vendor: update vishvananda/netlink v1.1.0

full diff: https://github.com/vishvananda/netlink/compare/v1.0.0...v1.1.0

also updated moby/ipvs, which is compatible with this version of netlink,
and update vishvananda/netns to current master (which added go.mod)

Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This commit is contained in:
Sebastiaan van Stijn 2020-03-12 13:23:26 +01:00
parent 3dc9ff93cb
commit d846c2b1ab
54 changed files with 3737 additions and 592 deletions

View file

@ -793,7 +793,7 @@ func (n *network) watchMiss(nlSock *nl.NetlinkSocket, nsPath string) {
return
}
for {
msgs, err := nlSock.Receive()
msgs, _, err := nlSock.Receive()
if err != nil {
n.Lock()
nlFd := nlSock.GetFd()

View file

@ -45,8 +45,8 @@ github.com/sirupsen/logrus 8bdbc7bcc01dcbb8ec23dc8a28e332258d25251f
github.com/konsorten/go-windows-terminal-sequences 5c8c8bd35d3832f5d134ae1e1e375b69a4d25242 # v1.0.1
github.com/ugorji/go b4c50a2b199d93b13dc15e78929cfb23bfdf21ab # v1.1.1
github.com/urfave/cli a65b733b303f0055f8d324d805f393cd3e7a7904
github.com/vishvananda/netlink a2ad57a690f3caf3015351d2d6e1c0b95c349752 # v1.0.0
github.com/vishvananda/netns 7109fa855b0ff1ebef7fbd2f6aa613e8db7cfbc0
github.com/vishvananda/netlink f049be6f391489d3f374498fe0c8df8449258372 # v1.1.0
github.com/vishvananda/netns 0a2b9b5464df8343199164a0321edf3313202f7e
golang.org/x/crypto b7391e95e576cacdcdd422573063bc057239113d
golang.org/x/net a680a1efc54dd51c040b3b5ce4939ea3cf2ea0d1
golang.org/x/sys d455e41777fca6e8a5a79e34a14b8368bc11d9ba
@ -58,4 +58,4 @@ go.opencensus.io 9c377598961b706d1542bd2d84d538b5094d596e
gotest.tools 1083505acf35a0bd8a696b26837e1fb3187a7a83 # v2.3.0
github.com/google/go-cmp 3af367b6b30c263d47e8895973edcca9a49cf029 # v0.2.0
github.com/moby/ipvs 45bd6e080830fe2f2a43320b154d82b8f3f5aa0c
github.com/moby/ipvs 8f137da6850a975020f4f739c589d293dd3a9d7b # v1.0.0

34
libnetwork/vendor/github.com/moby/ipvs/README.md generated vendored Normal file
View file

@ -0,0 +1,34 @@
# ipvs - networking for containers
![Test](https://github.com/moby/ipvs/workflows/Test/badge.svg) [![GoDoc](https://godoc.org/github.com/moby/ipvs?status.svg)](https://godoc.org/github.com/moby/ipvs) [![Go Report Card](https://goreportcard.com/badge/github.com/moby/ipvs)](https://goreportcard.com/report/github.com/moby/ipvs)
ipvs provides a native Go implementation for communicating with IPVS kernel module using a netlink socket.
#### Using ipvs
```go
import (
"log"
"github.com/moby/ipvs"
)
func main() {
handle, err := ipvs.New("")
if err != nil {
log.Fatalf("ipvs.New: %s", err)
}
svcs, err := handle.GetServices()
if err != nil {
log.Fatalf("handle.GetServices: %s", err)
}
}
```
## Contributing
Want to hack on ipvs? [Docker's contributions guidelines](https://github.com/docker/docker/blob/master/CONTRIBUTING.md) apply.
## Copyright and license
Code and documentation copyright 2015 Docker, inc. Code released under the Apache 2.0 license. Docs released under Creative commons.

View file

@ -3,10 +3,10 @@ module github.com/moby/ipvs
go 1.13
require (
github.com/pkg/errors v0.8.1 // indirect
github.com/sirupsen/logrus v1.4.1
github.com/vishvananda/netlink v1.0.0
github.com/vishvananda/netns v0.0.0-20190625233234-7109fa855b0f
golang.org/x/sys v0.0.0-20190303122642-d455e41777fc
gotest.tools v1.4.1-0.20181223230014-1083505acf35
github.com/pkg/errors v0.9.1 // indirect
github.com/sirupsen/logrus v1.4.2
github.com/vishvananda/netlink v1.1.0
github.com/vishvananda/netns v0.0.0-20191106174202-0a2b9b5464df
golang.org/x/sys v0.0.0-20200302150141-5c8b2ff67527
gotest.tools/v3 v3.0.2
)

View file

@ -217,7 +217,7 @@ func execute(s *nl.NetlinkSocket, req *nl.NetlinkRequest, resType uint16) ([][]b
done:
for {
msgs, err := s.Receive()
msgs, _, err := s.Receive()
if err != nil {
if s.GetFd() == -1 {
return nil, fmt.Errorf("Socket got closed on receive")

View file

@ -15,39 +15,62 @@ import (
const IFA_FLAGS = 0x8
// AddrAdd will add an IP address to a link device.
//
// Equivalent to: `ip addr add $addr dev $link`
//
// If `addr` is an IPv4 address and the broadcast address is not given, it
// will be automatically computed based on the IP mask if /30 or larger.
func AddrAdd(link Link, addr *Addr) error {
return pkgHandle.AddrAdd(link, addr)
}
// AddrAdd will add an IP address to a link device.
//
// Equivalent to: `ip addr add $addr dev $link`
//
// If `addr` is an IPv4 address and the broadcast address is not given, it
// will be automatically computed based on the IP mask if /30 or larger.
func (h *Handle) AddrAdd(link Link, addr *Addr) error {
req := h.newNetlinkRequest(unix.RTM_NEWADDR, unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK)
return h.addrHandle(link, addr, req)
}
// AddrReplace will replace (or, if not present, add) an IP address on a link device.
//
// Equivalent to: `ip addr replace $addr dev $link`
//
// If `addr` is an IPv4 address and the broadcast address is not given, it
// will be automatically computed based on the IP mask if /30 or larger.
func AddrReplace(link Link, addr *Addr) error {
return pkgHandle.AddrReplace(link, addr)
}
// AddrReplace will replace (or, if not present, add) an IP address on a link device.
//
// Equivalent to: `ip addr replace $addr dev $link`
//
// If `addr` is an IPv4 address and the broadcast address is not given, it
// will be automatically computed based on the IP mask if /30 or larger.
func (h *Handle) AddrReplace(link Link, addr *Addr) error {
req := h.newNetlinkRequest(unix.RTM_NEWADDR, unix.NLM_F_CREATE|unix.NLM_F_REPLACE|unix.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`
//
// If `addr` is an IPv4 address and the broadcast address is not given, it
// will be automatically computed based on the IP mask if /30 or larger.
func AddrDel(link Link, addr *Addr) error {
return pkgHandle.AddrDel(link, addr)
}
// AddrDel will delete an IP address from a link device.
// Equivalent to: `ip addr del $addr dev $link`
//
// If `addr` is an IPv4 address and the broadcast address is not given, it
// will be automatically computed based on the IP mask if /30 or larger.
func (h *Handle) AddrDel(link Link, addr *Addr) error {
req := h.newNetlinkRequest(unix.RTM_DELADDR, unix.NLM_F_ACK)
return h.addrHandle(link, addr, req)
@ -65,7 +88,11 @@ func (h *Handle) addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error
msg := nl.NewIfAddrmsg(family)
msg.Index = uint32(base.Index)
msg.Scope = uint8(addr.Scope)
prefixlen, masklen := addr.Mask.Size()
mask := addr.Mask
if addr.Peer != nil {
mask = addr.Peer.Mask
}
prefixlen, masklen := mask.Size()
msg.Prefixlen = uint8(prefixlen)
req.AddData(msg)
@ -104,14 +131,20 @@ func (h *Handle) addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error
}
if family == FAMILY_V4 {
if addr.Broadcast == nil {
// Automatically set the broadcast address if it is unset and the
// subnet is large enough to sensibly have one (/30 or larger).
// See: RFC 3021
if addr.Broadcast == nil && prefixlen < 31 {
calcBroadcast := make(net.IP, masklen/8)
for i := range localAddrData {
calcBroadcast[i] = localAddrData[i] | ^addr.Mask[i]
calcBroadcast[i] = localAddrData[i] | ^mask[i]
}
addr.Broadcast = calcBroadcast
}
req.AddData(nl.NewRtAttr(unix.IFA_BROADCAST, addr.Broadcast))
if addr.Broadcast != nil {
req.AddData(nl.NewRtAttr(unix.IFA_BROADCAST, addr.Broadcast))
}
if addr.Label != "" {
labelData := nl.NewRtAttr(unix.IFA_LABEL, nl.ZeroTerminated(addr.Label))
@ -206,13 +239,17 @@ func parseAddr(m []byte) (addr Addr, family, index int, err error) {
IP: attr.Value,
Mask: net.CIDRMask(int(msg.Prefixlen), 8*len(attr.Value)),
}
addr.Peer = dst
case unix.IFA_LOCAL:
// iproute2 manual:
// If a peer address is specified, the local address
// cannot have a prefix length. The network prefix is
// associated with the peer rather than with the local
// address.
n := 8 * len(attr.Value)
local = &net.IPNet{
IP: attr.Value,
Mask: net.CIDRMask(int(msg.Prefixlen), 8*len(attr.Value)),
Mask: net.CIDRMask(n, n),
}
addr.IPNet = local
case unix.IFA_BROADCAST:
addr.Broadcast = attr.Value
case unix.IFA_LABEL:
@ -226,12 +263,24 @@ func parseAddr(m []byte) (addr Addr, family, index int, err error) {
}
}
// IFA_LOCAL should be there but if not, fall back to IFA_ADDRESS
// libnl addr.c comment:
// IPv6 sends the local address as IFA_ADDRESS with no
// IFA_LOCAL, IPv4 sends both IFA_LOCAL and IFA_ADDRESS
// with IFA_ADDRESS being the peer address if they differ
//
// But obviously, as there are IPv6 PtP addresses, too,
// IFA_LOCAL should also be handled for IPv6.
if local != nil {
addr.IPNet = local
if family == FAMILY_V4 && local.IP.Equal(dst.IP) {
addr.IPNet = dst
} else {
addr.IPNet = local
addr.Peer = dst
}
} else {
addr.IPNet = dst
}
addr.Scope = int(msg.Scope)
return
@ -250,21 +299,22 @@ type AddrUpdate struct {
// 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 addrSubscribeAt(netns.None(), netns.None(), ch, done, nil, false)
return addrSubscribeAt(netns.None(), netns.None(), ch, done, nil, false, 0)
}
// 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 addrSubscribeAt(ns, netns.None(), ch, done, nil, false)
return addrSubscribeAt(ns, netns.None(), ch, done, nil, false, 0)
}
// AddrSubscribeOptions contains a set of options to use with
// AddrSubscribeWithOptions.
type AddrSubscribeOptions struct {
Namespace *netns.NsHandle
ErrorCallback func(error)
ListExisting bool
Namespace *netns.NsHandle
ErrorCallback func(error)
ListExisting bool
ReceiveBufferSize int
}
// AddrSubscribeWithOptions work like AddrSubscribe but enable to
@ -275,10 +325,10 @@ func AddrSubscribeWithOptions(ch chan<- AddrUpdate, done <-chan struct{}, option
none := netns.None()
options.Namespace = &none
}
return addrSubscribeAt(*options.Namespace, netns.None(), ch, done, options.ErrorCallback, options.ListExisting)
return addrSubscribeAt(*options.Namespace, netns.None(), ch, done, options.ErrorCallback, options.ListExisting, options.ReceiveBufferSize)
}
func addrSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-chan struct{}, cberr func(error), listExisting bool) error {
func addrSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-chan struct{}, cberr func(error), listExisting bool, rcvbuf int) error {
s, err := nl.SubscribeAt(newNs, curNs, unix.NETLINK_ROUTE, unix.RTNLGRP_IPV4_IFADDR, unix.RTNLGRP_IPV6_IFADDR)
if err != nil {
return err
@ -289,6 +339,12 @@ func addrSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-c
s.Close()
}()
}
if rcvbuf != 0 {
err = pkgHandle.SetSocketReceiveBufferSize(rcvbuf, false)
if err != nil {
return err
}
}
if listExisting {
req := pkgHandle.newNetlinkRequest(unix.RTM_GETADDR,
unix.NLM_F_DUMP)
@ -301,13 +357,19 @@ func addrSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-c
go func() {
defer close(ch)
for {
msgs, err := s.Receive()
msgs, from, err := s.Receive()
if err != nil {
if cberr != nil {
cberr(err)
}
return
}
if from.Pid != nl.PidKernel {
if cberr != nil {
cberr(fmt.Errorf("Wrong sender portid %d, expected %d", from.Pid, nl.PidKernel))
}
continue
}
for _, m := range msgs {
if m.Header.Type == unix.NLMSG_DONE {
continue
@ -319,16 +381,17 @@ func addrSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-c
continue
}
if cberr != nil {
cberr(syscall.Errno(-error))
cberr(fmt.Errorf("error message: %v",
syscall.Errno(-error)))
}
return
continue
}
msgType := m.Header.Type
if msgType != unix.RTM_NEWADDR && msgType != unix.RTM_DELADDR {
if cberr != nil {
cberr(fmt.Errorf("bad message type: %d", msgType))
}
return
continue
}
addr, _, ifindex, err := parseAddr(m.Data)
@ -336,7 +399,7 @@ func addrSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-c
if cberr != nil {
cberr(fmt.Errorf("could not parse address: %v", err))
}
return
continue
}
ch <- AddrUpdate{LinkAddress: *addr.IPNet,

View file

@ -96,7 +96,7 @@ func (h *Handle) bridgeVlanModify(cmd int, link Link, vid uint16, pvid, untagged
flags |= nl.BRIDGE_FLAGS_MASTER
}
if flags > 0 {
nl.NewRtAttrChild(br, nl.IFLA_BRIDGE_FLAGS, nl.Uint16Attr(flags))
br.AddRtAttr(nl.IFLA_BRIDGE_FLAGS, nl.Uint16Attr(flags))
}
vlanInfo := &nl.BridgeVlanInfo{Vid: vid}
if pvid {
@ -105,11 +105,8 @@ func (h *Handle) bridgeVlanModify(cmd int, link Link, vid uint16, pvid, untagged
if untagged {
vlanInfo.Flags |= nl.BRIDGE_VLAN_INFO_UNTAGGED
}
nl.NewRtAttrChild(br, nl.IFLA_BRIDGE_VLAN_INFO, vlanInfo.Serialize())
br.AddRtAttr(nl.IFLA_BRIDGE_VLAN_INFO, vlanInfo.Serialize())
req.AddData(br)
_, err := req.Execute(unix.NETLINK_ROUTE, 0)
if err != nil {
return err
}
return nil
return err
}

View file

@ -4,25 +4,76 @@ import (
"fmt"
)
// Class interfaces for all classes
type Class interface {
Attrs() *ClassAttrs
Type() string
}
// Generic networking statistics for netlink users.
// This file contains "gnet_" prefixed structs and relevant functions.
// See Documentation/networking/getn_stats.txt in Linux source code for more details.
// GnetStatsBasic Ref: struct gnet_stats_basic { ... }
type GnetStatsBasic struct {
Bytes uint64 // number of seen bytes
Packets uint32 // number of seen packets
}
// GnetStatsRateEst Ref: struct gnet_stats_rate_est { ... }
type GnetStatsRateEst struct {
Bps uint32 // current byte rate
Pps uint32 // current packet rate
}
// GnetStatsRateEst64 Ref: struct gnet_stats_rate_est64 { ... }
type GnetStatsRateEst64 struct {
Bps uint64 // current byte rate
Pps uint64 // current packet rate
}
// GnetStatsQueue Ref: struct gnet_stats_queue { ... }
type GnetStatsQueue struct {
Qlen uint32 // queue length
Backlog uint32 // backlog size of queue
Drops uint32 // number of dropped packets
Requeues uint32 // number of requues
Overlimits uint32 // number of enqueues over the limit
}
// ClassStatistics representation based on generic networking statistics for netlink.
// See Documentation/networking/gen_stats.txt in Linux source code for more details.
type ClassStatistics struct {
Basic *GnetStatsBasic
Queue *GnetStatsQueue
RateEst *GnetStatsRateEst
}
// NewClassStatistics Construct a ClassStatistics struct which fields are all initialized by 0.
func NewClassStatistics() *ClassStatistics {
return &ClassStatistics{
Basic: &GnetStatsBasic{},
Queue: &GnetStatsQueue{},
RateEst: &GnetStatsRateEst{},
}
}
// 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
LinkIndex int
Handle uint32
Parent uint32
Leaf uint32
Statistics *ClassStatistics
}
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)
}
// HtbClassAttrs stores the attributes of HTB class
type HtbClassAttrs struct {
// TODO handle all attributes
Rate uint64
@ -54,10 +105,12 @@ func (q HtbClass) String() string {
return fmt.Sprintf("{Rate: %d, Ceil: %d, Buffer: %d, Cbuffer: %d}", q.Rate, q.Ceil, q.Buffer, q.Cbuffer)
}
// Attrs returns the class attributes
func (q *HtbClass) Attrs() *ClassAttrs {
return &q.ClassAttrs
}
// Type return the class type
func (q *HtbClass) Type() string {
return "htb"
}
@ -69,10 +122,90 @@ type GenericClass struct {
ClassType string
}
// Attrs return the class attributes
func (class *GenericClass) Attrs() *ClassAttrs {
return &class.ClassAttrs
}
// Type return the class type
func (class *GenericClass) Type() string {
return class.ClassType
}
// ServiceCurve is the way the HFSC curve are represented
type ServiceCurve struct {
m1 uint32
d uint32
m2 uint32
}
// Attrs return the parameters of the service curve
func (c *ServiceCurve) Attrs() (uint32, uint32, uint32) {
return c.m1, c.d, c.m2
}
// HfscClass is a representation of the HFSC class
type HfscClass struct {
ClassAttrs
Rsc ServiceCurve
Fsc ServiceCurve
Usc ServiceCurve
}
// SetUsc sets the Usc curve
func (hfsc *HfscClass) SetUsc(m1 uint32, d uint32, m2 uint32) {
hfsc.Usc = ServiceCurve{m1: m1 / 8, d: d, m2: m2 / 8}
}
// SetFsc sets the Fsc curve
func (hfsc *HfscClass) SetFsc(m1 uint32, d uint32, m2 uint32) {
hfsc.Fsc = ServiceCurve{m1: m1 / 8, d: d, m2: m2 / 8}
}
// SetRsc sets the Rsc curve
func (hfsc *HfscClass) SetRsc(m1 uint32, d uint32, m2 uint32) {
hfsc.Rsc = ServiceCurve{m1: m1 / 8, d: d, m2: m2 / 8}
}
// SetSC implements the SC from the tc CLI
func (hfsc *HfscClass) SetSC(m1 uint32, d uint32, m2 uint32) {
hfsc.Rsc = ServiceCurve{m1: m1 / 8, d: d, m2: m2 / 8}
hfsc.Fsc = ServiceCurve{m1: m1 / 8, d: d, m2: m2 / 8}
}
// SetUL implements the UL from the tc CLI
func (hfsc *HfscClass) SetUL(m1 uint32, d uint32, m2 uint32) {
hfsc.Usc = ServiceCurve{m1: m1 / 8, d: d, m2: m2 / 8}
}
// SetLS implements the LS from the tc CLI
func (hfsc *HfscClass) SetLS(m1 uint32, d uint32, m2 uint32) {
hfsc.Fsc = ServiceCurve{m1: m1 / 8, d: d, m2: m2 / 8}
}
// NewHfscClass returns a new HFSC struct with the set parameters
func NewHfscClass(attrs ClassAttrs) *HfscClass {
return &HfscClass{
ClassAttrs: attrs,
Rsc: ServiceCurve{},
Fsc: ServiceCurve{},
Usc: ServiceCurve{},
}
}
func (hfsc *HfscClass) String() string {
return fmt.Sprintf(
"{%s -- {RSC: {m1=%d d=%d m2=%d}} {FSC: {m1=%d d=%d m2=%d}} {USC: {m1=%d d=%d m2=%d}}}",
hfsc.Attrs(), hfsc.Rsc.m1*8, hfsc.Rsc.d, hfsc.Rsc.m2*8, hfsc.Fsc.m1*8, hfsc.Fsc.d, hfsc.Fsc.m2*8, hfsc.Usc.m1*8, hfsc.Usc.d, hfsc.Usc.m2*8,
)
}
// Attrs return the Hfsc parameters
func (hfsc *HfscClass) Attrs() *ClassAttrs {
return &hfsc.ClassAttrs
}
// Type return the type of the class
func (hfsc *HfscClass) Type() string {
return "hfsc"
}

