls = append(policy.Tmpls, resTmpl)
 			}
+		case nl.XFRMA_MARK:
+			mark := nl.DeserializeXfrmMark(attr.Value[:])
+			policy.Mark = new(XfrmMark)
+			policy.Mark.Value = mark.Value
+			policy.Mark.Mask = mark.Mask
 		}
-		res = append(res, policy)
 	}
-	return res, nil
+
+	return &policy, nil
 }

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

@@ -1,29 +1,19 @@
 package netlink
 
 import (
+	"bytes"
 	"net"
 	"testing"
 )
 
-func TestXfrmPolicyAddDel(t *testing.T) {
+const zeroCIDR = "0.0.0.0/0"
+
+func TestXfrmPolicyAddUpdateDel(t *testing.T) {
 	tearDown := setUpNetlinkTest(t)
 	defer tearDown()
 
-	src, _ := ParseIPNet("127.1.1.1/32")
-	dst, _ := ParseIPNet("127.1.1.2/32")
-	policy := XfrmPolicy{
-		Src: src,
-		Dst: dst,
-		Dir: XFRM_DIR_OUT,
-	}
-	tmpl := XfrmPolicyTmpl{
-		Src:   net.ParseIP("127.0.0.1"),
-		Dst:   net.ParseIP("127.0.0.2"),
-		Proto: XFRM_PROTO_ESP,
-		Mode:  XFRM_MODE_TUNNEL,
-	}
-	policy.Tmpls = append(policy.Tmpls, tmpl)
-	if err := XfrmPolicyAdd(&policy); err != nil {
+	policy := getPolicy()
+	if err := XfrmPolicyAdd(policy); err != nil {
 		t.Fatal(err)
 	}
 	policies, err := XfrmPolicyList(FAMILY_ALL)
@@ -35,7 +25,34 @@ func TestXfrmPolicyAddDel(t *testing.T) {
 		t.Fatal("Policy not added properly")
 	}
 
-	if err = XfrmPolicyDel(&policy); err != nil {
+	if !comparePolicies(policy, &policies[0]) {
+		t.Fatalf("unexpected policy returned.\nExpected: %v.\nGot %v", policy, policies[0])
+	}
+
+	// Look for a specific policy
+	sp, err := XfrmPolicyGet(policy)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if !comparePolicies(policy, sp) {
+		t.Fatalf("unexpected policy returned")
+	}
+
+	// Modify the policy
+	policy.Priority = 100
+	if err := XfrmPolicyUpdate(policy); err != nil {
+		t.Fatal(err)
+	}
+	sp, err = XfrmPolicyGet(policy)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if sp.Priority != 100 {
+		t.Fatalf("failed to modify the policy")
+	}
+
+	if err = XfrmPolicyDel(policy); err != nil {
 		t.Fatal(err)
 	}
 
@@ -46,4 +63,135 @@ func TestXfrmPolicyAddDel(t *testing.T) {
 	if len(policies) != 0 {
 		t.Fatal("Policy not removed properly")
 	}
+
+	// Src and dst are not mandatory field. Creation should succeed
+	policy.Src = nil
+	policy.Dst = nil
+	if err = XfrmPolicyAdd(policy); err != nil {
+		t.Fatal(err)
+	}
+
+	sp, err = XfrmPolicyGet(policy)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if !comparePolicies(policy, sp) {
+		t.Fatalf("unexpected policy returned")
+	}
+
+	if err = XfrmPolicyDel(policy); err != nil {
+		t.Fatal(err)
+	}
+
+	if _, err := XfrmPolicyGet(policy); err == nil {
+		t.Fatalf("Unexpected success")
+	}
+}
+
+func TestXfrmPolicyFlush(t *testing.T) {
+	setUpNetlinkTest(t)()
+
+	p1 := getPolicy()
+	if err := XfrmPolicyAdd(p1); err != nil {
+		t.Fatal(err)
+	}
+
+	p1.Dir = XFRM_DIR_IN
+	s := p1.Src
+	p1.Src = p1.Dst
+	p1.Dst = s
+	if err := XfrmPolicyAdd(p1); err != nil {
+		t.Fatal(err)
+	}
+
+	policies, err := XfrmPolicyList(FAMILY_ALL)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(policies) != 2 {
+		t.Fatalf("unexpected number of policies: %d", len(policies))
+	}
+
+	if err := XfrmPolicyFlush(); err != nil {
+		t.Fatal(err)
+	}
+
+	policies, err = XfrmPolicyList(FAMILY_ALL)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(policies) != 0 {
+		t.Fatalf("unexpected number of policies: %d", len(policies))
+	}
+
+}
+
+func comparePolicies(a, b *XfrmPolicy) bool {
+	if a == b {
+		return true
+	}
+	if a == nil || b == nil {
+		return false
+	}
+	// Do not check Index which is assigned by kernel
+	return a.Dir == b.Dir && a.Priority == b.Priority &&
+		compareIPNet(a.Src, b.Src) && compareIPNet(a.Dst, b.Dst) &&
+		a.Mark.Value == b.Mark.Value && a.Mark.Mask == b.Mark.Mask &&
+		compareTemplates(a.Tmpls, b.Tmpls)
+}
+
+func compareTemplates(a, b []XfrmPolicyTmpl) bool {
+	if len(a) != len(b) {
+		return false
+	}
+	for i, ta := range a {
+		tb := b[i]
+		if !ta.Dst.Equal(tb.Dst) || !ta.Src.Equal(tb.Src) || ta.Spi != tb.Spi ||
+			ta.Mode != tb.Mode || ta.Reqid != tb.Reqid || ta.Proto != tb.Proto {
+			return false
+		}
+	}
+	return true
+}
+
+func compareIPNet(a, b *net.IPNet) bool {
+	if a == b {
+		return true
+	}
+	// For unspecified src/dst parseXfrmPolicy would set the zero address cidr
+	if (a == nil && b.String() == zeroCIDR) || (b == nil && a.String() == zeroCIDR) {
+		return true
+	}
+	if a == nil || b == nil {
+		return false
+	}
+	return a.IP.Equal(b.IP) && bytes.Equal(a.Mask, b.Mask)
+}
+
+func getPolicy() *XfrmPolicy {
+	src, _ := ParseIPNet("127.1.1.1/32")
+	dst, _ := ParseIPNet("127.1.1.2/32")
+	policy := &XfrmPolicy{
+		Src:     src,
+		Dst:     dst,
+		Proto:   17,
+		DstPort: 1234,
+		SrcPort: 5678,
+		Dir:     XFRM_DIR_OUT,
+		Mark: &XfrmMark{
+			Value: 0xabff22,
+			Mask:  0xffffffff,
+		},
+		Priority: 10,
+	}
+	tmpl := XfrmPolicyTmpl{
+		Src:   net.ParseIP("127.0.0.1"),
+		Dst:   net.ParseIP("127.0.0.2"),
+		Proto: XFRM_PROTO_ESP,
+		Mode:  XFRM_MODE_TUNNEL,
+		Spi:   0xabcdef99,
+	}
+	policy.Tmpls = append(policy.Tmpls, tmpl)
+	return policy
 }

