Update vishvananda/netlink and add vishvananda/netns

This commit changes vishvananda/netlink binds godep version to:
  49a735373919c4c9a53aff1f9f63da73a243f32d
And adds vishvananda/netns with version:
  8ba1072b58e0c2a240eb5f6120165c7776c3e7b8

Godeps/Godeps.json

@@ -2042,11 +2042,11 @@
 		},
 		{
 			"ImportPath": "github.com/vishvananda/netlink",
-			"Rev": "1e2e08e8a2dcdacaae3f14ac44c5cfa31361f270"
+			"Rev": "49a735373919c4c9a53aff1f9f63da73a243f32d"
 		},
 		{
 			"ImportPath": "github.com/vishvananda/netlink/nl",
-			"Rev": "1e2e08e8a2dcdacaae3f14ac44c5cfa31361f270"
+			"Rev": "49a735373919c4c9a53aff1f9f63da73a243f32d"
 		},
 		{
 			"ImportPath": "github.com/vmware/govmomi",
@@ -2361,6 +2361,10 @@
 			"ImportPath": "k8s.io/heapster/metrics/apis/metrics/v1alpha1",
 			"Comment": "v1.1.0-beta2",
 			"Rev": "9cb18ac0ceb193eb530a1fe339355c94ea454d85"
+		},
+		{
+			"ImportPath": "github.com/vishvananda/netns",
+			"Rev": "8ba1072b58e0c2a240eb5f6120165c7776c3e7b8"
 		}
 	]
 }

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

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

@@ -11,11 +11,13 @@ import (
 type Addr struct {
 	*net.IPNet
 	Label string
+	Flags int
+	Scope int
 }
 
 // String returns $ip/$netmask $label
 func (a Addr) String() string {
-	return fmt.Sprintf("%s %s", a.IPNet, a.Label)
+	return strings.TrimSpace(fmt.Sprintf("%s %s", a.IPNet, a.Label))
 }
 
 // ParseAddr parses the string representation of an address in the

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

@@ -2,39 +2,56 @@ package netlink
 
 import (
 	"fmt"
+	"log"
 	"net"
 	"strings"
 	"syscall"
 
 	"github.com/vishvananda/netlink/nl"
+	"github.com/vishvananda/netns"
 )
 
+// 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)
+// AddrAdd will add an IP address to a link device.
-	return addrHandle(link, addr, req)
+// 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 pkgHandle.AddrDel(link, addr)
-	return addrHandle(link, addr, req)
 }
 
-func addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error {
+// 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 (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 +69,17 @@ 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 +93,14 @@ 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)
 
@@ -74,29 +109,51 @@ func AddrList(link Link, family int) ([]Addr, error) {
 		return nil, err
 	}
 
-	index := 0
+	indexFilter := 0
 	if link != nil {
 		base := link.Attrs()
-		ensureIndex(base)
+		h.ensureIndex(base)
-		index = base.Index
+		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 family != FAMILY_ALL && msgFamily != family {
-		if err != nil {
+			continue
-			return nil, err
 		}
 
+		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
-		var addr Addr
 	for _, attr := range attrs {
 		switch attr.Attr.Type {
 		case syscall.IFA_ADDRESS:
@@ -111,6 +168,8 @@ func AddrList(link Link, family int) ([]Addr, error) {
 			}
 		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]))
 		}
 	}
 
@@ -120,9 +179,65 @@ func AddrList(link Link, family int) ([]Addr, error) {
 	} 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 {
+	return addrSubscribe(netns.None(), netns.None(), ch, done)
+}
+
+// AddrSubscribeAt works like AddrSubscribe plus it allows the caller
+// to choose the network namespace in which to subscribe (ns).
+func AddrSubscribeAt(ns netns.NsHandle, ch chan<- AddrUpdate, done <-chan struct{}) error {
+	return addrSubscribe(ns, netns.None(), ch, done)
+}
+
+func addrSubscribe(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-chan struct{}) error {
+	s, err := nl.SubscribeAt(newNs, curNs, 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/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

@@ -0,0 +1,78 @@
+package netlink
+
+import (
+	"fmt"
+)
+
+type Class interface {
+	Attrs() *ClassAttrs
+	Type() string
+}
+
+// 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 {
+	LinkIndex int
+	Handle    uint32
+	Parent    uint32
+	Leaf      uint32
+}
+
+func (q ClassAttrs) String() string {
+	return fmt.Sprintf("{LinkIndex: %d, Handle: %s, Parent: %s, Leaf: %d}", q.LinkIndex, HandleStr(q.Handle), HandleStr(q.Parent), q.Leaf)
+}
+
+type HtbClassAttrs struct {
+	// TODO handle all attributes
+	Rate    uint64
+	Ceil    uint64
+	Buffer  uint32
+	Cbuffer uint32
+	Quantum uint32
+	Level   uint32
+	Prio    uint32
+}
+
+func (q HtbClassAttrs) String() string {
+	return fmt.Sprintf("{Rate: %d, Ceil: %d, Buffer: %d, Cbuffer: %d}", q.Rate, q.Ceil, q.Buffer, q.Cbuffer)
+}
+
+// HtbClass represents an Htb class
+type HtbClass struct {
+	ClassAttrs
+	Rate    uint64
+	Ceil    uint64
+	Buffer  uint32
+	Cbuffer uint32
+	Quantum uint32
+	Level   uint32
+	Prio    uint32
+}
+
+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 (q *HtbClass) Attrs() *ClassAttrs {
+	return &q.ClassAttrs
+}
+
+func (q *HtbClass) Type() string {
+	return "htb"
+}
+
+// GenericClass classes represent types that are not currently understood
+// by this netlink library.
+type GenericClass struct {
+	ClassAttrs
+	ClassType string
+}
+
+func (class *GenericClass) Attrs() *ClassAttrs {
+	return &class.ClassAttrs
+}
+
+func (class *GenericClass) Type() string {
+	return class.ClassType
+}

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

@@ -0,0 +1,254 @@
+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 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 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.
+// 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 ClassReplace(class Class) error {
+	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 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 (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,
+		Ifindex: int32(base.LinkIndex),
+		Handle:  base.Handle,
+		Parent:  base.Parent,
+	}
+	req.AddData(msg)
+
+	if cmd != syscall.RTM_DELTCLASS {
+		if err := classPayload(req, class); err != nil {
+			return err
+		}
+	}
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
+func classPayload(req *nl.NetlinkRequest, class Class) error {
+	req.AddData(nl.NewRtAttr(nl.TCA_KIND, nl.ZeroTerminated(class.Type())))
+
+	options := nl.NewRtAttr(nl.TCA_OPTIONS, nil)
+	if htb, ok := class.(*HtbClass); ok {
+		opt := nl.TcHtbCopt{}
+		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
+}
+
+// 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 ClassList(link Link, parent uint32) ([]Class, error) {
+	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()
+		h.ensureIndex(base)
+		msg.Ifindex = int32(base.Index)
+	}
+	req.AddData(msg)
+
+	msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWTCLASS)
+	if err != nil {
+		return nil, err
+	}
+
+	var res []Class
+	for _, m := range msgs {
+		msg := nl.DeserializeTcMsg(m)
+
+		attrs, err := nl.ParseRouteAttr(m[msg.Len():])
+		if err != nil {
+			return nil, err
+		}
+
+		base := ClassAttrs{
+			LinkIndex: int(msg.Ifindex),
+			Handle:    msg.Handle,
+			Parent:    msg.Parent,
+		}
+
+		var class Class
+		classType := ""
+		for _, attr := range attrs {
+			switch attr.Attr.Type {
+			case nl.TCA_KIND:
+				classType = string(attr.Value[:len(attr.Value)-1])
+				switch classType {
+				case "htb":
+					class = &HtbClass{}
+				default:
+					class = &GenericClass{ClassType: classType}
+				}
+			case nl.TCA_OPTIONS:
+				switch classType {
+				case "htb":
+					data, err := nl.ParseRouteAttr(attr.Value)
+					if err != nil {
+						return nil, err
+					}
+					_, err = parseHtbClassData(class, data)
+					if err != nil {
+						return nil, err
+					}
+				}
+			}
+		}
+		*class.Attrs() = base
+		res = append(res, class)
+	}
+
+	return res, nil
+}
+
+func parseHtbClassData(class Class, data []syscall.NetlinkRouteAttr) (bool, error) {
+	htb := class.(*HtbClass)
+	detailed := false
+	for _, datum := range data {
+		switch datum.Attr.Type {
+		case nl.TCA_HTB_PARMS:
+			opt := nl.DeserializeTcHtbCopt(datum.Value)
+			htb.Rate = uint64(opt.Rate.Rate)
+			htb.Ceil = uint64(opt.Ceil.Rate)
+			htb.Buffer = opt.Buffer
+			htb.Cbuffer = opt.Cbuffer
+			htb.Quantum = opt.Quantum
+			htb.Level = opt.Level
+			htb.Prio = opt.Prio
+		}
+	}
+	return detailed, nil
+}