View file

@ -1,14 +1,34 @@
package netlink
import (
"bytes"
"encoding/binary"
"encoding/hex"
"errors"
"fmt"
"syscall"
"github.com/vishvananda/netlink/nl"
"golang.org/x/sys/unix"
)
// NOTE: function is in here because it uses other linux functions
// Internal tc_stats representation in Go struct.
// This is for internal uses only to deserialize the payload of rtattr.
// After the deserialization, this should be converted into the canonical stats
// struct, ClassStatistics, in case of statistics of a class.
// Ref: struct tc_stats { ... }
type tcStats struct {
Bytes uint64 // Number of enqueued bytes
Packets uint32 // Number of enqueued packets
Drops uint32 // Packets dropped because of lack of resources
Overlimits uint32 // Number of throttle events when this flow goes out of allocated bandwidth
Bps uint32 // Current flow byte rate
Pps uint32 // Current flow packet rate
Qlen uint32
Backlog uint32
}
// NewHtbClass NOTE: function is in here because it uses other linux functions
func NewHtbClass(attrs ClassAttrs, cattrs HtbClassAttrs) *HtbClass {
mtu := 1600
rate := cattrs.Rate / 8
@ -126,7 +146,9 @@ 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 {
switch class.Type() {
case "htb":
htb := class.(*HtbClass)
opt := nl.TcHtbCopt{}
opt.Buffer = htb.Buffer
opt.Cbuffer = htb.Cbuffer
@ -151,9 +173,18 @@ func classPayload(req *nl.NetlinkRequest, class Class) error {
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))
options.AddRtAttr(nl.TCA_HTB_PARMS, opt.Serialize())
options.AddRtAttr(nl.TCA_HTB_RTAB, SerializeRtab(rtab))
options.AddRtAttr(nl.TCA_HTB_CTAB, SerializeRtab(ctab))
case "hfsc":
hfsc := class.(*HfscClass)
opt := nl.HfscCopt{}
opt.Rsc.Set(hfsc.Rsc.Attrs())
opt.Fsc.Set(hfsc.Fsc.Attrs())
opt.Usc.Set(hfsc.Usc.Attrs())
options.AddRtAttr(nl.TCA_HFSC_RSC, nl.SerializeHfscCurve(&opt.Rsc))
options.AddRtAttr(nl.TCA_HFSC_FSC, nl.SerializeHfscCurve(&opt.Fsc))
options.AddRtAttr(nl.TCA_HFSC_USC, nl.SerializeHfscCurve(&opt.Usc))
}
req.AddData(options)
return nil
@ -197,9 +228,10 @@ func (h *Handle) ClassList(link Link, parent uint32) ([]Class, error) {
}
base := ClassAttrs{
LinkIndex: int(msg.Ifindex),
Handle: msg.Handle,
Parent: msg.Parent,
LinkIndex: int(msg.Ifindex),
Handle: msg.Handle,
Parent: msg.Parent,
Statistics: nil,
}
var class Class
@ -211,6 +243,8 @@ func (h *Handle) ClassList(link Link, parent uint32) ([]Class, error) {
switch classType {
case "htb":
class = &HtbClass{}
case "hfsc":
class = &HfscClass{}
default:
class = &GenericClass{ClassType: classType}
}
@ -225,6 +259,26 @@ func (h *Handle) ClassList(link Link, parent uint32) ([]Class, error) {
if err != nil {
return nil, err
}
case "hfsc":
data, err := nl.ParseRouteAttr(attr.Value)
if err != nil {
return nil, err
}
_, err = parseHfscClassData(class, data)
if err != nil {
return nil, err
}
}
// For backward compatibility.
case nl.TCA_STATS:
base.Statistics, err = parseTcStats(attr.Value)
if err != nil {
return nil, err
}
case nl.TCA_STATS2:
base.Statistics, err = parseTcStats2(attr.Value)
if err != nil {
return nil, err
}
}
}
@ -253,3 +307,78 @@ func parseHtbClassData(class Class, data []syscall.NetlinkRouteAttr) (bool, erro
}
return detailed, nil
}
func parseHfscClassData(class Class, data []syscall.NetlinkRouteAttr) (bool, error) {
hfsc := class.(*HfscClass)
detailed := false
for _, datum := range data {
m1, d, m2 := nl.DeserializeHfscCurve(datum.Value).Attrs()
switch datum.Attr.Type {
case nl.TCA_HFSC_RSC:
hfsc.Rsc = ServiceCurve{m1: m1, d: d, m2: m2}
case nl.TCA_HFSC_FSC:
hfsc.Fsc = ServiceCurve{m1: m1, d: d, m2: m2}
case nl.TCA_HFSC_USC:
hfsc.Usc = ServiceCurve{m1: m1, d: d, m2: m2}
}
}
return detailed, nil
}
func parseTcStats(data []byte) (*ClassStatistics, error) {
buf := &bytes.Buffer{}
buf.Write(data)
native := nl.NativeEndian()
tcStats := &tcStats{}
if err := binary.Read(buf, native, tcStats); err != nil {
return nil, err
}
stats := NewClassStatistics()
stats.Basic.Bytes = tcStats.Bytes
stats.Basic.Packets = tcStats.Packets
stats.Queue.Qlen = tcStats.Qlen
stats.Queue.Backlog = tcStats.Backlog
stats.Queue.Drops = tcStats.Drops
stats.Queue.Overlimits = tcStats.Overlimits
stats.RateEst.Bps = tcStats.Bps
stats.RateEst.Pps = tcStats.Pps
return stats, nil
}
func parseGnetStats(data []byte, gnetStats interface{}) error {
buf := &bytes.Buffer{}
buf.Write(data)
native := nl.NativeEndian()
return binary.Read(buf, native, gnetStats)
}
func parseTcStats2(data []byte) (*ClassStatistics, error) {
rtAttrs, err := nl.ParseRouteAttr(data)
if err != nil {
return nil, err
}
stats := NewClassStatistics()
for _, datum := range rtAttrs {
switch datum.Attr.Type {
case nl.TCA_STATS_BASIC:
if err := parseGnetStats(datum.Value, stats.Basic); err != nil {
return nil, fmt.Errorf("Failed to parse ClassStatistics.Basic with: %v\n%s",
err, hex.Dump(datum.Value))
}
case nl.TCA_STATS_QUEUE:
if err := parseGnetStats(datum.Value, stats.Queue); err != nil {
return nil, fmt.Errorf("Failed to parse ClassStatistics.Queue with: %v\n%s",
err, hex.Dump(datum.Value))
}
case nl.TCA_STATS_RATE_EST:
if err := parseGnetStats(datum.Value, stats.RateEst); err != nil {
return nil, fmt.Errorf("Failed to parse ClassStatistics.RateEst with: %v\n%s",
err, hex.Dump(datum.Value))
}
}
}
return stats, nil
}

View file

@ -22,11 +22,7 @@ const (
// https://github.com/torvalds/linux/blob/master/include/uapi/linux/netfilter/nfnetlink.h -> #define NFNL_SUBSYS_CTNETLINK_EXP 2
ConntrackExpectTable = 2
)
const (
// For Parsing Mark
TCP_PROTO = 6
UDP_PROTO = 17
)
const (
// backward compatibility with golang 1.6 which does not have io.SeekCurrent
seekCurrent = 1
@ -135,11 +131,13 @@ func (h *Handle) dumpConntrackTable(table ConntrackTableType, family InetFamily)
// http://git.netfilter.org/libnetfilter_conntrack/tree/include/internal/object.h
// For the time being, the structure below allows to parse and extract the base information of a flow
type ipTuple struct {
SrcIP net.IP
Bytes uint64
DstIP net.IP
Protocol uint8
SrcPort uint16
DstPort uint16
Packets uint64
Protocol uint8
SrcIP net.IP
SrcPort uint16
}
type ConntrackFlow struct {
@ -151,11 +149,12 @@ type ConntrackFlow struct {
func (s *ConntrackFlow) String() string {
// conntrack cmd output:
// udp 17 src=127.0.0.1 dst=127.0.0.1 sport=4001 dport=1234 [UNREPLIED] src=127.0.0.1 dst=127.0.0.1 sport=1234 dport=4001 mark=0
return fmt.Sprintf("%s\t%d src=%s dst=%s sport=%d dport=%d\tsrc=%s dst=%s sport=%d dport=%d mark=%d",
// udp 17 src=127.0.0.1 dst=127.0.0.1 sport=4001 dport=1234 packets=5 bytes=532 [UNREPLIED] src=127.0.0.1 dst=127.0.0.1 sport=1234 dport=4001 packets=10 bytes=1078 mark=0
return fmt.Sprintf("%s\t%d src=%s dst=%s sport=%d dport=%d packets=%d bytes=%d\tsrc=%s dst=%s sport=%d dport=%d packets=%d bytes=%d mark=%d",
nl.L4ProtoMap[s.Forward.Protocol], s.Forward.Protocol,
s.Forward.SrcIP.String(), s.Forward.DstIP.String(), s.Forward.SrcPort, s.Forward.DstPort,
s.Reverse.SrcIP.String(), s.Reverse.DstIP.String(), s.Reverse.SrcPort, s.Reverse.DstPort, s.Mark)
s.Forward.SrcIP.String(), s.Forward.DstIP.String(), s.Forward.SrcPort, s.Forward.DstPort, s.Forward.Packets, s.Forward.Bytes,
s.Reverse.SrcIP.String(), s.Reverse.DstIP.String(), s.Reverse.SrcPort, s.Reverse.DstPort, s.Reverse.Packets, s.Reverse.Bytes,
s.Mark)
}
// This method parse the ip tuple structure
@ -220,9 +219,35 @@ func parseBERaw16(r *bytes.Reader, v *uint16) {
binary.Read(r, binary.BigEndian, v)
}
func parseBERaw32(r *bytes.Reader, v *uint32) {
binary.Read(r, binary.BigEndian, v)
}
func parseBERaw64(r *bytes.Reader, v *uint64) {
binary.Read(r, binary.BigEndian, v)
}
func parseByteAndPacketCounters(r *bytes.Reader) (bytes, packets uint64) {
for i := 0; i < 2; i++ {
switch _, t, _ := parseNfAttrTL(r); t {
case nl.CTA_COUNTERS_BYTES:
parseBERaw64(r, &bytes)
case nl.CTA_COUNTERS_PACKETS:
parseBERaw64(r, &packets)
default:
return
}
}
return
}
func parseConnectionMark(r *bytes.Reader) (mark uint32) {
parseBERaw32(r, &mark)
return
}
func parseRawData(data []byte) *ConntrackFlow {
s := &ConntrackFlow{}
var proto uint8
// First there is the Nfgenmsg header
// consume only the family field
reader := bytes.NewReader(data)
@ -238,36 +263,31 @@ func parseRawData(data []byte) *ConntrackFlow {
// <len, NLA_F_NESTED|CTA_TUPLE_IP> 4 bytes
// flow information of the reverse flow
for reader.Len() > 0 {
nested, t, l := parseNfAttrTL(reader)
if nested && t == nl.CTA_TUPLE_ORIG {
if nested, t, _ = parseNfAttrTL(reader); nested && t == nl.CTA_TUPLE_IP {
proto = parseIpTuple(reader, &s.Forward)
if nested, t, l := parseNfAttrTL(reader); nested {
switch t {
case nl.CTA_TUPLE_ORIG:
if nested, t, _ = parseNfAttrTL(reader); nested && t == nl.CTA_TUPLE_IP {
parseIpTuple(reader, &s.Forward)
}
case nl.CTA_TUPLE_REPLY:
if nested, t, _ = parseNfAttrTL(reader); nested && t == nl.CTA_TUPLE_IP {
parseIpTuple(reader, &s.Reverse)
} else {
// Header not recognized skip it
reader.Seek(int64(l), seekCurrent)
}
case nl.CTA_COUNTERS_ORIG:
s.Forward.Bytes, s.Forward.Packets = parseByteAndPacketCounters(reader)
case nl.CTA_COUNTERS_REPLY:
s.Reverse.Bytes, s.Reverse.Packets = parseByteAndPacketCounters(reader)
}
} else if nested && t == nl.CTA_TUPLE_REPLY {
if nested, t, _ = parseNfAttrTL(reader); nested && t == nl.CTA_TUPLE_IP {
parseIpTuple(reader, &s.Reverse)
// Got all the useful information stop parsing
break
} else {
// Header not recognized skip it
reader.Seek(int64(l), seekCurrent)
} else {
switch t {
case nl.CTA_MARK:
s.Mark = parseConnectionMark(reader)
}
}
}
if proto == TCP_PROTO {
reader.Seek(64, seekCurrent)
_, t, _, v := parseNfAttrTLV(reader)
if t == nl.CTA_MARK {
s.Mark = uint32(v[3])
}
} else if proto == UDP_PROTO {
reader.Seek(16, seekCurrent)
_, t, _, v := parseNfAttrTLV(reader)
if t == nl.CTA_MARK {
s.Mark = uint32(v[3])
}
}
return s
}
@ -285,7 +305,7 @@ func parseRawData(data []byte) *ConntrackFlow {
// Common parameters and options:
// -s, --src, --orig-src ip Source address from original direction
// -d, --dst, --orig-dst ip Destination address from original direction
// -r, --reply-src ip Source addres from reply direction
// -r, --reply-src ip Source address from reply direction
// -q, --reply-dst ip Destination address from reply direction
// -p, --protonum proto Layer 4 Protocol, eg. 'tcp'
// -f, --family proto Layer 3 Protocol, eg. 'ipv6'
@ -302,11 +322,14 @@ func parseRawData(data []byte) *ConntrackFlow {
type ConntrackFilterType uint8
const (
ConntrackOrigSrcIP = iota // -orig-src ip Source address from original direction
ConntrackOrigDstIP // -orig-dst ip Destination address from original direction
ConntrackNatSrcIP // -src-nat ip Source NAT ip
ConntrackNatDstIP // -dst-nat ip Destination NAT ip
ConntrackNatAnyIP // -any-nat ip Source or destination NAT ip
ConntrackOrigSrcIP = iota // -orig-src ip Source address from original direction
ConntrackOrigDstIP // -orig-dst ip Destination address from original direction
ConntrackReplySrcIP // --reply-src ip Reply Source IP
ConntrackReplyDstIP // --reply-dst ip Reply Destination IP
ConntrackReplyAnyIP // Match source or destination reply IP
ConntrackNatSrcIP = ConntrackReplySrcIP // deprecated use instead ConntrackReplySrcIP
ConntrackNatDstIP = ConntrackReplyDstIP // deprecated use instead ConntrackReplyDstIP
ConntrackNatAnyIP = ConntrackReplyAnyIP // deprecated use instaed ConntrackReplyAnyIP
)
type CustomConntrackFilter interface {
@ -351,17 +374,17 @@ func (f *ConntrackFilter) MatchConntrackFlow(flow *ConntrackFlow) bool {
}
// -src-nat ip Source NAT ip
if elem, found := f.ipFilter[ConntrackNatSrcIP]; match && found {
if elem, found := f.ipFilter[ConntrackReplySrcIP]; match && found {
match = match && elem.Equal(flow.Reverse.SrcIP)
}
// -dst-nat ip Destination NAT ip
if elem, found := f.ipFilter[ConntrackNatDstIP]; match && found {
if elem, found := f.ipFilter[ConntrackReplyDstIP]; match && found {
match = match && elem.Equal(flow.Reverse.DstIP)
}
// -any-nat ip Source or destination NAT ip
if elem, found := f.ipFilter[ConntrackNatAnyIP]; match && found {
// Match source or destination reply IP
if elem, found := f.ipFilter[ConntrackReplyAnyIP]; match && found {
match = match && (elem.Equal(flow.Reverse.SrcIP) || elem.Equal(flow.Reverse.DstIP))
}

View file

@ -0,0 +1,272 @@
package netlink
import (
"syscall"
"fmt"
"github.com/vishvananda/netlink/nl"
"golang.org/x/sys/unix"
)
// DevlinkDevEswitchAttr represents device's eswitch attributes
type DevlinkDevEswitchAttr struct {
Mode string
InlineMode string
EncapMode string
}
// DevlinkDevAttrs represents device attributes
type DevlinkDevAttrs struct {
Eswitch DevlinkDevEswitchAttr
}
// DevlinkDevice represents device and its attributes
type DevlinkDevice struct {
BusName string
DeviceName string
Attrs DevlinkDevAttrs
}
func parseDevLinkDeviceList(msgs [][]byte) ([]*DevlinkDevice, error) {
devices := make([]*DevlinkDevice, 0, len(msgs))
for _, m := range msgs {
attrs, err := nl.ParseRouteAttr(m[nl.SizeofGenlmsg:])
if err != nil {
return nil, err
}
dev := &DevlinkDevice{}
if err = dev.parseAttributes(attrs); err != nil {
return nil, err
}
devices = append(devices, dev)
}
return devices, nil
}
func eswitchStringToMode(modeName string) (uint16, error) {
if modeName == "legacy" {
return nl.DEVLINK_ESWITCH_MODE_LEGACY, nil
} else if modeName == "switchdev" {
return nl.DEVLINK_ESWITCH_MODE_SWITCHDEV, nil
} else {
return 0xffff, fmt.Errorf("invalid switchdev mode")
}
}
func parseEswitchMode(mode uint16) string {
var eswitchMode = map[uint16]string{
nl.DEVLINK_ESWITCH_MODE_LEGACY: "legacy",
nl.DEVLINK_ESWITCH_MODE_SWITCHDEV: "switchdev",
}
if eswitchMode[mode] == "" {
return "unknown"
} else {
return eswitchMode[mode]
}
}
func parseEswitchInlineMode(inlinemode uint8) string {
var eswitchInlineMode = map[uint8]string{
nl.DEVLINK_ESWITCH_INLINE_MODE_NONE: "none",
nl.DEVLINK_ESWITCH_INLINE_MODE_LINK: "link",
nl.DEVLINK_ESWITCH_INLINE_MODE_NETWORK: "network",
nl.DEVLINK_ESWITCH_INLINE_MODE_TRANSPORT: "transport",
}
if eswitchInlineMode[inlinemode] == "" {
return "unknown"
} else {
return eswitchInlineMode[inlinemode]
}
}
func parseEswitchEncapMode(encapmode uint8) string {
var eswitchEncapMode = map[uint8]string{
nl.DEVLINK_ESWITCH_ENCAP_MODE_NONE: "disable",
nl.DEVLINK_ESWITCH_ENCAP_MODE_BASIC: "enable",
}
if eswitchEncapMode[encapmode] == "" {
return "unknown"
} else {
return eswitchEncapMode[encapmode]
}
}
func (d *DevlinkDevice) parseAttributes(attrs []syscall.NetlinkRouteAttr) error {
for _, a := range attrs {
switch a.Attr.Type {
case nl.DEVLINK_ATTR_BUS_NAME:
d.BusName = string(a.Value)
case nl.DEVLINK_ATTR_DEV_NAME:
d.DeviceName = string(a.Value)
case nl.DEVLINK_ATTR_ESWITCH_MODE:
d.Attrs.Eswitch.Mode = parseEswitchMode(native.Uint16(a.Value))
case nl.DEVLINK_ATTR_ESWITCH_INLINE_MODE:
d.Attrs.Eswitch.InlineMode = parseEswitchInlineMode(uint8(a.Value[0]))
case nl.DEVLINK_ATTR_ESWITCH_ENCAP_MODE:
d.Attrs.Eswitch.EncapMode = parseEswitchEncapMode(uint8(a.Value[0]))
}
}
return nil
}
func (dev *DevlinkDevice) parseEswitchAttrs(msgs [][]byte) {
m := msgs[0]
attrs, err := nl.ParseRouteAttr(m[nl.SizeofGenlmsg:])
if err != nil {
return
}
dev.parseAttributes(attrs)
}
func (h *Handle) getEswitchAttrs(family *GenlFamily, dev *DevlinkDevice) {
msg := &nl.Genlmsg{
Command: nl.DEVLINK_CMD_ESWITCH_GET,
Version: nl.GENL_DEVLINK_VERSION,
}
req := h.newNetlinkRequest(int(family.ID), unix.NLM_F_REQUEST|unix.NLM_F_ACK)
req.AddData(msg)
b := make([]byte, len(dev.BusName))
copy(b, dev.BusName)
data := nl.NewRtAttr(nl.DEVLINK_ATTR_BUS_NAME, b)
req.AddData(data)
b = make([]byte, len(dev.DeviceName))
copy(b, dev.DeviceName)
data = nl.NewRtAttr(nl.DEVLINK_ATTR_DEV_NAME, b)
req.AddData(data)
msgs, err := req.Execute(unix.NETLINK_GENERIC, 0)
if err != nil {
return
}
dev.parseEswitchAttrs(msgs)
}
// DevLinkGetDeviceList provides a pointer to devlink devices and nil error,
// otherwise returns an error code.
func (h *Handle) DevLinkGetDeviceList() ([]*DevlinkDevice, error) {
f, err := h.GenlFamilyGet(nl.GENL_DEVLINK_NAME)
if err != nil {
return nil, err
}
msg := &nl.Genlmsg{
Command: nl.DEVLINK_CMD_GET,
Version: nl.GENL_DEVLINK_VERSION,
}
req := h.newNetlinkRequest(int(f.ID),
unix.NLM_F_REQUEST|unix.NLM_F_ACK|unix.NLM_F_DUMP)
req.AddData(msg)
msgs, err := req.Execute(unix.NETLINK_GENERIC, 0)
if err != nil {
return nil, err
}
devices, err := parseDevLinkDeviceList(msgs)
if err != nil {
return nil, err
}
for _, d := range devices {
h.getEswitchAttrs(f, d)
}
return devices, nil
}
// DevLinkGetDeviceList provides a pointer to devlink devices and nil error,
// otherwise returns an error code.
func DevLinkGetDeviceList() ([]*DevlinkDevice, error) {
return pkgHandle.DevLinkGetDeviceList()
}
func parseDevlinkDevice(msgs [][]byte) (*DevlinkDevice, error) {
m := msgs[0]
attrs, err := nl.ParseRouteAttr(m[nl.SizeofGenlmsg:])
if err != nil {
return nil, err
}
dev := &DevlinkDevice{}
if err = dev.parseAttributes(attrs); err != nil {
return nil, err
}
return dev, nil
}
func (h *Handle) createCmdReq(cmd uint8, bus string, device string) (*GenlFamily, *nl.NetlinkRequest, error) {
f, err := h.GenlFamilyGet(nl.GENL_DEVLINK_NAME)
if err != nil {
return nil, nil, err
}
msg := &nl.Genlmsg{
Command: cmd,
Version: nl.GENL_DEVLINK_VERSION,
}
req := h.newNetlinkRequest(int(f.ID),
unix.NLM_F_REQUEST|unix.NLM_F_ACK)
req.AddData(msg)
b := make([]byte, len(bus)+1)
copy(b, bus)
data := nl.NewRtAttr(nl.DEVLINK_ATTR_BUS_NAME, b)
req.AddData(data)
b = make([]byte, len(device)+1)
copy(b, device)
data = nl.NewRtAttr(nl.DEVLINK_ATTR_DEV_NAME, b)
req.AddData(data)
return f, req, nil
}
// DevlinkGetDeviceByName provides a pointer to devlink device and nil error,
// otherwise returns an error code.
func (h *Handle) DevLinkGetDeviceByName(Bus string, Device string) (*DevlinkDevice, error) {
f, req, err := h.createCmdReq(nl.DEVLINK_CMD_GET, Bus, Device)
if err != nil {
return nil, err
}
respmsg, err := req.Execute(unix.NETLINK_GENERIC, 0)
if err != nil {
return nil, err
}
dev, err := parseDevlinkDevice(respmsg)
if err == nil {
h.getEswitchAttrs(f, dev)
}
return dev, err
}
// DevlinkGetDeviceByName provides a pointer to devlink device and nil error,
// otherwise returns an error code.
func DevLinkGetDeviceByName(Bus string, Device string) (*DevlinkDevice, error) {
return pkgHandle.DevLinkGetDeviceByName(Bus, Device)
}
// DevLinkSetEswitchMode sets eswitch mode if able to set successfully or
// returns an error code.
// Equivalent to: `devlink dev eswitch set $dev mode switchdev`
// Equivalent to: `devlink dev eswitch set $dev mode legacy`
func (h *Handle) DevLinkSetEswitchMode(Dev *DevlinkDevice, NewMode string) error {
mode, err := eswitchStringToMode(NewMode)
if err != nil {
return err
}
_, req, err := h.createCmdReq(nl.DEVLINK_CMD_ESWITCH_SET, Dev.BusName, Dev.DeviceName)
if err != nil {
return err
}
req.AddData(nl.NewRtAttr(nl.DEVLINK_ATTR_ESWITCH_MODE, nl.Uint16Attr(mode)))
_, err = req.Execute(unix.NETLINK_GENERIC, 0)
return err
}
// DevLinkSetEswitchMode sets eswitch mode if able to set successfully or
// returns an error code.
// Equivalent to: `devlink dev eswitch set $dev mode switchdev`
// Equivalent to: `devlink dev eswitch set $dev mode legacy`
func DevLinkSetEswitchMode(Dev *DevlinkDevice, NewMode string) error {
return pkgHandle.DevLinkSetEswitchMode(Dev, NewMode)
}