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

@@ -1,6 +1,7 @@
 package netlink
 
 import (
+	"fmt"
 	"net"
 )
 
@@ -11,7 +12,11 @@ type XfrmStateAlgo struct {
 	TruncateLen int // Auth only
 }
 
-// EncapType is an enum representing an ipsec template direction.
+func (a XfrmStateAlgo) String() string {
+	return fmt.Sprintf("{Name: %s, Key: 0x%x, TruncateLen: %d}", a.Name, a.Key, a.TruncateLen)
+}
+
+// EncapType is an enum representing the optional packet encapsulation.
 type EncapType uint8
 
 const (
@@ -22,14 +27,14 @@ const (
 func (e EncapType) String() string {
 	switch e {
 	case XFRM_ENCAP_ESPINUDP_NONIKE:
-		return "espinudp-nonike"
+		return "espinudp-non-ike"
 	case XFRM_ENCAP_ESPINUDP:
 		return "espinudp"
 	}
 	return "unknown"
 }
 
-// XfrmEncap represents the encapsulation to use for the ipsec encryption.
+// XfrmStateEncap represents the encapsulation to use for the ipsec encryption.
 type XfrmStateEncap struct {
 	Type            EncapType
 	SrcPort         int
@@ -37,6 +42,23 @@ type XfrmStateEncap struct {
 	OriginalAddress net.IP
 }
 
+func (e XfrmStateEncap) String() string {
+	return fmt.Sprintf("{Type: %s, Srcport: %d, DstPort: %d, OriginalAddress: %v}",
+		e.Type, e.SrcPort, e.DstPort, e.OriginalAddress)
+}
+
+// XfrmStateLimits represents the configured limits for the state.
+type XfrmStateLimits struct {
+	ByteSoft    uint64
+	ByteHard    uint64
+	PacketSoft  uint64
+	PacketHard  uint64
+	TimeSoft    uint64
+	TimeHard    uint64
+	TimeUseSoft uint64
+	TimeUseHard uint64
+}
+
 // XfrmState represents the state of an ipsec policy. It optionally
 // contains an XfrmStateAlgo for encryption and one for authentication.
 type XfrmState struct {
@@ -47,7 +69,30 @@ type XfrmState struct {
 	Spi          int
 	Reqid        int
 	ReplayWindow int
+	Limits       XfrmStateLimits
+	Mark         *XfrmMark
 	Auth         *XfrmStateAlgo
 	Crypt        *XfrmStateAlgo
 	Encap        *XfrmStateEncap
 }
+
+func (sa XfrmState) String() string {
+	return fmt.Sprintf("Dst: %v, Src: %v, Proto: %s, Mode: %s, SPI: 0x%x, ReqID: 0x%x, ReplayWindow: %d, Mark: %v, Auth: %v, Crypt: %v, Encap: %v",
+		sa.Dst, sa.Src, sa.Proto, sa.Mode, sa.Spi, sa.Reqid, sa.ReplayWindow, sa.Mark, sa.Auth, sa.Crypt, sa.Encap)
+}
+func (sa XfrmState) Print(stats bool) string {
+	if !stats {
+		return sa.String()
+	}
+
+	return fmt.Sprintf("%s, ByteSoft: %s, ByteHard: %s, PacketSoft: %s, PacketHard: %s, TimeSoft: %d, TimeHard: %d, TimeUseSoft: %d, TimeUseHard: %d",
+		sa.String(), printLimit(sa.Limits.ByteSoft), printLimit(sa.Limits.ByteHard), printLimit(sa.Limits.PacketSoft), printLimit(sa.Limits.PacketHard),
+		sa.Limits.TimeSoft, sa.Limits.TimeHard, sa.Limits.TimeUseSoft, sa.Limits.TimeUseHard)
+}
+
+func printLimit(lmt uint64) string {
+	if lmt == ^uint64(0) {
+		return "(INF)"
+	}
+	return fmt.Sprintf("%d", lmt)
+}

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

@@ -3,6 +3,7 @@ package netlink
 import (
 	"fmt"
 	"syscall"
+	"unsafe"
 
 	"github.com/vishvananda/netlink/nl"
 )
@@ -34,14 +35,47 @@ func writeStateAlgoAuth(a *XfrmStateAlgo) []byte {
 	return algo.Serialize()
 }
 
+func writeMark(m *XfrmMark) []byte {
+	mark := &nl.XfrmMark{
+		Value: m.Value,
+		Mask:  m.Mask,
+	}
+	if mark.Mask == 0 {
+		mark.Mask = ^uint32(0)
+	}
+	return mark.Serialize()
+}
+
 // XfrmStateAdd will add an xfrm state to the system.
 // Equivalent to: `ip xfrm state add $state`
 func XfrmStateAdd(state *XfrmState) error {
+	return pkgHandle.XfrmStateAdd(state)
+}
+
+// XfrmStateAdd will add an xfrm state to the system.
+// Equivalent to: `ip xfrm state add $state`
+func (h *Handle) XfrmStateAdd(state *XfrmState) error {
+	return h.xfrmStateAddOrUpdate(state, nl.XFRM_MSG_NEWSA)
+}
+
+// XfrmStateUpdate will update an xfrm state to the system.
+// Equivalent to: `ip xfrm state update $state`
+func XfrmStateUpdate(state *XfrmState) error {
+	return pkgHandle.XfrmStateUpdate(state)
+}
+
+// XfrmStateUpdate will update an xfrm state to the system.
+// Equivalent to: `ip xfrm state update $state`
+func (h *Handle) XfrmStateUpdate(state *XfrmState) error {
+	return h.xfrmStateAddOrUpdate(state, nl.XFRM_MSG_UPDSA)
+}
+
+func (h *Handle) xfrmStateAddOrUpdate(state *XfrmState, nlProto int) error {
 	// A state with spi 0 can't be deleted so don't allow it to be set
 	if state.Spi == 0 {
 		return fmt.Errorf("Spi must be set when adding xfrm state.")
 	}
-	req := nl.NewNetlinkRequest(nl.XFRM_MSG_NEWSA, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	req := h.newNetlinkRequest(nlProto, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
 