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

@@ -1,15 +1,13 @@
 package netlink
 
-import (
+import "fmt"
-	"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 {
@@ -24,11 +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 {
@@ -39,6 +196,38 @@ func (filter *U32) Type() string {
 	return "u32"
 }
 
+type FilterFwAttrs struct {
+	ClassId   uint32
+	InDev     string
+	Mask      uint32
+	Index     uint32
+	Buffer    uint32
+	Mtu       uint32
+	Mpu       uint16
+	Rate      uint32
+	AvRate    uint32
+	PeakRate  uint32
+	Action    TcPolAct
+	Overhead  uint16
+	LinkLayer int
+}
+
+type BpfFilter struct {
+	FilterAttrs
+	ClassId      uint32
+	Fd           int
+	Name         string
+	DirectAction bool
+}
+
+func (filter *BpfFilter) Type() string {
+	return "bpf"
+}
+
+func (filter *BpfFilter) Attrs() *FilterAttrs {
+	return &filter.FilterAttrs
+}
+
 // GenericFilter filters represent types that are not currently understood
 // by this netlink library.
 type GenericFilter struct {

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

@@ -1,16 +1,95 @@
 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,
@@ -28,7 +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 {
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWTFILTER, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	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 := 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,
@@ -49,18 +135,61 @@ 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)
+		if u32.ClassId != 0 {
-		table := nl.NewRtAttrChild(actions, nl.TCA_ACT_TAB, nil)
+			nl.NewRtAttrChild(options, nl.TCA_U32_CLASSID, nl.Uint32Attr(u32.ClassId))
-		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),
 		}
-		aopts := nl.NewRtAttrChild(table, nl.TCA_OPTIONS, nil)
+		actionsAttr := nl.NewRtAttrChild(options, nl.TCA_U32_ACT, nil)
-		nl.NewRtAttrChild(aopts, nl.TCA_MIRRED_PARMS, mir.Serialize())
+		// 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
+		}
+	} else if fw, ok := filter.(*Fw); ok {
+		if fw.Mask != 0 {
+			b := make([]byte, 4)
+			native.PutUint32(b, fw.Mask)
+			nl.NewRtAttrChild(options, nl.TCA_FW_MASK, b)
+		}
+		if fw.InDev != "" {
+			nl.NewRtAttrChild(options, nl.TCA_FW_INDEV, nl.ZeroTerminated(fw.InDev))
+		}
+		if (fw.Police != nl.TcPolice{}) {
+
+			police := nl.NewRtAttrChild(options, nl.TCA_FW_POLICE, nil)
+			nl.NewRtAttrChild(police, nl.TCA_POLICE_TBF, fw.Police.Serialize())
+			if (fw.Police.Rate != nl.TcRateSpec{}) {
+				payload := SerializeRtab(fw.Rtab)
+				nl.NewRtAttrChild(police, nl.TCA_POLICE_RATE, payload)
+			}
+			if (fw.Police.PeakRate != nl.TcRateSpec{}) {
+				payload := SerializeRtab(fw.Ptab)
+				nl.NewRtAttrChild(police, nl.TCA_POLICE_PEAKRATE, payload)
+			}
+		}
+		if fw.ClassId != 0 {
+			b := make([]byte, 4)
+			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)
 	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
 	return err
@@ -68,16 +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)
@@ -114,20 +250,36 @@ func FilterList(link Link, parent uint32) ([]Filter, error) {
 				switch filterType {
 				case "u32":
 					filter = &U32{}
+				case "fw":
+					filter = &Fw{}
+				case "bpf":
+					filter = &BpfFilter{}
 				default:
 					filter = &GenericFilter{FilterType: filterType}
 				}
 			case nl.TCA_OPTIONS:
-				switch filterType {
-				case "u32":
 				data, err := nl.ParseRouteAttr(attr.Value)
 				if err != nil {
 					return nil, err
 				}
+				switch filterType {
+				case "u32":
 					detailed, err = parseU32Data(filter, data)
 					if err != nil {
 						return nil, err
 					}
+				case "fw":
+					detailed, err = parseFwData(filter, data)
+					if err != nil {
+						return nil, err
+					}
+				case "bpf":
+					detailed, err = parseBpfData(filter, data)
+					if err != nil {
+						return nil, err
+					}
+				default:
+					detailed = true
 				}
 			}
 		}
@@ -141,6 +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)
@@ -155,37 +430,131 @@ 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 {
+			u32.Actions, err = parseActions(tables)
-				return detailed, fmt.Errorf("Action table not formed properly")
-			}
-			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 {
+			for _, action := range u32.Actions {
-						switch adatum.Attr.Type {
+				if action, ok := action.(*MirredAction); ok {
-						case nl.TCA_MIRRED_PARMS:
+					u32.RedirIndex = int(action.Ifindex)
-							mir := nl.DeserializeTcMirred(adatum.Value)
-							u32.RedirIndex = int(mir.Ifindex)
-						}
-					}
 				}
 			}
 		}
 	}
 	return detailed, nil
 }
+
+func parseFwData(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error) {
+	native = nl.NativeEndian()
+	fw := filter.(*Fw)
+	detailed := true
+	for _, datum := range data {
+		switch datum.Attr.Type {
+		case nl.TCA_FW_MASK:
+			fw.Mask = native.Uint32(datum.Value[0:4])
+		case nl.TCA_FW_CLASSID:
+			fw.ClassId = native.Uint32(datum.Value[0:4])
+		case nl.TCA_FW_INDEV:
+			fw.InDev = string(datum.Value[:len(datum.Value)-1])
+		case nl.TCA_FW_POLICE:
+			adata, _ := nl.ParseRouteAttr(datum.Value)
+			for _, aattr := range adata {
+				switch aattr.Attr.Type {
+				case nl.TCA_POLICE_TBF:
+					fw.Police = *nl.DeserializeTcPolice(aattr.Value)
+				case nl.TCA_POLICE_RATE:
+					fw.Rtab = DeserializeRtab(aattr.Value)
+				case nl.TCA_POLICE_PEAKRATE:
+					fw.Ptab = DeserializeRtab(aattr.Value)
+				}
+			}
+		}
+	}
+	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)
+	if (size % nl.ATM_CELL_PAYLOAD) > 0 {
+		cells++
+	}
+	linksize = cells * nl.ATM_CELL_SIZE
+	return uint(linksize)
+}
+
+func AdjustSize(sz uint, mpu uint, linklayer int) uint {
+	if sz < mpu {
+		sz = mpu
+	}
+	switch linklayer {
+	case nl.LINKLAYER_ATM:
+		return AlignToAtm(sz)
+	default:
+		return sz
+	}
+}
+
+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 cellLog < 0 {
+		cellLog = 0
+		for (mtu >> uint(cellLog)) > 255 {
+			cellLog++
+		}
+	}
+	for i := 0; i < 256; i++ {
+		sz = AdjustSize(uint((i+1)<<uint32(cellLog)), uint(mpu), linklayer)
+		rtab[i] = uint32(Xmittime(uint64(bps), uint32(sz)))
+	}
+	rate.CellAlign = -1
+	rate.CellLog = uint8(cellLog)
+	rate.Linklayer = uint8(linklayer & nl.TC_LINKLAYER_MASK)
+	return cellLog
+}
+
+func DeserializeRtab(b []byte) [256]uint32 {
+	var rtab [256]uint32
+	native := nl.NativeEndian()
+	r := bytes.NewReader(b)
+	_ = binary.Read(r, native, &rtab)
+	return rtab
+}
+
+func SerializeRtab(rtab [256]uint32) []byte {
+	native := nl.NativeEndian()
+	var w bytes.Buffer
+	_ = binary.Write(&w, native, rtab)
+	return w.Bytes()
+}