View file

@ -2,6 +2,7 @@ package netlink
import (
"fmt"
"net"
)
type Filter interface {
@ -135,6 +136,27 @@ func (action *BpfAction) Attrs() *ActionAttrs {
return &action.ActionAttrs
}
type ConnmarkAction struct {
ActionAttrs
Zone uint16
}
func (action *ConnmarkAction) Type() string {
return "connmark"
}
func (action *ConnmarkAction) Attrs() *ActionAttrs {
return &action.ActionAttrs
}
func NewConnmarkAction() *ConnmarkAction {
return &ConnmarkAction{
ActionAttrs: ActionAttrs{
Action: TC_ACT_PIPE,
},
}
}
type MirredAct uint8
func (a MirredAct) String() string {
@ -182,47 +204,59 @@ func NewMirredAction(redirIndex int) *MirredAction {
}
}
// Sel of the U32 filters that contains multiple TcU32Key. This is the copy
// and the frontend representation of nl.TcU32Sel. It is serialized into canonical
// nl.TcU32Sel with the appropriate endianness.
type TcU32Sel struct {
Flags uint8
Offshift uint8
Nkeys uint8
Pad uint8
Offmask uint16
Off uint16
Offoff int16
Hoff int16
Hmask uint32
Keys []TcU32Key
type TunnelKeyAct int8
const (
TCA_TUNNEL_KEY_SET TunnelKeyAct = 1 // set tunnel key
TCA_TUNNEL_KEY_UNSET TunnelKeyAct = 2 // unset tunnel key
)
type TunnelKeyAction struct {
ActionAttrs
Action TunnelKeyAct
SrcAddr net.IP
DstAddr net.IP
KeyID uint32
}
// TcU32Key contained of Sel in the U32 filters. This is the copy and the frontend
// representation of nl.TcU32Key. It is serialized into chanonical nl.TcU32Sel
// with the appropriate endianness.
type TcU32Key struct {
Mask uint32
Val uint32
Off int32
OffMask int32
func (action *TunnelKeyAction) Type() string {
return "tunnel_key"
}
// U32 filters on many packet related properties
type U32 struct {
FilterAttrs
ClassId uint32
RedirIndex int
Sel *TcU32Sel
Actions []Action
func (action *TunnelKeyAction) Attrs() *ActionAttrs {
return &action.ActionAttrs
}
func (filter *U32) Attrs() *FilterAttrs {
return &filter.FilterAttrs
func NewTunnelKeyAction() *TunnelKeyAction {
return &TunnelKeyAction{
ActionAttrs: ActionAttrs{
Action: TC_ACT_PIPE,
},
}
}
func (filter *U32) Type() string {
return "u32"
type SkbEditAction struct {
ActionAttrs
QueueMapping *uint16
PType *uint16
Priority *uint32
Mark *uint32
}
func (action *SkbEditAction) Type() string {
return "skbedit"
}
func (action *SkbEditAction) Attrs() *ActionAttrs {
return &action.ActionAttrs
}
func NewSkbEditAction() *SkbEditAction {
return &SkbEditAction{
ActionAttrs: ActionAttrs{
Action: TC_ACT_PIPE,
},
}
}
// MatchAll filters match all packets
@ -262,6 +296,8 @@ type BpfFilter struct {
Fd int
Name string
DirectAction bool
Id int
Tag string
}
func (filter *BpfFilter) Type() string {

View file

@ -3,10 +3,11 @@ package netlink
import (
"bytes"
"encoding/binary"
"encoding/hex"
"errors"
"fmt"
"net"
"syscall"
"unsafe"
"github.com/vishvananda/netlink/nl"
"golang.org/x/sys/unix"
@ -20,6 +21,35 @@ const (
TC_U32_EAT = nl.TC_U32_EAT
)
// Sel of the U32 filters that contains multiple TcU32Key. This is the type
// alias and the frontend representation of nl.TcU32Sel. It is serialized into
// canonical nl.TcU32Sel with the appropriate endianness.
type TcU32Sel = nl.TcU32Sel
// TcU32Key contained of Sel in the U32 filters. This is the type alias and the
// frontend representation of nl.TcU32Key. It is serialized into chanonical
// nl.TcU32Sel with the appropriate endianness.
type TcU32Key = nl.TcU32Key
// U32 filters on many packet related properties
type U32 struct {
FilterAttrs
ClassId uint32
Divisor uint32 // Divisor MUST be power of 2.
Hash uint32
RedirIndex int
Sel *TcU32Sel
Actions []Action
}
func (filter *U32) Attrs() *FilterAttrs {
return &filter.FilterAttrs
}
func (filter *U32) Type() string {
return "u32"
}
// 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
@ -123,8 +153,24 @@ func FilterAdd(filter Filter) error {
// FilterAdd will add a filter to the system.
// Equivalent to: `tc filter add $filter`
func (h *Handle) FilterAdd(filter Filter) error {
return h.filterModify(filter, unix.NLM_F_CREATE|unix.NLM_F_EXCL)
}
// FilterReplace will replace a filter.
// Equivalent to: `tc filter replace $filter`
func FilterReplace(filter Filter) error {
return pkgHandle.FilterReplace(filter)
}
// FilterReplace will replace a filter.
// Equivalent to: `tc filter replace $filter`
func (h *Handle) FilterReplace(filter Filter) error {
return h.filterModify(filter, unix.NLM_F_CREATE)
}
func (h *Handle) filterModify(filter Filter, flags int) error {
native = nl.NativeEndian()
req := h.newNetlinkRequest(unix.RTM_NEWTFILTER, unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK)
req := h.newNetlinkRequest(unix.RTM_NEWTFILTER, flags|unix.NLM_F_ACK)
base := filter.Attrs()
msg := &nl.TcMsg{
Family: nl.FAMILY_ALL,
@ -140,8 +186,7 @@ func (h *Handle) FilterAdd(filter Filter) error {
switch filter := filter.(type) {
case *U32:
// Convert TcU32Sel into nl.TcU32Sel as it is without copy.
sel := (*nl.TcU32Sel)(unsafe.Pointer(filter.Sel))
sel := filter.Sel
if sel == nil {
// match all
sel = &nl.TcU32Sel{
@ -168,11 +213,20 @@ func (h *Handle) FilterAdd(filter Filter) error {
}
}
sel.Nkeys = uint8(len(sel.Keys))
nl.NewRtAttrChild(options, nl.TCA_U32_SEL, sel.Serialize())
options.AddRtAttr(nl.TCA_U32_SEL, sel.Serialize())
if filter.ClassId != 0 {
nl.NewRtAttrChild(options, nl.TCA_U32_CLASSID, nl.Uint32Attr(filter.ClassId))
options.AddRtAttr(nl.TCA_U32_CLASSID, nl.Uint32Attr(filter.ClassId))
}
actionsAttr := nl.NewRtAttrChild(options, nl.TCA_U32_ACT, nil)
if filter.Divisor != 0 {
if (filter.Divisor-1)&filter.Divisor != 0 {
return fmt.Errorf("illegal divisor %d. Must be a power of 2", filter.Divisor)
}
options.AddRtAttr(nl.TCA_U32_DIVISOR, nl.Uint32Attr(filter.Divisor))
}
if filter.Hash != 0 {
options.AddRtAttr(nl.TCA_U32_HASH, nl.Uint32Attr(filter.Hash))
}
actionsAttr := options.AddRtAttr(nl.TCA_U32_ACT, nil)
// backwards compatibility
if filter.RedirIndex != 0 {
filter.Actions = append([]Action{NewMirredAction(filter.RedirIndex)}, filter.Actions...)
@ -184,51 +238,51 @@ func (h *Handle) FilterAdd(filter Filter) error {
if filter.Mask != 0 {
b := make([]byte, 4)
native.PutUint32(b, filter.Mask)
nl.NewRtAttrChild(options, nl.TCA_FW_MASK, b)
options.AddRtAttr(nl.TCA_FW_MASK, b)
}
if filter.InDev != "" {
nl.NewRtAttrChild(options, nl.TCA_FW_INDEV, nl.ZeroTerminated(filter.InDev))
options.AddRtAttr(nl.TCA_FW_INDEV, nl.ZeroTerminated(filter.InDev))
}
if (filter.Police != nl.TcPolice{}) {
police := nl.NewRtAttrChild(options, nl.TCA_FW_POLICE, nil)
nl.NewRtAttrChild(police, nl.TCA_POLICE_TBF, filter.Police.Serialize())
police := options.AddRtAttr(nl.TCA_FW_POLICE, nil)
police.AddRtAttr(nl.TCA_POLICE_TBF, filter.Police.Serialize())
if (filter.Police.Rate != nl.TcRateSpec{}) {
payload := SerializeRtab(filter.Rtab)
nl.NewRtAttrChild(police, nl.TCA_POLICE_RATE, payload)
police.AddRtAttr(nl.TCA_POLICE_RATE, payload)
}
if (filter.Police.PeakRate != nl.TcRateSpec{}) {
payload := SerializeRtab(filter.Ptab)
nl.NewRtAttrChild(police, nl.TCA_POLICE_PEAKRATE, payload)
police.AddRtAttr(nl.TCA_POLICE_PEAKRATE, payload)
}
}
if filter.ClassId != 0 {
b := make([]byte, 4)
native.PutUint32(b, filter.ClassId)
nl.NewRtAttrChild(options, nl.TCA_FW_CLASSID, b)
options.AddRtAttr(nl.TCA_FW_CLASSID, b)
}
case *BpfFilter:
var bpfFlags uint32
if filter.ClassId != 0 {
nl.NewRtAttrChild(options, nl.TCA_BPF_CLASSID, nl.Uint32Attr(filter.ClassId))
options.AddRtAttr(nl.TCA_BPF_CLASSID, nl.Uint32Attr(filter.ClassId))
}
if filter.Fd >= 0 {
nl.NewRtAttrChild(options, nl.TCA_BPF_FD, nl.Uint32Attr((uint32(filter.Fd))))
options.AddRtAttr(nl.TCA_BPF_FD, nl.Uint32Attr((uint32(filter.Fd))))
}
if filter.Name != "" {
nl.NewRtAttrChild(options, nl.TCA_BPF_NAME, nl.ZeroTerminated(filter.Name))
options.AddRtAttr(nl.TCA_BPF_NAME, nl.ZeroTerminated(filter.Name))
}
if filter.DirectAction {
bpfFlags |= nl.TCA_BPF_FLAG_ACT_DIRECT
}
nl.NewRtAttrChild(options, nl.TCA_BPF_FLAGS, nl.Uint32Attr(bpfFlags))
options.AddRtAttr(nl.TCA_BPF_FLAGS, nl.Uint32Attr(bpfFlags))
case *MatchAll:
actionsAttr := nl.NewRtAttrChild(options, nl.TCA_MATCHALL_ACT, nil)
actionsAttr := options.AddRtAttr(nl.TCA_MATCHALL_ACT, nil)
if err := EncodeActions(actionsAttr, filter.Actions); err != nil {
return err
}
if filter.ClassId != 0 {
nl.NewRtAttrChild(options, nl.TCA_MATCHALL_CLASSID, nl.Uint32Attr(filter.ClassId))
options.AddRtAttr(nl.TCA_MATCHALL_CLASSID, nl.Uint32Attr(filter.ClassId))
}
}
@ -366,34 +420,91 @@ func EncodeActions(attr *nl.RtAttr, actions []Action) error {
default:
return fmt.Errorf("unknown action type %s", action.Type())
case *MirredAction:
table := nl.NewRtAttrChild(attr, tabIndex, nil)
table := attr.AddRtAttr(tabIndex, nil)
tabIndex++
nl.NewRtAttrChild(table, nl.TCA_ACT_KIND, nl.ZeroTerminated("mirred"))
aopts := nl.NewRtAttrChild(table, nl.TCA_ACT_OPTIONS, nil)
table.AddRtAttr(nl.TCA_ACT_KIND, nl.ZeroTerminated("mirred"))
aopts := table.AddRtAttr(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())
aopts.AddRtAttr(nl.TCA_MIRRED_PARMS, mirred.Serialize())
case *TunnelKeyAction:
table := attr.AddRtAttr(tabIndex, nil)
tabIndex++
table.AddRtAttr(nl.TCA_ACT_KIND, nl.ZeroTerminated("tunnel_key"))
aopts := table.AddRtAttr(nl.TCA_ACT_OPTIONS, nil)
tun := nl.TcTunnelKey{
Action: int32(action.Action),
}
toTcGen(action.Attrs(), &tun.TcGen)
aopts.AddRtAttr(nl.TCA_TUNNEL_KEY_PARMS, tun.Serialize())
if action.Action == TCA_TUNNEL_KEY_SET {
aopts.AddRtAttr(nl.TCA_TUNNEL_KEY_ENC_KEY_ID, htonl(action.KeyID))
if v4 := action.SrcAddr.To4(); v4 != nil {
aopts.AddRtAttr(nl.TCA_TUNNEL_KEY_ENC_IPV4_SRC, v4[:])
} else if v6 := action.SrcAddr.To16(); v6 != nil {
aopts.AddRtAttr(nl.TCA_TUNNEL_KEY_ENC_IPV6_SRC, v6[:])
} else {
return fmt.Errorf("invalid src addr %s for tunnel_key action", action.SrcAddr)
}
if v4 := action.DstAddr.To4(); v4 != nil {
aopts.AddRtAttr(nl.TCA_TUNNEL_KEY_ENC_IPV4_DST, v4[:])
} else if v6 := action.DstAddr.To16(); v6 != nil {
aopts.AddRtAttr(nl.TCA_TUNNEL_KEY_ENC_IPV6_DST, v6[:])
} else {
return fmt.Errorf("invalid dst addr %s for tunnel_key action", action.DstAddr)
}
}
case *SkbEditAction:
table := attr.AddRtAttr(tabIndex, nil)
tabIndex++
table.AddRtAttr(nl.TCA_ACT_KIND, nl.ZeroTerminated("skbedit"))
aopts := table.AddRtAttr(nl.TCA_ACT_OPTIONS, nil)
skbedit := nl.TcSkbEdit{}
toTcGen(action.Attrs(), &skbedit.TcGen)
aopts.AddRtAttr(nl.TCA_SKBEDIT_PARMS, skbedit.Serialize())
if action.QueueMapping != nil {
aopts.AddRtAttr(nl.TCA_SKBEDIT_QUEUE_MAPPING, nl.Uint16Attr(*action.QueueMapping))
}
if action.Priority != nil {
aopts.AddRtAttr(nl.TCA_SKBEDIT_PRIORITY, nl.Uint32Attr(*action.Priority))
}
if action.PType != nil {
aopts.AddRtAttr(nl.TCA_SKBEDIT_PTYPE, nl.Uint16Attr(*action.PType))
}
if action.Mark != nil {
aopts.AddRtAttr(nl.TCA_SKBEDIT_MARK, nl.Uint32Attr(*action.Mark))
}
case *ConnmarkAction:
table := attr.AddRtAttr(tabIndex, nil)
tabIndex++
table.AddRtAttr(nl.TCA_ACT_KIND, nl.ZeroTerminated("connmark"))
aopts := table.AddRtAttr(nl.TCA_ACT_OPTIONS, nil)
connmark := nl.TcConnmark{
Zone: action.Zone,
}
toTcGen(action.Attrs(), &connmark.TcGen)
aopts.AddRtAttr(nl.TCA_CONNMARK_PARMS, connmark.Serialize())
case *BpfAction:
table := nl.NewRtAttrChild(attr, tabIndex, nil)
table := attr.AddRtAttr(tabIndex, nil)
tabIndex++
nl.NewRtAttrChild(table, nl.TCA_ACT_KIND, nl.ZeroTerminated("bpf"))
aopts := nl.NewRtAttrChild(table, nl.TCA_ACT_OPTIONS, nil)
table.AddRtAttr(nl.TCA_ACT_KIND, nl.ZeroTerminated("bpf"))
aopts := table.AddRtAttr(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))
aopts.AddRtAttr(nl.TCA_ACT_BPF_PARMS, gen.Serialize())
aopts.AddRtAttr(nl.TCA_ACT_BPF_FD, nl.Uint32Attr(uint32(action.Fd)))
aopts.AddRtAttr(nl.TCA_ACT_BPF_NAME, nl.ZeroTerminated(action.Name))
case *GenericAction:
table := nl.NewRtAttrChild(attr, tabIndex, nil)
table := attr.AddRtAttr(tabIndex, nil)
tabIndex++
nl.NewRtAttrChild(table, nl.TCA_ACT_KIND, nl.ZeroTerminated("gact"))
aopts := nl.NewRtAttrChild(table, nl.TCA_ACT_OPTIONS, nil)
table.AddRtAttr(nl.TCA_ACT_KIND, nl.ZeroTerminated("gact"))
aopts := table.AddRtAttr(nl.TCA_ACT_OPTIONS, nil)
gen := nl.TcGen{}
toTcGen(action.Attrs(), &gen)
nl.NewRtAttrChild(aopts, nl.TCA_GACT_PARMS, gen.Serialize())
aopts.AddRtAttr(nl.TCA_GACT_PARMS, gen.Serialize())
}
}
return nil
@ -419,8 +530,14 @@ func parseActions(tables []syscall.NetlinkRouteAttr) ([]Action, error) {
action = &MirredAction{}
case "bpf":
action = &BpfAction{}
case "connmark":
action = &ConnmarkAction{}
case "gact":
action = &GenericAction{}
case "tunnel_key":
action = &TunnelKeyAction{}
case "skbedit":
action = &SkbEditAction{}
default:
break nextattr
}
@ -435,11 +552,46 @@ func parseActions(tables []syscall.NetlinkRouteAttr) ([]Action, error) {
switch adatum.Attr.Type {
case nl.TCA_MIRRED_PARMS:
mirred := *nl.DeserializeTcMirred(adatum.Value)
toAttrs(&mirred.TcGen, action.Attrs())
action.(*MirredAction).ActionAttrs = ActionAttrs{}
toAttrs(&mirred.TcGen, action.Attrs())
action.(*MirredAction).Ifindex = int(mirred.Ifindex)
action.(*MirredAction).MirredAction = MirredAct(mirred.Eaction)
}
case "tunnel_key":
switch adatum.Attr.Type {
case nl.TCA_TUNNEL_KEY_PARMS:
tun := *nl.DeserializeTunnelKey(adatum.Value)
action.(*TunnelKeyAction).ActionAttrs = ActionAttrs{}
toAttrs(&tun.TcGen, action.Attrs())
action.(*TunnelKeyAction).Action = TunnelKeyAct(tun.Action)
case nl.TCA_TUNNEL_KEY_ENC_KEY_ID:
action.(*TunnelKeyAction).KeyID = networkOrder.Uint32(adatum.Value[0:4])
case nl.TCA_TUNNEL_KEY_ENC_IPV6_SRC:
case nl.TCA_TUNNEL_KEY_ENC_IPV4_SRC:
action.(*TunnelKeyAction).SrcAddr = net.IP(adatum.Value[:])
case nl.TCA_TUNNEL_KEY_ENC_IPV6_DST:
case nl.TCA_TUNNEL_KEY_ENC_IPV4_DST:
action.(*TunnelKeyAction).DstAddr = net.IP(adatum.Value[:])
}
case "skbedit":
switch adatum.Attr.Type {
case nl.TCA_SKBEDIT_PARMS:
skbedit := *nl.DeserializeSkbEdit(adatum.Value)
action.(*SkbEditAction).ActionAttrs = ActionAttrs{}
toAttrs(&skbedit.TcGen, action.Attrs())
case nl.TCA_SKBEDIT_MARK:
mark := native.Uint32(adatum.Value[0:4])
action.(*SkbEditAction).Mark = &mark
case nl.TCA_SKBEDIT_PRIORITY:
priority := native.Uint32(adatum.Value[0:4])
action.(*SkbEditAction).Priority = &priority
case nl.TCA_SKBEDIT_PTYPE:
ptype := native.Uint16(adatum.Value[0:2])
action.(*SkbEditAction).PType = &ptype
case nl.TCA_SKBEDIT_QUEUE_MAPPING:
mapping := native.Uint16(adatum.Value[0:2])
action.(*SkbEditAction).QueueMapping = &mapping
}
case "bpf":
switch adatum.Attr.Type {
case nl.TCA_ACT_BPF_PARMS:
@ -450,6 +602,14 @@ func parseActions(tables []syscall.NetlinkRouteAttr) ([]Action, error) {
case nl.TCA_ACT_BPF_NAME:
action.(*BpfAction).Name = string(adatum.Value[:len(adatum.Value)-1])
}
case "connmark":
switch adatum.Attr.Type {
case nl.TCA_CONNMARK_PARMS:
connmark := *nl.DeserializeTcConnmark(adatum.Value)
action.(*ConnmarkAction).ActionAttrs = ActionAttrs{}
toAttrs(&connmark.TcGen, action.Attrs())
action.(*ConnmarkAction).Zone = connmark.Zone
}
case "gact":
switch adatum.Attr.Type {
case nl.TCA_GACT_PARMS:
@ -474,7 +634,7 @@ func parseU32Data(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error)
case nl.TCA_U32_SEL:
detailed = true
sel := nl.DeserializeTcU32Sel(datum.Value)
u32.Sel = (*TcU32Sel)(unsafe.Pointer(sel))
u32.Sel = sel
if native != networkOrder {
// Handle the endianness of attributes
u32.Sel.Offmask = native.Uint16(htons(sel.Offmask))
@ -500,6 +660,10 @@ func parseU32Data(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error)
}
case nl.TCA_U32_CLASSID:
u32.ClassId = native.Uint32(datum.Value)
case nl.TCA_U32_DIVISOR:
u32.Divisor = native.Uint32(datum.Value)
case nl.TCA_U32_HASH:
u32.Hash = native.Uint32(datum.Value)
}
}
return detailed, nil
@ -551,6 +715,10 @@ func parseBpfData(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error)
if (flags & nl.TCA_BPF_FLAG_ACT_DIRECT) != 0 {
bpf.DirectAction = true
}
case nl.TCA_BPF_ID:
bpf.Id = int(native.Uint32(datum.Value[0:4]))
case nl.TCA_BPF_TAG:
bpf.Tag = hex.EncodeToString(datum.Value[:len(datum.Value)-1])
}
}
return detailed, nil