 	msg := &nl.XfrmUsersaInfo{}
 	msg.Family = uint16(nl.GetIPFamily(state.Dst))
@@ -52,10 +86,7 @@ func XfrmStateAdd(state *XfrmState) error {
 	msg.Id.Spi = nl.Swap32(uint32(state.Spi))
 	msg.Reqid = uint32(state.Reqid)
 	msg.ReplayWindow = uint8(state.ReplayWindow)
-	msg.Lft.SoftByteLimit = nl.XFRM_INF
-	msg.Lft.HardByteLimit = nl.XFRM_INF
-	msg.Lft.SoftPacketLimit = nl.XFRM_INF
-	msg.Lft.HardPacketLimit = nl.XFRM_INF
+	limitsToLft(state.Limits, &msg.Lft)
 	req.AddData(msg)
 
 	if state.Auth != nil {
@@ -76,6 +107,10 @@ func XfrmStateAdd(state *XfrmState) error {
 		out := nl.NewRtAttr(nl.XFRMA_ENCAP, encapData)
 		req.AddData(out)
 	}
+	if state.Mark != nil {
+		out := nl.NewRtAttr(nl.XFRMA_MARK, writeMark(state.Mark))
+		req.AddData(out)
+	}
 
 	_, err := req.Execute(syscall.NETLINK_XFRM, 0)
 	return err
@@ -85,33 +120,29 @@ func XfrmStateAdd(state *XfrmState) error {
 // the Algos are ignored when matching the state to delete.
 // Equivalent to: `ip xfrm state del $state`
 func XfrmStateDel(state *XfrmState) error {
-	req := nl.NewNetlinkRequest(nl.XFRM_MSG_DELSA, syscall.NLM_F_ACK)
-
-	msg := &nl.XfrmUsersaId{}
-	msg.Daddr.FromIP(state.Dst)
-	msg.Family = uint16(nl.GetIPFamily(state.Dst))
-	msg.Proto = uint8(state.Proto)
-	msg.Spi = nl.Swap32(uint32(state.Spi))
-	req.AddData(msg)
-
-	saddr := nl.XfrmAddress{}
-	saddr.FromIP(state.Src)
-	srcdata := nl.NewRtAttr(nl.XFRMA_SRCADDR, saddr.Serialize())
-
-	req.AddData(srcdata)
+	return pkgHandle.XfrmStateDel(state)
+}
 
-	_, err := req.Execute(syscall.NETLINK_XFRM, 0)
+// XfrmStateDel will delete an xfrm state from the system. Note that
+// the Algos are ignored when matching the state to delete.
+// Equivalent to: `ip xfrm state del $state`
+func (h *Handle) XfrmStateDel(state *XfrmState) error {
+	_, err := h.xfrmStateGetOrDelete(state, nl.XFRM_MSG_DELSA)
 	return err
 }
 
 // XfrmStateList gets a list of xfrm states in the system.
-// Equivalent to: `ip xfrm state show`.
+// Equivalent to: `ip [-4|-6] xfrm state show`.
 // The list can be filtered by ip family.
 func XfrmStateList(family int) ([]XfrmState, error) {
-	req := nl.NewNetlinkRequest(nl.XFRM_MSG_GETSA, syscall.NLM_F_DUMP)
+	return pkgHandle.XfrmStateList(family)
+}
 
-	msg := nl.NewIfInfomsg(family)
-	req.AddData(msg)
+// XfrmStateList gets a list of xfrm states in the system.
+// Equivalent to: `ip xfrm state show`.
+// The list can be filtered by ip family.
+func (h *Handle) XfrmStateList(family int) ([]XfrmState, error) {
+	req := h.newNetlinkRequest(nl.XFRM_MSG_GETSA, syscall.NLM_F_DUMP)
 
 	msgs, err := req.Execute(syscall.NETLINK_XFRM, nl.XFRM_MSG_NEWSA)
 	if err != nil {
@@ -120,62 +151,194 @@ func XfrmStateList(family int) ([]XfrmState, error) {
 
 	var res []XfrmState
 	for _, m := range msgs {
-		msg := nl.DeserializeXfrmUsersaInfo(m)
-
-		if family != FAMILY_ALL && family != int(msg.Family) {
+		if state, err := parseXfrmState(m, family); err == nil {
+			res = append(res, *state)
+		} else if err == familyError {
 			continue
+		} else {
+			return nil, err
 		}
+	}
+	return res, nil
+}
 
-		var state XfrmState
+// XfrmStateGet gets the xfrm state described by the ID, if found.
+// Equivalent to: `ip xfrm state get ID [ mark MARK [ mask MASK ] ]`.
+// Only the fields which constitue the SA ID must be filled in:
+// ID := [ src ADDR ] [ dst ADDR ] [ proto XFRM-PROTO ] [ spi SPI ]
+// mark is optional
+func XfrmStateGet(state *XfrmState) (*XfrmState, error) {
+	return pkgHandle.XfrmStateGet(state)
+}
 