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

@@ -0,0 +1,86 @@
+package netlink
+
+import (
+	"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 on the
+// same netlink family share the same netlink socket,
+// which gets released when the handle is deleted.
+type Handle struct {
+	sockets      map[int]*nl.SocketHandle
+	lookupByDump bool
+}
+
+// SupportsNetlinkFamily reports whether the passed netlink family is supported by this Handle
+func (h *Handle) SupportsNetlinkFamily(nlFamily int) bool {
+	_, ok := h.sockets[nlFamily]
+	return ok
+}
+
+// NewHandle returns a netlink handle on the current network namespace.
+// Caller may specify the netlink families the handle should support.
+// If no families are specified, all the families the netlink package
+// supports will be automatically added.
+func NewHandle(nlFamilies ...int) (*Handle, error) {
+	return newHandle(netns.None(), netns.None(), nlFamilies...)
+}
+
+// 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, nlFamilies ...int) (*Handle, error) {
+	return newHandle(ns, netns.None(), nlFamilies...)
+}
+
+// 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, nlFamilies ...int) (*Handle, error) {
+	h := &Handle{sockets: map[int]*nl.SocketHandle{}}
+	fams := nl.SupportedNlFamilies
+	if len(nlFamilies) != 0 {
+		fams = nlFamilies
+	}
+	for _, f := range fams {
+		s, err := nl.GetNetlinkSocketAt(newNs, curNs, f)
+		if err != nil {
+			return nil, err
+		}
+		h.sockets[f] = &nl.SocketHandle{Socket: s}
+	}
+	return h, nil
+}
+
+// Delete releases the resources allocated to this handle
+func (h *Handle) Delete() {
+	for _, sh := range h.sockets {
+		sh.Close()
+	}
+	h.sockets = nil
+}
+
+func (h *Handle) newNetlinkRequest(proto, flags int) *nl.NetlinkRequest {
+	// Do this so that package API still use nl package variable nextSeqNr
+	if h.sockets == 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),
+		},
+		Sockets: h.sockets,
+	}
+}

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

@@ -1,6 +1,9 @@
 package netlink
 
-import "net"
+import (
+	"fmt"
+	"net"
+)
 
 // Link represents a link device from netlink. Shared link attributes
 // like name may be retrieved using the Attrs() method. Unique data
@@ -26,6 +29,9 @@ 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
+	Promisc      int
 }
 
 // NewLinkAttrs returns LinkAttrs structure filled with default values
@@ -35,6 +41,35 @@ 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 {
@@ -136,6 +171,24 @@ func (macvtap Macvtap) Type() string {
 	return "macvtap"
 }
 
+type TuntapMode uint16
+type TuntapFlag uint16
+
+// Tuntap links created via /dev/tun/tap, but can be destroyed via netlink
+type Tuntap struct {
+	LinkAttrs
+	Mode  TuntapMode
+	Flags TuntapFlag
+}
+
+func (tuntap *Tuntap) Attrs() *LinkAttrs {
+	return &tuntap.LinkAttrs
+}
+
+func (tuntap *Tuntap) Type() string {
+	return "tuntap"
+}
+
 // Veth devices must specify PeerName on create
 type Veth struct {
 	LinkAttrs
@@ -178,6 +231,7 @@ type Vxlan struct {
 	RSC          bool
 	L2miss       bool
 	L3miss       bool
+	UDPCSum      bool
 	NoAge        bool
 	GBP          bool
 	Age          int
@@ -216,6 +270,316 @@ 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_tuntap_linux.go

@@ -0,0 +1,14 @@
+package netlink
+
+// ideally golang.org/x/sys/unix would define IfReq but it only has
+// IFNAMSIZ, hence this minimalistic implementation
+const (
+	SizeOfIfReq = 40
+	IFNAMSIZ    = 16
+)
+
+type ifReq struct {
+	Name  [IFNAMSIZ]byte
+	Flags uint16
+	pad   [SizeOfIfReq - IFNAMSIZ - 2]byte
+}

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 or replace an IP to MAC mapping to the ARP table
+// 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)
+	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)
 }
 
@@ -130,9 +162,17 @@ 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),
 	}
 	req.AddData(&msg)
 

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

@@ -8,18 +8,7 @@
 // interface that is loosly modeled on the iproute2 cli.
 package netlink
 
-import (
+import "net"
-	"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
-)
 
 // 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 {
+	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) {
+func NeighDeserialize(m []byte) (*Neigh, error) {
-	return nil, nil, ErrNotImplemented
+	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/nl_linux.go

@@ -6,9 +6,13 @@ import (
 	"encoding/binary"
 	"fmt"
 	"net"
+	"runtime"
+	"sync"
 	"sync/atomic"
 	"syscall"
 	"unsafe"
+
+	"github.com/vishvananda/netns"
 )
 
 const (
@@ -18,6 +22,9 @@ const (
 	FAMILY_V6  = syscall.AF_INET6
 )
 
+// SupportedNlFamilies contains the list of netlink families this netlink package supports
+var SupportedNlFamilies = []int{syscall.NETLINK_ROUTE, syscall.NETLINK_XFRM}
+
 var nextSeqNr uint32
 
 // GetIPFamily returns the family type of a net.IP.
@@ -149,10 +156,12 @@ func (a *RtAttr) Serialize() []byte {
 	length := a.Len()
 	buf := make([]byte, rtaAlignOf(length))
 
-	if a.Data != nil {
-		copy(buf[4:], a.Data)
-	} else {
 	next := 4
+	if a.Data != nil {
+		copy(buf[next:], a.Data)
+		next += rtaAlignOf(len(a.Data))
+	}
+	if len(a.children) > 0 {
 		for _, child := range a.children {
 			childBuf := child.Serialize()
 			copy(buf[next:], childBuf)
@@ -170,6 +179,7 @@ func (a *RtAttr) Serialize() []byte {
 type NetlinkRequest struct {
 	syscall.NlMsghdr
 	Data    []NetlinkRequestData
+	Sockets map[int]*SocketHandle
 }
 
 // Serialize the Netlink Request into a byte array
@@ -201,14 +211,32 @@ 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)
+	var (
+		s   *NetlinkSocket
+		err error
+	)
+
+	if req.Sockets != nil {
+		if sh, ok := req.Sockets[sockType]; ok {
+			s = sh.Socket
+			req.Seq = atomic.AddUint32(&sh.Seq, 1)
+		}
+	}
+	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()
+	}
 
 	if err := s.Send(req); err != nil {
 		return nil, err
@@ -229,7 +257,10 @@ 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)
@@ -274,6 +305,7 @@ func NewNetlinkRequest(proto, flags int) *NetlinkRequest {
 type NetlinkSocket struct {
 	fd  int
 	lsa syscall.SockaddrNetlink
+	sync.Mutex
 }
 
 func getNetlinkSocket(protocol int) (*NetlinkSocket, error) {
@@ -293,6 +325,71 @@ 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) {
+	c, err := executeInNetns(newNs, curNs)
+	if err != nil {
+		return nil, err
+	}
+	defer c()
+	return getNetlinkSocket(protocol)
+}
+
+// executeInNetns sets execution of the code following this call to the
+// network namespace newNs, then moves the thread back to curNs if open,
+// otherwise to the current netns at the time the function was invoked
+// In case of success, the caller is expected to execute the returned function
+// at the end of the code that needs to be executed in the network namespace.
+// Example:
+// func jobAt(...) error {
+//      d, err := executeInNetns(...)
+//      if err != nil { return err}
+//      defer d()
+//      < code which needs to be executed in specific netns>
+//  }
+// TODO: his function probably belongs to netns pkg.
+func executeInNetns(newNs, curNs netns.NsHandle) (func(), error) {
+	var (
+		err       error
+		moveBack  func(netns.NsHandle) error
+		closeNs   func() error
+		unlockThd func()
+	)
+	restore := func() {
+		// order matters
+		if moveBack != nil {
+			moveBack(curNs)
+		}
+		if closeNs != nil {
+			closeNs()
+		}
+		if unlockThd != nil {
+			unlockThd()
+		}
+	}
+	if newNs.IsOpen() {
+		runtime.LockOSThread()
+		unlockThd = runtime.UnlockOSThread
+		if !curNs.IsOpen() {
+			if curNs, err = netns.Get(); err != nil {
+				restore()
+				return nil, fmt.Errorf("could not get current namespace while creating netlink socket: %v", err)
+			}
+			closeNs = curNs.Close
+		}
+		if err := netns.Set(newNs); err != nil {
+			restore()
+			return nil, fmt.Errorf("failed to set into network namespace %d while creating netlink socket: %v", newNs, err)
+		}
+		moveBack = netns.Set
+	}
+	return restore, nil
+}
+
 // 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
@@ -319,11 +416,31 @@ func Subscribe(protocol int, groups ...uint) (*NetlinkSocket, error) {
 	return s, nil
 }
 