View file

@ -90,11 +90,7 @@ func (h *Handle) FouAdd(f Fou) error {
req.AddRawData(raw)
_, err = req.Execute(unix.NETLINK_GENERIC, 0)
if err != nil {
return err
}
return nil
return err
}
func FouDel(f Fou) error {

View file

@ -157,6 +157,9 @@ func (h *Handle) GenlFamilyGet(name string) (*GenlFamily, error) {
return nil, err
}
families, err := parseFamilies(msgs)
if err != nil {
return nil, err
}
if len(families) != 1 {
return nil, fmt.Errorf("invalid response for GENL_CTRL_CMD_GETFAMILY")
}

View file

@ -0,0 +1,8 @@
module github.com/vishvananda/netlink
go 1.12
require (
github.com/vishvananda/netns v0.0.0-20191106174202-0a2b9b5464df
golang.org/x/sys v0.0.0-20190606203320-7fc4e5ec1444
)

View file

@ -91,7 +91,7 @@ func (h *Handle) GetSocketReceiveBufferSize() ([]int, error) {
return results, nil
}
// NewHandle returns a netlink handle on the network namespace
// NewHandleAt 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) {

View file

@ -73,10 +73,18 @@ func (h *Handle) LinkSetVfVlan(link Link, vf, vlan int) error {
return ErrNotImplemented
}
func (h *Handle) LinkSetVfVlanQos(link Link, vf, vlan, qos int) error {
return ErrNotImplemented
}
func (h *Handle) LinkSetVfTxRate(link Link, vf, rate int) error {
return ErrNotImplemented
}
func (h *Handle) LinkSetVfRate(link Link, vf, minRate, maxRate int) error {
return ErrNotImplemented
}
func (h *Handle) LinkSetMaster(link Link, master *Bridge) error {
return ErrNotImplemented
}
@ -149,6 +157,10 @@ func (h *Handle) LinkSetTxQLen(link Link, qlen int) error {
return ErrNotImplemented
}
func (h *Handle) LinkSetGroup(link Link, group int) error {
return ErrNotImplemented
}
func (h *Handle) setProtinfoAttr(link Link, mode bool, attr int) error {
return ErrNotImplemented
}

View file

@ -56,18 +56,10 @@ type ethtoolSset struct {
data [1]uint32
}
// ethtoolGstrings is string set for data tagging
type ethtoolGstrings struct {
cmd uint32
stringSet uint32
length uint32
data [32]byte
}
type ethtoolStats struct {
cmd uint32
nStats uint32
data [1]uint64
// Followed by nStats * []uint64.
}
// newIocltSlaveReq returns filled IfreqSlave with proper interface names

View file

@ -4,6 +4,7 @@ import (
"fmt"
"net"
"os"
"strconv"
)
// Link represents a link device from netlink. Shared link attributes
@ -41,6 +42,29 @@ type LinkAttrs struct {
NetNsID int
NumTxQueues int
NumRxQueues int
GSOMaxSize uint32
GSOMaxSegs uint32
Vfs []VfInfo // virtual functions available on link
Group uint32
Slave LinkSlave
}
// LinkSlave represents a slave device.
type LinkSlave interface {
SlaveType() string
}
// VfInfo represents configuration of virtual function
type VfInfo struct {
ID int
Mac net.HardwareAddr
Vlan int
Qos int
TxRate int // IFLA_VF_TX_RATE Max TxRate
Spoofchk bool
LinkState uint32
MaxTxRate uint32 // IFLA_VF_RATE Max TxRate
MinTxRate uint32 // IFLA_VF_RATE Min TxRate
}
// LinkOperState represents the values of the IFLA_OPERSTATE link
@ -223,6 +247,7 @@ type Bridge struct {
LinkAttrs
MulticastSnooping *bool
HelloTime *uint32
VlanFiltering *bool
}
func (bridge *Bridge) Attrs() *LinkAttrs {
@ -236,7 +261,8 @@ func (bridge *Bridge) Type() string {
// Vlan links have ParentIndex set in their Attrs()
type Vlan struct {
LinkAttrs
VlanId int
VlanId int
VlanProtocol VlanProtocol
}
func (vlan *Vlan) Attrs() *LinkAttrs {
@ -290,10 +316,13 @@ type TuntapFlag uint16
// Tuntap links created via /dev/tun/tap, but can be destroyed via netlink
type Tuntap struct {
LinkAttrs
Mode TuntapMode
Flags TuntapFlag
Queues int
Fds []*os.File
Mode TuntapMode
Flags TuntapFlag
NonPersist bool
Queues int
Fds []*os.File
Owner uint32
Group uint32
}
func (tuntap *Tuntap) Attrs() *LinkAttrs {
@ -307,7 +336,8 @@ func (tuntap *Tuntap) Type() string {
// Veth devices must specify PeerName on create
type Veth struct {
LinkAttrs
PeerName string // veth on create only
PeerName string // veth on create only
PeerHardwareAddr net.HardwareAddr
}
func (veth *Veth) Attrs() *LinkAttrs {
@ -376,9 +406,18 @@ const (
IPVLAN_MODE_MAX
)
type IPVlanFlag uint16
const (
IPVLAN_FLAG_BRIDGE IPVlanFlag = iota
IPVLAN_FLAG_PRIVATE
IPVLAN_FLAG_VEPA
)
type IPVlan struct {
LinkAttrs
Mode IPVlanMode
Flag IPVlanFlag
}
func (ipvlan *IPVlan) Attrs() *LinkAttrs {
@ -389,6 +428,43 @@ func (ipvlan *IPVlan) Type() string {
return "ipvlan"
}
// VlanProtocol type
type VlanProtocol int
func (p VlanProtocol) String() string {
s, ok := VlanProtocolToString[p]
if !ok {
return fmt.Sprintf("VlanProtocol(%d)", p)
}
return s
}
// StringToVlanProtocol returns vlan protocol, or unknown is the s is invalid.
func StringToVlanProtocol(s string) VlanProtocol {
mode, ok := StringToVlanProtocolMap[s]
if !ok {
return VLAN_PROTOCOL_UNKNOWN
}
return mode
}
// VlanProtocol possible values
const (
VLAN_PROTOCOL_UNKNOWN VlanProtocol = 0
VLAN_PROTOCOL_8021Q VlanProtocol = 0x8100
VLAN_PROTOCOL_8021AD VlanProtocol = 0x88A8
)
var VlanProtocolToString = map[VlanProtocol]string{
VLAN_PROTOCOL_8021Q: "802.1q",
VLAN_PROTOCOL_8021AD: "802.1ad",
}
var StringToVlanProtocolMap = map[string]VlanProtocol{
"802.1q": VLAN_PROTOCOL_8021Q,
"802.1ad": VLAN_PROTOCOL_8021AD,
}
// BondMode type
type BondMode int
@ -400,7 +476,7 @@ func (b BondMode) String() string {
return s
}
// StringToBondMode returns bond mode, or uknonw is the s is invalid.
// StringToBondMode returns bond mode, or unknown is the s is invalid.
func StringToBondMode(s string) BondMode {
mode, ok := StringToBondModeMap[s]
if !ok {
@ -491,7 +567,7 @@ func (b BondXmitHashPolicy) String() string {
return s
}
// StringToBondXmitHashPolicy returns bond lacp arte, or uknonw is the s is invalid.
// StringToBondXmitHashPolicy returns bond lacp arte, or unknown is the s is invalid.
func StringToBondXmitHashPolicy(s string) BondXmitHashPolicy {
lacp, ok := StringToBondXmitHashPolicyMap[s]
if !ok {
@ -536,7 +612,7 @@ func (b BondLacpRate) String() string {
return s
}
// StringToBondLacpRate returns bond lacp arte, or uknonw is the s is invalid.
// StringToBondLacpRate returns bond lacp arte, or unknown is the s is invalid.
func StringToBondLacpRate(s string) BondLacpRate {
lacp, ok := StringToBondLacpRateMap[s]
if !ok {
@ -680,6 +756,67 @@ func (bond *Bond) Type() string {
return "bond"
}
// BondSlaveState represents the values of the IFLA_BOND_SLAVE_STATE bond slave
// attribute, which contains the state of the bond slave.
type BondSlaveState uint8
const (
BondStateActive = iota // Link is active.
BondStateBackup // Link is backup.
)
func (s BondSlaveState) String() string {
switch s {
case BondStateActive:
return "ACTIVE"
case BondStateBackup:
return "BACKUP"
default:
return strconv.Itoa(int(s))
}
}
// BondSlaveState represents the values of the IFLA_BOND_SLAVE_MII_STATUS bond slave
// attribute, which contains the status of MII link monitoring
type BondSlaveMiiStatus uint8
const (
BondLinkUp = iota // link is up and running.
BondLinkFail // link has just gone down.
BondLinkDown // link has been down for too long time.
BondLinkBack // link is going back.
)
func (s BondSlaveMiiStatus) String() string {
switch s {
case BondLinkUp:
return "UP"
case BondLinkFail:
return "GOING_DOWN"
case BondLinkDown:
return "DOWN"
case BondLinkBack:
return "GOING_BACK"
default:
return strconv.Itoa(int(s))
}
}
type BondSlave struct {
State BondSlaveState
MiiStatus BondSlaveMiiStatus
LinkFailureCount uint32
PermHardwareAddr net.HardwareAddr
QueueId uint16
AggregatorId uint16
AdActorOperPortState uint8
AdPartnerOperPortState uint16
}
func (b *BondSlave) SlaveType() string {
return "bond"
}
// Gretap devices must specify LocalIP and RemoteIP on create
type Gretap struct {
LinkAttrs
@ -734,6 +871,27 @@ func (iptun *Iptun) Type() string {
return "ipip"
}
type Ip6tnl struct {
LinkAttrs
Link uint32
Local net.IP
Remote net.IP
Ttl uint8
Tos uint8
EncapLimit uint8
Flags uint32
Proto uint8
FlowInfo uint32
}
func (ip6tnl *Ip6tnl) Attrs() *LinkAttrs {
return &ip6tnl.LinkAttrs
}
func (ip6tnl *Ip6tnl) Type() string {
return "ip6tnl"
}
type Sittun struct {
LinkAttrs
Link uint32
@ -769,7 +927,10 @@ func (vti *Vti) Attrs() *LinkAttrs {
return &vti.LinkAttrs
}
func (iptun *Vti) Type() string {
func (vti *Vti) Type() string {
if vti.Local.To4() == nil {
return "vti6"
}
return "vti"
}
@ -831,11 +992,68 @@ func (gtp *GTP) Type() string {
return "gtp"
}
// Virtual XFRM Interfaces
// Named "xfrmi" to prevent confusion with XFRM objects
type Xfrmi struct {
LinkAttrs
Ifid uint32
}
func (xfrm *Xfrmi) Attrs() *LinkAttrs {
return &xfrm.LinkAttrs
}
func (xfrm *Xfrmi) Type() string {
return "xfrm"
}
// IPoIB interface
type IPoIBMode uint16
func (m *IPoIBMode) String() string {
str, ok := iPoIBModeToString[*m]
if !ok {
return fmt.Sprintf("mode(%d)", *m)
}
return str
}
const (
IPOIB_MODE_DATAGRAM = iota
IPOIB_MODE_CONNECTED
)
var iPoIBModeToString = map[IPoIBMode]string{
IPOIB_MODE_DATAGRAM: "datagram",
IPOIB_MODE_CONNECTED: "connected",
}
var StringToIPoIBMode = map[string]IPoIBMode{
"datagram": IPOIB_MODE_DATAGRAM,
"connected": IPOIB_MODE_CONNECTED,
}
type IPoIB struct {
LinkAttrs
Pkey uint16
Mode IPoIBMode
Umcast uint16
}
func (ipoib *IPoIB) Attrs() *LinkAttrs {
return &ipoib.LinkAttrs
}
func (ipoib *IPoIB) Type() string {
return "ipoib"
}
// iproute2 supported devices;
// vlan | veth | vcan | dummy | ifb | macvlan | macvtap |
// bridge | bond | ipoib | ip6tnl | ipip | sit | vxlan |
// gre | gretap | ip6gre | ip6gretap | vti | nlmon |
// bond_slave | ipvlan
// gre | gretap | ip6gre | ip6gretap | vti | vti6 | nlmon |
// bond_slave | ipvlan | xfrm
// LinkNotFoundError wraps the various not found errors when
// getting/reading links. This is intended for better error

File diff suppressed because it is too large Load diff

View file

@ -17,9 +17,16 @@ type Neigh struct {
LLIPAddr net.IP //Used in the case of NHRP
Vlan int
VNI int
MasterIndex int
}
// String returns $ip/$hwaddr $label
func (neigh *Neigh) String() string {
return fmt.Sprintf("%s %s", neigh.IP, neigh.HardwareAddr)
}
// NeighUpdate is sent when a neighbor changes - type is RTM_NEWNEIGH or RTM_DELNEIGH.
type NeighUpdate struct {
Type uint16
Neigh
}

View file

@ -1,10 +1,13 @@
package netlink
import (
"fmt"
"net"
"syscall"
"unsafe"
"github.com/vishvananda/netlink/nl"
"github.com/vishvananda/netns"
"golang.org/x/sys/unix"
)
@ -18,7 +21,10 @@ const (
NDA_PORT
NDA_VNI
NDA_IFINDEX
NDA_MAX = NDA_IFINDEX
NDA_MASTER
NDA_LINK_NETNSID
NDA_SRC_VNI
NDA_MAX = NDA_SRC_VNI
)
// Neighbor Cache Entry States.
@ -43,6 +49,7 @@ const (
NTF_ROUTER = 0x80
)
// Ndmsg is for adding, removing or receiving information about a neighbor table entry
type Ndmsg struct {
Family uint8
Index uint32
@ -170,45 +177,58 @@ func neighHandle(neigh *Neigh, req *nl.NetlinkRequest) error {
req.AddData(vniData)
}
if neigh.MasterIndex != 0 {
masterData := nl.NewRtAttr(NDA_MASTER, nl.Uint32Attr(uint32(neigh.MasterIndex)))
req.AddData(masterData)
}
_, err := req.Execute(unix.NETLINK_ROUTE, 0)
return err
}
// NeighList gets a list of IP-MAC mappings in the system (ARP table).
// NeighList returns 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 NeighList(linkIndex, family int) ([]Neigh, error) {
return pkgHandle.NeighList(linkIndex, family)
}
// NeighProxyList gets a list of neighbor proxies in the system.
// NeighProxyList returns a list of neighbor proxies in the system.
// Equivalent to: `ip neighbor show proxy`.
// The list can be filtered by link and ip family.
func NeighProxyList(linkIndex, family int) ([]Neigh, error) {
return pkgHandle.NeighProxyList(linkIndex, family)
}
// NeighList gets a list of IP-MAC mappings in the system (ARP table).
// NeighList returns a list of IP-MAC mappings in the system (ARP table).
// Equivalent to: `ip neighbor show`.
// The list can be filtered by link and ip family.
func (h *Handle) NeighList(linkIndex, family int) ([]Neigh, error) {
return h.neighList(linkIndex, family, 0)
return h.NeighListExecute(Ndmsg{
Family: uint8(family),
Index: uint32(linkIndex),
})
}
// NeighProxyList gets a list of neighbor proxies in the system.
// NeighProxyList returns a list of neighbor proxies in the system.
// Equivalent to: `ip neighbor show proxy`.
// The list can be filtered by link, ip family.
func (h *Handle) NeighProxyList(linkIndex, family int) ([]Neigh, error) {
return h.neighList(linkIndex, family, NTF_PROXY)
}
func (h *Handle) neighList(linkIndex, family, flags int) ([]Neigh, error) {
req := h.newNetlinkRequest(unix.RTM_GETNEIGH, unix.NLM_F_DUMP)
msg := Ndmsg{
return h.NeighListExecute(Ndmsg{
Family: uint8(family),
Index: uint32(linkIndex),
Flags: uint8(flags),
}
Flags: NTF_PROXY,
})
}
// NeighListExecute returns a list of neighbour entries filtered by link, ip family, flag and state.
func NeighListExecute(msg Ndmsg) ([]Neigh, error) {
return pkgHandle.NeighListExecute(msg)
}
// NeighListExecute returns a list of neighbour entries filtered by link, ip family, flag and state.
func (h *Handle) NeighListExecute(msg Ndmsg) ([]Neigh, error) {
req := h.newNetlinkRequest(unix.RTM_GETNEIGH, unix.NLM_F_DUMP)
req.AddData(&msg)
msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWNEIGH)
@ -219,7 +239,7 @@ func (h *Handle) neighList(linkIndex, family, flags int) ([]Neigh, error) {
var res []Neigh
for _, m := range msgs {
ndm := deserializeNdmsg(m)
if linkIndex != 0 && int(ndm.Index) != linkIndex {
if msg.Index != 0 && ndm.Index != msg.Index {
// Ignore messages from other interfaces
continue
}
@ -251,14 +271,6 @@ func NeighDeserialize(m []byte) (*Neigh, error) {
return nil, err
}
// This should be cached for perfomance
// once per table dump
link, err := LinkByIndex(neigh.LinkIndex)
if err != nil {
return nil, err
}
encapType := link.Attrs().EncapType
for _, attr := range attrs {
switch attr.Attr.Type {
case NDA_DST:
@ -268,13 +280,16 @@ func NeighDeserialize(m []byte) (*Neigh, error) {
// #define RTA_LENGTH(len) (RTA_ALIGN(sizeof(struct rtattr)) + (len))
// #define RTA_PAYLOAD(rta) ((int)((rta)->rta_len) - RTA_LENGTH(0))
attrLen := attr.Attr.Len - unix.SizeofRtAttr
if attrLen == 4 && (encapType == "ipip" ||
encapType == "sit" ||
encapType == "gre") {
if attrLen == 4 {
neigh.LLIPAddr = net.IP(attr.Value)
} else if attrLen == 16 &&
encapType == "tunnel6" {
neigh.IP = net.IP(attr.Value)
} else if attrLen == 16 {
// Can be IPv6 or FireWire HWAddr
link, err := LinkByIndex(neigh.LinkIndex)
if err == nil && link.Attrs().EncapType == "tunnel6" {
neigh.IP = net.IP(attr.Value)
} else {
neigh.HardwareAddr = net.HardwareAddr(attr.Value)
}
} else {
neigh.HardwareAddr = net.HardwareAddr(attr.Value)
}
@ -282,8 +297,126 @@ func NeighDeserialize(m []byte) (*Neigh, error) {
neigh.Vlan = int(native.Uint16(attr.Value[0:2]))
case NDA_VNI:
neigh.VNI = int(native.Uint32(attr.Value[0:4]))
case NDA_MASTER:
neigh.MasterIndex = int(native.Uint32(attr.Value[0:4]))
}
}
return &neigh, nil
}
// NeighSubscribe takes a chan down which notifications will be sent
// when neighbors are added or deleted. Close the 'done' chan to stop subscription.
func NeighSubscribe(ch chan<- NeighUpdate, done <-chan struct{}) error {
return neighSubscribeAt(netns.None(), netns.None(), ch, done, nil, false)
}
// NeighSubscribeAt works like NeighSubscribe plus it allows the caller
// to choose the network namespace in which to subscribe (ns).
func NeighSubscribeAt(ns netns.NsHandle, ch chan<- NeighUpdate, done <-chan struct{}) error {
return neighSubscribeAt(ns, netns.None(), ch, done, nil, false)
}
// NeighSubscribeOptions contains a set of options to use with
// NeighSubscribeWithOptions.
type NeighSubscribeOptions struct {
Namespace *netns.NsHandle
ErrorCallback func(error)
ListExisting bool
}
// NeighSubscribeWithOptions work like NeighSubscribe but enable to
// provide additional options to modify the behavior. Currently, the
// namespace can be provided as well as an error callback.
func NeighSubscribeWithOptions(ch chan<- NeighUpdate, done <-chan struct{}, options NeighSubscribeOptions) error {
if options.Namespace == nil {
none := netns.None()
options.Namespace = &none
}
return neighSubscribeAt(*options.Namespace, netns.None(), ch, done, options.ErrorCallback, options.ListExisting)
}
func neighSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- NeighUpdate, done <-chan struct{}, cberr func(error), listExisting bool) error {
s, err := nl.SubscribeAt(newNs, curNs, unix.NETLINK_ROUTE, unix.RTNLGRP_NEIGH)
makeRequest := func(family int) error {
req := pkgHandle.newNetlinkRequest(unix.RTM_GETNEIGH,
unix.NLM_F_DUMP)
infmsg := nl.NewIfInfomsg(family)
req.AddData(infmsg)
if err := s.Send(req); err != nil {
return err
}
return nil
}
if err != nil {
return err
}
if done != nil {
go func() {
<-done
s.Close()
}()
}
if listExisting {
if err := makeRequest(unix.AF_UNSPEC); err != nil {
return err
}
// We have to wait for NLMSG_DONE before making AF_BRIDGE request
}
go func() {
defer close(ch)
for {
msgs, from, err := s.Receive()
if err != nil {
if cberr != nil {
cberr(err)
}
return
}
if from.Pid != nl.PidKernel {
if cberr != nil {
cberr(fmt.Errorf("Wrong sender portid %d, expected %d", from.Pid, nl.PidKernel))
}
continue
}
for _, m := range msgs {
if m.Header.Type == unix.NLMSG_DONE {
if listExisting {
// This will be called after handling AF_UNSPEC
// list request, we have to wait for NLMSG_DONE
// before making another request
if err := makeRequest(unix.AF_BRIDGE); err != nil {
if cberr != nil {
cberr(err)
}
return
}
listExisting = false
}
continue
}
if m.Header.Type == unix.NLMSG_ERROR {
native := nl.NativeEndian()
error := int32(native.Uint32(m.Data[0:4]))
if error == 0 {
continue
}
if cberr != nil {
cberr(syscall.Errno(-error))
}
return
}
neigh, err := NeighDeserialize(m.Data)
if err != nil {
if cberr != nil {
cberr(err)
}
return
}
ch <- NeighUpdate{Type: m.Header.Type, Neigh: *neigh}
}
}
}()
return nil
}