-		state.Dst = msg.Id.Daddr.ToIP()
-		state.Src = msg.Saddr.ToIP()
-		state.Proto = Proto(msg.Id.Proto)
-		state.Mode = Mode(msg.Mode)
-		state.Spi = int(nl.Swap32(msg.Id.Spi))
-		state.Reqid = int(msg.Reqid)
-		state.ReplayWindow = int(msg.ReplayWindow)
+// XfrmStateGet gets the xfrm state described by the ID, if found.
+// Equivalent to: `ip xfrm state get ID [ mark MARK [ mask MASK ] ]`.
+// Only the fields which constitue the SA ID must be filled in:
+// ID := [ src ADDR ] [ dst ADDR ] [ proto XFRM-PROTO ] [ spi SPI ]
+// mark is optional
+func (h *Handle) XfrmStateGet(state *XfrmState) (*XfrmState, error) {
+	return h.xfrmStateGetOrDelete(state, nl.XFRM_MSG_GETSA)
+}
 
-		attrs, err := nl.ParseRouteAttr(m[msg.Len():])
-		if err != nil {
-			return nil, err
-		}
+func (h *Handle) xfrmStateGetOrDelete(state *XfrmState, nlProto int) (*XfrmState, error) {
+	req := h.newNetlinkRequest(nlProto, syscall.NLM_F_ACK)
 
-		for _, attr := range attrs {
-			switch attr.Attr.Type {
-			case nl.XFRMA_ALG_AUTH, nl.XFRMA_ALG_CRYPT:
-				var resAlgo *XfrmStateAlgo
-				if attr.Attr.Type == nl.XFRMA_ALG_AUTH {
-					if state.Auth == nil {
-						state.Auth = new(XfrmStateAlgo)
-					}
-					resAlgo = state.Auth
-				} else {
-					state.Crypt = new(XfrmStateAlgo)
-					resAlgo = state.Crypt
-				}
-				algo := nl.DeserializeXfrmAlgo(attr.Value[:])
-				(*resAlgo).Name = nl.BytesToString(algo.AlgName[:])
-				(*resAlgo).Key = algo.AlgKey
-			case nl.XFRMA_ALG_AUTH_TRUNC:
+	msg := &nl.XfrmUsersaId{}
+	msg.Family = uint16(nl.GetIPFamily(state.Dst))
+	msg.Daddr.FromIP(state.Dst)
+	msg.Proto = uint8(state.Proto)
+	msg.Spi = nl.Swap32(uint32(state.Spi))
+	req.AddData(msg)
+
+	if state.Mark != nil {
+		out := nl.NewRtAttr(nl.XFRMA_MARK, writeMark(state.Mark))
+		req.AddData(out)
+	}
+	if state.Src != nil {
+		out := nl.NewRtAttr(nl.XFRMA_SRCADDR, state.Src)
+		req.AddData(out)
+	}
+
+	resType := nl.XFRM_MSG_NEWSA
+	if nlProto == nl.XFRM_MSG_DELSA {
+		resType = 0
+	}
+
+	msgs, err := req.Execute(syscall.NETLINK_XFRM, uint16(resType))
+	if err != nil {
+		return nil, err
+	}
+
+	if nlProto == nl.XFRM_MSG_DELSA {
+		return nil, nil
+	}
+
+	s, err := parseXfrmState(msgs[0], FAMILY_ALL)
+	if err != nil {
+		return nil, err
+	}
+
+	return s, nil
+}
+
+var familyError = fmt.Errorf("family error")
+
+func parseXfrmState(m []byte, family int) (*XfrmState, error) {
+	msg := nl.DeserializeXfrmUsersaInfo(m)
+
+	// This is mainly for the state dump
+	if family != FAMILY_ALL && family != int(msg.Family) {
+		return nil, familyError
+	}
+
+	var state XfrmState
+
+	state.Dst = msg.Id.Daddr.ToIP()
+	state.Src = msg.Saddr.ToIP()
+	state.Proto = Proto(msg.Id.Proto)
+	state.Mode = Mode(msg.Mode)
+	state.Spi = int(nl.Swap32(msg.Id.Spi))
+	state.Reqid = int(msg.Reqid)
+	state.ReplayWindow = int(msg.ReplayWindow)
+	lftToLimits(&msg.Lft, &state.Limits)
+
+	attrs, err := nl.ParseRouteAttr(m[nl.SizeofXfrmUsersaInfo:])
+	if err != nil {
+		return nil, err
+	}
+
+	for _, attr := range attrs {
+		switch attr.Attr.Type {
+		case nl.XFRMA_ALG_AUTH, nl.XFRMA_ALG_CRYPT:
+			var resAlgo *XfrmStateAlgo
+			if attr.Attr.Type == nl.XFRMA_ALG_AUTH {
 				if state.Auth == nil {
 					state.Auth = new(XfrmStateAlgo)
 				}
-				algo := nl.DeserializeXfrmAlgoAuth(attr.Value[:])
-				state.Auth.Name = nl.BytesToString(algo.AlgName[:])
-				state.Auth.Key = algo.AlgKey
-				state.Auth.TruncateLen = int(algo.AlgTruncLen)
-			case nl.XFRMA_ENCAP:
-				encap := nl.DeserializeXfrmEncapTmpl(attr.Value[:])
-				state.Encap = new(XfrmStateEncap)
-				state.Encap.Type = EncapType(encap.EncapType)
-				state.Encap.SrcPort = int(nl.Swap16(encap.EncapSport))
-				state.Encap.DstPort = int(nl.Swap16(encap.EncapDport))
-				state.Encap.OriginalAddress = encap.EncapOa.ToIP()
+				resAlgo = state.Auth
+			} else {
+				state.Crypt = new(XfrmStateAlgo)
+				resAlgo = state.Crypt
 			}
-
+			algo := nl.DeserializeXfrmAlgo(attr.Value[:])
+			(*resAlgo).Name = nl.BytesToString(algo.AlgName[:])
+			(*resAlgo).Key = algo.AlgKey
+		case nl.XFRMA_ALG_AUTH_TRUNC:
+			if state.Auth == nil {
+				state.Auth = new(XfrmStateAlgo)
+			}
+			algo := nl.DeserializeXfrmAlgoAuth(attr.Value[:])
+			state.Auth.Name = nl.BytesToString(algo.AlgName[:])
+			state.Auth.Key = algo.AlgKey
+			state.Auth.TruncateLen = int(algo.AlgTruncLen)
+		case nl.XFRMA_ENCAP:
+			encap := nl.DeserializeXfrmEncapTmpl(attr.Value[:])
+			state.Encap = new(XfrmStateEncap)
+			state.Encap.Type = EncapType(encap.EncapType)
+			state.Encap.SrcPort = int(nl.Swap16(encap.EncapSport))
+			state.Encap.DstPort = int(nl.Swap16(encap.EncapDport))
+			state.Encap.OriginalAddress = encap.EncapOa.ToIP()
+		case nl.XFRMA_MARK:
+			mark := nl.DeserializeXfrmMark(attr.Value[:])
+			state.Mark = new(XfrmMark)
+			state.Mark.Value = mark.Value
+			state.Mark.Mask = mark.Mask
 		}
-		res = append(res, state)
 	}
-	return res, nil
+
+	return &state, nil
+}
+
+// XfrmStateFlush will flush the xfrm state on the system.
+// proto = 0 means any transformation protocols
+// Equivalent to: `ip xfrm state flush [ proto XFRM-PROTO ]`
+func XfrmStateFlush(proto Proto) error {
+	return pkgHandle.XfrmStateFlush(proto)
+}
+
+// XfrmStateFlush will flush the xfrm state on the system.
+// proto = 0 means any transformation protocols
+// Equivalent to: `ip xfrm state flush [ proto XFRM-PROTO ]`
+func (h *Handle) XfrmStateFlush(proto Proto) error {
+	req := h.newNetlinkRequest(nl.XFRM_MSG_FLUSHSA, syscall.NLM_F_ACK)
+
+	req.AddData(&nl.XfrmUsersaFlush{Proto: uint8(proto)})
+
+	_, err := req.Execute(syscall.NETLINK_XFRM, 0)
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func limitsToLft(lmts XfrmStateLimits, lft *nl.XfrmLifetimeCfg) {
+	if lmts.ByteSoft != 0 {
+		lft.SoftByteLimit = lmts.ByteSoft
+	} else {
+		lft.SoftByteLimit = nl.XFRM_INF
+	}
+	if lmts.ByteHard != 0 {
+		lft.HardByteLimit = lmts.ByteHard
+	} else {
+		lft.HardByteLimit = nl.XFRM_INF
+	}
+	if lmts.PacketSoft != 0 {
+		lft.SoftPacketLimit = lmts.PacketSoft
+	} else {
+		lft.SoftPacketLimit = nl.XFRM_INF
+	}
+	if lmts.PacketHard != 0 {
+		lft.HardPacketLimit = lmts.PacketHard
+	} else {
+		lft.HardPacketLimit = nl.XFRM_INF
+	}
+	lft.SoftAddExpiresSeconds = lmts.TimeSoft
+	lft.HardAddExpiresSeconds = lmts.TimeHard
+	lft.SoftUseExpiresSeconds = lmts.TimeUseSoft
+	lft.HardUseExpiresSeconds = lmts.TimeUseHard
+}
+
+func lftToLimits(lft *nl.XfrmLifetimeCfg, lmts *XfrmStateLimits) {
+	*lmts = *(*XfrmStateLimits)(unsafe.Pointer(lft))
 }