+// SubscribeAt works like Subscribe plus let's the caller choose the network
+// namespace in which the socket would be opened (newNs). Then control goes back
+// to curNs if open, otherwise to the netns at the time this function was called.
+func SubscribeAt(newNs, curNs netns.NsHandle, protocol int, groups ...uint) (*NetlinkSocket, error) {
+	c, err := executeInNetns(newNs, curNs)
+	if err != nil {
+		return nil, err
+	}
+	defer c()
+	return Subscribe(protocol, groups...)
+}
+
 func (s *NetlinkSocket) Close() {
 	syscall.Close(s.fd)
+	s.fd = -1
+}
+
+func (s *NetlinkSocket) GetFd() int {
+	return s.fd
 }
 
 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
 	}
@@ -331,6 +448,9 @@ 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 {
@@ -416,3 +536,17 @@ func netlinkRouteAttrAndValue(b []byte) (*syscall.RtAttr, []byte, int, error) {
 	}
 	return a, b[syscall.SizeofRtAttr:], rtaAlignOf(int(a.Len)), nil
 }
+
+// SocketHandle contains the netlink socket and the associated
+// sequence counter for a specific netlink family
+type SocketHandle struct {
+	Seq    uint32
+	Socket *NetlinkSocket
+}
+
+// Close closes the netlink socket
+func (sh *SocketHandle) Close() {
+	if sh.Socket != nil {
+		sh.Socket.Close()
+	}
+}

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

@@ -4,6 +4,32 @@ import (
 	"unsafe"
 )
 
+// LinkLayer
+const (
+	LINKLAYER_UNSPEC = iota
+	LINKLAYER_ETHERNET
+	LINKLAYER_ATM
+)
+
+// ATM
+const (
+	ATM_CELL_PAYLOAD = 48
+	ATM_CELL_SIZE    = 53
+)
+
+const TC_LINKLAYER_MASK = 0x0F
+
+// Police
+const (
+	TCA_POLICE_UNSPEC = iota
+	TCA_POLICE_TBF
+	TCA_POLICE_RATE
+	TCA_POLICE_PEAKRATE
+	TCA_POLICE_AVRATE
+	TCA_POLICE_RESULT
+	TCA_POLICE_MAX = TCA_POLICE_RESULT
+)
+
 // Message types
 const (
 	TCA_UNSPEC = iota
@@ -23,6 +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
@@ -34,10 +69,18 @@ const (
 	SizeofTcActionMsg    = 0x04
 	SizeofTcPrioMap      = 0x14
 	SizeofTcRateSpec     = 0x0c
+	SizeofTcNetemQopt    = 0x18
+	SizeofTcNetemCorr    = 0x0c
+	SizeofTcNetemReorder = 0x08
+	SizeofTcNetemCorrupt = 0x08
 	SizeofTcTbfQopt      = 2*SizeofTcRateSpec + 0x0c
+	SizeofTcHtbCopt      = 2*SizeofTcRateSpec + 0x14
+	SizeofTcHtbGlob      = 0x14
 	SizeofTcU32Key       = 0x10
 	SizeofTcU32Sel       = 0x10 // without keys
-	SizeofTcMirred    = 0x1c
+	SizeofTcGen          = 0x14
+	SizeofTcMirred       = SizeofTcGen + 0x08
+	SizeofTcPolice       = 2*SizeofTcRateSpec + 0x20
 )
 
 // struct tcmsg {
@@ -162,6 +205,121 @@ func (x *TcRateSpec) Serialize() []byte {
 	return (*(*[SizeofTcRateSpec]byte)(unsafe.Pointer(x)))[:]
 }
 
+/**
+* NETEM
+ */
+
+const (
+	TCA_NETEM_UNSPEC = iota
+	TCA_NETEM_CORR
+	TCA_NETEM_DELAY_DIST
+	TCA_NETEM_REORDER
+	TCA_NETEM_CORRUPT
+	TCA_NETEM_LOSS
+	TCA_NETEM_RATE
+	TCA_NETEM_ECN
+	TCA_NETEM_RATE64
+	TCA_NETEM_MAX = TCA_NETEM_RATE64
+)
+
+// struct tc_netem_qopt {
+//	__u32	latency;	/* added delay (us) */
+//	__u32   limit;		/* fifo limit (packets) */
+//	__u32	loss;		/* random packet loss (0=none ~0=100%) */
+//	__u32	gap;		/* re-ordering gap (0 for none) */
+//	__u32   duplicate;	/* random packet dup  (0=none ~0=100%) */
+// 	__u32	jitter;		/* random jitter in latency (us) */
+// };
+
+type TcNetemQopt struct {
+	Latency   uint32
+	Limit     uint32
+	Loss      uint32
+	Gap       uint32
+	Duplicate uint32
+	Jitter    uint32
+}
+
+func (msg *TcNetemQopt) Len() int {
+	return SizeofTcNetemQopt
+}
+
+func DeserializeTcNetemQopt(b []byte) *TcNetemQopt {
+	return (*TcNetemQopt)(unsafe.Pointer(&b[0:SizeofTcNetemQopt][0]))
+}
+
+func (x *TcNetemQopt) Serialize() []byte {
+	return (*(*[SizeofTcNetemQopt]byte)(unsafe.Pointer(x)))[:]
+}
+
+// struct tc_netem_corr {
+//  __u32   delay_corr; /* delay correlation */
+//  __u32   loss_corr;  /* packet loss correlation */
+//  __u32   dup_corr;   /* duplicate correlation  */
+// };
+
+type TcNetemCorr struct {
+	DelayCorr uint32
+	LossCorr  uint32
+	DupCorr   uint32
+}
+
+func (msg *TcNetemCorr) Len() int {
+	return SizeofTcNetemCorr
+}
+
+func DeserializeTcNetemCorr(b []byte) *TcNetemCorr {
+	return (*TcNetemCorr)(unsafe.Pointer(&b[0:SizeofTcNetemCorr][0]))
+}
+
+func (x *TcNetemCorr) Serialize() []byte {
+	return (*(*[SizeofTcNetemCorr]byte)(unsafe.Pointer(x)))[:]
+}
+
+// struct tc_netem_reorder {
+//  __u32   probability;
+//  __u32   correlation;
+// };
+
+type TcNetemReorder struct {
+	Probability uint32
+	Correlation uint32
+}
+
+func (msg *TcNetemReorder) Len() int {
+	return SizeofTcNetemReorder
+}
+
+func DeserializeTcNetemReorder(b []byte) *TcNetemReorder {
+	return (*TcNetemReorder)(unsafe.Pointer(&b[0:SizeofTcNetemReorder][0]))
+}
+
+func (x *TcNetemReorder) Serialize() []byte {
+	return (*(*[SizeofTcNetemReorder]byte)(unsafe.Pointer(x)))[:]
+}
+
+// struct tc_netem_corrupt {
+//  __u32   probability;
+//  __u32   correlation;
+// };
+
+type TcNetemCorrupt struct {
+	Probability uint32
+	Correlation uint32
+}
+
+func (msg *TcNetemCorrupt) Len() int {
+	return SizeofTcNetemCorrupt
+}
+
+func DeserializeTcNetemCorrupt(b []byte) *TcNetemCorrupt {
+	return (*TcNetemCorrupt)(unsafe.Pointer(&b[0:SizeofTcNetemCorrupt][0]))
+}
+
+func (x *TcNetemCorrupt) Serialize() []byte {
+	return (*(*[SizeofTcNetemCorrupt]byte)(unsafe.Pointer(x)))[:]
+}
+
 // struct tc_tbf_qopt {
 //   struct tc_ratespec rate;
 //   struct tc_ratespec peakrate;
@@ -190,6 +348,70 @@ func (x *TcTbfQopt) Serialize() []byte {
 	return (*(*[SizeofTcTbfQopt]byte)(unsafe.Pointer(x)))[:]
 }
 