View file

@ -27,7 +27,8 @@ func ParseIPNet(s string) (*net.IPNet, error) {
if err != nil {
return nil, err
}
return &net.IPNet{IP: ip, Mask: ipNet.Mask}, nil
ipNet.IP = ip
return ipNet, nil
}
// NewIPNet generates an IPNet from an ip address using a netmask of 32 or 128.

View file

@ -48,10 +48,18 @@ func LinkSetVfVlan(link Link, vf, vlan int) error {
return ErrNotImplemented
}
func LinkSetVfVlanQos(link Link, vf, vlan, qos int) error {
return ErrNotImplemented
}
func LinkSetVfTxRate(link Link, vf, rate int) error {
return ErrNotImplemented
}
func LinkSetVfRate(link Link, vf, minRate, maxRate int) error {
return ErrNotImplemented
}
func LinkSetNoMaster(link Link) error {
return ErrNotImplemented
}
@ -152,6 +160,10 @@ func AddrAdd(link Link, addr *Addr) error {
return ErrNotImplemented
}
func AddrReplace(link Link, addr *Addr) error {
return ErrNotImplemented
}
func AddrDel(link Link, addr *Addr) error {
return ErrNotImplemented
}

View file

@ -0,0 +1,141 @@
package netlink
// Network namespace ID functions
//
// The kernel has a weird concept called the network namespace ID.
// This is different from the file reference in proc (and any bind-mounted
// namespaces, etc.)
//
// Instead, namespaces can be assigned a numeric ID at any time. Once set,
// the ID is fixed. The ID can either be set manually by the user, or
// automatically, triggered by certain kernel actions. The most common kernel
// action that triggers namespace ID creation is moving one end of a veth pair
// in to that namespace.
import (
"fmt"
"github.com/vishvananda/netlink/nl"
"golang.org/x/sys/unix"
)
// These can be replaced by the values from sys/unix when it is next released.
const (
_ = iota
NETNSA_NSID
NETNSA_PID
NETNSA_FD
)
// GetNetNsIdByPid looks up the network namespace ID for a given pid (really thread id).
// Returns -1 if the namespace does not have an ID set.
func (h *Handle) GetNetNsIdByPid(pid int) (int, error) {
return h.getNetNsId(NETNSA_PID, uint32(pid))
}
// GetNetNsIdByPid looks up the network namespace ID for a given pid (really thread id).
// Returns -1 if the namespace does not have an ID set.
func GetNetNsIdByPid(pid int) (int, error) {
return pkgHandle.GetNetNsIdByPid(pid)
}
// SetNetNSIdByPid sets the ID of the network namespace for a given pid (really thread id).
// The ID can only be set for namespaces without an ID already set.
func (h *Handle) SetNetNsIdByPid(pid, nsid int) error {
return h.setNetNsId(NETNSA_PID, uint32(pid), uint32(nsid))
}
// SetNetNSIdByPid sets the ID of the network namespace for a given pid (really thread id).
// The ID can only be set for namespaces without an ID already set.
func SetNetNsIdByPid(pid, nsid int) error {
return pkgHandle.SetNetNsIdByPid(pid, nsid)
}
// GetNetNsIdByFd looks up the network namespace ID for a given fd.
// fd must be an open file descriptor to a namespace file.
// Returns -1 if the namespace does not have an ID set.
func (h *Handle) GetNetNsIdByFd(fd int) (int, error) {
return h.getNetNsId(NETNSA_FD, uint32(fd))
}
// GetNetNsIdByFd looks up the network namespace ID for a given fd.
// fd must be an open file descriptor to a namespace file.
// Returns -1 if the namespace does not have an ID set.
func GetNetNsIdByFd(fd int) (int, error) {
return pkgHandle.GetNetNsIdByFd(fd)
}
// SetNetNSIdByFd sets the ID of the network namespace for a given fd.
// fd must be an open file descriptor to a namespace file.
// The ID can only be set for namespaces without an ID already set.
func (h *Handle) SetNetNsIdByFd(fd, nsid int) error {
return h.setNetNsId(NETNSA_FD, uint32(fd), uint32(nsid))
}
// SetNetNSIdByFd sets the ID of the network namespace for a given fd.
// fd must be an open file descriptor to a namespace file.
// The ID can only be set for namespaces without an ID already set.
func SetNetNsIdByFd(fd, nsid int) error {
return pkgHandle.SetNetNsIdByFd(fd, nsid)
}
// getNetNsId requests the netnsid for a given type-val pair
// type should be either NETNSA_PID or NETNSA_FD
func (h *Handle) getNetNsId(attrType int, val uint32) (int, error) {
req := h.newNetlinkRequest(unix.RTM_GETNSID, unix.NLM_F_REQUEST)
rtgen := nl.NewRtGenMsg()
req.AddData(rtgen)
b := make([]byte, 4, 4)
native.PutUint32(b, val)
attr := nl.NewRtAttr(attrType, b)
req.AddData(attr)
msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWNSID)
if err != nil {
return 0, err
}
for _, m := range msgs {
msg := nl.DeserializeRtGenMsg(m)
attrs, err := nl.ParseRouteAttr(m[msg.Len():])
if err != nil {
return 0, err
}
for _, attr := range attrs {
switch attr.Attr.Type {
case NETNSA_NSID:
return int(int32(native.Uint32(attr.Value))), nil
}
}
}
return 0, fmt.Errorf("unexpected empty result")
}
// setNetNsId sets the netnsid for a given type-val pair
// type should be either NETNSA_PID or NETNSA_FD
// The ID can only be set for namespaces without an ID already set
func (h *Handle) setNetNsId(attrType int, val uint32, newnsid uint32) error {
req := h.newNetlinkRequest(unix.RTM_NEWNSID, unix.NLM_F_REQUEST|unix.NLM_F_ACK)
rtgen := nl.NewRtGenMsg()
req.AddData(rtgen)
b := make([]byte, 4, 4)
native.PutUint32(b, val)
attr := nl.NewRtAttr(attrType, b)
req.AddData(attr)
b1 := make([]byte, 4, 4)
native.PutUint32(b1, newnsid)
attr1 := nl.NewRtAttr(NETNSA_NSID, b1)
req.AddData(attr1)
_, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWNSID)
return err
}

View file

@ -0,0 +1,19 @@
// +build !linux
package netlink
func GetNetNsIdByPid(pid int) (int, error) {
return 0, ErrNotImplemented
}
func SetNetNsIdByPid(pid, nsid int) error {
return ErrNotImplemented
}
func GetNetNsIdByFd(fd int) (int, error) {
return 0, ErrNotImplemented
}
func SetNetNsIdByFd(fd, nsid int) error {
return ErrNotImplemented
}

View file

@ -11,8 +11,8 @@ const (
/* Bridge Flags */
const (
BRIDGE_FLAGS_MASTER = iota /* Bridge command to/from master */
BRIDGE_FLAGS_SELF /* Bridge command to/from lowerdev */
BRIDGE_FLAGS_MASTER = iota + 1 /* Bridge command to/from master */
BRIDGE_FLAGS_SELF /* Bridge command to/from lowerdev */
)
/* Bridge management nested attributes

View file

@ -76,12 +76,17 @@ const (
// __CTA_MAX
// };
const (
CTA_TUPLE_ORIG = 1
CTA_TUPLE_REPLY = 2
CTA_STATUS = 3
CTA_TIMEOUT = 7
CTA_MARK = 8
CTA_PROTOINFO = 4
CTA_TUPLE_ORIG = 1
CTA_TUPLE_REPLY = 2
CTA_STATUS = 3
CTA_PROTOINFO = 4
CTA_TIMEOUT = 7
CTA_MARK = 8
CTA_COUNTERS_ORIG = 9
CTA_COUNTERS_REPLY = 10
CTA_USE = 11
CTA_ID = 12
CTA_TIMESTAMP = 20
)
// enum ctattr_tuple {
@ -163,6 +168,29 @@ const (
CTA_PROTOINFO_TCP_FLAGS_REPLY = 5
)
// enum ctattr_counters {
// CTA_COUNTERS_UNSPEC,
// CTA_COUNTERS_PACKETS, /* 64bit counters */
// CTA_COUNTERS_BYTES, /* 64bit counters */
// CTA_COUNTERS32_PACKETS, /* old 32bit counters, unused */
// CTA_COUNTERS32_BYTES, /* old 32bit counters, unused */
// CTA_COUNTERS_PAD,
// __CTA_COUNTERS_M
// };
// #define CTA_COUNTERS_MAX (__CTA_COUNTERS_MAX - 1)
const (
CTA_COUNTERS_PACKETS = 1
CTA_COUNTERS_BYTES = 2
)
// enum CTA TIMESTAMP TLVs
// CTA_TIMESTAMP_START /* 64bit value */
// CTA_TIMESTAMP_STOP /* 64bit value */
const (
CTA_TIMESTAMP_START = 1
CTA_TIMESTAMP_STOP = 2
)
// /* General form of address family dependent message.
// */
// struct nfgenmsg {

View file

@ -0,0 +1,40 @@
package nl
// All the following constants are coming from:
// https://github.com/torvalds/linux/blob/master/include/uapi/linux/devlink.h
const (
GENL_DEVLINK_VERSION = 1
GENL_DEVLINK_NAME = "devlink"
)
const (
DEVLINK_CMD_GET = 1
DEVLINK_CMD_ESWITCH_GET = 29
DEVLINK_CMD_ESWITCH_SET = 30
)
const (
DEVLINK_ATTR_BUS_NAME = 1
DEVLINK_ATTR_DEV_NAME = 2
DEVLINK_ATTR_ESWITCH_MODE = 25
DEVLINK_ATTR_ESWITCH_INLINE_MODE = 26
DEVLINK_ATTR_ESWITCH_ENCAP_MODE = 62
)
const (
DEVLINK_ESWITCH_MODE_LEGACY = 0
DEVLINK_ESWITCH_MODE_SWITCHDEV = 1
)
const (
DEVLINK_ESWITCH_INLINE_MODE_NONE = 0
DEVLINK_ESWITCH_INLINE_MODE_LINK = 1
DEVLINK_ESWITCH_INLINE_MODE_NETWORK = 2
DEVLINK_ESWITCH_INLINE_MODE_TRANSPORT = 3
)
const (
DEVLINK_ESWITCH_ENCAP_MODE_NONE = 0
DEVLINK_ESWITCH_ENCAP_MODE_BASIC = 1
)

View file

@ -13,7 +13,9 @@ const (
IFLA_INFO_KIND
IFLA_INFO_DATA
IFLA_INFO_XSTATS
IFLA_INFO_MAX = IFLA_INFO_XSTATS
IFLA_INFO_SLAVE_KIND
IFLA_INFO_SLAVE_DATA
IFLA_INFO_MAX = IFLA_INFO_SLAVE_DATA
)
const (
@ -87,7 +89,8 @@ const (
const (
IFLA_IPVLAN_UNSPEC = iota
IFLA_IPVLAN_MODE
IFLA_IPVLAN_MAX = IFLA_IPVLAN_MODE
IFLA_IPVLAN_FLAG
IFLA_IPVLAN_MAX = IFLA_IPVLAN_FLAG
)
const (
@ -164,6 +167,8 @@ const (
IFLA_BOND_SLAVE_PERM_HWADDR
IFLA_BOND_SLAVE_QUEUE_ID
IFLA_BOND_SLAVE_AD_AGGREGATOR_ID
IFLA_BOND_SLAVE_AD_ACTOR_OPER_PORT_STATE
IFLA_BOND_SLAVE_AD_PARTNER_OPER_PORT_STATE
)
const (
@ -217,9 +222,11 @@ const (
IFLA_VF_RSS_QUERY_EN /* RSS Redirection Table and Hash Key query
* on/off switch
*/
IFLA_VF_STATS /* network device statistics */
IFLA_VF_TRUST /* Trust state of VF */
IFLA_VF_MAX = IFLA_VF_TRUST
IFLA_VF_STATS /* network device statistics */
IFLA_VF_TRUST /* Trust state of VF */
IFLA_VF_IB_NODE_GUID /* VF Infiniband node GUID */
IFLA_VF_IB_PORT_GUID /* VF Infiniband port GUID */
IFLA_VF_MAX = IFLA_VF_IB_PORT_GUID
)
const (
@ -248,6 +255,7 @@ const (
SizeofVfLinkState = 0x08
SizeofVfRssQueryEn = 0x08
SizeofVfTrust = 0x08
SizeofVfGUID = 0x10
)
// struct ifla_vf_mac {
@ -430,6 +438,30 @@ func (msg *VfTrust) Serialize() []byte {
return (*(*[SizeofVfTrust]byte)(unsafe.Pointer(msg)))[:]
}
// struct ifla_vf_guid {
// __u32 vf;
// __u32 rsvd;
// __u64 guid;
// };
type VfGUID struct {
Vf uint32
Rsvd uint32
GUID uint64
}
func (msg *VfGUID) Len() int {
return SizeofVfGUID
}
func DeserializeVfGUID(b []byte) *VfGUID {
return (*VfGUID)(unsafe.Pointer(&b[0:SizeofVfGUID][0]))
}
func (msg *VfGUID) Serialize() []byte {
return (*(*[SizeofVfGUID]byte)(unsafe.Pointer(msg)))[:]
}
const (
XDP_FLAGS_UPDATE_IF_NOEXIST = 1 << iota
XDP_FLAGS_SKB_MODE
@ -546,3 +578,33 @@ const (
GTP_ROLE_GGSN = iota
GTP_ROLE_SGSN
)
const (
IFLA_XFRM_UNSPEC = iota
IFLA_XFRM_LINK
IFLA_XFRM_IF_ID
IFLA_XFRM_MAX = iota - 1
)
const (
IFLA_TUN_UNSPEC = iota
IFLA_TUN_OWNER
IFLA_TUN_GROUP
IFLA_TUN_TYPE
IFLA_TUN_PI
IFLA_TUN_VNET_HDR
IFLA_TUN_PERSIST
IFLA_TUN_MULTI_QUEUE
IFLA_TUN_NUM_QUEUES
IFLA_TUN_NUM_DISABLED_QUEUES
IFLA_TUN_MAX = IFLA_TUN_NUM_DISABLED_QUEUES
)
const (
IFLA_IPOIB_UNSPEC = iota
IFLA_IPOIB_PKEY
IFLA_IPOIB_MODE
IFLA_IPOIB_UMCAST
IFLA_IPOIB_MAX = IFLA_IPOIB_UMCAST
)

View file

@ -21,7 +21,13 @@ const (
FAMILY_ALL = unix.AF_UNSPEC
FAMILY_V4 = unix.AF_INET
FAMILY_V6 = unix.AF_INET6
FAMILY_MPLS = AF_MPLS
FAMILY_MPLS = unix.AF_MPLS
// Arbitrary set value (greater than default 4k) to allow receiving
// from kernel more verbose messages e.g. for statistics,
// tc rules or filters, or other more memory requiring data.
RECEIVE_BUFFER_SIZE = 65536
// Kernel netlink pid
PidKernel uint32 = 0
)
// SupportedNlFamilies contains the list of netlink families this netlink package supports
@ -42,7 +48,7 @@ func GetIPFamily(ip net.IP) int {
var nativeEndian binary.ByteOrder
// Get native endianness for the system
// NativeEndian gets native endianness for the system
func NativeEndian() binary.ByteOrder {
if nativeEndian == nil {
var x uint32 = 0x01020304
@ -271,15 +277,22 @@ func NewRtAttr(attrType int, data []byte) *RtAttr {
}
}
// Create a new RtAttr obj anc add it as a child of an existing object
// NewRtAttrChild adds an RtAttr as a child to the parent and returns the new attribute
//
// Deprecated: Use AddRtAttr() on the parent object
func NewRtAttrChild(parent *RtAttr, attrType int, data []byte) *RtAttr {
return parent.AddRtAttr(attrType, data)
}
// AddRtAttr adds an RtAttr as a child and returns the new attribute
func (a *RtAttr) AddRtAttr(attrType int, data []byte) *RtAttr {
attr := NewRtAttr(attrType, data)
parent.children = append(parent.children, attr)
a.children = append(a.children, attr)
return attr
}
// AddChild adds an existing RtAttr as a child.
func (a *RtAttr) AddChild(attr *RtAttr) {
// AddChild adds an existing NetlinkRequestData as a child.
func (a *RtAttr) AddChild(attr NetlinkRequestData) {
a.children = append(a.children, attr)
}
@ -360,16 +373,12 @@ func (req *NetlinkRequest) Serialize() []byte {
}
func (req *NetlinkRequest) AddData(data NetlinkRequestData) {
if data != nil {
req.Data = append(req.Data, data)
}
req.Data = append(req.Data, data)
}
// AddRawData adds raw bytes to the end of the NetlinkRequest object during serialization
func (req *NetlinkRequest) AddRawData(data []byte) {
if data != nil {
req.RawData = append(req.RawData, data...)
}
req.RawData = append(req.RawData, data...)
}
// Execute the request against a the given sockType.
@ -413,10 +422,13 @@ func (req *NetlinkRequest) Execute(sockType int, resType uint16) ([][]byte, erro
done:
for {
msgs, err := s.Receive()
msgs, from, err := s.Receive()
if err != nil {
return nil, err
}
if from.Pid != PidKernel {
return nil, fmt.Errorf("Wrong sender portid %d, expected %d", from.Pid, PidKernel)
}
for _, m := range msgs {
if m.Header.Seq != req.Seq {
if sharedSocket {
@ -425,7 +437,7 @@ done:
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)
continue
}
if m.Header.Type == unix.NLMSG_DONE {
break done
@ -610,21 +622,31 @@ func (s *NetlinkSocket) Send(request *NetlinkRequest) error {
return nil
}
func (s *NetlinkSocket) Receive() ([]syscall.NetlinkMessage, error) {
func (s *NetlinkSocket) Receive() ([]syscall.NetlinkMessage, *unix.SockaddrNetlink, error) {
fd := int(atomic.LoadInt32(&s.fd))
if fd < 0 {
return nil, fmt.Errorf("Receive called on a closed socket")
return nil, nil, fmt.Errorf("Receive called on a closed socket")
}
rb := make([]byte, unix.Getpagesize())
nr, _, err := unix.Recvfrom(fd, rb, 0)
var fromAddr *unix.SockaddrNetlink
var rb [RECEIVE_BUFFER_SIZE]byte
nr, from, err := unix.Recvfrom(fd, rb[:], 0)
if err != nil {
return nil, err
return nil, nil, err
}
fromAddr, ok := from.(*unix.SockaddrNetlink)
if !ok {
return nil, nil, fmt.Errorf("Error converting to netlink sockaddr")
}
if nr < unix.NLMSG_HDRLEN {
return nil, fmt.Errorf("Got short response from netlink")
return nil, nil, fmt.Errorf("Got short response from netlink")
}
rb = rb[:nr]
return syscall.ParseNetlinkMessage(rb)
rb2 := make([]byte, nr)
copy(rb2, rb[:nr])
nl, err := syscall.ParseNetlinkMessage(rb2)
if err != nil {
return nil, nil, err
}
return nl, fromAddr, nil
}
// SetSendTimeout allows to set a send timeout on the socket