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

@@ -1,15 +1,161 @@
 package netlink
 
 import (
+	"bytes"
 	"net"
 	"testing"
 )
 
-func TestXfrmStateAddDel(t *testing.T) {
+func TestXfrmStateAddGetDel(t *testing.T) {
 	tearDown := setUpNetlinkTest(t)
 	defer tearDown()
 
-	state := XfrmState{
+	state := getBaseState()
+
+	if err := XfrmStateAdd(state); err != nil {
+		t.Fatal(err)
+	}
+	states, err := XfrmStateList(FAMILY_ALL)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if len(states) != 1 {
+		t.Fatal("State not added properly")
+	}
+
+	if !compareStates(state, &states[0]) {
+		t.Fatalf("unexpected states returned")
+	}
+
+	// Get specific state
+	sa, err := XfrmStateGet(state)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if !compareStates(state, sa) {
+		t.Fatalf("unexpected state returned")
+	}
+
+	if err = XfrmStateDel(state); err != nil {
+		t.Fatal(err)
+	}
+
+	states, err = XfrmStateList(FAMILY_ALL)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(states) != 0 {
+		t.Fatal("State not removed properly")
+	}
+
+	if _, err := XfrmStateGet(state); err == nil {
+		t.Fatalf("Unexpected success")
+	}
+}
+
+func TestXfrmStateFlush(t *testing.T) {
+	setUpNetlinkTest(t)()
+
+	state1 := getBaseState()
+	state2 := getBaseState()
+	state2.Src = net.ParseIP("127.1.0.1")
+	state2.Dst = net.ParseIP("127.1.0.2")
+	state2.Proto = XFRM_PROTO_AH
+	state2.Mode = XFRM_MODE_TUNNEL
+	state2.Spi = 20
+	state2.Mark = nil
+	state2.Crypt = nil
+
+	if err := XfrmStateAdd(state1); err != nil {
+		t.Fatal(err)
+	}
+	if err := XfrmStateAdd(state2); err != nil {
+		t.Fatal(err)
+	}
+
+	// flushing proto for which no state is present should return silently
+	if err := XfrmStateFlush(XFRM_PROTO_COMP); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := XfrmStateFlush(XFRM_PROTO_AH); err != nil {
+		t.Fatal(err)
+	}
+
+	if _, err := XfrmStateGet(state2); err == nil {
+		t.Fatalf("Unexpected success")
+	}
+
+	if err := XfrmStateAdd(state2); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := XfrmStateFlush(0); err != nil {
+		t.Fatal(err)
+	}
+
+	states, err := XfrmStateList(FAMILY_ALL)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(states) != 0 {
+		t.Fatal("State not flushed properly")
+	}
+
+}
+
+func TestXfrmStateUpdateLimits(t *testing.T) {
+	setUpNetlinkTest(t)()
+
+	// Program state with limits
+	state := getBaseState()
+	state.Limits.TimeHard = 3600
+	state.Limits.TimeSoft = 60
+	state.Limits.PacketHard = 1000
+	state.Limits.PacketSoft = 50
+	state.Limits.ByteHard = 1000000
+	state.Limits.ByteSoft = 50000
+	state.Limits.TimeUseHard = 3000
+	state.Limits.TimeUseSoft = 1500
+	if err := XfrmStateAdd(state); err != nil {
+		t.Fatal(err)
+	}
+	// Verify limits
+	s, err := XfrmStateGet(state)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !compareLimits(state, s) {
+		t.Fatalf("Incorrect time hard/soft retrieved: %s", s.Print(true))
+	}
+
+	// Update limits
+	state.Limits.TimeHard = 1800
+	state.Limits.TimeSoft = 30
+	state.Limits.PacketHard = 500
+	state.Limits.PacketSoft = 25
+	state.Limits.ByteHard = 500000
+	state.Limits.ByteSoft = 25000
+	state.Limits.TimeUseHard = 2000
+	state.Limits.TimeUseSoft = 1000
+	if err := XfrmStateUpdate(state); err != nil {
+		t.Fatal(err)
+	}
+
+	// Verify new limits
+	s, err = XfrmStateGet(state)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if s.Limits.TimeHard != 1800 || s.Limits.TimeSoft != 30 {
+		t.Fatalf("Incorrect time hard retrieved: (%d, %d)", s.Limits.TimeHard, s.Limits.TimeSoft)
+	}
+}
+
+func getBaseState() *XfrmState {
+	return &XfrmState{
 		Src:   net.ParseIP("127.0.0.1"),
 		Dst:   net.ParseIP("127.0.0.2"),
 		Proto: XFRM_PROTO_ESP,
@@ -23,28 +169,34 @@ func TestXfrmStateAddDel(t *testing.T) {
 			Name: "cbc(aes)",
 			Key:  []byte("abcdefghijklmnopqrstuvwzyzABCDEF"),
 		},
+		Mark: &XfrmMark{
+			Value: 0x12340000,
+			Mask:  0xffff0000,
+		},
 	}
-	if err := XfrmStateAdd(&state); err != nil {
-		t.Fatal(err)
-	}
-	policies, err := XfrmStateList(FAMILY_ALL)
-	if err != nil {
-		t.Fatal(err)
-	}
+}
 