+const (
+	TCA_HTB_UNSPEC = iota
+	TCA_HTB_PARMS
+	TCA_HTB_INIT
+	TCA_HTB_CTAB
+	TCA_HTB_RTAB
+	TCA_HTB_DIRECT_QLEN
+	TCA_HTB_RATE64
+	TCA_HTB_CEIL64
+	TCA_HTB_MAX = TCA_HTB_CEIL64
+)
+
+//struct tc_htb_opt {
+//	struct tc_ratespec	rate;
+//	struct tc_ratespec	ceil;
+//	__u32	buffer;
+//	__u32	cbuffer;
+//	__u32	quantum;
+//	__u32	level;		/* out only */
+//	__u32	prio;
+//};
+
+type TcHtbCopt struct {
+	Rate    TcRateSpec
+	Ceil    TcRateSpec
+	Buffer  uint32
+	Cbuffer uint32
+	Quantum uint32
+	Level   uint32
+	Prio    uint32
+}
+
+func (msg *TcHtbCopt) Len() int {
+	return SizeofTcHtbCopt
+}
+
+func DeserializeTcHtbCopt(b []byte) *TcHtbCopt {
+	return (*TcHtbCopt)(unsafe.Pointer(&b[0:SizeofTcHtbCopt][0]))
+}
+
+func (x *TcHtbCopt) Serialize() []byte {
+	return (*(*[SizeofTcHtbCopt]byte)(unsafe.Pointer(x)))[:]
+}
+
+type TcHtbGlob struct {
+	Version      uint32
+	Rate2Quantum uint32
+	Defcls       uint32
+	Debug        uint32
+	DirectPkts   uint32
+}
+
+func (msg *TcHtbGlob) Len() int {
+	return SizeofTcHtbGlob
+}
+
+func DeserializeTcHtbGlob(b []byte) *TcHtbGlob {
+	return (*TcHtbGlob)(unsafe.Pointer(&b[0:SizeofTcHtbGlob][0]))
+}
+
+func (x *TcHtbGlob) Serialize() []byte {
+	return (*(*[SizeofTcHtbGlob]byte)(unsafe.Pointer(x)))[:]
+}
+
 const (
 	TCA_U32_UNSPEC = iota
 	TCA_U32_CLASSID
@@ -294,6 +516,81 @@ 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_GACT = 5
+)
+
+const (
+	TCA_GACT_UNSPEC = iota
+	TCA_GACT_TM
+	TCA_GACT_PARMS
+	TCA_GACT_PROB
+	TCA_GACT_MAX = TCA_GACT_PROB
+)
+
+type TcGact TcGen
+
+const (
+	TCA_ACT_BPF = 13
+)
+
+const (
+	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
 )
@@ -305,31 +602,6 @@ const (
 	TCA_MIRRED_MAX = TCA_MIRRED_PARMS
 )
 
-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 */
-)
-
-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
-)
-
-// #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 */
@@ -337,11 +609,7 @@ const (
 // };
 
 type TcMirred struct {
-	Index   uint32
+	TcGen
-	Capab   uint32
-	Action  int32
-	Refcnt  int32
-	Bindcnt int32
 	Eaction int32
 	Ifindex uint32
 }
@@ -357,3 +625,51 @@ func DeserializeTcMirred(b []byte) *TcMirred {
 func (x *TcMirred) Serialize() []byte {
 	return (*(*[SizeofTcMirred]byte)(unsafe.Pointer(x)))[:]
 }
+
+// struct tc_police {
+// 	__u32			index;
+// 	int			action;
+// 	__u32			limit;
+// 	__u32			burst;
+// 	__u32			mtu;
+// 	struct tc_ratespec	rate;
+// 	struct tc_ratespec	peakrate;
+// 	int				refcnt;
+// 	int				bindcnt;
+// 	__u32			capab;
+// };
+
+type TcPolice struct {
+	Index    uint32
+	Action   int32
+	Limit    uint32
+	Burst    uint32
+	Mtu      uint32
+	Rate     TcRateSpec
+	PeakRate TcRateSpec
+	Refcnt   int32
+	Bindcnt  int32
+	Capab    uint32
+}
+
+func (msg *TcPolice) Len() int {
+	return SizeofTcPolice
+}
+
+func DeserializeTcPolice(b []byte) *TcPolice {
+	return (*TcPolice)(unsafe.Pointer(&b[0:SizeofTcPolice][0]))
+}
+
+func (x *TcPolice) Serialize() []byte {
+	return (*(*[SizeofTcPolice]byte)(unsafe.Pointer(x)))[:]
+}
+
+const (
+	TCA_FW_UNSPEC = iota
+	TCA_FW_CLASSID
+	TCA_FW_POLICE
+	TCA_FW_INDEV
+	TCA_FW_ACT
+	TCA_FW_MASK
+	TCA_FW_MAX = TCA_FW_MASK
+)

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

@@ -10,7 +10,9 @@ const (
 	SizeofXfrmUsersaInfo  = 0xe0
 	SizeofXfrmAlgo        = 0x44
 	SizeofXfrmAlgoAuth    = 0x48
+	SizeofXfrmAlgoAEAD    = 0x48
 	SizeofXfrmEncapTmpl   = 0x18
+	SizeofXfrmUsersaFlush = 0x8
 )
 
 // struct xfrm_usersa_id {
@@ -193,6 +195,35 @@ func (msg *XfrmAlgoAuth) Serialize() []byte {
 //   char    alg_key[0];
 // }
 
+type XfrmAlgoAEAD struct {
+	AlgName   [64]byte
+	AlgKeyLen uint32
+	AlgICVLen uint32
+	AlgKey    []byte
+}
+
+func (msg *XfrmAlgoAEAD) Len() int {
+	return SizeofXfrmAlgoAEAD + int(msg.AlgKeyLen/8)
+}
+
+func DeserializeXfrmAlgoAEAD(b []byte) *XfrmAlgoAEAD {
+	ret := XfrmAlgoAEAD{}
+	copy(ret.AlgName[:], b[0:64])
+	ret.AlgKeyLen = *(*uint32)(unsafe.Pointer(&b[64]))
+	ret.AlgICVLen = *(*uint32)(unsafe.Pointer(&b[68]))
+	ret.AlgKey = b[72:ret.Len()]
+	return &ret
+}
+
+func (msg *XfrmAlgoAEAD) Serialize() []byte {
+	b := make([]byte, msg.Len())
+	copy(b[0:64], msg.AlgName[:])
+	copy(b[64:68], (*(*[4]byte)(unsafe.Pointer(&msg.AlgKeyLen)))[:])
+	copy(b[68:72], (*(*[4]byte)(unsafe.Pointer(&msg.AlgICVLen)))[:])
+	copy(b[72:msg.Len()], msg.AlgKey[:])
+	return b
+}
+
 // struct xfrm_encap_tmpl {
 //   __u16   encap_type;
 //   __be16    encap_sport;
@@ -219,3 +250,23 @@ 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/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

@@ -2,21 +2,27 @@ package netlink
 
 import (
 	"fmt"
+	"math"
 )
 
 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 {
@@ -27,7 +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 {
@@ -52,6 +58,14 @@ func HandleStr(handle uint32) string {
 	}
 }
 
+func Percentage2u32(percentage float32) uint32 {
+	// FIXME this is most likely not the best way to convert from % to uint32
+	if percentage == 100 {
+		return math.MaxUint32
+	}
+	return uint32(math.MaxUint32 * (percentage / 100))
+}
+
 // PfifoFast is the default qdisc created by the kernel if one has not
 // been defined for the interface
 type PfifoFast struct {
@@ -91,7 +105,86 @@ func (qdisc *Prio) Type() string {
 	return "prio"
 }
 
-// Tbf is a classful qdisc that rate limits based on tokens
+// Htb is a classful qdisc that rate limits based on tokens
+type Htb struct {
+	QdiscAttrs
+	Version      uint32
+	Rate2Quantum uint32
+	Defcls       uint32
+	Debug        uint32
+	DirectPkts   uint32
+}
+
+func NewHtb(attrs QdiscAttrs) *Htb {
+	return &Htb{
+		QdiscAttrs:   attrs,
+		Version:      3,
+		Defcls:       0,
+		Rate2Quantum: 10,
+		Debug:        0,
+		DirectPkts:   0,
+	}
+}
+
+func (qdisc *Htb) Attrs() *QdiscAttrs {
+	return &qdisc.QdiscAttrs
+}
+
+func (qdisc *Htb) Type() string {
+	return "htb"
+}
+
+// Netem is a classless qdisc that rate limits based on tokens
+
+type NetemQdiscAttrs struct {
+	Latency       uint32  // in us
+	DelayCorr     float32 // in %
+	Limit         uint32
+	Loss          float32 // in %
+	LossCorr      float32 // in %
+	Gap           uint32
+	Duplicate     float32 // in %
+	DuplicateCorr float32 // in %
+	Jitter        uint32  // in us
+	ReorderProb   float32 // in %
+	ReorderCorr   float32 // in %
+	CorruptProb   float32 // in %
+	CorruptCorr   float32 // in %
+}
+
+func (q NetemQdiscAttrs) String() string {
+	return fmt.Sprintf(
+		"{Latency: %d, Limit: %d, Loss: %f, Gap: %d, Duplicate: %f, Jitter: %d}",
+		q.Latency, q.Limit, q.Loss, q.Gap, q.Duplicate, q.Jitter,
+	)
+}
+
+type Netem struct {
+	QdiscAttrs
+	Latency       uint32
+	DelayCorr     uint32
+	Limit         uint32
+	Loss          uint32
+	LossCorr      uint32
+	Gap           uint32
+	Duplicate     uint32
+	DuplicateCorr uint32
+	Jitter        uint32
+	ReorderProb   uint32
+	ReorderCorr   uint32
+	CorruptProb   uint32
+	CorruptCorr   uint32
+}
+
+func (qdisc *Netem) Attrs() *QdiscAttrs {
+	return &qdisc.QdiscAttrs
+}
+
+func (qdisc *Netem) Type() string {
+	return "netem"
+}
+
+// Tbf is a classless qdisc that rate limits based on tokens
 type Tbf struct {
 	QdiscAttrs
 	// TODO: handle 64bit rate properly