View file

@ -0,0 +1,35 @@
package nl
const (
RDMA_NL_GET_CLIENT_SHIFT = 10
)
const (
RDMA_NL_NLDEV = 5
)
const (
RDMA_NLDEV_CMD_GET = 1
RDMA_NLDEV_CMD_SET = 2
RDMA_NLDEV_CMD_SYS_GET = 6
RDMA_NLDEV_CMD_SYS_SET = 7
)
const (
RDMA_NLDEV_ATTR_DEV_INDEX = 1
RDMA_NLDEV_ATTR_DEV_NAME = 2
RDMA_NLDEV_ATTR_PORT_INDEX = 3
RDMA_NLDEV_ATTR_CAP_FLAGS = 4
RDMA_NLDEV_ATTR_FW_VERSION = 5
RDMA_NLDEV_ATTR_NODE_GUID = 6
RDMA_NLDEV_ATTR_SYS_IMAGE_GUID = 7
RDMA_NLDEV_ATTR_SUBNET_PREFIX = 8
RDMA_NLDEV_ATTR_LID = 9
RDMA_NLDEV_ATTR_SM_LID = 10
RDMA_NLDEV_ATTR_LMC = 11
RDMA_NLDEV_ATTR_PORT_STATE = 12
RDMA_NLDEV_ATTR_PORT_PHYS_STATE = 13
RDMA_NLDEV_ATTR_DEV_NODE_TYPE = 14
RDMA_NLDEV_SYS_ATTR_NETNS_MODE = 66
RDMA_NLDEV_NET_NS_FD = 68
)

View file

@ -79,3 +79,29 @@ func (msg *RtNexthop) Serialize() []byte {
}
return buf
}
type RtGenMsg struct {
unix.RtGenmsg
}
func NewRtGenMsg() *RtGenMsg {
return &RtGenMsg{
RtGenmsg: unix.RtGenmsg{
Family: unix.AF_UNSPEC,
},
}
}
func (msg *RtGenMsg) Len() int {
return rtaAlignOf(unix.SizeofRtGenmsg)
}
func DeserializeRtGenMsg(b []byte) *RtGenMsg {
return &RtGenMsg{RtGenmsg: unix.RtGenmsg{Family: b[0]}}
}
func (msg *RtGenMsg) Serialize() []byte {
out := make([]byte, msg.Len())
out[0] = msg.Family
return out
}

View file

@ -99,6 +99,49 @@ func DecodeSEG6Encap(buf []byte) (int, []net.IP, error) {
return mode, srh.Segments, nil
}
func DecodeSEG6Srh(buf []byte) ([]net.IP, error) {
native := NativeEndian()
srh := IPv6SrHdr{
nextHdr: buf[0],
hdrLen: buf[1],
routingType: buf[2],
segmentsLeft: buf[3],
firstSegment: buf[4],
flags: buf[5],
reserved: native.Uint16(buf[6:8]),
}
buf = buf[8:]
if len(buf)%16 != 0 {
err := fmt.Errorf("DecodeSEG6Srh: error parsing Segment List (buf len: %d)", len(buf))
return nil, err
}
for len(buf) > 0 {
srh.Segments = append(srh.Segments, net.IP(buf[:16]))
buf = buf[16:]
}
return srh.Segments, nil
}
func EncodeSEG6Srh(segments []net.IP) ([]byte, error) {
nsegs := len(segments) // nsegs: number of segments
if nsegs == 0 {
return nil, errors.New("EncodeSEG6Srh: No Segments")
}
b := make([]byte, 8, 8+len(segments)*16)
native := NativeEndian()
b[0] = 0 // srh.nextHdr (0 when calling netlink)
b[1] = uint8(16 * nsegs >> 3) // srh.hdrLen (in 8-octets unit)
b[2] = IPV6_SRCRT_TYPE_4 // srh.routingType (assigned by IANA)
b[3] = uint8(nsegs - 1) // srh.segmentsLeft
b[4] = uint8(nsegs - 1) // srh.firstSegment
b[5] = 0 // srh.flags (SR6_FLAG1_HMAC for srh_hmac)
// srh.reserved: Defined as "Tag" in draft-ietf-6man-segment-routing-header-07
native.PutUint16(b[6:], 0) // srh.reserved
for _, netIP := range segments {
b = append(b, netIP...) // srh.Segments
}
return b, nil
}
// Helper functions
func SEG6EncapModeString(mode int) string {
switch mode {

View file

@ -0,0 +1,76 @@
package nl
import ()
// seg6local parameters
const (
SEG6_LOCAL_UNSPEC = iota
SEG6_LOCAL_ACTION
SEG6_LOCAL_SRH
SEG6_LOCAL_TABLE
SEG6_LOCAL_NH4
SEG6_LOCAL_NH6
SEG6_LOCAL_IIF
SEG6_LOCAL_OIF
__SEG6_LOCAL_MAX
)
const (
SEG6_LOCAL_MAX = __SEG6_LOCAL_MAX
)
// seg6local actions
const (
SEG6_LOCAL_ACTION_END = iota + 1 // 1
SEG6_LOCAL_ACTION_END_X // 2
SEG6_LOCAL_ACTION_END_T // 3
SEG6_LOCAL_ACTION_END_DX2 // 4
SEG6_LOCAL_ACTION_END_DX6 // 5
SEG6_LOCAL_ACTION_END_DX4 // 6
SEG6_LOCAL_ACTION_END_DT6 // 7
SEG6_LOCAL_ACTION_END_DT4 // 8
SEG6_LOCAL_ACTION_END_B6 // 9
SEG6_LOCAL_ACTION_END_B6_ENCAPS // 10
SEG6_LOCAL_ACTION_END_BM // 11
SEG6_LOCAL_ACTION_END_S // 12
SEG6_LOCAL_ACTION_END_AS // 13
SEG6_LOCAL_ACTION_END_AM // 14
__SEG6_LOCAL_ACTION_MAX
)
const (
SEG6_LOCAL_ACTION_MAX = __SEG6_LOCAL_ACTION_MAX - 1
)
// Helper functions
func SEG6LocalActionString(action int) string {
switch action {
case SEG6_LOCAL_ACTION_END:
return "End"
case SEG6_LOCAL_ACTION_END_X:
return "End.X"
case SEG6_LOCAL_ACTION_END_T:
return "End.T"
case SEG6_LOCAL_ACTION_END_DX2:
return "End.DX2"
case SEG6_LOCAL_ACTION_END_DX6:
return "End.DX6"
case SEG6_LOCAL_ACTION_END_DX4:
return "End.DX4"
case SEG6_LOCAL_ACTION_END_DT6:
return "End.DT6"
case SEG6_LOCAL_ACTION_END_DT4:
return "End.DT4"
case SEG6_LOCAL_ACTION_END_B6:
return "End.B6"
case SEG6_LOCAL_ACTION_END_B6_ENCAPS:
return "End.B6.Encaps"
case SEG6_LOCAL_ACTION_END_BM:
return "End.BM"
case SEG6_LOCAL_ACTION_END_S:
return "End.S"
case SEG6_LOCAL_ACTION_END_AS:
return "End.AS"
case SEG6_LOCAL_ACTION_END_AM:
return "End.AM"
}
return "unknown"
}

View file

@ -42,16 +42,6 @@ const (
TCPDIAG_NOCOOKIE = 0xFFFFFFFF /* TCPDIAG_NOCOOKIE in net/ipv4/tcp_diag.h*/
)
const (
AF_MPLS = 28
)
const (
RTA_NEWDST = 0x13
RTA_ENCAP_TYPE = 0x15
RTA_ENCAP = 0x16
)
// RTA_ENCAP subtype
const (
MPLS_IPTUNNEL_UNSPEC = iota
@ -67,6 +57,7 @@ const (
LWTUNNEL_ENCAP_IP6
LWTUNNEL_ENCAP_SEG6
LWTUNNEL_ENCAP_BPF
LWTUNNEL_ENCAP_SEG6_LOCAL
)
// routing header types

View file

@ -1,6 +1,7 @@
package nl
import (
"encoding/binary"
"unsafe"
)
@ -64,6 +65,15 @@ const (
TCA_PRIO_MAX = TCA_PRIO_MQ
)
const (
TCA_STATS_UNSPEC = iota
TCA_STATS_BASIC
TCA_STATS_RATE_EST
TCA_STATS_QUEUE
TCA_STATS_APP
TCA_STATS_MAX = TCA_STATS_APP
)
const (
SizeofTcMsg = 0x14
SizeofTcActionMsg = 0x04
@ -79,7 +89,10 @@ const (
SizeofTcU32Key = 0x10
SizeofTcU32Sel = 0x10 // without keys
SizeofTcGen = 0x14
SizeofTcConnmark = SizeofTcGen + 0x04
SizeofTcMirred = SizeofTcGen + 0x08
SizeofTcTunnelKey = SizeofTcGen + 0x04
SizeofTcSkbEdit = SizeofTcGen
SizeofTcPolice = 2*SizeofTcRateSpec + 0x20
)
@ -412,6 +425,57 @@ func (x *TcHtbGlob) Serialize() []byte {
return (*(*[SizeofTcHtbGlob]byte)(unsafe.Pointer(x)))[:]
}
// HFSC
type Curve struct {
m1 uint32
d uint32
m2 uint32
}
type HfscCopt struct {
Rsc Curve
Fsc Curve
Usc Curve
}
func (c *Curve) Attrs() (uint32, uint32, uint32) {
return c.m1, c.d, c.m2
}
func (c *Curve) Set(m1 uint32, d uint32, m2 uint32) {
c.m1 = m1
c.d = d
c.m2 = m2
}
func DeserializeHfscCurve(b []byte) *Curve {
return &Curve{
m1: binary.LittleEndian.Uint32(b[0:4]),
d: binary.LittleEndian.Uint32(b[4:8]),
m2: binary.LittleEndian.Uint32(b[8:12]),
}
}
func SerializeHfscCurve(c *Curve) (b []byte) {
t := make([]byte, binary.MaxVarintLen32)
binary.LittleEndian.PutUint32(t, c.m1)
b = append(b, t[:4]...)
binary.LittleEndian.PutUint32(t, c.d)
b = append(b, t[:4]...)
binary.LittleEndian.PutUint32(t, c.m2)
b = append(b, t[:4]...)
return b
}
type TcHfscOpt struct {
Defcls uint16
}
func (x *TcHfscOpt) Serialize() []byte {
return (*(*[2]byte)(unsafe.Pointer(x)))[:]
}
const (
TCA_U32_UNSPEC = iota
TCA_U32_CLASSID
@ -586,11 +650,47 @@ const (
TCA_BPF_FD
TCA_BPF_NAME
TCA_BPF_FLAGS
TCA_BPF_MAX = TCA_BPF_FLAGS
TCA_BPF_FLAGS_GEN
TCA_BPF_TAG
TCA_BPF_ID
TCA_BPF_MAX = TCA_BPF_ID
)
type TcBpf TcGen
const (
TCA_ACT_CONNMARK = 14
)
const (
TCA_CONNMARK_UNSPEC = iota
TCA_CONNMARK_PARMS
TCA_CONNMARK_TM
TCA_CONNMARK_MAX = TCA_CONNMARK_TM
)
// struct tc_connmark {
// tc_gen;
// __u16 zone;
// };
type TcConnmark struct {
TcGen
Zone uint16
}
func (msg *TcConnmark) Len() int {
return SizeofTcConnmark
}
func DeserializeTcConnmark(b []byte) *TcConnmark {
return (*TcConnmark)(unsafe.Pointer(&b[0:SizeofTcConnmark][0]))
}
func (x *TcConnmark) Serialize() []byte {
return (*(*[SizeofTcConnmark]byte)(unsafe.Pointer(x)))[:]
}
const (
TCA_ACT_MIRRED = 8
)
@ -626,6 +726,63 @@ func (x *TcMirred) Serialize() []byte {
return (*(*[SizeofTcMirred]byte)(unsafe.Pointer(x)))[:]
}
const (
TCA_TUNNEL_KEY_UNSPEC = iota
TCA_TUNNEL_KEY_TM
TCA_TUNNEL_KEY_PARMS
TCA_TUNNEL_KEY_ENC_IPV4_SRC
TCA_TUNNEL_KEY_ENC_IPV4_DST
TCA_TUNNEL_KEY_ENC_IPV6_SRC
TCA_TUNNEL_KEY_ENC_IPV6_DST
TCA_TUNNEL_KEY_ENC_KEY_ID
TCA_TUNNEL_KEY_MAX = TCA_TUNNEL_KEY_ENC_KEY_ID
)
type TcTunnelKey struct {
TcGen
Action int32
}
func (x *TcTunnelKey) Len() int {
return SizeofTcTunnelKey
}
func DeserializeTunnelKey(b []byte) *TcTunnelKey {
return (*TcTunnelKey)(unsafe.Pointer(&b[0:SizeofTcTunnelKey][0]))
}
func (x *TcTunnelKey) Serialize() []byte {
return (*(*[SizeofTcTunnelKey]byte)(unsafe.Pointer(x)))[:]
}
const (
TCA_SKBEDIT_UNSPEC = iota
TCA_SKBEDIT_TM
TCA_SKBEDIT_PARMS
TCA_SKBEDIT_PRIORITY
TCA_SKBEDIT_QUEUE_MAPPING
TCA_SKBEDIT_MARK
TCA_SKBEDIT_PAD
TCA_SKBEDIT_PTYPE
TCA_SKBEDIT_MAX = TCA_SKBEDIT_MARK
)
type TcSkbEdit struct {
TcGen
}
func (x *TcSkbEdit) Len() int {
return SizeofTcSkbEdit
}
func DeserializeSkbEdit(b []byte) *TcSkbEdit {
return (*TcSkbEdit)(unsafe.Pointer(&b[0:SizeofTcSkbEdit][0]))
}
func (x *TcSkbEdit) Serialize() []byte {
return (*(*[SizeofTcSkbEdit]byte)(unsafe.Pointer(x)))[:]
}
// struct tc_police {
// __u32 index;
// int action;
@ -708,3 +865,10 @@ const (
TCA_FQ_CODEL_DROP_BATCH_SIZE
TCA_FQ_CODEL_MEMORY_LIMIT
)
const (
TCA_HFSC_UNSPEC = iota
TCA_HFSC_RSC
TCA_HFSC_FSC
TCA_HFSC_USC
)

View file

@ -50,34 +50,44 @@ const (
// Attribute types
const (
/* Netlink message attributes. */
XFRMA_UNSPEC = 0x00
XFRMA_ALG_AUTH = 0x01 /* struct xfrm_algo */
XFRMA_ALG_CRYPT = 0x02 /* struct xfrm_algo */
XFRMA_ALG_COMP = 0x03 /* struct xfrm_algo */
XFRMA_ENCAP = 0x04 /* struct xfrm_algo + struct xfrm_encap_tmpl */
XFRMA_TMPL = 0x05 /* 1 or more struct xfrm_user_tmpl */
XFRMA_SA = 0x06 /* struct xfrm_usersa_info */
XFRMA_POLICY = 0x07 /* struct xfrm_userpolicy_info */
XFRMA_SEC_CTX = 0x08 /* struct xfrm_sec_ctx */
XFRMA_LTIME_VAL = 0x09
XFRMA_REPLAY_VAL = 0x0a
XFRMA_REPLAY_THRESH = 0x0b
XFRMA_ETIMER_THRESH = 0x0c
XFRMA_SRCADDR = 0x0d /* xfrm_address_t */
XFRMA_COADDR = 0x0e /* xfrm_address_t */
XFRMA_LASTUSED = 0x0f /* unsigned long */
XFRMA_POLICY_TYPE = 0x10 /* struct xfrm_userpolicy_type */
XFRMA_MIGRATE = 0x11
XFRMA_ALG_AEAD = 0x12 /* struct xfrm_algo_aead */
XFRMA_KMADDRESS = 0x13 /* struct xfrm_user_kmaddress */
XFRMA_ALG_AUTH_TRUNC = 0x14 /* struct xfrm_algo_auth */
XFRMA_MARK = 0x15 /* struct xfrm_mark */
XFRMA_TFCPAD = 0x16 /* __u32 */
XFRMA_REPLAY_ESN_VAL = 0x17 /* struct xfrm_replay_esn */
XFRMA_SA_EXTRA_FLAGS = 0x18 /* __u32 */
XFRMA_MAX = 0x18
XFRMA_UNSPEC = iota
XFRMA_ALG_AUTH /* struct xfrm_algo */
XFRMA_ALG_CRYPT /* struct xfrm_algo */
XFRMA_ALG_COMP /* struct xfrm_algo */
XFRMA_ENCAP /* struct xfrm_algo + struct xfrm_encap_tmpl */
XFRMA_TMPL /* 1 or more struct xfrm_user_tmpl */
XFRMA_SA /* struct xfrm_usersa_info */
XFRMA_POLICY /* struct xfrm_userpolicy_info */
XFRMA_SEC_CTX /* struct xfrm_sec_ctx */
XFRMA_LTIME_VAL
XFRMA_REPLAY_VAL
XFRMA_REPLAY_THRESH
XFRMA_ETIMER_THRESH
XFRMA_SRCADDR /* xfrm_address_t */
XFRMA_COADDR /* xfrm_address_t */
XFRMA_LASTUSED /* unsigned long */
XFRMA_POLICY_TYPE /* struct xfrm_userpolicy_type */
XFRMA_MIGRATE
XFRMA_ALG_AEAD /* struct xfrm_algo_aead */
XFRMA_KMADDRESS /* struct xfrm_user_kmaddress */
XFRMA_ALG_AUTH_TRUNC /* struct xfrm_algo_auth */
XFRMA_MARK /* struct xfrm_mark */
XFRMA_TFCPAD /* __u32 */
XFRMA_REPLAY_ESN_VAL /* struct xfrm_replay_esn */
XFRMA_SA_EXTRA_FLAGS /* __u32 */
XFRMA_PROTO /* __u8 */
XFRMA_ADDRESS_FILTER /* struct xfrm_address_filter */
XFRMA_PAD
XFRMA_OFFLOAD_DEV /* struct xfrm_state_offload */
XFRMA_SET_MARK /* __u32 */
XFRMA_SET_MARK_MASK /* __u32 */
XFRMA_IF_ID /* __u32 */
XFRMA_MAX = iota - 1
)
const XFRMA_OUTPUT_MARK = XFRMA_SET_MARK
const (
SizeofXfrmAddress = 0x10
SizeofXfrmSelector = 0x38

View file

@ -18,6 +18,10 @@ type Protinfo struct {
// String returns a list of enabled flags
func (prot *Protinfo) String() string {
if prot == nil {
return "<nil>"
}
var boolStrings []string
if prot.Hairpin {
boolStrings = append(boolStrings, "Hairpin")

View file

@ -41,7 +41,7 @@ func (h *Handle) LinkGetProtinfo(link Link) (Protinfo, error) {
if err != nil {
return pi, err
}
pi = *parseProtinfo(infos)
pi = parseProtinfo(infos)
return pi, nil
}
@ -49,8 +49,7 @@ func (h *Handle) LinkGetProtinfo(link Link) (Protinfo, error) {
return pi, fmt.Errorf("Device with index %d not found", base.Index)
}
func parseProtinfo(infos []syscall.NetlinkRouteAttr) *Protinfo {
var pi Protinfo
func parseProtinfo(infos []syscall.NetlinkRouteAttr) (pi Protinfo) {
for _, info := range infos {
switch info.Attr.Type {
case nl.IFLA_BRPORT_MODE:
@ -71,5 +70,5 @@ func parseProtinfo(infos []syscall.NetlinkRouteAttr) *Protinfo {
pi.ProxyArpWiFi = byteToBool(info.Value[0])
}
}
return &pi
return
}

View file

@ -176,6 +176,13 @@ type Netem struct {
CorruptCorr uint32
}
func (netem *Netem) String() string {
return fmt.Sprintf(
"{Latency: %v, Limit: %v, Loss: %v, Gap: %v, Duplicate: %v, Jitter: %v}",
netem.Latency, netem.Limit, netem.Loss, netem.Gap, netem.Duplicate, netem.Jitter,
)
}
func (qdisc *Netem) Attrs() *QdiscAttrs {
return &qdisc.QdiscAttrs
}
@ -231,6 +238,33 @@ func (qdisc *GenericQdisc) Type() string {
return qdisc.QdiscType
}
type Hfsc struct {
QdiscAttrs
Defcls uint16
}
func NewHfsc(attrs QdiscAttrs) *Hfsc {
return &Hfsc{
QdiscAttrs: attrs,
Defcls: 1,
}
}
func (hfsc *Hfsc) Attrs() *QdiscAttrs {
return &hfsc.QdiscAttrs
}
func (hfsc *Hfsc) Type() string {
return "hfsc"
}
func (hfsc *Hfsc) String() string {
return fmt.Sprintf(
"{%v -- default: %d}",
hfsc.Attrs(), hfsc.Defcls,
)
}
// Fq is a classless packet scheduler meant to be mostly used for locally generated traffic.
type Fq struct {
QdiscAttrs
@ -249,6 +283,13 @@ type Fq struct {
LowRateThreshold uint32
}
func (fq *Fq) String() string {
return fmt.Sprintf(
"{PacketLimit: %v, FlowPacketLimit: %v, Quantum: %v, InitialQuantum: %v, Pacing: %v, FlowDefaultRate: %v, FlowMaxRate: %v, Buckets: %v, FlowRefillDelay: %v, LowRateThreshold: %v}",
fq.PacketLimit, fq.FlowPacketLimit, fq.Quantum, fq.InitialQuantum, fq.Pacing, fq.FlowDefaultRate, fq.FlowMaxRate, fq.Buckets, fq.FlowRefillDelay, fq.LowRateThreshold,
)
}
func NewFq(attrs QdiscAttrs) *Fq {
return &Fq{
QdiscAttrs: attrs,
@ -276,6 +317,13 @@ type FqCodel struct {
// There are some more attributes here, but support for them seems not ubiquitous
}
func (fqcodel *FqCodel) String() string {
return fmt.Sprintf(
"{%v -- Target: %v, Limit: %v, Interval: %v, ECM: %v, Flows: %v, Quantum: %v}",
fqcodel.Attrs(), fqcodel.Target, fqcodel.Limit, fqcodel.Interval, fqcodel.ECN, fqcodel.Flows, fqcodel.Quantum,
)
}
func NewFqCodel(attrs QdiscAttrs) *FqCodel {
return &FqCodel{
QdiscAttrs: attrs,