-	if len(policies) != 1 {
-		t.Fatal("State not added properly")
+func compareStates(a, b *XfrmState) bool {
+	if a == b {
+		return true
 	}
-
-	if err = XfrmStateDel(&state); err != nil {
-		t.Fatal(err)
+	if a == nil || b == nil {
+		return false
 	}
+	return a.Src.Equal(b.Src) && a.Dst.Equal(b.Dst) &&
+		a.Mode == b.Mode && a.Spi == b.Spi && a.Proto == b.Proto &&
+		a.Auth.Name == b.Auth.Name && bytes.Equal(a.Auth.Key, b.Auth.Key) &&
+		a.Crypt.Name == b.Crypt.Name && bytes.Equal(a.Crypt.Key, b.Crypt.Key) &&
+		a.Mark.Value == b.Mark.Value && a.Mark.Mask == b.Mark.Mask
+}
 
-	policies, err = XfrmStateList(FAMILY_ALL)
-	if err != nil {
-		t.Fatal(err)
-	}
-	if len(policies) != 0 {
-		t.Fatal("State not removed properly")
-	}
+func compareLimits(a, b *XfrmState) bool {
+	return a.Limits.TimeHard == b.Limits.TimeHard &&
+		a.Limits.TimeSoft == b.Limits.TimeSoft &&
+		a.Limits.PacketHard == b.Limits.PacketHard &&
+		a.Limits.PacketSoft == b.Limits.PacketSoft &&
+		a.Limits.ByteHard == b.Limits.ByteHard &&
+		a.Limits.ByteSoft == b.Limits.ByteSoft &&
+		a.Limits.TimeUseHard == b.Limits.TimeUseHard &&
+		a.Limits.TimeUseSoft == b.Limits.TimeUseSoft
 }

vendor/github.com/vishvananda/netns/LICENSE

@@ -0,0 +1,192 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   Copyright 2014 Vishvananda Ishaya.
+   Copyright 2014 Docker, Inc.
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

vendor/github.com/vishvananda/netns/README.md

@@ -0,0 +1,50 @@
+# netns - network namespaces in go #
+
+The netns package provides an ultra-simple interface for handling
+network namespaces in go. Changing namespaces requires elevated
+privileges, so in most cases this code needs to be run as root.
+
+## Local Build and Test ##
+
+You can use go get command:
+
+    go get github.com/vishvananda/netns
+
+Testing (requires root):
+
+    sudo -E go test github.com/vishvananda/netns
+
+## Example ##
+
+```go
+package main
+
+import (
+    "fmt"
+    "net"
+    "runtime"
+    "github.com/vishvananda/netns"
+)
+
+func main() {
+    // Lock the OS Thread so we don't accidentally switch namespaces
+    runtime.LockOSThread()
+    defer runtime.UnlockOSThread()
+
+    // Save the current network namespace
+    origns, _ := netns.Get()
+    defer origns.Close()
+
+    // Create a new network namespace
+    newns, _ := netns.New()
+    defer newns.Close()
+
+    // Do something with the network namespace
+    ifaces, _ := net.Interfaces()
+    fmt.Printf("Interfaces: %v\n", ifaces)
+
+    // Switch back to the original namespace
+    netns.Set(origns)
+}
+
+```