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

@@ -10,27 +10,131 @@ 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 {
-	req := nl.NewNetlinkRequest(syscall.RTM_DELQDISC, syscall.NLM_F_ACK)
+	return pkgHandle.QdiscDel(qdisc)
-	base := qdisc.Attrs()
+}
-	msg := &nl.TcMsg{
-		Family:  nl.FAMILY_ALL,
-		Ifindex: int32(base.LinkIndex),
-		Handle:  base.Handle,
-		Parent:  base.Parent,
-	}
-	req.AddData(msg)
 
-	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+// QdiscDel will delete a qdisc from the system.
-	return err
+// 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 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 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)
 }
 
 // QdiscAdd will add a qdisc to the system.
 // Equivalent to: `tc qdisc add $qdisc`
 func QdiscAdd(qdisc Qdisc) error {
-	req := nl.NewNetlinkRequest(syscall.RTM_NEWQDISC, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK)
+	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 (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,
@@ -39,6 +143,20 @@ func QdiscAdd(qdisc Qdisc) error {
 		Parent:  base.Parent,
 	}
 	req.AddData(msg)
+
+	// When deleting don't bother building the rest of the netlink payload
+	if cmd != syscall.RTM_DELQDISC {
+		if err := qdiscPayload(req, qdisc); err != nil {
+			return err
+		}
+	}
+
+	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return err
+}
+
+func qdiscPayload(req *nl.NetlinkRequest, qdisc Qdisc) error {
+
 	req.AddData(nl.NewRtAttr(nl.TCA_KIND, nl.ZeroTerminated(qdisc.Type())))
 
 	options := nl.NewRtAttr(nl.TCA_OPTIONS, nil)
@@ -55,26 +173,75 @@ func QdiscAdd(qdisc Qdisc) error {
 		opt.Limit = tbf.Limit
 		opt.Buffer = tbf.Buffer
 		nl.NewRtAttrChild(options, nl.TCA_TBF_PARMS, opt.Serialize())
+	} else if htb, ok := qdisc.(*Htb); ok {
+		opt := nl.TcHtbGlob{}
+		opt.Version = htb.Version
+		opt.Rate2Quantum = htb.Rate2Quantum
+		opt.Defcls = htb.Defcls
+		// TODO: Handle Debug properly. For now default to 0
+		opt.Debug = htb.Debug
+		opt.DirectPkts = htb.DirectPkts
+		nl.NewRtAttrChild(options, nl.TCA_HTB_INIT, opt.Serialize())
+		// nl.NewRtAttrChild(options, nl.TCA_HTB_DIRECT_QLEN, opt.Serialize())
+	} else if netem, ok := qdisc.(*Netem); ok {
+		opt := nl.TcNetemQopt{}
+		opt.Latency = netem.Latency
+		opt.Limit = netem.Limit
+		opt.Loss = netem.Loss
+		opt.Gap = netem.Gap
+		opt.Duplicate = netem.Duplicate
+		opt.Jitter = netem.Jitter
+		options = nl.NewRtAttr(nl.TCA_OPTIONS, opt.Serialize())
+		// Correlation
+		corr := nl.TcNetemCorr{}
+		corr.DelayCorr = netem.DelayCorr
+		corr.LossCorr = netem.LossCorr
+		corr.DupCorr = netem.DuplicateCorr
+
+		if corr.DelayCorr > 0 || corr.LossCorr > 0 || corr.DupCorr > 0 {
+			nl.NewRtAttrChild(options, nl.TCA_NETEM_CORR, corr.Serialize())
+		}
+		// Corruption
+		corruption := nl.TcNetemCorrupt{}
+		corruption.Probability = netem.CorruptProb
+		corruption.Correlation = netem.CorruptCorr
+		if corruption.Probability > 0 {
+			nl.NewRtAttrChild(options, nl.TCA_NETEM_CORRUPT, corruption.Serialize())
+		}
+		// Reorder
+		reorder := nl.TcNetemReorder{}
+		reorder.Probability = netem.ReorderProb
+		reorder.Correlation = netem.ReorderCorr
+		if reorder.Probability > 0 {
+			nl.NewRtAttrChild(options, nl.TCA_NETEM_REORDER, reorder.Serialize())
+		}
 	} else if _, ok := qdisc.(*Ingress); ok {
 		// ingress filters must use the proper handle
-		if msg.Parent != HANDLE_INGRESS {
+		if qdisc.Attrs().Parent != HANDLE_INGRESS {
 			return fmt.Errorf("Ingress filters must set Parent to HANDLE_INGRESS")
 		}
 	}
+
 	req.AddData(options)
-	_, err := req.Execute(syscall.NETLINK_ROUTE, 0)
+	return nil
-	return err
 }
 
 // QdiscList gets a list of qdiscs in the system.
 // 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{
@@ -123,6 +290,10 @@ func QdiscList(link Link) ([]Qdisc, error) {
 					qdisc = &Tbf{}
 				case "ingress":
 					qdisc = &Ingress{}
+				case "htb":
+					qdisc = &Htb{}
+				case "netem":
+					qdisc = &Netem{}
 				default:
 					qdisc = &GenericQdisc{QdiscType: qdiscType}
 				}
@@ -146,6 +317,19 @@ func QdiscList(link Link) ([]Qdisc, error) {
 					if err := parseTbfData(qdisc, data); err != nil {
 						return nil, err
 					}
+				case "htb":
+					data, err := nl.ParseRouteAttr(attr.Value)
+					if err != nil {
+						return nil, err
+					}
+					if err := parseHtbData(qdisc, data); err != nil {
+						return nil, err
+					}
+				case "netem":
+					if err := parseNetemData(qdisc, attr.Value); err != nil {
+						return nil, err
+					}
+
 					// no options for ingress
 				}
 			}
@@ -173,6 +357,59 @@ func parsePrioData(qdisc Qdisc, value []byte) error {
 	return nil
 }
 
+func parseHtbData(qdisc Qdisc, data []syscall.NetlinkRouteAttr) error {
+	native = nl.NativeEndian()
+	htb := qdisc.(*Htb)
+	for _, datum := range data {
+		switch datum.Attr.Type {
+		case nl.TCA_HTB_INIT:
+			opt := nl.DeserializeTcHtbGlob(datum.Value)
+			htb.Version = opt.Version
+			htb.Rate2Quantum = opt.Rate2Quantum
+			htb.Defcls = opt.Defcls
+			htb.Debug = opt.Debug
+			htb.DirectPkts = opt.DirectPkts
+		case nl.TCA_HTB_DIRECT_QLEN:
+			// TODO
+			//htb.DirectQlen = native.uint32(datum.Value)
+		}
+	}
+	return nil
+}
+
+func parseNetemData(qdisc Qdisc, value []byte) error {
+	netem := qdisc.(*Netem)
+	opt := nl.DeserializeTcNetemQopt(value)
+	netem.Latency = opt.Latency
+	netem.Limit = opt.Limit
+	netem.Loss = opt.Loss
+	netem.Gap = opt.Gap
+	netem.Duplicate = opt.Duplicate
+	netem.Jitter = opt.Jitter
+	data, err := nl.ParseRouteAttr(value[nl.SizeofTcNetemQopt:])
+	if err != nil {
+		return err
+	}
+	for _, datum := range data {
+		switch datum.Attr.Type {
+		case nl.TCA_NETEM_CORR:
+			opt := nl.DeserializeTcNetemCorr(datum.Value)
+			netem.DelayCorr = opt.DelayCorr
+			netem.LossCorr = opt.LossCorr
+			netem.DuplicateCorr = opt.DupCorr
+		case nl.TCA_NETEM_CORRUPT:
+			opt := nl.DeserializeTcNetemCorrupt(datum.Value)
+			netem.CorruptProb = opt.Probability
+			netem.CorruptCorr = opt.Correlation
+		case nl.TCA_NETEM_REORDER:
+			opt := nl.DeserializeTcNetemReorder(datum.Value)
+			netem.ReorderProb = opt.Probability
+			netem.ReorderCorr = opt.Correlation
+		}
+	}
+	return nil
+}
+
 func parseTbfData(qdisc Qdisc, data []syscall.NetlinkRouteAttr) error {
 	native = nl.NativeEndian()
 	tbf := qdisc.(*Tbf)
@@ -195,8 +432,9 @@ const (
 )
 