View file

@ -175,15 +175,15 @@ func qdiscPayload(req *nl.NetlinkRequest, qdisc Qdisc) error {
opt.Peakrate.Rate = uint32(qdisc.Peakrate)
opt.Limit = qdisc.Limit
opt.Buffer = qdisc.Buffer
nl.NewRtAttrChild(options, nl.TCA_TBF_PARMS, opt.Serialize())
options.AddRtAttr(nl.TCA_TBF_PARMS, opt.Serialize())
if qdisc.Rate >= uint64(1<<32) {
nl.NewRtAttrChild(options, nl.TCA_TBF_RATE64, nl.Uint64Attr(qdisc.Rate))
options.AddRtAttr(nl.TCA_TBF_RATE64, nl.Uint64Attr(qdisc.Rate))
}
if qdisc.Peakrate >= uint64(1<<32) {
nl.NewRtAttrChild(options, nl.TCA_TBF_PRATE64, nl.Uint64Attr(qdisc.Peakrate))
options.AddRtAttr(nl.TCA_TBF_PRATE64, nl.Uint64Attr(qdisc.Peakrate))
}
if qdisc.Peakrate > 0 {
nl.NewRtAttrChild(options, nl.TCA_TBF_PBURST, nl.Uint32Attr(qdisc.Minburst))
options.AddRtAttr(nl.TCA_TBF_PBURST, nl.Uint32Attr(qdisc.Minburst))
}
case *Htb:
opt := nl.TcHtbGlob{}
@ -193,8 +193,12 @@ func qdiscPayload(req *nl.NetlinkRequest, qdisc Qdisc) error {
// TODO: Handle Debug properly. For now default to 0
opt.Debug = qdisc.Debug
opt.DirectPkts = qdisc.DirectPkts
nl.NewRtAttrChild(options, nl.TCA_HTB_INIT, opt.Serialize())
// nl.NewRtAttrChild(options, nl.TCA_HTB_DIRECT_QLEN, opt.Serialize())
options.AddRtAttr(nl.TCA_HTB_INIT, opt.Serialize())
// options.AddRtAttr(nl.TCA_HTB_DIRECT_QLEN, opt.Serialize())
case *Hfsc:
opt := nl.TcHfscOpt{}
opt.Defcls = qdisc.Defcls
options = nl.NewRtAttr(nl.TCA_OPTIONS, opt.Serialize())
case *Netem:
opt := nl.TcNetemQopt{}
opt.Latency = qdisc.Latency
@ -211,21 +215,21 @@ func qdiscPayload(req *nl.NetlinkRequest, qdisc Qdisc) error {
corr.DupCorr = qdisc.DuplicateCorr
if corr.DelayCorr > 0 || corr.LossCorr > 0 || corr.DupCorr > 0 {
nl.NewRtAttrChild(options, nl.TCA_NETEM_CORR, corr.Serialize())
options.AddRtAttr(nl.TCA_NETEM_CORR, corr.Serialize())
}
// Corruption
corruption := nl.TcNetemCorrupt{}
corruption.Probability = qdisc.CorruptProb
corruption.Correlation = qdisc.CorruptCorr
if corruption.Probability > 0 {
nl.NewRtAttrChild(options, nl.TCA_NETEM_CORRUPT, corruption.Serialize())
options.AddRtAttr(nl.TCA_NETEM_CORRUPT, corruption.Serialize())
}
// Reorder
reorder := nl.TcNetemReorder{}
reorder.Probability = qdisc.ReorderProb
reorder.Correlation = qdisc.ReorderCorr
if reorder.Probability > 0 {
nl.NewRtAttrChild(options, nl.TCA_NETEM_REORDER, reorder.Serialize())
options.AddRtAttr(nl.TCA_NETEM_REORDER, reorder.Serialize())
}
case *Ingress:
// ingress filters must use the proper handle
@ -233,50 +237,54 @@ func qdiscPayload(req *nl.NetlinkRequest, qdisc Qdisc) error {
return fmt.Errorf("Ingress filters must set Parent to HANDLE_INGRESS")
}
case *FqCodel:
nl.NewRtAttrChild(options, nl.TCA_FQ_CODEL_ECN, nl.Uint32Attr((uint32(qdisc.ECN))))
options.AddRtAttr(nl.TCA_FQ_CODEL_ECN, nl.Uint32Attr((uint32(qdisc.ECN))))
if qdisc.Limit > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_CODEL_LIMIT, nl.Uint32Attr((uint32(qdisc.Limit))))
options.AddRtAttr(nl.TCA_FQ_CODEL_LIMIT, nl.Uint32Attr((uint32(qdisc.Limit))))
}
if qdisc.Interval > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_CODEL_INTERVAL, nl.Uint32Attr((uint32(qdisc.Interval))))
options.AddRtAttr(nl.TCA_FQ_CODEL_INTERVAL, nl.Uint32Attr((uint32(qdisc.Interval))))
}
if qdisc.Flows > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_CODEL_FLOWS, nl.Uint32Attr((uint32(qdisc.Flows))))
options.AddRtAttr(nl.TCA_FQ_CODEL_FLOWS, nl.Uint32Attr((uint32(qdisc.Flows))))
}
if qdisc.Quantum > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_CODEL_QUANTUM, nl.Uint32Attr((uint32(qdisc.Quantum))))
options.AddRtAttr(nl.TCA_FQ_CODEL_QUANTUM, nl.Uint32Attr((uint32(qdisc.Quantum))))
}
case *Fq:
nl.NewRtAttrChild(options, nl.TCA_FQ_RATE_ENABLE, nl.Uint32Attr((uint32(qdisc.Pacing))))
options.AddRtAttr(nl.TCA_FQ_RATE_ENABLE, nl.Uint32Attr((uint32(qdisc.Pacing))))
if qdisc.Buckets > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_BUCKETS_LOG, nl.Uint32Attr((uint32(qdisc.Buckets))))
options.AddRtAttr(nl.TCA_FQ_BUCKETS_LOG, nl.Uint32Attr((uint32(qdisc.Buckets))))
}
if qdisc.LowRateThreshold > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_LOW_RATE_THRESHOLD, nl.Uint32Attr((uint32(qdisc.LowRateThreshold))))
options.AddRtAttr(nl.TCA_FQ_LOW_RATE_THRESHOLD, nl.Uint32Attr((uint32(qdisc.LowRateThreshold))))
}
if qdisc.Quantum > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_QUANTUM, nl.Uint32Attr((uint32(qdisc.Quantum))))
options.AddRtAttr(nl.TCA_FQ_QUANTUM, nl.Uint32Attr((uint32(qdisc.Quantum))))
}
if qdisc.InitialQuantum > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_INITIAL_QUANTUM, nl.Uint32Attr((uint32(qdisc.InitialQuantum))))
options.AddRtAttr(nl.TCA_FQ_INITIAL_QUANTUM, nl.Uint32Attr((uint32(qdisc.InitialQuantum))))
}
if qdisc.FlowRefillDelay > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_FLOW_REFILL_DELAY, nl.Uint32Attr((uint32(qdisc.FlowRefillDelay))))
options.AddRtAttr(nl.TCA_FQ_FLOW_REFILL_DELAY, nl.Uint32Attr((uint32(qdisc.FlowRefillDelay))))
}
if qdisc.FlowPacketLimit > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_FLOW_PLIMIT, nl.Uint32Attr((uint32(qdisc.FlowPacketLimit))))
options.AddRtAttr(nl.TCA_FQ_FLOW_PLIMIT, nl.Uint32Attr((uint32(qdisc.FlowPacketLimit))))
}
if qdisc.FlowMaxRate > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_FLOW_MAX_RATE, nl.Uint32Attr((uint32(qdisc.FlowMaxRate))))
options.AddRtAttr(nl.TCA_FQ_FLOW_MAX_RATE, nl.Uint32Attr((uint32(qdisc.FlowMaxRate))))
}
if qdisc.FlowDefaultRate > 0 {
nl.NewRtAttrChild(options, nl.TCA_FQ_FLOW_DEFAULT_RATE, nl.Uint32Attr((uint32(qdisc.FlowDefaultRate))))
options.AddRtAttr(nl.TCA_FQ_FLOW_DEFAULT_RATE, nl.Uint32Attr((uint32(qdisc.FlowDefaultRate))))
}
default:
options = nil
}
req.AddData(options)
if options != nil {
req.AddData(options)
}
return nil
}
@ -348,6 +356,8 @@ func (h *Handle) QdiscList(link Link) ([]Qdisc, error) {
qdisc = &Htb{}
case "fq":
qdisc = &Fq{}
case "hfsc":
qdisc = &Hfsc{}
case "fq_codel":
qdisc = &FqCodel{}
case "netem":
@ -375,6 +385,10 @@ func (h *Handle) QdiscList(link Link) ([]Qdisc, error) {
if err := parseTbfData(qdisc, data); err != nil {
return nil, err
}
case "hfsc":
if err := parseHfscData(qdisc, attr.Value); err != nil {
return nil, err
}
case "htb":
data, err := nl.ParseRouteAttr(attr.Value)
if err != nil {
@ -474,6 +488,13 @@ func parseFqCodelData(qdisc Qdisc, data []syscall.NetlinkRouteAttr) error {
return nil
}
func parseHfscData(qdisc Qdisc, data []byte) error {
Hfsc := qdisc.(*Hfsc)
native = nl.NativeEndian()
Hfsc.Defcls = native.Uint16(data)
return nil
}
func parseFqData(qdisc Qdisc, data []syscall.NetlinkRouteAttr) error {
native = nl.NativeEndian()
fq := qdisc.(*Fq)

View file

@ -0,0 +1,264 @@
package netlink
import (
"bytes"
"encoding/binary"
"fmt"
"net"
"github.com/vishvananda/netlink/nl"
"golang.org/x/sys/unix"
)
// LinkAttrs represents data shared by most link types
type RdmaLinkAttrs struct {
Index uint32
Name string
FirmwareVersion string
NodeGuid string
SysImageGuid string
}
// Link represents a rdma device from netlink.
type RdmaLink struct {
Attrs RdmaLinkAttrs
}
func getProtoField(clientType int, op int) int {
return ((clientType << nl.RDMA_NL_GET_CLIENT_SHIFT) | op)
}
func uint64ToGuidString(guid uint64) string {
//Convert to byte array
sysGuidBytes := new(bytes.Buffer)
binary.Write(sysGuidBytes, binary.LittleEndian, guid)
//Convert to HardwareAddr
sysGuidNet := net.HardwareAddr(sysGuidBytes.Bytes())
//Get the String
return sysGuidNet.String()
}
func executeOneGetRdmaLink(data []byte) (*RdmaLink, error) {
link := RdmaLink{}
reader := bytes.NewReader(data)
for reader.Len() >= 4 {
_, attrType, len, value := parseNfAttrTLV(reader)
switch attrType {
case nl.RDMA_NLDEV_ATTR_DEV_INDEX:
var Index uint32
r := bytes.NewReader(value)
binary.Read(r, nl.NativeEndian(), &Index)
link.Attrs.Index = Index
case nl.RDMA_NLDEV_ATTR_DEV_NAME:
link.Attrs.Name = string(value[0 : len-1])
case nl.RDMA_NLDEV_ATTR_FW_VERSION:
link.Attrs.FirmwareVersion = string(value[0 : len-1])
case nl.RDMA_NLDEV_ATTR_NODE_GUID:
var guid uint64
r := bytes.NewReader(value)
binary.Read(r, nl.NativeEndian(), &guid)
link.Attrs.NodeGuid = uint64ToGuidString(guid)
case nl.RDMA_NLDEV_ATTR_SYS_IMAGE_GUID:
var sysGuid uint64
r := bytes.NewReader(value)
binary.Read(r, nl.NativeEndian(), &sysGuid)
link.Attrs.SysImageGuid = uint64ToGuidString(sysGuid)
}
if (len % 4) != 0 {
// Skip pad bytes
reader.Seek(int64(4-(len%4)), seekCurrent)
}
}
return &link, nil
}
func execRdmaGetLink(req *nl.NetlinkRequest, name string) (*RdmaLink, error) {
msgs, err := req.Execute(unix.NETLINK_RDMA, 0)
if err != nil {
return nil, err
}
for _, m := range msgs {
link, err := executeOneGetRdmaLink(m)
if err != nil {
return nil, err
}
if link.Attrs.Name == name {
return link, nil
}
}
return nil, fmt.Errorf("Rdma device %v not found", name)
}
func execRdmaSetLink(req *nl.NetlinkRequest) error {
_, err := req.Execute(unix.NETLINK_RDMA, 0)
return err
}
// RdmaLinkByName finds a link by name and returns a pointer to the object if
// found and nil error, otherwise returns error code.
func RdmaLinkByName(name string) (*RdmaLink, error) {
return pkgHandle.RdmaLinkByName(name)
}
// RdmaLinkByName finds a link by name and returns a pointer to the object if
// found and nil error, otherwise returns error code.
func (h *Handle) RdmaLinkByName(name string) (*RdmaLink, error) {
proto := getProtoField(nl.RDMA_NL_NLDEV, nl.RDMA_NLDEV_CMD_GET)
req := h.newNetlinkRequest(proto, unix.NLM_F_ACK|unix.NLM_F_DUMP)
return execRdmaGetLink(req, name)
}
// RdmaLinkSetName sets the name of the rdma link device. Return nil on success
// or error otherwise.
// Equivalent to: `rdma dev set $old_devname name $name`
func RdmaLinkSetName(link *RdmaLink, name string) error {
return pkgHandle.RdmaLinkSetName(link, name)
}
// RdmaLinkSetName sets the name of the rdma link device. Return nil on success
// or error otherwise.
// Equivalent to: `rdma dev set $old_devname name $name`
func (h *Handle) RdmaLinkSetName(link *RdmaLink, name string) error {
proto := getProtoField(nl.RDMA_NL_NLDEV, nl.RDMA_NLDEV_CMD_SET)
req := h.newNetlinkRequest(proto, unix.NLM_F_ACK)
b := make([]byte, 4)
native.PutUint32(b, uint32(link.Attrs.Index))
data := nl.NewRtAttr(nl.RDMA_NLDEV_ATTR_DEV_INDEX, b)
req.AddData(data)
b = make([]byte, len(name)+1)
copy(b, name)
data = nl.NewRtAttr(nl.RDMA_NLDEV_ATTR_DEV_NAME, b)
req.AddData(data)
return execRdmaSetLink(req)
}
func netnsModeToString(mode uint8) string {
switch mode {
case 0:
return "exclusive"
case 1:
return "shared"
default:
return "unknown"
}
}
func executeOneGetRdmaNetnsMode(data []byte) (string, error) {
reader := bytes.NewReader(data)
for reader.Len() >= 4 {
_, attrType, len, value := parseNfAttrTLV(reader)
switch attrType {
case nl.RDMA_NLDEV_SYS_ATTR_NETNS_MODE:
var mode uint8
r := bytes.NewReader(value)
binary.Read(r, nl.NativeEndian(), &mode)
return netnsModeToString(mode), nil
}
if (len % 4) != 0 {
// Skip pad bytes
reader.Seek(int64(4-(len%4)), seekCurrent)
}
}
return "", fmt.Errorf("Invalid netns mode")
}
// RdmaSystemGetNetnsMode gets the net namespace mode for RDMA subsystem
// Returns mode string and error status as nil on success or returns error
// otherwise.
// Equivalent to: `rdma system show netns'
func RdmaSystemGetNetnsMode() (string, error) {
return pkgHandle.RdmaSystemGetNetnsMode()
}
// RdmaSystemGetNetnsMode gets the net namespace mode for RDMA subsystem
// Returns mode string and error status as nil on success or returns error
// otherwise.
// Equivalent to: `rdma system show netns'
func (h *Handle) RdmaSystemGetNetnsMode() (string, error) {
proto := getProtoField(nl.RDMA_NL_NLDEV, nl.RDMA_NLDEV_CMD_SYS_GET)
req := h.newNetlinkRequest(proto, unix.NLM_F_ACK)
msgs, err := req.Execute(unix.NETLINK_RDMA, 0)
if err != nil {
return "", err
}
if len(msgs) == 0 {
return "", fmt.Errorf("No valid response from kernel")
}
return executeOneGetRdmaNetnsMode(msgs[0])
}
func netnsModeStringToUint8(mode string) (uint8, error) {
switch mode {
case "exclusive":
return 0, nil
case "shared":
return 1, nil
default:
return 0, fmt.Errorf("Invalid mode; %q", mode)
}
}
// RdmaSystemSetNetnsMode sets the net namespace mode for RDMA subsystem
// Returns nil on success or appropriate error code.
// Equivalent to: `rdma system set netns { shared | exclusive }'
func RdmaSystemSetNetnsMode(NewMode string) error {
return pkgHandle.RdmaSystemSetNetnsMode(NewMode)
}
// RdmaSystemSetNetnsMode sets the net namespace mode for RDMA subsystem
// Returns nil on success or appropriate error code.
// Equivalent to: `rdma system set netns { shared | exclusive }'
func (h *Handle) RdmaSystemSetNetnsMode(NewMode string) error {
value, err := netnsModeStringToUint8(NewMode)
if err != nil {
return err
}
proto := getProtoField(nl.RDMA_NL_NLDEV, nl.RDMA_NLDEV_CMD_SYS_SET)
req := h.newNetlinkRequest(proto, unix.NLM_F_ACK)
data := nl.NewRtAttr(nl.RDMA_NLDEV_SYS_ATTR_NETNS_MODE, []byte{value})
req.AddData(data)
_, err = req.Execute(unix.NETLINK_RDMA, 0)
return err
}
// RdmaLinkSetNsFd puts the RDMA device into a new network namespace. The
// fd must be an open file descriptor to a network namespace.
// Similar to: `rdma dev set $dev netns $ns`
func RdmaLinkSetNsFd(link *RdmaLink, fd uint32) error {
return pkgHandle.RdmaLinkSetNsFd(link, fd)
}
// RdmaLinkSetNsFd puts the RDMA device into a new network namespace. The
// fd must be an open file descriptor to a network namespace.
// Similar to: `rdma dev set $dev netns $ns`
func (h *Handle) RdmaLinkSetNsFd(link *RdmaLink, fd uint32) error {
proto := getProtoField(nl.RDMA_NL_NLDEV, nl.RDMA_NLDEV_CMD_SET)
req := h.newNetlinkRequest(proto, unix.NLM_F_ACK)
data := nl.NewRtAttr(nl.RDMA_NLDEV_ATTR_DEV_INDEX,
nl.Uint32Attr(link.Attrs.Index))
req.AddData(data)
data = nl.NewRtAttr(nl.RDMA_NLDEV_NET_NS_FD, nl.Uint32Attr(fd))
req.AddData(data)
return execRdmaSetLink(req)
}

View file

@ -47,6 +47,7 @@ type Route struct {
Encap Encap
MTU int
AdvMSS int
Hoplimit int
}
func (r Route) String() string {
@ -89,6 +90,7 @@ func (r Route) Equal(x Route) bool {
r.Table == x.Table &&
r.Type == x.Type &&
r.Tos == x.Tos &&
r.Hoplimit == x.Hoplimit &&
r.Flags == x.Flags &&
(r.MPLSDst == x.MPLSDst || (r.MPLSDst != nil && x.MPLSDst != nil && *r.MPLSDst == *x.MPLSDst)) &&
(r.NewDst == x.NewDst || (r.NewDst != nil && r.NewDst.Equal(x.NewDst))) &&