vendor/github.com/vishvananda/netns/netns.go

@@ -0,0 +1,80 @@
+// Package netns allows ultra-simple network namespace handling. NsHandles
+// can be retrieved and set. Note that the current namespace is thread
+// local so actions that set and reset namespaces should use LockOSThread
+// to make sure the namespace doesn't change due to a goroutine switch.
+// It is best to close NsHandles when you are done with them. This can be
+// accomplished via a `defer ns.Close()` on the handle. Changing namespaces
+// requires elevated privileges, so in most cases this code needs to be run
+// as root.
+package netns
+
+import (
+	"fmt"
+	"syscall"
+)
+
+// NsHandle is a handle to a network namespace. It can be cast directly
+// to an int and used as a file descriptor.
+type NsHandle int
+
+// Equal determines if two network handles refer to the same network
+// namespace. This is done by comparing the device and inode that the
+// file descripors point to.
+func (ns NsHandle) Equal(other NsHandle) bool {
+	if ns == other {
+		return true
+	}
+	var s1, s2 syscall.Stat_t
+	if err := syscall.Fstat(int(ns), &s1); err != nil {
+		return false
+	}
+	if err := syscall.Fstat(int(other), &s2); err != nil {
+		return false
+	}
+	return (s1.Dev == s2.Dev) && (s1.Ino == s2.Ino)
+}
+
+// String shows the file descriptor number and its dev and inode.
+func (ns NsHandle) String() string {
+	var s syscall.Stat_t
+	if ns == -1 {
+		return "NS(None)"
+	}
+	if err := syscall.Fstat(int(ns), &s); err != nil {
+		return fmt.Sprintf("NS(%d: unknown)", ns)
+	}
+	return fmt.Sprintf("NS(%d: %d, %d)", ns, s.Dev, s.Ino)
+}
+
+// UniqueId returns a string which uniquely identifies the namespace
+// associated with the network handle.
+func (ns NsHandle) UniqueId() string {
+	var s syscall.Stat_t
+	if ns == -1 {
+		return "NS(none)"
+	}
+	if err := syscall.Fstat(int(ns), &s); err != nil {
+		return "NS(unknown)"
+	}
+	return fmt.Sprintf("NS(%d:%d)", s.Dev, s.Ino)
+}
+
+// IsOpen returns true if Close() has not been called.
+func (ns NsHandle) IsOpen() bool {
+	return ns != -1
+}
+
+// Close closes the NsHandle and resets its file descriptor to -1.
+// It is not safe to use an NsHandle after Close() is called.
+func (ns *NsHandle) Close() error {
+	if err := syscall.Close(int(*ns)); err != nil {
+		return err
+	}
+	(*ns) = -1
+	return nil
+}
+
+// None gets an empty (closed) NsHandle.
+func None() NsHandle {
+	return NsHandle(-1)
+}

vendor/github.com/vishvananda/netns/netns_linux.go

@@ -0,0 +1,220 @@
+// +build linux
+
+package netns
+
+import (
+	"fmt"
+	"io/ioutil"
+	"os"
+	"path/filepath"
+	"runtime"
+	"strconv"
+	"strings"
+	"syscall"
+)
+
+// SYS_SETNS syscall allows changing the namespace of the current process.
+var SYS_SETNS = map[string]uintptr{
+	"386":     346,
+	"amd64":   308,
+	"arm64":   268,
+	"arm":     375,
+	"ppc64":   350,
+	"ppc64le": 350,
+	"s390x":   339,
+}[runtime.GOARCH]
+
+// Deprecated: use syscall pkg instead (go >= 1.5 needed).
+const (
+	CLONE_NEWUTS  = 0x04000000 /* New utsname group? */
+	CLONE_NEWIPC  = 0x08000000 /* New ipcs */
+	CLONE_NEWUSER = 0x10000000 /* New user namespace */
+	CLONE_NEWPID  = 0x20000000 /* New pid namespace */
+	CLONE_NEWNET  = 0x40000000 /* New network namespace */
+	CLONE_IO      = 0x80000000 /* Get io context */
+)
+
+// Setns sets namespace using syscall. Note that this should be a method
+// in syscall but it has not been added.
+func Setns(ns NsHandle, nstype int) (err error) {
+	_, _, e1 := syscall.Syscall(SYS_SETNS, uintptr(ns), uintptr(nstype), 0)
+	if e1 != 0 {
+		err = e1
+	}
+	return
+}
+
+// Set sets the current network namespace to the namespace represented
+// by NsHandle.
+func Set(ns NsHandle) (err error) {
+	return Setns(ns, CLONE_NEWNET)
+}
+
+// New creates a new network namespace and returns a handle to it.
+func New() (ns NsHandle, err error) {
+	if err := syscall.Unshare(CLONE_NEWNET); err != nil {
+		return -1, err
+	}
+	return Get()
+}
+
+// Get gets a handle to the current threads network namespace.
+func Get() (NsHandle, error) {
+	return GetFromThread(os.Getpid(), syscall.Gettid())
+}
+
+// GetFromPath gets a handle to a network namespace
+// identified by the path
+func GetFromPath(path string) (NsHandle, error) {
+	fd, err := syscall.Open(path, syscall.O_RDONLY, 0)
+	if err != nil {
+		return -1, err
+	}
+	return NsHandle(fd), nil
+}
+
+// GetFromName gets a handle to a named network namespace such as one
+// created by `ip netns add`.
+func GetFromName(name string) (NsHandle, error) {
+	return GetFromPath(fmt.Sprintf("/var/run/netns/%s", name))
+}
+
+// GetFromPid gets a handle to the network namespace of a given pid.
+func GetFromPid(pid int) (NsHandle, error) {
+	return GetFromPath(fmt.Sprintf("/proc/%d/ns/net", pid))
+}
+
+// GetFromThread gets a handle to the network namespace of a given pid and tid.
+func GetFromThread(pid, tid int) (NsHandle, error) {
+	return GetFromPath(fmt.Sprintf("/proc/%d/task/%d/ns/net", pid, tid))
+}
+
+// GetFromDocker gets a handle to the network namespace of a docker container.
+// Id is prefixed matched against the running docker containers, so a short
+// identifier can be used as long as it isn't ambiguous.
+func GetFromDocker(id string) (NsHandle, error) {
+	pid, err := getPidForContainer(id)
+	if err != nil {
+		return -1, err
+	}
+	return GetFromPid(pid)
+}
+
+// borrowed from docker/utils/utils.go
+func findCgroupMountpoint(cgroupType string) (string, error) {
+	output, err := ioutil.ReadFile("/proc/mounts")
+	if err != nil {
+		return "", err
+	}
+
+	// /proc/mounts has 6 fields per line, one mount per line, e.g.
+	// cgroup /sys/fs/cgroup/devices cgroup rw,relatime,devices 0 0
+	for _, line := range strings.Split(string(output), "\n") {
+		parts := strings.Split(line, " ")
+		if len(parts) == 6 && parts[2] == "cgroup" {
+			for _, opt := range strings.Split(parts[3], ",") {
+				if opt == cgroupType {
+					return parts[1], nil
+				}
+			}
+		}
+	}
+
+	return "", fmt.Errorf("cgroup mountpoint not found for %s", cgroupType)
+}
+
+// Returns the relative path to the cgroup docker is running in.
+// borrowed from docker/utils/utils.go
+// modified to get the docker pid instead of using /proc/self
+func getThisCgroup(cgroupType string) (string, error) {
+	dockerpid, err := ioutil.ReadFile("/var/run/docker.pid")
+	if err != nil {
+		return "", err
+	}
+	result := strings.Split(string(dockerpid), "\n")
+	if len(result) == 0 || len(result[0]) == 0 {
+		return "", fmt.Errorf("docker pid not found in /var/run/docker.pid")
+	}
+	pid, err := strconv.Atoi(result[0])
+
+	output, err := ioutil.ReadFile(fmt.Sprintf("/proc/%d/cgroup", pid))
+	if err != nil {
+		return "", err
+	}
+	for _, line := range strings.Split(string(output), "\n") {
+		parts := strings.Split(line, ":")
+		// any type used by docker should work
+		if parts[1] == cgroupType {
+			return parts[2], nil
+		}
+	}
+	return "", fmt.Errorf("cgroup '%s' not found in /proc/%d/cgroup", cgroupType, pid)
+}
+
+// Returns the first pid in a container.
+// borrowed from docker/utils/utils.go
+// modified to only return the first pid
+// modified to glob with id
+// modified to search for newer docker containers
+func getPidForContainer(id string) (int, error) {
+	pid := 0
+
+	// memory is chosen randomly, any cgroup used by docker works
+	cgroupType := "memory"
+
+	cgroupRoot, err := findCgroupMountpoint(cgroupType)
+	if err != nil {
+		return pid, err
+	}
+
+	cgroupThis, err := getThisCgroup(cgroupType)
+	if err != nil {
+		return pid, err
+	}
+
+	id += "*"
+
+	attempts := []string{
+		filepath.Join(cgroupRoot, cgroupThis, id, "tasks"),
+		// With more recent lxc versions use, cgroup will be in lxc/
+		filepath.Join(cgroupRoot, cgroupThis, "lxc", id, "tasks"),
+		// With more recent docker, cgroup will be in docker/
+		filepath.Join(cgroupRoot, cgroupThis, "docker", id, "tasks"),
+		// Even more recent docker versions under systemd use docker-<id>.scope/
+		filepath.Join(cgroupRoot, "system.slice", "docker-"+id+".scope", "tasks"),
+		// Even more recent docker versions under cgroup/systemd/docker/<id>/
+		filepath.Join(cgroupRoot, "..", "systemd", "docker", id, "tasks"),
+	}
+
+	var filename string
+	for _, attempt := range attempts {
+		filenames, _ := filepath.Glob(attempt)
+		if len(filenames) > 1 {
+			return pid, fmt.Errorf("Ambiguous id supplied: %v", filenames)
+		} else if len(filenames) == 1 {
+			filename = filenames[0]
+			break
+		}
+	}
+
+	if filename == "" {
+		return pid, fmt.Errorf("Unable to find container: %v", id[:len(id)-1])
+	}
+
+	output, err := ioutil.ReadFile(filename)
+	if err != nil {
+		return pid, err
+	}
+
+	result := strings.Split(string(output), "\n")
+	if len(result) == 0 || len(result[0]) == 0 {
+		return pid, fmt.Errorf("No pid found for container")
+	}
+
+	pid, err = strconv.Atoi(result[0])
+	if err != nil {
+		return pid, fmt.Errorf("Invalid pid '%s': %s", result[0], err)
+	}
+
+	return pid, nil
+}