 var (
-	tickInUsec  float64 = 0.0
+	tickInUsec  float64
-	clockFactor float64 = 0.0
+	clockFactor float64
+	hz          float64
 )
 
 func initClock() {
@@ -222,6 +460,7 @@ func initClock() {
 	}
 	clockFactor = float64(vals[2]) / TIME_UNITS_PER_SEC
 	tickInUsec = float64(vals[0]) / float64(vals[1]) * clockFactor
+	hz = float64(vals[0])
 }
 
 func TickInUsec() float64 {
@@ -238,6 +477,13 @@ func ClockFactor() float64 {
 	return clockFactor
 }
 
+func Hz() float64 {
+	if hz == 0.0 {
+		initClock()
+	}
+	return hz
+}
+
 func time2Tick(time uint32) uint32 {
 	return uint32(float64(time) * TickInUsec())
 }
@@ -261,3 +507,7 @@ func burst(rate uint64, buffer uint32) uint32 {
 func latency(rate uint64, limit, buffer uint32) float64 {
 	return TIME_UNITS_PER_SEC*(float64(limit)/float64(rate)) - float64(tick2Time(buffer))
 }
+
+func Xmittime(rate uint64, size uint32) float64 {
+	return TickInUsec() * TIME_UNITS_PER_SEC * (float64(size) / float64(rate))
+}

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

@@ -3,33 +3,64 @@ package netlink
 import (
 	"fmt"
 	"net"
-	"syscall"
 )
 
 // Scope is an enum representing a route scope.
 type Scope uint8
 
-const (
+type NextHopFlag int
-	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
-)
 
-// Route represents a netlink route. A route is associated with a link,
+// Route represents a netlink route.
-// has a destination network, an optional source ip, and optional
-// gateway. Advanced route parameters and non-main routing tables are
-// currently not supported.
 type Route struct {
 	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}", r.LinkIndex, r.Dst,
+	if len(r.MultiPath) > 0 {
-		r.Src, r.Gw)
+		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) {
+	r.Flags |= int(flag)
+}
+
+func (r *Route) ClearFlag(flag NextHopFlag) {
+	r.Flags &^= int(flag)
+}
+
+type flagString struct {
+	f NextHopFlag
+	s string
+}
+
+// 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

@@ -6,30 +6,83 @@ import (
 	"syscall"
 
 	"github.com/vishvananda/netlink/nl"
+	"github.com/vishvananda/netns"
 )
 
 // 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 pkgHandle.RouteAdd(route)
-	return routeHandle(route, req, nl.NewRtMsg())
 }
 
-// RouteAdd will delete a route from the system.
+// 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 pkgHandle.RouteDel(route)
-	return routeHandle(route, req, nl.NewRtDelMsg())
 }
 
-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)
 	family := -1
 	var rtAttrs []*nl.RtAttr
 
@@ -78,8 +131,66 @@ 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)
@@ -101,44 +212,110 @@ 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)
+	return pkgHandle.RouteList(link, family)
-	msg := nl.NewIfInfomsg(family)
+}
-	req.AddData(msg)
+
+// 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
-	}
-
-	native := nl.NativeEndian()
 	var res []Route
-MsgLoop:
 	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 {
+			if filter == nil || filter != nil && filterMask&RT_FILTER_TABLE == 0 {
 				// Ignore non-main tables
 				continue
 			}
-
+		}
-		attrs, err := nl.ParseRouteAttr(m[msg.Len():])
+		route, err := deserializeRoute(m)
 		if err != nil {
 			return nil, err
 		}
+		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
+}
 
-		route := Route{Scope: Scope(msg.Scope)}
+// deserializeRoute decodes a binary netlink message into a Route struct
+func deserializeRoute(m []byte) (Route, error) {
+	msg := nl.DeserializeRtMsg(m)
+	attrs, err := nl.ParseRouteAttr(m[msg.Len():])
+	if err != nil {
+		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),
+	}
+
+	native := nl.NativeEndian()
 	for _, attr := range attrs {
 		switch attr.Attr.Type {
 		case syscall.RTA_GATEWAY:
@@ -151,24 +328,62 @@ MsgLoop:
 				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 = int(native.Uint32(attr.Value[0:4]))
-				if link != nil && routeIndex != index {
+		case syscall.RTA_IIF:
-					// Ignore routes from other interfaces
+			route.ILinkIndex = int(native.Uint32(attr.Value[0:4]))
-					continue MsgLoop
+		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")
 				}
-				route.LinkIndex = routeIndex
+				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)
 					}
 				}
-		res = append(res, route)
+				return info, value[int(nh.RtNexthop.Len):], nil
 			}
-
+			rest := attr.Value
-	return res, nil
+			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
 }
 
 // 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
@@ -192,34 +407,57 @@ func RouteGet(destination net.IP) ([]Route, error) {
 		return nil, err
 	}
 
-	native := nl.NativeEndian()
 	var res []Route
 	for _, m := range msgs {
-		msg := nl.DeserializeRtMsg(m)
+		route, err := deserializeRoute(m)
-		attrs, err := nl.ParseRouteAttr(m[msg.Len():])
 		if err != nil {
 			return nil, err
 		}
-
-		route := Route{}
-		for _, attr := range attrs {
-			switch attr.Attr.Type {
-			case syscall.RTA_GATEWAY:
-				route.Gw = net.IP(attr.Value)
-			case syscall.RTA_PREFSRC:
-				route.Src = net.IP(attr.Value)
-			case syscall.RTA_DST:
-				route.Dst = &net.IPNet{
-					IP:   attr.Value,
-					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
-			}
-		}
 		res = append(res, route)
 	}
 	return res, nil
 
 }
+
+// RouteSubscribe takes a chan down which notifications will be sent
+// when routes are added or deleted. Close the 'done' chan to stop subscription.
+func RouteSubscribe(ch chan<- RouteUpdate, done <-chan struct{}) error {
+	return routeSubscribeAt(netns.None(), netns.None(), ch, done)
+}
+
+// RouteSubscribeAt works like RouteSubscribe plus it allows the caller
+// to choose the network namespace in which to subscribe (ns).
+func RouteSubscribeAt(ns netns.NsHandle, ch chan<- RouteUpdate, done <-chan struct{}) error {
+	return routeSubscribeAt(ns, netns.None(), ch, done)
+}
+
+func routeSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- RouteUpdate, done <-chan struct{}) error {
+	s, err := nl.SubscribeAt(newNs, curNs, syscall.NETLINK_ROUTE, syscall.RTNLGRP_IPV4_ROUTE, syscall.RTNLGRP_IPV6_ROUTE)
+	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 {
+				return
+			}
+			for _, m := range msgs {
+				route, err := deserializeRoute(m.Data)
+				if err != nil {
+					return
+				}
+				ch <- RouteUpdate{Type: m.Header.Type, Route: route}
+			}
+		}
+	}()
+
+	return nil
+}

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/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.FAMILY_V4)
+	if policy.Dst != nil {
 		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)
+	}
+	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)
+	return pkgHandle.XfrmPolicyDel(policy)
+}
 