View file

@ -32,6 +32,7 @@ const (
RT_FILTER_SRC
RT_FILTER_GW
RT_FILTER_TABLE
RT_FILTER_HOPLIMIT
)
const (
@ -207,6 +208,7 @@ func (e *SEG6Encap) Decode(buf []byte) error {
}
buf = buf[:l] // make sure buf size upper limit is Length
typ := native.Uint16(buf[2:])
// LWTUNNEL_ENCAP_SEG6 has only one attr type SEG6_IPTUNNEL_SRH
if typ != nl.SEG6_IPTUNNEL_SRH {
return fmt.Errorf("unknown SEG6 Type: %d", typ)
}
@ -259,6 +261,188 @@ func (e *SEG6Encap) Equal(x Encap) bool {
return true
}
// SEG6LocalEncap definitions
type SEG6LocalEncap struct {
Flags [nl.SEG6_LOCAL_MAX]bool
Action int
Segments []net.IP // from SRH in seg6_local_lwt
Table int // table id for End.T and End.DT6
InAddr net.IP
In6Addr net.IP
Iif int
Oif int
}
func (e *SEG6LocalEncap) Type() int {
return nl.LWTUNNEL_ENCAP_SEG6_LOCAL
}
func (e *SEG6LocalEncap) Decode(buf []byte) error {
attrs, err := nl.ParseRouteAttr(buf)
if err != nil {
return err
}
native := nl.NativeEndian()
for _, attr := range attrs {
switch attr.Attr.Type {
case nl.SEG6_LOCAL_ACTION:
e.Action = int(native.Uint32(attr.Value[0:4]))
e.Flags[nl.SEG6_LOCAL_ACTION] = true
case nl.SEG6_LOCAL_SRH:
e.Segments, err = nl.DecodeSEG6Srh(attr.Value[:])
e.Flags[nl.SEG6_LOCAL_SRH] = true
case nl.SEG6_LOCAL_TABLE:
e.Table = int(native.Uint32(attr.Value[0:4]))
e.Flags[nl.SEG6_LOCAL_TABLE] = true
case nl.SEG6_LOCAL_NH4:
e.InAddr = net.IP(attr.Value[0:4])
e.Flags[nl.SEG6_LOCAL_NH4] = true
case nl.SEG6_LOCAL_NH6:
e.In6Addr = net.IP(attr.Value[0:16])
e.Flags[nl.SEG6_LOCAL_NH6] = true
case nl.SEG6_LOCAL_IIF:
e.Iif = int(native.Uint32(attr.Value[0:4]))
e.Flags[nl.SEG6_LOCAL_IIF] = true
case nl.SEG6_LOCAL_OIF:
e.Oif = int(native.Uint32(attr.Value[0:4]))
e.Flags[nl.SEG6_LOCAL_OIF] = true
}
}
return err
}
func (e *SEG6LocalEncap) Encode() ([]byte, error) {
var err error
native := nl.NativeEndian()
res := make([]byte, 8)
native.PutUint16(res, 8) // length
native.PutUint16(res[2:], nl.SEG6_LOCAL_ACTION)
native.PutUint32(res[4:], uint32(e.Action))
if e.Flags[nl.SEG6_LOCAL_SRH] {
srh, err := nl.EncodeSEG6Srh(e.Segments)
if err != nil {
return nil, err
}
attr := make([]byte, 4)
native.PutUint16(attr, uint16(len(srh)+4))
native.PutUint16(attr[2:], nl.SEG6_LOCAL_SRH)
attr = append(attr, srh...)
res = append(res, attr...)
}
if e.Flags[nl.SEG6_LOCAL_TABLE] {
attr := make([]byte, 8)
native.PutUint16(attr, 8)
native.PutUint16(attr[2:], nl.SEG6_LOCAL_TABLE)
native.PutUint32(attr[4:], uint32(e.Table))
res = append(res, attr...)
}
if e.Flags[nl.SEG6_LOCAL_NH4] {
attr := make([]byte, 4)
native.PutUint16(attr, 8)
native.PutUint16(attr[2:], nl.SEG6_LOCAL_NH4)
ipv4 := e.InAddr.To4()
if ipv4 == nil {
err = fmt.Errorf("SEG6_LOCAL_NH4 has invalid IPv4 address")
return nil, err
}
attr = append(attr, ipv4...)
res = append(res, attr...)
}
if e.Flags[nl.SEG6_LOCAL_NH6] {
attr := make([]byte, 4)
native.PutUint16(attr, 20)
native.PutUint16(attr[2:], nl.SEG6_LOCAL_NH6)
attr = append(attr, e.In6Addr...)
res = append(res, attr...)
}
if e.Flags[nl.SEG6_LOCAL_IIF] {
attr := make([]byte, 8)
native.PutUint16(attr, 8)
native.PutUint16(attr[2:], nl.SEG6_LOCAL_IIF)
native.PutUint32(attr[4:], uint32(e.Iif))
res = append(res, attr...)
}
if e.Flags[nl.SEG6_LOCAL_OIF] {
attr := make([]byte, 8)
native.PutUint16(attr, 8)
native.PutUint16(attr[2:], nl.SEG6_LOCAL_OIF)
native.PutUint32(attr[4:], uint32(e.Oif))
res = append(res, attr...)
}
return res, err
}
func (e *SEG6LocalEncap) String() string {
strs := make([]string, 0, nl.SEG6_LOCAL_MAX)
strs = append(strs, fmt.Sprintf("action %s", nl.SEG6LocalActionString(e.Action)))
if e.Flags[nl.SEG6_LOCAL_TABLE] {
strs = append(strs, fmt.Sprintf("table %d", e.Table))
}
if e.Flags[nl.SEG6_LOCAL_NH4] {
strs = append(strs, fmt.Sprintf("nh4 %s", e.InAddr))
}
if e.Flags[nl.SEG6_LOCAL_NH6] {
strs = append(strs, fmt.Sprintf("nh6 %s", e.In6Addr))
}
if e.Flags[nl.SEG6_LOCAL_IIF] {
link, err := LinkByIndex(e.Iif)
if err != nil {
strs = append(strs, fmt.Sprintf("iif %d", e.Iif))
} else {
strs = append(strs, fmt.Sprintf("iif %s", link.Attrs().Name))
}
}
if e.Flags[nl.SEG6_LOCAL_OIF] {
link, err := LinkByIndex(e.Oif)
if err != nil {
strs = append(strs, fmt.Sprintf("oif %d", e.Oif))
} else {
strs = append(strs, fmt.Sprintf("oif %s", link.Attrs().Name))
}
}
if e.Flags[nl.SEG6_LOCAL_SRH] {
segs := make([]string, 0, len(e.Segments))
//append segment backwards (from n to 0) since seg#0 is the last segment.
for i := len(e.Segments); i > 0; i-- {
segs = append(segs, fmt.Sprintf("%s", e.Segments[i-1]))
}
strs = append(strs, fmt.Sprintf("segs %d [ %s ]", len(e.Segments), strings.Join(segs, " ")))
}
return strings.Join(strs, " ")
}
func (e *SEG6LocalEncap) Equal(x Encap) bool {
o, ok := x.(*SEG6LocalEncap)
if !ok {
return false
}
if e == o {
return true
}
if e == nil || o == nil {
return false
}
// compare all arrays first
for i := range e.Flags {
if e.Flags[i] != o.Flags[i] {
return false
}
}
if len(e.Segments) != len(o.Segments) {
return false
}
for i := range e.Segments {
if !e.Segments[i].Equal(o.Segments[i]) {
return false
}
}
// compare values
if !e.InAddr.Equal(o.InAddr) || !e.In6Addr.Equal(o.In6Addr) {
return false
}
if e.Action != o.Action || e.Table != o.Table || e.Iif != o.Iif || e.Oif != o.Oif {
return false
}
return true
}
// RouteAdd will add a route to the system.
// Equivalent to: `ip route add $route`
func RouteAdd(route *Route) error {
@ -335,18 +519,18 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg
if err != nil {
return err
}
rtAttrs = append(rtAttrs, nl.NewRtAttr(nl.RTA_NEWDST, buf))
rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_NEWDST, buf))
}
if route.Encap != nil {
buf := make([]byte, 2)
native.PutUint16(buf, uint16(route.Encap.Type()))
rtAttrs = append(rtAttrs, nl.NewRtAttr(nl.RTA_ENCAP_TYPE, buf))
rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_ENCAP_TYPE, buf))
buf, err := route.Encap.Encode()
if err != nil {
return err
}
rtAttrs = append(rtAttrs, nl.NewRtAttr(nl.RTA_ENCAP, buf))
rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_ENCAP, buf))
}
if route.Src != nil {
@ -410,17 +594,17 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg
if err != nil {
return err
}
children = append(children, nl.NewRtAttr(nl.RTA_NEWDST, buf))
children = append(children, nl.NewRtAttr(unix.RTA_NEWDST, buf))
}
if nh.Encap != nil {
buf := make([]byte, 2)
native.PutUint16(buf, uint16(nh.Encap.Type()))
rtAttrs = append(rtAttrs, nl.NewRtAttr(nl.RTA_ENCAP_TYPE, buf))
children = append(children, nl.NewRtAttr(unix.RTA_ENCAP_TYPE, buf))
buf, err := nh.Encap.Encode()
if err != nil {
return err
}
children = append(children, nl.NewRtAttr(nl.RTA_ENCAP, buf))
children = append(children, nl.NewRtAttr(unix.RTA_ENCAP, buf))
}
rtnh.Children = children
buf = append(buf, rtnh.Serialize()...)
@ -464,6 +648,10 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg
b := nl.Uint32Attr(uint32(route.AdvMSS))
metrics = append(metrics, nl.NewRtAttr(unix.RTAX_ADVMSS, b))
}
if route.Hoplimit > 0 {
b := nl.Uint32Attr(uint32(route.Hoplimit))
metrics = append(metrics, nl.NewRtAttr(unix.RTAX_HOPLIMIT, b))
}
if metrics != nil {
attr := nl.NewRtAttr(unix.RTA_METRICS, nil)
@ -574,6 +762,8 @@ func (h *Handle) RouteListFiltered(family int, filter *Route, filterMask uint64)
continue
}
}
case filterMask&RT_FILTER_HOPLIMIT != 0 && route.Hoplimit != filter.Hoplimit:
continue
}
}
res = append(res, route)
@ -649,7 +839,7 @@ func deserializeRoute(m []byte) (Route, error) {
switch attr.Attr.Type {
case unix.RTA_GATEWAY:
info.Gw = net.IP(attr.Value)
case nl.RTA_NEWDST:
case unix.RTA_NEWDST:
var d Destination
switch msg.Family {
case nl.FAMILY_MPLS:
@ -659,9 +849,9 @@ func deserializeRoute(m []byte) (Route, error) {
return nil, nil, err
}
info.NewDst = d
case nl.RTA_ENCAP_TYPE:
case unix.RTA_ENCAP_TYPE:
encapType = attr
case nl.RTA_ENCAP:
case unix.RTA_ENCAP:
encap = attr
}
}
@ -690,7 +880,7 @@ func deserializeRoute(m []byte) (Route, error) {
route.MultiPath = append(route.MultiPath, info)
rest = buf
}
case nl.RTA_NEWDST:
case unix.RTA_NEWDST:
var d Destination
switch msg.Family {
case nl.FAMILY_MPLS:
@ -700,9 +890,9 @@ func deserializeRoute(m []byte) (Route, error) {
return route, err
}
route.NewDst = d
case nl.RTA_ENCAP_TYPE:
case unix.RTA_ENCAP_TYPE:
encapType = attr
case nl.RTA_ENCAP:
case unix.RTA_ENCAP:
encap = attr
case unix.RTA_METRICS:
metrics, err := nl.ParseRouteAttr(attr.Value)
@ -715,6 +905,8 @@ func deserializeRoute(m []byte) (Route, error) {
route.MTU = int(native.Uint32(metric.Value[0:4]))
case unix.RTAX_ADVMSS:
route.AdvMSS = int(native.Uint32(metric.Value[0:4]))
case unix.RTAX_HOPLIMIT:
route.Hoplimit = int(native.Uint32(metric.Value[0:4]))
}
}
}
@ -734,6 +926,11 @@ func deserializeRoute(m []byte) (Route, error) {
if err := e.Decode(encap.Value); err != nil {
return route, err
}
case nl.LWTUNNEL_ENCAP_SEG6_LOCAL:
e = &SEG6LocalEncap{}
if err := e.Decode(encap.Value); err != nil {
return route, err
}
}
route.Encap = e
}
@ -840,13 +1037,19 @@ func routeSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- RouteUpdate, done <
go func() {
defer close(ch)
for {
msgs, err := s.Receive()
msgs, from, err := s.Receive()
if err != nil {
if cberr != nil {
cberr(err)
}
return
}
if from.Pid != nl.PidKernel {
if cberr != nil {
cberr(fmt.Errorf("Wrong sender portid %d, expected %d", from.Pid, nl.PidKernel))
}
continue
}
for _, m := range msgs {
if m.Header.Type == unix.NLMSG_DONE {
continue

View file

@ -144,7 +144,7 @@ func ruleHandle(rule *Rule, req *nl.NetlinkRequest) error {
req.AddData(nl.NewRtAttr(nl.FRA_OIFNAME, []byte(rule.OifName)))
}
if rule.Goto >= 0 {
msg.Type = nl.FR_ACT_NOP
msg.Type = nl.FR_ACT_GOTO
b := make([]byte, 4)
native.PutUint32(b, uint32(rule.Goto))
req.AddData(nl.NewRtAttr(nl.FRA_GOTO, b))

View file

@ -141,10 +141,13 @@ func SocketGet(local, remote net.Addr) (*Socket, error) {
},
})
s.Send(req)
msgs, err := s.Receive()
msgs, from, err := s.Receive()
if err != nil {
return nil, err
}
if from.Pid != nl.PidKernel {
return nil, fmt.Errorf("Wrong sender portid %d, expected %d", from.Pid, nl.PidKernel)
}
if len(msgs) == 0 {
return nil, errors.New("no message nor error from netlink")
}

View file

@ -54,11 +54,15 @@ func XfrmMonitor(ch chan<- XfrmMsg, done <-chan struct{}, errorChan chan<- error
go func() {
defer close(ch)
for {
msgs, err := s.Receive()
msgs, from, err := s.Receive()
if err != nil {
errorChan <- err
return
}
if from.Pid != nl.PidKernel {
errorChan <- fmt.Errorf("Wrong sender portid %d, expected %d", from.Pid, nl.PidKernel)
return
}
for _, m := range msgs {
switch m.Header.Type {
case nl.XFRM_MSG_EXPIRE:

View file

@ -35,6 +35,25 @@ func (d Dir) String() string {
return fmt.Sprintf("socket %d", d-XFRM_SOCKET_IN)
}
// PolicyAction is an enum representing an ipsec policy action.
type PolicyAction uint8
const (
XFRM_POLICY_ALLOW PolicyAction = 0
XFRM_POLICY_BLOCK PolicyAction = 1
)
func (a PolicyAction) String() string {
switch a {
case XFRM_POLICY_ALLOW:
return "allow"
case XFRM_POLICY_BLOCK:
return "block"
default:
return fmt.Sprintf("action %d", a)
}
}
// XfrmPolicyTmpl encapsulates a rule for the base addresses of an ipsec
// policy. These rules are matched with XfrmState to determine encryption
// and authentication algorithms.
@ -64,11 +83,14 @@ type XfrmPolicy struct {
Dir Dir
Priority int
Index int
Action PolicyAction
Ifindex int
Ifid 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)
return fmt.Sprintf("{Dst: %v, Src: %v, Proto: %s, DstPort: %d, SrcPort: %d, Dir: %s, Priority: %d, Index: %d, Action: %s, Ifindex: %d, Ifid: %d, Mark: %s, Tmpls: %s}",
p.Dst, p.Src, p.Proto, p.DstPort, p.SrcPort, p.Dir, p.Priority, p.Index, p.Action, p.Ifindex, p.Ifid, p.Mark, p.Tmpls)
}

View file

@ -27,6 +27,7 @@ func selFromPolicy(sel *nl.XfrmSelector, policy *XfrmPolicy) {
if sel.Sport != 0 {
sel.SportMask = ^uint16(0)
}
sel.Ifindex = int32(policy.Ifindex)
}
// XfrmPolicyAdd will add an xfrm policy to the system.
@ -61,6 +62,7 @@ func (h *Handle) xfrmPolicyAddOrUpdate(policy *XfrmPolicy, nlProto int) error {
msg.Priority = uint32(policy.Priority)
msg.Index = uint32(policy.Index)
msg.Dir = uint8(policy.Dir)
msg.Action = uint8(policy.Action)
msg.Lft.SoftByteLimit = nl.XFRM_INF
msg.Lft.HardByteLimit = nl.XFRM_INF
msg.Lft.SoftPacketLimit = nl.XFRM_INF
@ -90,6 +92,9 @@ func (h *Handle) xfrmPolicyAddOrUpdate(policy *XfrmPolicy, nlProto int) error {
req.AddData(out)
}
ifId := nl.NewRtAttr(nl.XFRMA_IF_ID, nl.Uint32Attr(uint32(policy.Ifid)))
req.AddData(ifId)
_, err := req.Execute(unix.NETLINK_XFRM, 0)
return err
}
@ -183,6 +188,9 @@ func (h *Handle) xfrmPolicyGetOrDelete(policy *XfrmPolicy, nlProto int) (*XfrmPo
req.AddData(out)
}
ifId := nl.NewRtAttr(nl.XFRMA_IF_ID, nl.Uint32Attr(uint32(policy.Ifid)))
req.AddData(ifId)
resType := nl.XFRM_MSG_NEWPOLICY
if nlProto == nl.XFRM_MSG_DELPOLICY {
resType = 0
@ -197,12 +205,7 @@ func (h *Handle) xfrmPolicyGetOrDelete(policy *XfrmPolicy, nlProto int) (*XfrmPo
return nil, err
}
p, err := parseXfrmPolicy(msgs[0], FAMILY_ALL)
if err != nil {
return nil, err
}
return p, nil
return parseXfrmPolicy(msgs[0], FAMILY_ALL)
}
func parseXfrmPolicy(m []byte, family int) (*XfrmPolicy, error) {
@ -220,9 +223,11 @@ func parseXfrmPolicy(m []byte, family int) (*XfrmPolicy, error) {
policy.Proto = Proto(msg.Sel.Proto)
policy.DstPort = int(nl.Swap16(msg.Sel.Dport))
policy.SrcPort = int(nl.Swap16(msg.Sel.Sport))
policy.Ifindex = int(msg.Sel.Ifindex)
policy.Priority = int(msg.Priority)
policy.Index = int(msg.Index)
policy.Dir = Dir(msg.Dir)
policy.Action = PolicyAction(msg.Action)
attrs, err := nl.ParseRouteAttr(m[msg.Len():])
if err != nil {
@ -249,6 +254,8 @@ func parseXfrmPolicy(m []byte, family int) (*XfrmPolicy, error) {
policy.Mark = new(XfrmMark)
policy.Mark.Value = mark.Value
policy.Mark.Mask = mark.Mask
case nl.XFRMA_IF_ID:
policy.Ifid = int(native.Uint32(attr.Value))
}
}

View file

@ -94,6 +94,8 @@ type XfrmState struct {
Limits XfrmStateLimits
Statistics XfrmStateStats
Mark *XfrmMark
OutputMark int
Ifid int
Auth *XfrmStateAlgo
Crypt *XfrmStateAlgo
Aead *XfrmStateAlgo
@ -102,8 +104,8 @@ type XfrmState struct {
}
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, ESN: %t",
sa.Dst, sa.Src, sa.Proto, sa.Mode, sa.Spi, sa.Reqid, sa.ReplayWindow, sa.Mark, sa.Auth, sa.Crypt, sa.Aead, sa.Encap, sa.ESN)
return fmt.Sprintf("Dst: %v, Src: %v, Proto: %s, Mode: %s, SPI: 0x%x, ReqID: 0x%x, ReplayWindow: %d, Mark: %v, OutputMark: %d, Ifid: %d, Auth: %v, Crypt: %v, Aead: %v, Encap: %v, ESN: %t",
sa.Dst, sa.Src, sa.Proto, sa.Mode, sa.Spi, sa.Reqid, sa.ReplayWindow, sa.Mark, sa.OutputMark, sa.Ifid, sa.Auth, sa.Crypt, sa.Aead, sa.Encap, sa.ESN)
}
func (sa XfrmState) Print(stats bool) string {
if !stats {

View file

@ -158,6 +158,13 @@ func (h *Handle) xfrmStateAddOrUpdate(state *XfrmState, nlProto int) error {
out := nl.NewRtAttr(nl.XFRMA_REPLAY_ESN_VAL, writeReplayEsn(state.ReplayWindow))
req.AddData(out)
}
if state.OutputMark != 0 {
out := nl.NewRtAttr(nl.XFRMA_OUTPUT_MARK, nl.Uint32Attr(uint32(state.OutputMark)))
req.AddData(out)
}
ifId := nl.NewRtAttr(nl.XFRMA_IF_ID, nl.Uint32Attr(uint32(state.Ifid)))
req.AddData(ifId)
_, err := req.Execute(unix.NETLINK_XFRM, 0)
return err
@ -184,12 +191,7 @@ func (h *Handle) xfrmStateAllocSpi(state *XfrmState) (*XfrmState, error) {
return nil, err
}
s, err := parseXfrmState(msgs[0], FAMILY_ALL)
if err != nil {
return nil, err
}
return s, err
return parseXfrmState(msgs[0], FAMILY_ALL)
}
// XfrmStateDel will delete an xfrm state from the system. Note that
@ -275,6 +277,9 @@ func (h *Handle) xfrmStateGetOrDelete(state *XfrmState, nlProto int) (*XfrmState
req.AddData(out)
}
ifId := nl.NewRtAttr(nl.XFRMA_IF_ID, nl.Uint32Attr(uint32(state.Ifid)))
req.AddData(ifId)
resType := nl.XFRM_MSG_NEWSA
if nlProto == nl.XFRM_MSG_DELSA {
resType = 0
@ -372,6 +377,10 @@ func parseXfrmState(m []byte, family int) (*XfrmState, error) {
state.Mark = new(XfrmMark)
state.Mark.Value = mark.Value
state.Mark.Mask = mark.Mask
case nl.XFRMA_OUTPUT_MARK:
state.OutputMark = int(native.Uint32(attr.Value))
case nl.XFRMA_IF_ID:
state.Ifid = int(native.Uint32(attr.Value))
}
}
@ -394,11 +403,7 @@ func (h *Handle) XfrmStateFlush(proto Proto) error {
req.AddData(&nl.XfrmUsersaFlush{Proto: uint8(proto)})
_, err := req.Execute(unix.NETLINK_XFRM, 0)
if err != nil {
return err
}
return nil
return err
}
func limitsToLft(lmts XfrmStateLimits, lft *nl.XfrmLifetimeCfg) {

View file

@ -0,0 +1,3 @@
module github.com/vishvananda/netns
go 1.12