vendor/github.com/vishvananda/netns/netns_test.go

@@ -0,0 +1,66 @@
+package netns
+
+import (
+	"runtime"
+	"sync"
+	"testing"
+)
+
+func TestGetNewSetDelete(t *testing.T) {
+	runtime.LockOSThread()
+	defer runtime.UnlockOSThread()
+
+	origns, err := Get()
+	if err != nil {
+		t.Fatal(err)
+	}
+	newns, err := New()
+	if err != nil {
+		t.Fatal(err)
+	}
+	if origns.Equal(newns) {
+		t.Fatal("New ns failed")
+	}
+	if err := Set(origns); err != nil {
+		t.Fatal(err)
+	}
+	newns.Close()
+	if newns.IsOpen() {
+		t.Fatal("newns still open after close", newns)
+	}
+	ns, err := Get()
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !ns.Equal(origns) {
+		t.Fatal("Reset ns failed", origns, newns, ns)
+	}
+}
+
+func TestNone(t *testing.T) {
+	ns := None()
+	if ns.IsOpen() {
+		t.Fatal("None ns is open", ns)
+	}
+}
+
+func TestThreaded(t *testing.T) {
+	ncpu := runtime.GOMAXPROCS(-1)
+	if ncpu < 2 {
+		t.Skip("-cpu=2 or larger required")
+	}
+
+	// Lock this thread simply to ensure other threads get used.
+	runtime.LockOSThread()
+	defer runtime.UnlockOSThread()
+
+	wg := &sync.WaitGroup{}
+	for i := 0; i < ncpu; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			TestGetNewSetDelete(t)
+		}()
+	}
+	wg.Wait()
+}

vendor/github.com/vishvananda/netns/netns_unspecified.go

@@ -0,0 +1,35 @@
+// +build !linux
+
+package netns
+
+import (
+	"errors"
+)
+
+var (
+	ErrNotImplemented = errors.New("not implemented")
+)
+
+func Set(ns NsHandle) (err error) {
+	return ErrNotImplemented
+}
+
+func New() (ns NsHandle, err error) {
+	return -1, ErrNotImplemented
+}
+
+func Get() (NsHandle, error) {
+	return -1, ErrNotImplemented
+}
+
+func GetFromName(name string) (NsHandle, error) {
+	return -1, ErrNotImplemented
+}
+
+func GetFromPid(pid int) (NsHandle, error) {
+	return -1, ErrNotImplemented
+}
+
+func GetFromDocker(id string) (NsHandle, error) {
+	return -1, ErrNotImplemented
+}