-	msg := &nl.XfrmUserpolicyId{}
+// XfrmPolicyDel will delete an xfrm policy from the system. Note that
-	selFromPolicy(&msg.Sel, policy)
+// the Tmpls are ignored when matching the policy to delete.
-	msg.Index = uint32(policy.Index)
+// Equivalent to: `ip xfrm policy del $policy`
-	msg.Dir = uint8(policy.Dir)
+func (h *Handle) XfrmPolicyDel(policy *XfrmPolicy) error {
-	req.AddData(msg)
+	_, err := h.xfrmPolicyGetOrDelete(policy, nl.XFRM_MSG_DELPOLICY)
-
-	_, err := req.Execute(syscall.NETLINK_XFRM, 0)
 	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)
@@ -86,16 +133,94 @@ func XfrmPolicyList(family int) ([]XfrmPolicy, error) {
 
 	var res []XfrmPolicy
 	for _, m := range msgs {
+		if policy, err := parseXfrmPolicy(m, family); err == nil {
+			res = append(res, *policy)
+		} else if err == familyError {
+			continue
+		} else {
+			return nil, err
+		}
+	}
+	return res, nil
+}
+
+// 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)
+}
+
+// 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)
+}
+
+// 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
+	}
+
+	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) {
-			continue
+		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)
@@ -116,12 +241,17 @@ func XfrmPolicyList(family int) ([]XfrmPolicy, error) {
 				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 &policy, nil
-	return res, nil
 }

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

@@ -1,6 +1,7 @@
 package netlink
 
 import (
+	"fmt"
 	"net"
 )
 
@@ -9,9 +10,21 @@ type XfrmStateAlgo struct {
 	Name        string
 	Key         []byte
 	TruncateLen int // Auth only
+	ICVLen      int // AEAD only
 }
 
-// EncapType is an enum representing an ipsec template direction.
+func (a XfrmStateAlgo) String() string {
+	base := fmt.Sprintf("{Name: %s, Key: 0x%x", a.Name, a.Key)
+	if a.TruncateLen != 0 {
+		base = fmt.Sprintf("%s, Truncate length: %d", base, a.TruncateLen)
+	}
+	if a.ICVLen != 0 {
+		base = fmt.Sprintf("%s, ICV length: %d", base, a.ICVLen)
+	}
+	return fmt.Sprintf("%s}", base)
+}
+
+// EncapType is an enum representing the optional packet encapsulation.
 type EncapType uint8
 
 const (
@@ -22,14 +35,14 @@ 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 +50,23 @@ 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 +77,31 @@ type XfrmState struct {
 	Spi          int
 	Reqid        int
 	ReplayWindow int
+	Limits       XfrmStateLimits
+	Mark         *XfrmMark
 	Auth         *XfrmStateAlgo
 	Crypt        *XfrmStateAlgo
+	Aead         *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, Aead: %v,Encap: %v",
+		sa.Dst, sa.Src, sa.Proto, sa.Mode, sa.Spi, sa.Reqid, sa.ReplayWindow, sa.Mark, sa.Auth, sa.Crypt, sa.Aead, 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 +35,61 @@ func writeStateAlgoAuth(a *XfrmStateAlgo) []byte {
 	return algo.Serialize()
 }
 
+func writeStateAlgoAead(a *XfrmStateAlgo) []byte {
+	algo := nl.XfrmAlgoAEAD{
+		AlgKeyLen: uint32(len(a.Key) * 8),
+		AlgICVLen: uint32(a.ICVLen),
+		AlgKey:    a.Key,
+	}
+	end := len(a.Name)
+	if end > 64 {
+		end = 64
+	}
+	copy(algo.AlgName[:end], a.Name)
+	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 +100,7 @@ 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
+	limitsToLft(state.Limits, &msg.Lft)
-	msg.Lft.HardByteLimit = nl.XFRM_INF
-	msg.Lft.SoftPacketLimit = nl.XFRM_INF
-	msg.Lft.HardPacketLimit = nl.XFRM_INF
 	req.AddData(msg)
 
 	if state.Auth != nil {
@@ -66,6 +111,10 @@ func XfrmStateAdd(state *XfrmState) error {
 		out := nl.NewRtAttr(nl.XFRMA_ALG_CRYPT, writeStateAlgo(state.Crypt))
 		req.AddData(out)
 	}
+	if state.Aead != nil {
+		out := nl.NewRtAttr(nl.XFRMA_ALG_AEAD, writeStateAlgoAead(state.Aead))
+		req.AddData(out)
+	}
 	if state.Encap != nil {
 		encapData := make([]byte, nl.SizeofXfrmEncapTmpl)
 		encap := nl.DeserializeXfrmEncapTmpl(encapData)
@@ -76,6 +125,10 @@ 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 +138,29 @@ 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)
+	return pkgHandle.XfrmStateDel(state)
+}
 
-	msg := &nl.XfrmUsersaId{}
+// XfrmStateDel will delete an xfrm state from the system. Note that
-	msg.Daddr.FromIP(state.Dst)
+// the Algos are ignored when matching the state to delete.
-	msg.Family = uint16(nl.GetIPFamily(state.Dst))
+// Equivalent to: `ip xfrm state del $state`
-	msg.Proto = uint8(state.Proto)
+func (h *Handle) XfrmStateDel(state *XfrmState) error {
-	msg.Spi = nl.Swap32(uint32(state.Spi))
+	_, err := h.xfrmStateGetOrDelete(state, nl.XFRM_MSG_DELSA)
-	req.AddData(msg)
-
-	saddr := nl.XfrmAddress{}
-	saddr.FromIP(state.Src)
-	srcdata := nl.NewRtAttr(nl.XFRMA_SRCADDR, saddr.Serialize())
-
-	req.AddData(srcdata)
-
-	_, err := req.Execute(syscall.NETLINK_XFRM, 0)
 	return err
 }
 
+// XfrmStateList gets a list of xfrm states in the system.
+// Equivalent to: `ip [-4|-6] xfrm state show`.
+// The list can be filtered by ip family.
+func XfrmStateList(family int) ([]XfrmState, error) {
+	return pkgHandle.XfrmStateList(family)
+}
+
 // 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 XfrmStateList(family int) ([]XfrmState, error) {
+func (h *Handle) XfrmStateList(family int) ([]XfrmState, error) {
-	req := nl.NewNetlinkRequest(nl.XFRM_MSG_GETSA, syscall.NLM_F_DUMP)
+	req := h.newNetlinkRequest(nl.XFRM_MSG_GETSA, syscall.NLM_F_DUMP)
-
-	msg := nl.NewIfInfomsg(family)
-	req.AddData(msg)
 
 	msgs, err := req.Execute(syscall.NETLINK_XFRM, nl.XFRM_MSG_NEWSA)
 	if err != nil {
@@ -120,10 +169,84 @@ func XfrmStateList(family int) ([]XfrmState, error) {
 
 	var res []XfrmState
 	for _, m := range msgs {
+		if state, err := parseXfrmState(m, family); err == nil {
+			res = append(res, *state)
+		} else if err == familyError {
+			continue
+		} else {
+			return nil, err
+		}
+	}
+	return res, nil
+}
+
+// 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)
+}
+
+// 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)
+}
+
+func (h *Handle) xfrmStateGetOrDelete(state *XfrmState, nlProto int) (*XfrmState, error) {
+	req := h.newNetlinkRequest(nlProto, syscall.NLM_F_ACK)
+
+	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.To16())
+		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) {
-			continue
+		return nil, familyError
 	}
 
 	var state XfrmState
@@ -135,8 +258,9 @@ func XfrmStateList(family int) ([]XfrmState, error) {
 	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[msg.Len():])
+	attrs, err := nl.ParseRouteAttr(m[nl.SizeofXfrmUsersaInfo:])
 	if err != nil {
 		return nil, err
 	}
@@ -165,6 +289,12 @@ func XfrmStateList(family int) ([]XfrmState, error) {
 			state.Auth.Name = nl.BytesToString(algo.AlgName[:])
 			state.Auth.Key = algo.AlgKey
 			state.Auth.TruncateLen = int(algo.AlgTruncLen)
+		case nl.XFRMA_ALG_AEAD:
+			state.Aead = new(XfrmStateAlgo)
+			algo := nl.DeserializeXfrmAlgoAEAD(attr.Value[:])
+			state.Aead.Name = nl.BytesToString(algo.AlgName[:])
+			state.Aead.Key = algo.AlgKey
+			state.Aead.ICVLen = int(algo.AlgICVLen)
 		case nl.XFRMA_ENCAP:
 			encap := nl.DeserializeXfrmEncapTmpl(attr.Value[:])
 			state.Encap = new(XfrmStateEncap)
@@ -172,10 +302,67 @@ func XfrmStateList(family int) ([]XfrmState, error) {
 			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
+		}
 	}
 
-		}
+	return &state, nil
-		res = append(res, state)
+}
-	}
+
-	return res, 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/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_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
+}